diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/bc.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/bc.lua new file mode 100644 index 00000000..e58a3fef --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/bc.lua @@ -0,0 +1,190 @@ +---------------------------------------------------------------------------- +-- LuaJIT bytecode listing module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module lists the bytecode of a Lua function. If it's loaded by -jbc +-- it hooks into the parser and lists all functions of a chunk as they +-- are parsed. +-- +-- Example usage: +-- +-- luajit -jbc -e 'local x=0; for i=1,1e6 do x=x+i end; print(x)' +-- luajit -jbc=- foo.lua +-- luajit -jbc=foo.list foo.lua +-- +-- Default output is to stderr. To redirect the output to a file, pass a +-- filename as an argument (use '-' for stdout) or set the environment +-- variable LUAJIT_LISTFILE. The file is overwritten every time the module +-- is started. +-- +-- This module can also be used programmatically: +-- +-- local bc = require("jit.bc") +-- +-- local function foo() print("hello") end +-- +-- bc.dump(foo) --> -- BYTECODE -- [...] +-- print(bc.line(foo, 2)) --> 0002 KSTR 1 1 ; "hello" +-- +-- local out = { +-- -- Do something with each line: +-- write = function(t, ...) io.write(...) end, +-- close = function(t) end, +-- flush = function(t) end, +-- } +-- bc.dump(foo, out) +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local bit = require("bit") +local sub, gsub, format = string.sub, string.gsub, string.format +local byte, band, shr = string.byte, bit.band, bit.rshift +local funcinfo, funcbc, funck = jutil.funcinfo, jutil.funcbc, jutil.funck +local funcuvname = jutil.funcuvname +local bcnames = vmdef.bcnames +local stdout, stderr = io.stdout, io.stderr + +------------------------------------------------------------------------------ + +local function ctlsub(c) + if c == "\n" then return "\\n" + elseif c == "\r" then return "\\r" + elseif c == "\t" then return "\\t" + else return format("\\%03d", byte(c)) + end +end + +-- Return one bytecode line. +local function bcline(func, pc, prefix) + local ins, m = funcbc(func, pc) + if not ins then return end + local ma, mb, mc = band(m, 7), band(m, 15*8), band(m, 15*128) + local a = band(shr(ins, 8), 0xff) + local oidx = 6*band(ins, 0xff) + local op = sub(bcnames, oidx+1, oidx+6) + local s = format("%04d %s %-6s %3s ", + pc, prefix or " ", op, ma == 0 and "" or a) + local d = shr(ins, 16) + if mc == 13*128 then -- BCMjump + return format("%s=> %04d\n", s, pc+d-0x7fff) + end + if mb ~= 0 then + d = band(d, 0xff) + elseif mc == 0 then + return s.."\n" + end + local kc + if mc == 10*128 then -- BCMstr + kc = funck(func, -d-1) + kc = format(#kc > 40 and '"%.40s"~' or '"%s"', gsub(kc, "%c", ctlsub)) + elseif mc == 9*128 then -- BCMnum + kc = funck(func, d) + if op == "TSETM " then kc = kc - 2^52 end + elseif mc == 12*128 then -- BCMfunc + local fi = funcinfo(funck(func, -d-1)) + if fi.ffid then + kc = vmdef.ffnames[fi.ffid] + else + kc = fi.loc + end + elseif mc == 5*128 then -- BCMuv + kc = funcuvname(func, d) + end + if ma == 5 then -- BCMuv + local ka = funcuvname(func, a) + if kc then kc = ka.." ; "..kc else kc = ka end + end + if mb ~= 0 then + local b = shr(ins, 24) + if kc then return format("%s%3d %3d ; %s\n", s, b, d, kc) end + return format("%s%3d %3d\n", s, b, d) + end + if kc then return format("%s%3d ; %s\n", s, d, kc) end + if mc == 7*128 and d > 32767 then d = d - 65536 end -- BCMlits + return format("%s%3d\n", s, d) +end + +-- Collect branch targets of a function. +local function bctargets(func) + local target = {} + for pc=1,1000000000 do + local ins, m = funcbc(func, pc) + if not ins then break end + if band(m, 15*128) == 13*128 then target[pc+shr(ins, 16)-0x7fff] = true end + end + return target +end + +-- Dump bytecode instructions of a function. +local function bcdump(func, out, all) + if not out then out = stdout end + local fi = funcinfo(func) + if all and fi.children then + for n=-1,-1000000000,-1 do + local k = funck(func, n) + if not k then break end + if type(k) == "proto" then bcdump(k, out, true) end + end + end + out:write(format("-- BYTECODE -- %s-%d\n", fi.loc, fi.lastlinedefined)) + local target = bctargets(func) + for pc=1,1000000000 do + local s = bcline(func, pc, target[pc] and "=>") + if not s then break end + out:write(s) + end + out:write("\n") + out:flush() +end + +------------------------------------------------------------------------------ + +-- Active flag and output file handle. +local active, out + +-- List handler. +local function h_list(func) + return bcdump(func, out) +end + +-- Detach list handler. +local function bclistoff() + if active then + active = false + jit.attach(h_list) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach list handler. +local function bcliston(outfile) + if active then bclistoff() end + if not outfile then outfile = os.getenv("LUAJIT_LISTFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stderr + end + jit.attach(h_list, "bc") + active = true +end + +-- Public module functions. +return { + line = bcline, + dump = bcdump, + targets = bctargets, + on = bcliston, + off = bclistoff, + start = bcliston -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/bcsave.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/bcsave.lua new file mode 100644 index 00000000..ab13667a --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/bcsave.lua @@ -0,0 +1,665 @@ +---------------------------------------------------------------------------- +-- LuaJIT module to save/list bytecode. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module saves or lists the bytecode for an input file. +-- It's run by the -b command line option. +-- +------------------------------------------------------------------------------ + +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local bit = require("bit") + +-- Symbol name prefix for LuaJIT bytecode. +local LJBC_PREFIX = "luaJIT_BC_" + +local type, assert = type, assert +local format = string.format +local tremove, tconcat = table.remove, table.concat + +------------------------------------------------------------------------------ + +local function usage() + io.stderr:write[[ +Save LuaJIT bytecode: luajit -b[options] input output + -l Only list bytecode. + -s Strip debug info (default). + -g Keep debug info. + -n name Set module name (default: auto-detect from input name). + -t type Set output file type (default: auto-detect from output name). + -a arch Override architecture for object files (default: native). + -o os Override OS for object files (default: native). + -e chunk Use chunk string as input. + -- Stop handling options. + - Use stdin as input and/or stdout as output. + +File types: c h obj o raw (default) +]] + os.exit(1) +end + +local function check(ok, ...) + if ok then return ok, ... end + io.stderr:write("luajit: ", ...) + io.stderr:write("\n") + os.exit(1) +end + +local function readfile(input) + if type(input) == "function" then return input end + if input == "-" then input = nil end + return check(loadfile(input)) +end + +local function savefile(name, mode) + if name == "-" then return io.stdout end + return check(io.open(name, mode)) +end + +------------------------------------------------------------------------------ + +local map_type = { + raw = "raw", c = "c", h = "h", o = "obj", obj = "obj", +} + +local map_arch = { + x86 = { e = "le", b = 32, m = 3, p = 0x14c, }, + x64 = { e = "le", b = 64, m = 62, p = 0x8664, }, + arm = { e = "le", b = 32, m = 40, p = 0x1c0, }, + arm64 = { e = "le", b = 64, m = 183, p = 0xaa64, }, + arm64be = { e = "be", b = 64, m = 183, }, + ppc = { e = "be", b = 32, m = 20, }, + mips = { e = "be", b = 32, m = 8, f = 0x50001006, }, + mipsel = { e = "le", b = 32, m = 8, f = 0x50001006, }, + mips64 = { e = "be", b = 64, m = 8, f = 0x80000007, }, + mips64el = { e = "le", b = 64, m = 8, f = 0x80000007, }, + mips64r6 = { e = "be", b = 64, m = 8, f = 0xa0000407, }, + mips64r6el = { e = "le", b = 64, m = 8, f = 0xa0000407, }, +} + +local map_os = { + linux = true, windows = true, osx = true, freebsd = true, netbsd = true, + openbsd = true, dragonfly = true, solaris = true, +} + +local function checkarg(str, map, err) + str = str:lower() + local s = check(map[str], "unknown ", err) + return type(s) == "string" and s or str +end + +local function detecttype(str) + local ext = str:lower():match("%.(%a+)$") + return map_type[ext] or "raw" +end + +local function checkmodname(str) + check(str:match("^[%w_.%-]+$"), "bad module name") + return str:gsub("[%.%-]", "_") +end + +local function detectmodname(str) + if type(str) == "string" then + local tail = str:match("[^/\\]+$") + if tail then str = tail end + local head = str:match("^(.*)%.[^.]*$") + if head then str = head end + str = str:match("^[%w_.%-]+") + else + str = nil + end + check(str, "cannot derive module name, use -n name") + return str:gsub("[%.%-]", "_") +end + +------------------------------------------------------------------------------ + +local function bcsave_tail(fp, output, s) + local ok, err = fp:write(s) + if ok and output ~= "-" then ok, err = fp:close() end + check(ok, "cannot write ", output, ": ", err) +end + +local function bcsave_raw(output, s) + local fp = savefile(output, "wb") + bcsave_tail(fp, output, s) +end + +local function bcsave_c(ctx, output, s) + local fp = savefile(output, "w") + if ctx.type == "c" then + fp:write(format([[ +#ifdef __cplusplus +extern "C" +#endif +#ifdef _WIN32 +__declspec(dllexport) +#endif +const unsigned char %s%s[] = { +]], LJBC_PREFIX, ctx.modname)) + else + fp:write(format([[ +#define %s%s_SIZE %d +static const unsigned char %s%s[] = { +]], LJBC_PREFIX, ctx.modname, #s, LJBC_PREFIX, ctx.modname)) + end + local t, n, m = {}, 0, 0 + for i=1,#s do + local b = tostring(string.byte(s, i)) + m = m + #b + 1 + if m > 78 then + fp:write(tconcat(t, ",", 1, n), ",\n") + n, m = 0, #b + 1 + end + n = n + 1 + t[n] = b + end + bcsave_tail(fp, output, tconcat(t, ",", 1, n).."\n};\n") +end + +local function bcsave_elfobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct { + uint8_t emagic[4], eclass, eendian, eversion, eosabi, eabiversion, epad[7]; + uint16_t type, machine; + uint32_t version; + uint32_t entry, phofs, shofs; + uint32_t flags; + uint16_t ehsize, phentsize, phnum, shentsize, shnum, shstridx; +} ELF32header; +typedef struct { + uint8_t emagic[4], eclass, eendian, eversion, eosabi, eabiversion, epad[7]; + uint16_t type, machine; + uint32_t version; + uint64_t entry, phofs, shofs; + uint32_t flags; + uint16_t ehsize, phentsize, phnum, shentsize, shnum, shstridx; +} ELF64header; +typedef struct { + uint32_t name, type, flags, addr, ofs, size, link, info, align, entsize; +} ELF32sectheader; +typedef struct { + uint32_t name, type; + uint64_t flags, addr, ofs, size; + uint32_t link, info; + uint64_t align, entsize; +} ELF64sectheader; +typedef struct { + uint32_t name, value, size; + uint8_t info, other; + uint16_t sectidx; +} ELF32symbol; +typedef struct { + uint32_t name; + uint8_t info, other; + uint16_t sectidx; + uint64_t value, size; +} ELF64symbol; +typedef struct { + ELF32header hdr; + ELF32sectheader sect[6]; + ELF32symbol sym[2]; + uint8_t space[4096]; +} ELF32obj; +typedef struct { + ELF64header hdr; + ELF64sectheader sect[6]; + ELF64symbol sym[2]; + uint8_t space[4096]; +} ELF64obj; +]] + local symname = LJBC_PREFIX..ctx.modname + local ai = assert(map_arch[ctx.arch]) + local is64, isbe = ai.b == 64, ai.e == "be" + + -- Handle different host/target endianess. + local function f32(x) return x end + local f16, fofs = f32, f32 + if ffi.abi("be") ~= isbe then + f32 = bit.bswap + function f16(x) return bit.rshift(bit.bswap(x), 16) end + if is64 then + local two32 = ffi.cast("int64_t", 2^32) + function fofs(x) return bit.bswap(x)*two32 end + else + fofs = f32 + end + end + + -- Create ELF object and fill in header. + local o = ffi.new(is64 and "ELF64obj" or "ELF32obj") + local hdr = o.hdr + if ctx.os == "bsd" or ctx.os == "other" then -- Determine native hdr.eosabi. + local bf = assert(io.open("/bin/ls", "rb")) + local bs = bf:read(9) + bf:close() + ffi.copy(o, bs, 9) + check(hdr.emagic[0] == 127, "no support for writing native object files") + else + hdr.emagic = "\127ELF" + hdr.eosabi = ({ freebsd=9, netbsd=2, openbsd=12, solaris=6 })[ctx.os] or 0 + end + hdr.eclass = is64 and 2 or 1 + hdr.eendian = isbe and 2 or 1 + hdr.eversion = 1 + hdr.type = f16(1) + hdr.machine = f16(ai.m) + hdr.flags = f32(ai.f or 0) + hdr.version = f32(1) + hdr.shofs = fofs(ffi.offsetof(o, "sect")) + hdr.ehsize = f16(ffi.sizeof(hdr)) + hdr.shentsize = f16(ffi.sizeof(o.sect[0])) + hdr.shnum = f16(6) + hdr.shstridx = f16(2) + + -- Fill in sections and symbols. + local sofs, ofs = ffi.offsetof(o, "space"), 1 + for i,name in ipairs{ + ".symtab", ".shstrtab", ".strtab", ".rodata", ".note.GNU-stack", + } do + local sect = o.sect[i] + sect.align = fofs(1) + sect.name = f32(ofs) + ffi.copy(o.space+ofs, name) + ofs = ofs + #name+1 + end + o.sect[1].type = f32(2) -- .symtab + o.sect[1].link = f32(3) + o.sect[1].info = f32(1) + o.sect[1].align = fofs(8) + o.sect[1].ofs = fofs(ffi.offsetof(o, "sym")) + o.sect[1].entsize = fofs(ffi.sizeof(o.sym[0])) + o.sect[1].size = fofs(ffi.sizeof(o.sym)) + o.sym[1].name = f32(1) + o.sym[1].sectidx = f16(4) + o.sym[1].size = fofs(#s) + o.sym[1].info = 17 + o.sect[2].type = f32(3) -- .shstrtab + o.sect[2].ofs = fofs(sofs) + o.sect[2].size = fofs(ofs) + o.sect[3].type = f32(3) -- .strtab + o.sect[3].ofs = fofs(sofs + ofs) + o.sect[3].size = fofs(#symname+2) + ffi.copy(o.space+ofs+1, symname) + ofs = ofs + #symname + 2 + o.sect[4].type = f32(1) -- .rodata + o.sect[4].flags = fofs(2) + o.sect[4].ofs = fofs(sofs + ofs) + o.sect[4].size = fofs(#s) + o.sect[5].type = f32(1) -- .note.GNU-stack + o.sect[5].ofs = fofs(sofs + ofs + #s) + + -- Write ELF object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, ffi.sizeof(o)-4096+ofs)) + bcsave_tail(fp, output, s) +end + +local function bcsave_peobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct { + uint16_t arch, nsects; + uint32_t time, symtabofs, nsyms; + uint16_t opthdrsz, flags; +} PEheader; +typedef struct { + char name[8]; + uint32_t vsize, vaddr, size, ofs, relocofs, lineofs; + uint16_t nreloc, nline; + uint32_t flags; +} PEsection; +typedef struct __attribute((packed)) { + union { + char name[8]; + uint32_t nameref[2]; + }; + uint32_t value; + int16_t sect; + uint16_t type; + uint8_t scl, naux; +} PEsym; +typedef struct __attribute((packed)) { + uint32_t size; + uint16_t nreloc, nline; + uint32_t cksum; + uint16_t assoc; + uint8_t comdatsel, unused[3]; +} PEsymaux; +typedef struct { + PEheader hdr; + PEsection sect[2]; + // Must be an even number of symbol structs. + PEsym sym0; + PEsymaux sym0aux; + PEsym sym1; + PEsymaux sym1aux; + PEsym sym2; + PEsym sym3; + uint32_t strtabsize; + uint8_t space[4096]; +} PEobj; +]] + local symname = LJBC_PREFIX..ctx.modname + local ai = assert(map_arch[ctx.arch]) + local is64 = ai.b == 64 + local symexport = " /EXPORT:"..symname..",DATA " + + -- The file format is always little-endian. Swap if the host is big-endian. + local function f32(x) return x end + local f16 = f32 + if ffi.abi("be") then + f32 = bit.bswap + function f16(x) return bit.rshift(bit.bswap(x), 16) end + end + + -- Create PE object and fill in header. + local o = ffi.new("PEobj") + local hdr = o.hdr + hdr.arch = f16(assert(ai.p)) + hdr.nsects = f16(2) + hdr.symtabofs = f32(ffi.offsetof(o, "sym0")) + hdr.nsyms = f32(6) + + -- Fill in sections and symbols. + o.sect[0].name = ".drectve" + o.sect[0].size = f32(#symexport) + o.sect[0].flags = f32(0x00100a00) + o.sym0.sect = f16(1) + o.sym0.scl = 3 + o.sym0.name = ".drectve" + o.sym0.naux = 1 + o.sym0aux.size = f32(#symexport) + o.sect[1].name = ".rdata" + o.sect[1].size = f32(#s) + o.sect[1].flags = f32(0x40300040) + o.sym1.sect = f16(2) + o.sym1.scl = 3 + o.sym1.name = ".rdata" + o.sym1.naux = 1 + o.sym1aux.size = f32(#s) + o.sym2.sect = f16(2) + o.sym2.scl = 2 + o.sym2.nameref[1] = f32(4) + o.sym3.sect = f16(-1) + o.sym3.scl = 2 + o.sym3.value = f32(1) + o.sym3.name = "@feat.00" -- Mark as SafeSEH compliant. + ffi.copy(o.space, symname) + local ofs = #symname + 1 + o.strtabsize = f32(ofs + 4) + o.sect[0].ofs = f32(ffi.offsetof(o, "space") + ofs) + ffi.copy(o.space + ofs, symexport) + ofs = ofs + #symexport + o.sect[1].ofs = f32(ffi.offsetof(o, "space") + ofs) + + -- Write PE object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, ffi.sizeof(o)-4096+ofs)) + bcsave_tail(fp, output, s) +end + +local function bcsave_machobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct +{ + uint32_t magic, cputype, cpusubtype, filetype, ncmds, sizeofcmds, flags; +} mach_header; +typedef struct +{ + mach_header; uint32_t reserved; +} mach_header_64; +typedef struct { + uint32_t cmd, cmdsize; + char segname[16]; + uint32_t vmaddr, vmsize, fileoff, filesize; + uint32_t maxprot, initprot, nsects, flags; +} mach_segment_command; +typedef struct { + uint32_t cmd, cmdsize; + char segname[16]; + uint64_t vmaddr, vmsize, fileoff, filesize; + uint32_t maxprot, initprot, nsects, flags; +} mach_segment_command_64; +typedef struct { + char sectname[16], segname[16]; + uint32_t addr, size; + uint32_t offset, align, reloff, nreloc, flags; + uint32_t reserved1, reserved2; +} mach_section; +typedef struct { + char sectname[16], segname[16]; + uint64_t addr, size; + uint32_t offset, align, reloff, nreloc, flags; + uint32_t reserved1, reserved2, reserved3; +} mach_section_64; +typedef struct { + uint32_t cmd, cmdsize, symoff, nsyms, stroff, strsize; +} mach_symtab_command; +typedef struct { + int32_t strx; + uint8_t type, sect; + int16_t desc; + uint32_t value; +} mach_nlist; +typedef struct { + uint32_t strx; + uint8_t type, sect; + uint16_t desc; + uint64_t value; +} mach_nlist_64; +typedef struct +{ + uint32_t magic, nfat_arch; +} mach_fat_header; +typedef struct +{ + uint32_t cputype, cpusubtype, offset, size, align; +} mach_fat_arch; +typedef struct { + struct { + mach_header hdr; + mach_segment_command seg; + mach_section sec; + mach_symtab_command sym; + } arch[1]; + mach_nlist sym_entry; + uint8_t space[4096]; +} mach_obj; +typedef struct { + struct { + mach_header_64 hdr; + mach_segment_command_64 seg; + mach_section_64 sec; + mach_symtab_command sym; + } arch[1]; + mach_nlist_64 sym_entry; + uint8_t space[4096]; +} mach_obj_64; +typedef struct { + mach_fat_header fat; + mach_fat_arch fat_arch[2]; + struct { + mach_header hdr; + mach_segment_command seg; + mach_section sec; + mach_symtab_command sym; + } arch[2]; + mach_nlist sym_entry; + uint8_t space[4096]; +} mach_fat_obj; +]] + local symname = '_'..LJBC_PREFIX..ctx.modname + local isfat, is64, align, mobj = false, false, 4, "mach_obj" + if ctx.arch == "x64" then + is64, align, mobj = true, 8, "mach_obj_64" + elseif ctx.arch == "arm" then + isfat, mobj = true, "mach_fat_obj" + elseif ctx.arch == "arm64" then + is64, align, isfat, mobj = true, 8, true, "mach_fat_obj" + else + check(ctx.arch == "x86", "unsupported architecture for OSX") + end + local function aligned(v, a) return bit.band(v+a-1, -a) end + local be32 = bit.bswap -- Mach-O FAT is BE, supported archs are LE. + + -- Create Mach-O object and fill in header. + local o = ffi.new(mobj) + local mach_size = aligned(ffi.offsetof(o, "space")+#symname+2, align) + local cputype = ({ x86={7}, x64={0x01000007}, arm={7,12}, arm64={0x01000007,0x0100000c} })[ctx.arch] + local cpusubtype = ({ x86={3}, x64={3}, arm={3,9}, arm64={3,0} })[ctx.arch] + if isfat then + o.fat.magic = be32(0xcafebabe) + o.fat.nfat_arch = be32(#cpusubtype) + end + + -- Fill in sections and symbols. + for i=0,#cpusubtype-1 do + local ofs = 0 + if isfat then + local a = o.fat_arch[i] + a.cputype = be32(cputype[i+1]) + a.cpusubtype = be32(cpusubtype[i+1]) + -- Subsequent slices overlap each other to share data. + ofs = ffi.offsetof(o, "arch") + i*ffi.sizeof(o.arch[0]) + a.offset = be32(ofs) + a.size = be32(mach_size-ofs+#s) + end + local a = o.arch[i] + a.hdr.magic = is64 and 0xfeedfacf or 0xfeedface + a.hdr.cputype = cputype[i+1] + a.hdr.cpusubtype = cpusubtype[i+1] + a.hdr.filetype = 1 + a.hdr.ncmds = 2 + a.hdr.sizeofcmds = ffi.sizeof(a.seg)+ffi.sizeof(a.sec)+ffi.sizeof(a.sym) + a.seg.cmd = is64 and 0x19 or 0x1 + a.seg.cmdsize = ffi.sizeof(a.seg)+ffi.sizeof(a.sec) + a.seg.vmsize = #s + a.seg.fileoff = mach_size-ofs + a.seg.filesize = #s + a.seg.maxprot = 1 + a.seg.initprot = 1 + a.seg.nsects = 1 + ffi.copy(a.sec.sectname, "__data") + ffi.copy(a.sec.segname, "__DATA") + a.sec.size = #s + a.sec.offset = mach_size-ofs + a.sym.cmd = 2 + a.sym.cmdsize = ffi.sizeof(a.sym) + a.sym.symoff = ffi.offsetof(o, "sym_entry")-ofs + a.sym.nsyms = 1 + a.sym.stroff = ffi.offsetof(o, "sym_entry")+ffi.sizeof(o.sym_entry)-ofs + a.sym.strsize = aligned(#symname+2, align) + end + o.sym_entry.type = 0xf + o.sym_entry.sect = 1 + o.sym_entry.strx = 1 + ffi.copy(o.space+1, symname) + + -- Write Macho-O object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, mach_size)) + bcsave_tail(fp, output, s) +end + +local function bcsave_obj(ctx, output, s) + local ok, ffi = pcall(require, "ffi") + check(ok, "FFI library required to write this file type") + if ctx.os == "windows" then + return bcsave_peobj(ctx, output, s, ffi) + elseif ctx.os == "osx" then + return bcsave_machobj(ctx, output, s, ffi) + else + return bcsave_elfobj(ctx, output, s, ffi) + end +end + +------------------------------------------------------------------------------ + +local function bclist(input, output) + local f = readfile(input) + require("jit.bc").dump(f, savefile(output, "w"), true) +end + +local function bcsave(ctx, input, output) + local f = readfile(input) + local s = string.dump(f, ctx.strip) + local t = ctx.type + if not t then + t = detecttype(output) + ctx.type = t + end + if t == "raw" then + bcsave_raw(output, s) + else + if not ctx.modname then ctx.modname = detectmodname(input) end + if t == "obj" then + bcsave_obj(ctx, output, s) + else + bcsave_c(ctx, output, s) + end + end +end + +local function docmd(...) + local arg = {...} + local n = 1 + local list = false + local ctx = { + strip = true, arch = jit.arch, os = jit.os:lower(), + type = false, modname = false, + } + while n <= #arg do + local a = arg[n] + if type(a) == "string" and a:sub(1, 1) == "-" and a ~= "-" then + tremove(arg, n) + if a == "--" then break end + for m=2,#a do + local opt = a:sub(m, m) + if opt == "l" then + list = true + elseif opt == "s" then + ctx.strip = true + elseif opt == "g" then + ctx.strip = false + else + if arg[n] == nil or m ~= #a then usage() end + if opt == "e" then + if n ~= 1 then usage() end + arg[1] = check(loadstring(arg[1])) + elseif opt == "n" then + ctx.modname = checkmodname(tremove(arg, n)) + elseif opt == "t" then + ctx.type = checkarg(tremove(arg, n), map_type, "file type") + elseif opt == "a" then + ctx.arch = checkarg(tremove(arg, n), map_arch, "architecture") + elseif opt == "o" then + ctx.os = checkarg(tremove(arg, n), map_os, "OS name") + else + usage() + end + end + end + else + n = n + 1 + end + end + if list then + if #arg == 0 or #arg > 2 then usage() end + bclist(arg[1], arg[2] or "-") + else + if #arg ~= 2 then usage() end + bcsave(ctx, arg[1], arg[2]) + end +end + +------------------------------------------------------------------------------ + +-- Public module functions. +return { + start = docmd -- Process -b command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm.lua new file mode 100644 index 00000000..ba79c47e --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm.lua @@ -0,0 +1,689 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles most user-mode ARMv7 instructions +-- NYI: Advanced SIMD and VFP instructions. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch = string.match, string.gmatch +local concat = table.concat +local bit = require("bit") +local band, bor, ror, tohex = bit.band, bit.bor, bit.ror, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Opcode maps +------------------------------------------------------------------------------ + +local map_loadc = { + shift = 8, mask = 15, + [10] = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "vmovFmDN", "vstmFNdr", + _ = { + shift = 21, mask = 1, + [0] = "vstrFdl", + { shift = 16, mask = 15, [13] = "vpushFdr", _ = "vstmdbFNdr", } + }, + }, + { + shift = 23, mask = 3, + [0] = "vmovFDNm", + { shift = 16, mask = 15, [13] = "vpopFdr", _ = "vldmFNdr", }, + _ = { + shift = 21, mask = 1, + [0] = "vldrFdl", "vldmdbFNdr", + }, + }, + }, + [11] = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "vmovGmDN", "vstmGNdr", + _ = { + shift = 21, mask = 1, + [0] = "vstrGdl", + { shift = 16, mask = 15, [13] = "vpushGdr", _ = "vstmdbGNdr", } + }, + }, + { + shift = 23, mask = 3, + [0] = "vmovGDNm", + { shift = 16, mask = 15, [13] = "vpopGdr", _ = "vldmGNdr", }, + _ = { + shift = 21, mask = 1, + [0] = "vldrGdl", "vldmdbGNdr", + }, + }, + }, + _ = { + shift = 0, mask = 0 -- NYI ldc, mcrr, mrrc. + }, +} + +local map_vfps = { + shift = 6, mask = 0x2c001, + [0] = "vmlaF.dnm", "vmlsF.dnm", + [0x04000] = "vnmlsF.dnm", [0x04001] = "vnmlaF.dnm", + [0x08000] = "vmulF.dnm", [0x08001] = "vnmulF.dnm", + [0x0c000] = "vaddF.dnm", [0x0c001] = "vsubF.dnm", + [0x20000] = "vdivF.dnm", + [0x24000] = "vfnmsF.dnm", [0x24001] = "vfnmaF.dnm", + [0x28000] = "vfmaF.dnm", [0x28001] = "vfmsF.dnm", + [0x2c000] = "vmovF.dY", + [0x2c001] = { + shift = 7, mask = 0x1e01, + [0] = "vmovF.dm", "vabsF.dm", + [0x0200] = "vnegF.dm", [0x0201] = "vsqrtF.dm", + [0x0800] = "vcmpF.dm", [0x0801] = "vcmpeF.dm", + [0x0a00] = "vcmpzF.d", [0x0a01] = "vcmpzeF.d", + [0x0e01] = "vcvtG.dF.m", + [0x1000] = "vcvt.f32.u32Fdm", [0x1001] = "vcvt.f32.s32Fdm", + [0x1800] = "vcvtr.u32F.dm", [0x1801] = "vcvt.u32F.dm", + [0x1a00] = "vcvtr.s32F.dm", [0x1a01] = "vcvt.s32F.dm", + }, +} + +local map_vfpd = { + shift = 6, mask = 0x2c001, + [0] = "vmlaG.dnm", "vmlsG.dnm", + [0x04000] = "vnmlsG.dnm", [0x04001] = "vnmlaG.dnm", + [0x08000] = "vmulG.dnm", [0x08001] = "vnmulG.dnm", + [0x0c000] = "vaddG.dnm", [0x0c001] = "vsubG.dnm", + [0x20000] = "vdivG.dnm", + [0x24000] = "vfnmsG.dnm", [0x24001] = "vfnmaG.dnm", + [0x28000] = "vfmaG.dnm", [0x28001] = "vfmsG.dnm", + [0x2c000] = "vmovG.dY", + [0x2c001] = { + shift = 7, mask = 0x1e01, + [0] = "vmovG.dm", "vabsG.dm", + [0x0200] = "vnegG.dm", [0x0201] = "vsqrtG.dm", + [0x0800] = "vcmpG.dm", [0x0801] = "vcmpeG.dm", + [0x0a00] = "vcmpzG.d", [0x0a01] = "vcmpzeG.d", + [0x0e01] = "vcvtF.dG.m", + [0x1000] = "vcvt.f64.u32GdFm", [0x1001] = "vcvt.f64.s32GdFm", + [0x1800] = "vcvtr.u32FdG.m", [0x1801] = "vcvt.u32FdG.m", + [0x1a00] = "vcvtr.s32FdG.m", [0x1a01] = "vcvt.s32FdG.m", + }, +} + +local map_datac = { + shift = 24, mask = 1, + [0] = { + shift = 4, mask = 1, + [0] = { + shift = 8, mask = 15, + [10] = map_vfps, + [11] = map_vfpd, + -- NYI cdp, mcr, mrc. + }, + { + shift = 8, mask = 15, + [10] = { + shift = 20, mask = 15, + [0] = "vmovFnD", "vmovFDn", + [14] = "vmsrD", + [15] = { shift = 12, mask = 15, [15] = "vmrs", _ = "vmrsD", }, + }, + }, + }, + "svcT", +} + +local map_loadcu = { + shift = 0, mask = 0, -- NYI unconditional CP load/store. +} + +local map_datacu = { + shift = 0, mask = 0, -- NYI unconditional CP data. +} + +local map_simddata = { + shift = 0, mask = 0, -- NYI SIMD data. +} + +local map_simdload = { + shift = 0, mask = 0, -- NYI SIMD load/store, preload. +} + +local map_preload = { + shift = 0, mask = 0, -- NYI preload. +} + +local map_media = { + shift = 20, mask = 31, + [0] = false, + { --01 + shift = 5, mask = 7, + [0] = "sadd16DNM", "sasxDNM", "ssaxDNM", "ssub16DNM", + "sadd8DNM", false, false, "ssub8DNM", + }, + { --02 + shift = 5, mask = 7, + [0] = "qadd16DNM", "qasxDNM", "qsaxDNM", "qsub16DNM", + "qadd8DNM", false, false, "qsub8DNM", + }, + { --03 + shift = 5, mask = 7, + [0] = "shadd16DNM", "shasxDNM", "shsaxDNM", "shsub16DNM", + "shadd8DNM", false, false, "shsub8DNM", + }, + false, + { --05 + shift = 5, mask = 7, + [0] = "uadd16DNM", "uasxDNM", "usaxDNM", "usub16DNM", + "uadd8DNM", false, false, "usub8DNM", + }, + { --06 + shift = 5, mask = 7, + [0] = "uqadd16DNM", "uqasxDNM", "uqsaxDNM", "uqsub16DNM", + "uqadd8DNM", false, false, "uqsub8DNM", + }, + { --07 + shift = 5, mask = 7, + [0] = "uhadd16DNM", "uhasxDNM", "uhsaxDNM", "uhsub16DNM", + "uhadd8DNM", false, false, "uhsub8DNM", + }, + { --08 + shift = 5, mask = 7, + [0] = "pkhbtDNMU", false, "pkhtbDNMU", + { shift = 16, mask = 15, [15] = "sxtb16DMU", _ = "sxtab16DNMU", }, + "pkhbtDNMU", "selDNM", "pkhtbDNMU", + }, + false, + { --0a + shift = 5, mask = 7, + [0] = "ssatDxMu", "ssat16DxM", "ssatDxMu", + { shift = 16, mask = 15, [15] = "sxtbDMU", _ = "sxtabDNMU", }, + "ssatDxMu", false, "ssatDxMu", + }, + { --0b + shift = 5, mask = 7, + [0] = "ssatDxMu", "revDM", "ssatDxMu", + { shift = 16, mask = 15, [15] = "sxthDMU", _ = "sxtahDNMU", }, + "ssatDxMu", "rev16DM", "ssatDxMu", + }, + { --0c + shift = 5, mask = 7, + [3] = { shift = 16, mask = 15, [15] = "uxtb16DMU", _ = "uxtab16DNMU", }, + }, + false, + { --0e + shift = 5, mask = 7, + [0] = "usatDwMu", "usat16DwM", "usatDwMu", + { shift = 16, mask = 15, [15] = "uxtbDMU", _ = "uxtabDNMU", }, + "usatDwMu", false, "usatDwMu", + }, + { --0f + shift = 5, mask = 7, + [0] = "usatDwMu", "rbitDM", "usatDwMu", + { shift = 16, mask = 15, [15] = "uxthDMU", _ = "uxtahDNMU", }, + "usatDwMu", "revshDM", "usatDwMu", + }, + { --10 + shift = 12, mask = 15, + [15] = { + shift = 5, mask = 7, + "smuadNMS", "smuadxNMS", "smusdNMS", "smusdxNMS", + }, + _ = { + shift = 5, mask = 7, + [0] = "smladNMSD", "smladxNMSD", "smlsdNMSD", "smlsdxNMSD", + }, + }, + false, false, false, + { --14 + shift = 5, mask = 7, + [0] = "smlaldDNMS", "smlaldxDNMS", "smlsldDNMS", "smlsldxDNMS", + }, + { --15 + shift = 5, mask = 7, + [0] = { shift = 12, mask = 15, [15] = "smmulNMS", _ = "smmlaNMSD", }, + { shift = 12, mask = 15, [15] = "smmulrNMS", _ = "smmlarNMSD", }, + false, false, false, false, + "smmlsNMSD", "smmlsrNMSD", + }, + false, false, + { --18 + shift = 5, mask = 7, + [0] = { shift = 12, mask = 15, [15] = "usad8NMS", _ = "usada8NMSD", }, + }, + false, + { --1a + shift = 5, mask = 3, [2] = "sbfxDMvw", + }, + { --1b + shift = 5, mask = 3, [2] = "sbfxDMvw", + }, + { --1c + shift = 5, mask = 3, + [0] = { shift = 0, mask = 15, [15] = "bfcDvX", _ = "bfiDMvX", }, + }, + { --1d + shift = 5, mask = 3, + [0] = { shift = 0, mask = 15, [15] = "bfcDvX", _ = "bfiDMvX", }, + }, + { --1e + shift = 5, mask = 3, [2] = "ubfxDMvw", + }, + { --1f + shift = 5, mask = 3, [2] = "ubfxDMvw", + }, +} + +local map_load = { + shift = 21, mask = 9, + { + shift = 20, mask = 5, + [0] = "strtDL", "ldrtDL", [4] = "strbtDL", [5] = "ldrbtDL", + }, + _ = { + shift = 20, mask = 5, + [0] = "strDL", "ldrDL", [4] = "strbDL", [5] = "ldrbDL", + } +} + +local map_load1 = { + shift = 4, mask = 1, + [0] = map_load, map_media, +} + +local map_loadm = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "stmdaNR", "stmNR", + { shift = 16, mask = 63, [45] = "pushR", _ = "stmdbNR", }, "stmibNR", + }, + { + shift = 23, mask = 3, + [0] = "ldmdaNR", { shift = 16, mask = 63, [61] = "popR", _ = "ldmNR", }, + "ldmdbNR", "ldmibNR", + }, +} + +local map_data = { + shift = 21, mask = 15, + [0] = "andDNPs", "eorDNPs", "subDNPs", "rsbDNPs", + "addDNPs", "adcDNPs", "sbcDNPs", "rscDNPs", + "tstNP", "teqNP", "cmpNP", "cmnNP", + "orrDNPs", "movDPs", "bicDNPs", "mvnDPs", +} + +local map_mul = { + shift = 21, mask = 7, + [0] = "mulNMSs", "mlaNMSDs", "umaalDNMS", "mlsDNMS", + "umullDNMSs", "umlalDNMSs", "smullDNMSs", "smlalDNMSs", +} + +local map_sync = { + shift = 20, mask = 15, -- NYI: brackets around N. R(D+1) for ldrexd/strexd. + [0] = "swpDMN", false, false, false, + "swpbDMN", false, false, false, + "strexDMN", "ldrexDN", "strexdDN", "ldrexdDN", + "strexbDMN", "ldrexbDN", "strexhDN", "ldrexhDN", +} + +local map_mulh = { + shift = 21, mask = 3, + [0] = { shift = 5, mask = 3, + [0] = "smlabbNMSD", "smlatbNMSD", "smlabtNMSD", "smlattNMSD", }, + { shift = 5, mask = 3, + [0] = "smlawbNMSD", "smulwbNMS", "smlawtNMSD", "smulwtNMS", }, + { shift = 5, mask = 3, + [0] = "smlalbbDNMS", "smlaltbDNMS", "smlalbtDNMS", "smlalttDNMS", }, + { shift = 5, mask = 3, + [0] = "smulbbNMS", "smultbNMS", "smulbtNMS", "smulttNMS", }, +} + +local map_misc = { + shift = 4, mask = 7, + -- NYI: decode PSR bits of msr. + [0] = { shift = 21, mask = 1, [0] = "mrsD", "msrM", }, + { shift = 21, mask = 3, "bxM", false, "clzDM", }, + { shift = 21, mask = 3, "bxjM", }, + { shift = 21, mask = 3, "blxM", }, + false, + { shift = 21, mask = 3, [0] = "qaddDMN", "qsubDMN", "qdaddDMN", "qdsubDMN", }, + false, + { shift = 21, mask = 3, "bkptK", }, +} + +local map_datar = { + shift = 4, mask = 9, + [9] = { + shift = 5, mask = 3, + [0] = { shift = 24, mask = 1, [0] = map_mul, map_sync, }, + { shift = 20, mask = 1, [0] = "strhDL", "ldrhDL", }, + { shift = 20, mask = 1, [0] = "ldrdDL", "ldrsbDL", }, + { shift = 20, mask = 1, [0] = "strdDL", "ldrshDL", }, + }, + _ = { + shift = 20, mask = 25, + [16] = { shift = 7, mask = 1, [0] = map_misc, map_mulh, }, + _ = { + shift = 0, mask = 0xffffffff, + [bor(0xe1a00000)] = "nop", + _ = map_data, + } + }, +} + +local map_datai = { + shift = 20, mask = 31, -- NYI: decode PSR bits of msr. Decode imm12. + [16] = "movwDW", [20] = "movtDW", + [18] = { shift = 0, mask = 0xf00ff, [0] = "nopv6", _ = "msrNW", }, + [22] = "msrNW", + _ = map_data, +} + +local map_branch = { + shift = 24, mask = 1, + [0] = "bB", "blB" +} + +local map_condins = { + [0] = map_datar, map_datai, map_load, map_load1, + map_loadm, map_branch, map_loadc, map_datac +} + +-- NYI: setend. +local map_uncondins = { + [0] = false, map_simddata, map_simdload, map_preload, + false, "blxB", map_loadcu, map_datacu, +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "sp", "lr", "pc", +} + +local map_cond = { + [0] = "eq", "ne", "hs", "lo", "mi", "pl", "vs", "vc", + "hi", "ls", "ge", "lt", "gt", "le", "al", +} + +local map_shift = { [0] = "lsl", "lsr", "asr", "ror", } + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then + extra = "\t->"..sym + elseif band(ctx.op, 0x0e000000) ~= 0x0a000000 then + extra = "\t; 0x"..tohex(ctx.rel) + end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-5s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-5s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +-- Format operand 2 of load/store opcodes. +local function fmtload(ctx, op, pos) + local base = map_gpr[band(rshift(op, 16), 15)] + local x, ofs + local ext = (band(op, 0x04000000) == 0) + if not ext and band(op, 0x02000000) == 0 then + ofs = band(op, 4095) + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + ofs = "#"..ofs + elseif ext and band(op, 0x00400000) ~= 0 then + ofs = band(op, 15) + band(rshift(op, 4), 0xf0) + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + ofs = "#"..ofs + else + ofs = map_gpr[band(op, 15)] + if ext or band(op, 0xfe0) == 0 then + elseif band(op, 0xfe0) == 0x60 then + ofs = format("%s, rrx", ofs) + else + local sh = band(rshift(op, 7), 31) + if sh == 0 then sh = 32 end + ofs = format("%s, %s #%d", ofs, map_shift[band(rshift(op, 5), 3)], sh) + end + if band(op, 0x00800000) == 0 then ofs = "-"..ofs end + end + if ofs == "#0" then + x = format("[%s]", base) + elseif band(op, 0x01000000) == 0 then + x = format("[%s], %s", base, ofs) + else + x = format("[%s, %s]", base, ofs) + end + if band(op, 0x01200000) == 0x01200000 then x = x.."!" end + return x +end + +-- Format operand 2 of vector load/store opcodes. +local function fmtvload(ctx, op, pos) + local base = map_gpr[band(rshift(op, 16), 15)] + local ofs = band(op, 255)*4 + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + if ofs == 0 then + return format("[%s]", base) + else + return format("[%s, #%d]", base, ofs) + end +end + +local function fmtvr(op, vr, sh0, sh1) + if vr == "s" then + return format("s%d", 2*band(rshift(op, sh0), 15)+band(rshift(op, sh1), 1)) + else + return format("d%d", band(rshift(op, sh0), 15)+band(rshift(op, sh1-4), 16)) + end +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) + local operands = {} + local suffix = "" + local last, name, pat + local vr + ctx.op = op + ctx.rel = nil + + local cond = rshift(op, 28) + local opat + if cond == 15 then + opat = map_uncondins[band(rshift(op, 25), 7)] + else + if cond ~= 14 then suffix = map_cond[cond] end + opat = map_condins[band(rshift(op, 25), 7)] + end + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + name, pat = match(opat, "^([a-z0-9]*)(.*)") + if sub(pat, 1, 1) == "." then + local s2, p2 = match(pat, "^([a-z0-9.]*)(.*)") + suffix = suffix..s2 + pat = p2 + end + + for p in gmatch(pat, ".") do + local x = nil + if p == "D" then + x = map_gpr[band(rshift(op, 12), 15)] + elseif p == "N" then + x = map_gpr[band(rshift(op, 16), 15)] + elseif p == "S" then + x = map_gpr[band(rshift(op, 8), 15)] + elseif p == "M" then + x = map_gpr[band(op, 15)] + elseif p == "d" then + x = fmtvr(op, vr, 12, 22) + elseif p == "n" then + x = fmtvr(op, vr, 16, 7) + elseif p == "m" then + x = fmtvr(op, vr, 0, 5) + elseif p == "P" then + if band(op, 0x02000000) ~= 0 then + x = ror(band(op, 255), 2*band(rshift(op, 8), 15)) + else + x = map_gpr[band(op, 15)] + if band(op, 0xff0) ~= 0 then + operands[#operands+1] = x + local s = map_shift[band(rshift(op, 5), 3)] + local r = nil + if band(op, 0xf90) == 0 then + if s == "ror" then s = "rrx" else r = "#32" end + elseif band(op, 0x10) == 0 then + r = "#"..band(rshift(op, 7), 31) + else + r = map_gpr[band(rshift(op, 8), 15)] + end + if name == "mov" then name = s; x = r + elseif r then x = format("%s %s", s, r) + else x = s end + end + end + elseif p == "L" then + x = fmtload(ctx, op, pos) + elseif p == "l" then + x = fmtvload(ctx, op, pos) + elseif p == "B" then + local addr = ctx.addr + pos + 8 + arshift(lshift(op, 8), 6) + if cond == 15 then addr = addr + band(rshift(op, 23), 2) end + ctx.rel = addr + x = "0x"..tohex(addr) + elseif p == "F" then + vr = "s" + elseif p == "G" then + vr = "d" + elseif p == "." then + suffix = suffix..(vr == "s" and ".f32" or ".f64") + elseif p == "R" then + if band(op, 0x00200000) ~= 0 and #operands == 1 then + operands[1] = operands[1].."!" + end + local t = {} + for i=0,15 do + if band(rshift(op, i), 1) == 1 then t[#t+1] = map_gpr[i] end + end + x = "{"..concat(t, ", ").."}" + elseif p == "r" then + if band(op, 0x00200000) ~= 0 and #operands == 2 then + operands[1] = operands[1].."!" + end + local s = tonumber(sub(last, 2)) + local n = band(op, 255) + if vr == "d" then n = rshift(n, 1) end + operands[#operands] = format("{%s-%s%d}", last, vr, s+n-1) + elseif p == "W" then + x = band(op, 0x0fff) + band(rshift(op, 4), 0xf000) + elseif p == "T" then + x = "#0x"..tohex(band(op, 0x00ffffff), 6) + elseif p == "U" then + x = band(rshift(op, 7), 31) + if x == 0 then x = nil end + elseif p == "u" then + x = band(rshift(op, 7), 31) + if band(op, 0x40) == 0 then + if x == 0 then x = nil else x = "lsl #"..x end + else + if x == 0 then x = "asr #32" else x = "asr #"..x end + end + elseif p == "v" then + x = band(rshift(op, 7), 31) + elseif p == "w" then + x = band(rshift(op, 16), 31) + elseif p == "x" then + x = band(rshift(op, 16), 31) + 1 + elseif p == "X" then + x = band(rshift(op, 16), 31) - last + 1 + elseif p == "Y" then + x = band(rshift(op, 12), 0xf0) + band(op, 0x0f) + elseif p == "K" then + x = "#0x"..tohex(band(rshift(op, 4), 0x0000fff0) + band(op, 15), 4) + elseif p == "s" then + if band(op, 0x00100000) ~= 0 then suffix = "s"..suffix end + else + assert(false) + end + if x then + last = x + if type(x) == "number" then x = "#"..x end + operands[#operands+1] = x + end + end + + return putop(ctx, name..suffix, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ctx.pos = ofs + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 16 then return map_gpr[r] end + return "d"..(r-16) +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm64.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm64.lua new file mode 100644 index 00000000..ad909fbd --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm64.lua @@ -0,0 +1,1216 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM64 disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +-- +-- Contributed by Djordje Kovacevic and Stefan Pejic from RT-RK.com. +-- Sponsored by Cisco Systems, Inc. +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles most user-mode AArch64 instructions. +-- NYI: Advanced SIMD and VFP instructions. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local concat = table.concat +local bit = require("bit") +local band, bor, bxor, tohex = bit.band, bit.bor, bit.bxor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift +local ror = bit.ror + +------------------------------------------------------------------------------ +-- Opcode maps +------------------------------------------------------------------------------ + +local map_adr = { -- PC-relative addressing. + shift = 31, mask = 1, + [0] = "adrDBx", "adrpDBx" +} + +local map_addsubi = { -- Add/subtract immediate. + shift = 29, mask = 3, + [0] = "add|movDNIg", "adds|cmnD0NIg", "subDNIg", "subs|cmpD0NIg", +} + +local map_logi = { -- Logical immediate. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "andDNig", "orr|movDN0ig", "eorDNig", "ands|tstD0Nig" + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = "andDNig", "orr|movDN0ig", "eorDNig", "ands|tstD0Nig" + } +} + +local map_movwi = { -- Move wide immediate. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "movnDWRg", false, "movz|movDYRg", "movkDWRg" + }, false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = "movnDWRg", false, "movz|movDYRg", "movkDWRg" + }, +} + +local map_bitf = { -- Bitfield. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "sbfm|sbfiz|sbfx|asr|sxtw|sxth|sxtbDN12w", + "bfm|bfi|bfxilDN13w", + "ubfm|ubfiz|ubfx|lsr|lsl|uxth|uxtbDN12w" + } + }, + { + shift = 22, mask = 1, + { + shift = 29, mask = 3, + [0] = "sbfm|sbfiz|sbfx|asr|sxtw|sxth|sxtbDN12x", + "bfm|bfi|bfxilDN13x", + "ubfm|ubfiz|ubfx|lsr|lsl|uxth|uxtbDN12x" + } + } +} + +local map_datai = { -- Data processing - immediate. + shift = 23, mask = 7, + [0] = map_adr, map_adr, map_addsubi, false, + map_logi, map_movwi, map_bitf, + { + shift = 15, mask = 0x1c0c1, + [0] = "extr|rorDNM4w", [0x10080] = "extr|rorDNM4x", + [0x10081] = "extr|rorDNM4x" + } +} + +local map_logsr = { -- Logical, shifted register. + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = { + shift = 21, mask = 7, + [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg", + "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg" + }, + { + shift = 21, mask = 7, + [0] ="orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg", + "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg" + }, + { + shift = 21, mask = 7, + [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg", + "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg" + }, + { + shift = 21, mask = 7, + [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg", + "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg" + } + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = { + shift = 21, mask = 7, + [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg", + "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg" + }, + { + shift = 21, mask = 7, + [0] = "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg", + "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg" + }, + { + shift = 21, mask = 7, + [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg", + "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg" + }, + { + shift = 21, mask = 7, + [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg", + "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg" + } + } +} + +local map_assh = { + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "addDNMSg", "addDNMSg", "addDNMSg", "addDNMg" + }, + { + shift = 22, mask = 3, + [0] = "adds|cmnD0NMSg", "adds|cmnD0NMSg", + "adds|cmnD0NMSg", "adds|cmnD0NMg" + }, + { + shift = 22, mask = 3, + [0] = "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0Mg" + }, + { + shift = 22, mask = 3, + [0] = "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0MzSg", + "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0Mzg" + }, + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "addDNMSg", "addDNMSg", "addDNMSg", "addDNMg" + }, + { + shift = 22, mask = 3, + [0] = "adds|cmnD0NMSg", "adds|cmnD0NMSg", "adds|cmnD0NMSg", + "adds|cmnD0NMg" + }, + { + shift = 22, mask = 3, + [0] = "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0Mg" + }, + { + shift = 22, mask = 3, + [0] = "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0MzSg", + "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0Mzg" + } + } +} + +local map_addsubsh = { -- Add/subtract, shifted register. + shift = 22, mask = 3, + [0] = map_assh, map_assh, map_assh +} + +local map_addsubex = { -- Add/subtract, extended register. + shift = 22, mask = 3, + [0] = { + shift = 29, mask = 3, + [0] = "addDNMXg", "adds|cmnD0NMXg", "subDNMXg", "subs|cmpD0NMzXg", + } +} + +local map_addsubc = { -- Add/subtract, with carry. + shift = 10, mask = 63, + [0] = { + shift = 29, mask = 3, + [0] = "adcDNMg", "adcsDNMg", "sbc|ngcDN0Mg", "sbcs|ngcsDN0Mg", + } +} + +local map_ccomp = { + shift = 4, mask = 1, + [0] = { + shift = 10, mask = 3, + [0] = { -- Conditional compare register. + shift = 29, mask = 3, + "ccmnNMVCg", false, "ccmpNMVCg", + }, + [2] = { -- Conditional compare immediate. + shift = 29, mask = 3, + "ccmnN5VCg", false, "ccmpN5VCg", + } + } +} + +local map_csel = { -- Conditional select. + shift = 11, mask = 1, + [0] = { + shift = 10, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "cselDNMzCg", false, "csinv|cinv|csetmDNMcg", false, + }, + { + shift = 29, mask = 3, + [0] = "csinc|cinc|csetDNMcg", false, "csneg|cnegDNMcg", false, + } + } +} + +local map_data1s = { -- Data processing, 1 source. + shift = 29, mask = 1, + [0] = { + shift = 31, mask = 1, + [0] = { + shift = 10, mask = 0x7ff, + [0] = "rbitDNg", "rev16DNg", "revDNw", false, "clzDNg", "clsDNg" + }, + { + shift = 10, mask = 0x7ff, + [0] = "rbitDNg", "rev16DNg", "rev32DNx", "revDNx", "clzDNg", "clsDNg" + } + } +} + +local map_data2s = { -- Data processing, 2 sources. + shift = 29, mask = 1, + [0] = { + shift = 10, mask = 63, + false, "udivDNMg", "sdivDNMg", false, false, false, false, "lslDNMg", + "lsrDNMg", "asrDNMg", "rorDNMg" + } +} + +local map_data3s = { -- Data processing, 3 sources. + shift = 29, mask = 7, + [0] = { + shift = 21, mask = 7, + [0] = { + shift = 15, mask = 1, + [0] = "madd|mulDNMA0g", "msub|mnegDNMA0g" + } + }, false, false, false, + { + shift = 15, mask = 1, + [0] = { + shift = 21, mask = 7, + [0] = "madd|mulDNMA0g", "smaddl|smullDxNMwA0x", "smulhDNMx", false, + false, "umaddl|umullDxNMwA0x", "umulhDNMx" + }, + { + shift = 21, mask = 7, + [0] = "msub|mnegDNMA0g", "smsubl|smneglDxNMwA0x", false, false, + false, "umsubl|umneglDxNMwA0x" + } + } +} + +local map_datar = { -- Data processing, register. + shift = 28, mask = 1, + [0] = { + shift = 24, mask = 1, + [0] = map_logsr, + { + shift = 21, mask = 1, + [0] = map_addsubsh, map_addsubex + } + }, + { + shift = 21, mask = 15, + [0] = map_addsubc, false, map_ccomp, false, map_csel, false, + { + shift = 30, mask = 1, + [0] = map_data2s, map_data1s + }, + false, map_data3s, map_data3s, map_data3s, map_data3s, map_data3s, + map_data3s, map_data3s, map_data3s + } +} + +local map_lrl = { -- Load register, literal. + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = "ldrDwB", "ldrDxB", "ldrswDxB" + }, + { + shift = 30, mask = 3, + [0] = "ldrDsB", "ldrDdB" + } +} + +local map_lsriind = { -- Load/store register, immediate pre/post-indexed. + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwzL", "ldrbDwzL", "ldrsbDxzL", "ldrsbDwzL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strhDwzL", "ldrhDwzL", "ldrshDxzL", "ldrshDwzL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strDwzL", "ldrDwzL", "ldrswDxzL" + }, + { + shift = 22, mask = 3, + [0] = "strDszL", "ldrDszL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strDxzL", "ldrDxzL" + }, + { + shift = 22, mask = 3, + [0] = "strDdzL", "ldrDdzL" + } + } +} + +local map_lsriro = { + shift = 21, mask = 1, + [0] = { -- Load/store register immediate. + shift = 10, mask = 3, + [0] = { -- Unscaled immediate. + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "sturbDwK", "ldurbDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturhDwK", "ldurhDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturDwK", "ldurDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturDxK", "ldurDxK" + } + } + }, map_lsriind, false, map_lsriind + }, + { -- Load/store register, register offset. + shift = 10, mask = 3, + [2] = { + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwO", "ldrbDwO", "ldrsbDxO", "ldrsbDwO" + }, + { + shift = 22, mask = 3, + [0] = "strhDwO", "ldrhDwO", "ldrshDxO", "ldrshDwO" + }, + { + shift = 22, mask = 3, + [0] = "strDwO", "ldrDwO", "ldrswDxO" + }, + { + shift = 22, mask = 3, + [0] = "strDxO", "ldrDxO" + } + }, + { + shift = 30, mask = 3, + [2] = { + shift = 22, mask = 3, + [0] = "strDsO", "ldrDsO" + }, + [3] = { + shift = 22, mask = 3, + [0] = "strDdO", "ldrDdO" + } + } + } + } +} + +local map_lsp = { -- Load/store register pair, offset. + shift = 22, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = "stpDzAzwP", "stpDzAzsP", + }, + { + shift = 26, mask = 1, + "stpDzAzdP" + }, + { + shift = 26, mask = 1, + [0] = "stpDzAzxP" + } + }, + { + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = "ldpDzAzwP", "ldpDzAzsP", + }, + { + shift = 26, mask = 1, + [0] = "ldpswDAxP", "ldpDzAzdP" + }, + { + shift = 26, mask = 1, + [0] = "ldpDzAzxP" + } + } +} + +local map_ls = { -- Loads and stores. + shift = 24, mask = 0x31, + [0x10] = map_lrl, [0x30] = map_lsriro, + [0x20] = { + shift = 23, mask = 3, + map_lsp, map_lsp, map_lsp + }, + [0x21] = { + shift = 23, mask = 3, + map_lsp, map_lsp, map_lsp + }, + [0x31] = { + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwzU", "ldrbDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strhDwzU", "ldrhDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strDwzU", "ldrDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strDxzU", "ldrDxzU" + } + }, + { + shift = 30, mask = 3, + [2] = { + shift = 22, mask = 3, + [0] = "strDszU", "ldrDszU" + }, + [3] = { + shift = 22, mask = 3, + [0] = "strDdzU", "ldrDdzU" + } + } + }, +} + +local map_datafp = { -- Data processing, SIMD and FP. + shift = 28, mask = 7, + { -- 001 + shift = 24, mask = 1, + [0] = { + shift = 21, mask = 1, + { + shift = 10, mask = 3, + [0] = { + shift = 12, mask = 1, + [0] = { + shift = 13, mask = 1, + [0] = { + shift = 14, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { -- FP/int conversion. + shift = 31, mask = 1, + [0] = { + shift = 16, mask = 0xff, + [0x20] = "fcvtnsDwNs", [0x21] = "fcvtnuDwNs", + [0x22] = "scvtfDsNw", [0x23] = "ucvtfDsNw", + [0x24] = "fcvtasDwNs", [0x25] = "fcvtauDwNs", + [0x26] = "fmovDwNs", [0x27] = "fmovDsNw", + [0x28] = "fcvtpsDwNs", [0x29] = "fcvtpuDwNs", + [0x30] = "fcvtmsDwNs", [0x31] = "fcvtmuDwNs", + [0x38] = "fcvtzsDwNs", [0x39] = "fcvtzuDwNs", + [0x60] = "fcvtnsDwNd", [0x61] = "fcvtnuDwNd", + [0x62] = "scvtfDdNw", [0x63] = "ucvtfDdNw", + [0x64] = "fcvtasDwNd", [0x65] = "fcvtauDwNd", + [0x68] = "fcvtpsDwNd", [0x69] = "fcvtpuDwNd", + [0x70] = "fcvtmsDwNd", [0x71] = "fcvtmuDwNd", + [0x78] = "fcvtzsDwNd", [0x79] = "fcvtzuDwNd" + }, + { + shift = 16, mask = 0xff, + [0x20] = "fcvtnsDxNs", [0x21] = "fcvtnuDxNs", + [0x22] = "scvtfDsNx", [0x23] = "ucvtfDsNx", + [0x24] = "fcvtasDxNs", [0x25] = "fcvtauDxNs", + [0x28] = "fcvtpsDxNs", [0x29] = "fcvtpuDxNs", + [0x30] = "fcvtmsDxNs", [0x31] = "fcvtmuDxNs", + [0x38] = "fcvtzsDxNs", [0x39] = "fcvtzuDxNs", + [0x60] = "fcvtnsDxNd", [0x61] = "fcvtnuDxNd", + [0x62] = "scvtfDdNx", [0x63] = "ucvtfDdNx", + [0x64] = "fcvtasDxNd", [0x65] = "fcvtauDxNd", + [0x66] = "fmovDxNd", [0x67] = "fmovDdNx", + [0x68] = "fcvtpsDxNd", [0x69] = "fcvtpuDxNd", + [0x70] = "fcvtmsDxNd", [0x71] = "fcvtmuDxNd", + [0x78] = "fcvtzsDxNd", [0x79] = "fcvtzuDxNd" + } + } + }, + { -- FP data-processing, 1 source. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = { + shift = 15, mask = 63, + [0] = "fmovDNf", "fabsDNf", "fnegDNf", + "fsqrtDNf", false, "fcvtDdNs", false, false, + "frintnDNf", "frintpDNf", "frintmDNf", "frintzDNf", + "frintaDNf", false, "frintxDNf", "frintiDNf", + }, + { + shift = 15, mask = 63, + [0] = "fmovDNf", "fabsDNf", "fnegDNf", + "fsqrtDNf", "fcvtDsNd", false, false, false, + "frintnDNf", "frintpDNf", "frintmDNf", "frintzDNf", + "frintaDNf", false, "frintxDNf", "frintiDNf", + } + } + } + }, + { -- FP compare. + shift = 31, mask = 1, + [0] = { + shift = 14, mask = 3, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 0, mask = 31, + [0] = "fcmpNMf", [8] = "fcmpNZf", + [16] = "fcmpeNMf", [24] = "fcmpeNZf", + } + } + } + } + }, + { -- FP immediate. + shift = 31, mask = 1, + [0] = { + shift = 5, mask = 31, + [0] = { + shift = 23, mask = 1, + [0] = "fmovDFf" + } + } + } + }, + { -- FP conditional compare. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 4, mask = 1, + [0] = "fccmpNMVCf", "fccmpeNMVCf" + } + } + }, + { -- FP data-processing, 2 sources. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 12, mask = 15, + [0] = "fmulDNMf", "fdivDNMf", "faddDNMf", "fsubDNMf", + "fmaxDNMf", "fminDNMf", "fmaxnmDNMf", "fminnmDNMf", + "fnmulDNMf" + } + } + }, + { -- FP conditional select. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = "fcselDNMCf" + } + } + } + }, + { -- FP data-processing, 3 sources. + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 21, mask = 5, + [0] = "fmaddDNMAf", "fnmaddDNMAf" + }, + { + shift = 21, mask = 5, + [0] = "fmsubDNMAf", "fnmsubDNMAf" + } + } + } + } +} + +local map_br = { -- Branches, exception generating and system instructions. + shift = 29, mask = 7, + [0] = "bB", + { -- Compare & branch, immediate. + shift = 24, mask = 3, + [0] = "cbzDBg", "cbnzDBg", "tbzDTBw", "tbnzDTBw" + }, + { -- Conditional branch, immediate. + shift = 24, mask = 3, + [0] = { + shift = 4, mask = 1, + [0] = { + shift = 0, mask = 15, + [0] = "beqB", "bneB", "bhsB", "bloB", "bmiB", "bplB", "bvsB", "bvcB", + "bhiB", "blsB", "bgeB", "bltB", "bgtB", "bleB", "balB" + } + } + }, false, "blB", + { -- Compare & branch, immediate. + shift = 24, mask = 3, + [0] = "cbzDBg", "cbnzDBg", "tbzDTBx", "tbnzDTBx" + }, + { + shift = 24, mask = 3, + [0] = { -- Exception generation. + shift = 0, mask = 0xe0001f, + [0x200000] = "brkW" + }, + { -- System instructions. + shift = 0, mask = 0x3fffff, + [0x03201f] = "nop" + }, + { -- Unconditional branch, register. + shift = 0, mask = 0xfffc1f, + [0x1f0000] = "brNx", [0x3f0000] = "blrNx", + [0x5f0000] = "retNx" + }, + } +} + +local map_init = { + shift = 25, mask = 15, + [0] = false, false, false, false, map_ls, map_datar, map_ls, map_datafp, + map_datai, map_datai, map_br, map_br, map_ls, map_datar, map_ls, map_datafp +} + +------------------------------------------------------------------------------ + +local map_regs = { x = {}, w = {}, d = {}, s = {} } + +for i=0,30 do + map_regs.x[i] = "x"..i + map_regs.w[i] = "w"..i + map_regs.d[i] = "d"..i + map_regs.s[i] = "s"..i +end +map_regs.x[31] = "sp" +map_regs.w[31] = "wsp" +map_regs.d[31] = "d31" +map_regs.s[31] = "s31" + +local map_cond = { + [0] = "eq", "ne", "cs", "cc", "mi", "pl", "vs", "vc", + "hi", "ls", "ge", "lt", "gt", "le", "al", +} + +local map_shift = { [0] = "lsl", "lsr", "asr", } + +local map_extend = { + [0] = "uxtb", "uxth", "uxtw", "uxtx", "sxtb", "sxth", "sxtw", "sxtx", +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then + extra = "\t->"..sym + end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-5s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-5s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +local function match_reg(p, pat, regnum) + return map_regs[match(pat, p.."%w-([xwds])")][regnum] +end + +local function fmt_hex32(x) + if x < 0 then + return tohex(x) + else + return format("%x", x) + end +end + +local imm13_rep = { 0x55555555, 0x11111111, 0x01010101, 0x00010001, 0x00000001 } + +local function decode_imm13(op) + local imms = band(rshift(op, 10), 63) + local immr = band(rshift(op, 16), 63) + if band(op, 0x00400000) == 0 then + local len = 5 + if imms >= 56 then + if imms >= 60 then len = 1 else len = 2 end + elseif imms >= 48 then len = 3 elseif imms >= 32 then len = 4 end + local l = lshift(1, len)-1 + local s = band(imms, l) + local r = band(immr, l) + local imm = ror(rshift(-1, 31-s), r) + if len ~= 5 then imm = band(imm, lshift(1, l)-1) + rshift(imm, 31-l) end + imm = imm * imm13_rep[len] + local ix = fmt_hex32(imm) + if rshift(op, 31) ~= 0 then + return ix..tohex(imm) + else + return ix + end + else + local lo, hi = -1, 0 + if imms < 32 then lo = rshift(-1, 31-imms) else hi = rshift(-1, 63-imms) end + if immr ~= 0 then + lo, hi = ror(lo, immr), ror(hi, immr) + local x = immr == 32 and 0 or band(bxor(lo, hi), lshift(-1, 32-immr)) + lo, hi = bxor(lo, x), bxor(hi, x) + if immr >= 32 then lo, hi = hi, lo end + end + if hi ~= 0 then + return fmt_hex32(hi)..tohex(lo) + else + return fmt_hex32(lo) + end + end +end + +local function parse_immpc(op, name) + if name == "b" or name == "bl" then + return arshift(lshift(op, 6), 4) + elseif name == "adr" or name == "adrp" then + local immlo = band(rshift(op, 29), 3) + local immhi = lshift(arshift(lshift(op, 8), 13), 2) + return bor(immhi, immlo) + elseif name == "tbz" or name == "tbnz" then + return lshift(arshift(lshift(op, 13), 18), 2) + else + return lshift(arshift(lshift(op, 8), 13), 2) + end +end + +local function parse_fpimm8(op) + local sign = band(op, 0x100000) == 0 and 1 or -1 + local exp = bxor(rshift(arshift(lshift(op, 12), 5), 24), 0x80) - 131 + local frac = 16+band(rshift(op, 13), 15) + return sign * frac * 2^exp +end + +local function prefer_bfx(sf, uns, imms, immr) + if imms < immr or imms == 31 or imms == 63 then + return false + end + if immr == 0 then + if sf == 0 and (imms == 7 or imms == 15) then + return false + end + if sf ~= 0 and uns == 0 and (imms == 7 or imms == 15 or imms == 31) then + return false + end + end + return true +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) + local operands = {} + local suffix = "" + local last, name, pat + local map_reg + ctx.op = op + ctx.rel = nil + last = nil + local opat + opat = map_init[band(rshift(op, 25), 15)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + name, pat = match(opat, "^([a-z0-9]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.|]*)(.*)") + if altname then pat = pat2 end + if sub(pat, 1, 1) == "." then + local s2, p2 = match(pat, "^([a-z0-9.]*)(.*)") + suffix = suffix..s2 + pat = p2 + end + + local rt = match(pat, "[gf]") + if rt then + if rt == "g" then + map_reg = band(op, 0x80000000) ~= 0 and map_regs.x or map_regs.w + else + map_reg = band(op, 0x400000) ~= 0 and map_regs.d or map_regs.s + end + end + + local second0, immr + + for p in gmatch(pat, ".") do + local x = nil + if p == "D" then + local regnum = band(op, 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "N" then + local regnum = band(rshift(op, 5), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "M" then + local regnum = band(rshift(op, 16), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "A" then + local regnum = band(rshift(op, 10), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "B" then + local addr = ctx.addr + pos + parse_immpc(op, name) + ctx.rel = addr + x = "0x"..tohex(addr) + elseif p == "T" then + x = bor(band(rshift(op, 26), 32), band(rshift(op, 19), 31)) + elseif p == "V" then + x = band(op, 15) + elseif p == "C" then + x = map_cond[band(rshift(op, 12), 15)] + elseif p == "c" then + local rn = band(rshift(op, 5), 31) + local rm = band(rshift(op, 16), 31) + local cond = band(rshift(op, 12), 15) + local invc = bxor(cond, 1) + x = map_cond[cond] + if altname and cond ~= 14 and cond ~= 15 then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if rn == rm then + local n = #operands + operands[n] = nil + x = map_cond[invc] + if rn ~= 31 then + if a1 then name = a1 else name = altname end + else + operands[n-1] = nil + name = a2 + end + end + end + elseif p == "W" then + x = band(rshift(op, 5), 0xffff) + elseif p == "Y" then + x = band(rshift(op, 5), 0xffff) + local hw = band(rshift(op, 21), 3) + if altname and (hw == 0 or x ~= 0) then + name = altname + end + elseif p == "L" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local imm9 = arshift(lshift(op, 11), 23) + if band(op, 0x800) ~= 0 then + x = "["..rn..", #"..imm9.."]!" + else + x = "["..rn.."], #"..imm9 + end + elseif p == "U" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local sz = band(rshift(op, 30), 3) + local imm12 = lshift(arshift(lshift(op, 10), 20), sz) + if imm12 ~= 0 then + x = "["..rn..", #"..imm12.."]" + else + x = "["..rn.."]" + end + elseif p == "K" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local imm9 = arshift(lshift(op, 11), 23) + if imm9 ~= 0 then + x = "["..rn..", #"..imm9.."]" + else + x = "["..rn.."]" + end + elseif p == "O" then + local rn, rm = map_regs.x[band(rshift(op, 5), 31)] + local m = band(rshift(op, 13), 1) + if m == 0 then + rm = map_regs.w[band(rshift(op, 16), 31)] + else + rm = map_regs.x[band(rshift(op, 16), 31)] + end + x = "["..rn..", "..rm + local opt = band(rshift(op, 13), 7) + local s = band(rshift(op, 12), 1) + local sz = band(rshift(op, 30), 3) + -- extension to be applied + if opt == 3 then + if s == 0 then x = x.."]" + else x = x..", lsl #"..sz.."]" end + elseif opt == 2 or opt == 6 or opt == 7 then + if s == 0 then x = x..", "..map_extend[opt].."]" + else x = x..", "..map_extend[opt].." #"..sz.."]" end + else + x = x.."]" + end + elseif p == "P" then + local opcv, sh = rshift(op, 26), 2 + if opcv >= 0x2a then sh = 4 elseif opcv >= 0x1b then sh = 3 end + local imm7 = lshift(arshift(lshift(op, 10), 25), sh) + local rn = map_regs.x[band(rshift(op, 5), 31)] + local ind = band(rshift(op, 23), 3) + if ind == 1 then + x = "["..rn.."], #"..imm7 + elseif ind == 2 then + if imm7 == 0 then + x = "["..rn.."]" + else + x = "["..rn..", #"..imm7.."]" + end + elseif ind == 3 then + x = "["..rn..", #"..imm7.."]!" + end + elseif p == "I" then + local shf = band(rshift(op, 22), 3) + local imm12 = band(rshift(op, 10), 0x0fff) + local rn, rd = band(rshift(op, 5), 31), band(op, 31) + if altname == "mov" and shf == 0 and imm12 == 0 and (rn == 31 or rd == 31) then + name = altname + x = nil + elseif shf == 0 then + x = imm12 + elseif shf == 1 then + x = imm12..", lsl #12" + end + elseif p == "i" then + x = "#0x"..decode_imm13(op) + elseif p == "1" then + immr = band(rshift(op, 16), 63) + x = immr + elseif p == "2" then + x = band(rshift(op, 10), 63) + if altname then + local a1, a2, a3, a4, a5, a6 = + match(altname, "([^|]*)|([^|]*)|([^|]*)|([^|]*)|([^|]*)|(.*)") + local sf = band(rshift(op, 26), 32) + local uns = band(rshift(op, 30), 1) + if prefer_bfx(sf, uns, x, immr) then + name = a2 + x = x - immr + 1 + elseif immr == 0 and x == 7 then + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + name = a6 + x = nil + elseif immr == 0 and x == 15 then + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + name = a5 + x = nil + elseif x == 31 or x == 63 then + if x == 31 and immr == 0 and name == "sbfm" then + name = a4 + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + else + name = a3 + end + x = nil + elseif band(x, 31) ~= 31 and immr == x+1 and name == "ubfm" then + name = a4 + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = nil + elseif x < immr then + name = a1 + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = x + 1 + end + end + elseif p == "3" then + x = band(rshift(op, 10), 63) + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if x < immr then + name = a1 + local sf = band(rshift(op, 26), 32) + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = x + 1 + else + name = a2 + x = x - immr + 1 + end + end + elseif p == "4" then + x = band(rshift(op, 10), 63) + local rn = band(rshift(op, 5), 31) + local rm = band(rshift(op, 16), 31) + if altname and rn == rm then + local n = #operands + operands[n] = nil + last = operands[n-1] + name = altname + end + elseif p == "5" then + x = band(rshift(op, 16), 31) + elseif p == "S" then + x = band(rshift(op, 10), 63) + if x == 0 then x = nil + else x = map_shift[band(rshift(op, 22), 3)].." #"..x end + elseif p == "X" then + local opt = band(rshift(op, 13), 7) + -- Width specifier . + if opt ~= 3 and opt ~= 7 then + last = map_regs.w[band(rshift(op, 16), 31)] + operands[#operands] = last + end + x = band(rshift(op, 10), 7) + -- Extension. + if opt == 2 + band(rshift(op, 31), 1) and + band(rshift(op, second0 and 5 or 0), 31) == 31 then + if x == 0 then x = nil + else x = "lsl #"..x end + else + if x == 0 then x = map_extend[band(rshift(op, 13), 7)] + else x = map_extend[band(rshift(op, 13), 7)].." #"..x end + end + elseif p == "R" then + x = band(rshift(op,21), 3) + if x == 0 then x = nil + else x = "lsl #"..x*16 end + elseif p == "z" then + local n = #operands + if operands[n] == "sp" then operands[n] = "xzr" + elseif operands[n] == "wsp" then operands[n] = "wzr" + end + elseif p == "Z" then + x = 0 + elseif p == "F" then + x = parse_fpimm8(op) + elseif p == "g" or p == "f" or p == "x" or p == "w" or + p == "d" or p == "s" then + -- These are handled in D/N/M/A. + elseif p == "0" then + if last == "sp" or last == "wsp" then + local n = #operands + operands[n] = nil + last = operands[n-1] + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if not a1 then + name = altname + elseif second0 then + name, altname = a2, a1 + else + name, altname = a1, a2 + end + end + end + second0 = true + else + assert(false) + end + if x then + last = x + if type(x) == "number" then x = "#"..x end + operands[#operands+1] = x + end + end + + return putop(ctx, name..suffix, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ctx.pos = ofs + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_regs.x[r] end + return map_regs.d[r-32] +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm64be.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm64be.lua new file mode 100644 index 00000000..edcbffa8 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_arm64be.lua @@ -0,0 +1,12 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM64BE disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- ARM64 instructions are always little-endian. So just forward to the +-- common ARM64 disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +return require((string.match(..., ".*%.") or "").."dis_arm64") + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips.lua new file mode 100644 index 00000000..6ad17f54 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips.lua @@ -0,0 +1,694 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT/X license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles all standard MIPS32R1/R2 instructions. +-- Default mode is big-endian, but see: dis_mipsel.lua +------------------------------------------------------------------------------ + +local type = type +local byte, format = string.byte, string.format +local match, gmatch = string.match, string.gmatch +local concat = table.concat +local bit = require("bit") +local band, bor, tohex = bit.band, bit.bor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Extended opcode maps common to all MIPS releases +------------------------------------------------------------------------------ + +local map_srl = { shift = 21, mask = 1, [0] = "srlDTA", "rotrDTA", } +local map_srlv = { shift = 6, mask = 1, [0] = "srlvDTS", "rotrvDTS", } + +local map_cop0 = { + shift = 25, mask = 1, + [0] = { + shift = 21, mask = 15, + [0] = "mfc0TDW", [4] = "mtc0TDW", + [10] = "rdpgprDT", + [11] = { shift = 5, mask = 1, [0] = "diT0", "eiT0", }, + [14] = "wrpgprDT", + }, { + shift = 0, mask = 63, + [1] = "tlbr", [2] = "tlbwi", [6] = "tlbwr", [8] = "tlbp", + [24] = "eret", [31] = "deret", + [32] = "wait", + }, +} + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps for MIPS R1-R5 +------------------------------------------------------------------------------ + +local map_movci = { shift = 16, mask = 1, [0] = "movfDSC", "movtDSC", } + +local map_special = { + shift = 0, mask = 63, + [0] = { shift = 0, mask = -1, [0] = "nop", _ = "sllDTA" }, + map_movci, map_srl, "sraDTA", + "sllvDTS", false, map_srlv, "sravDTS", + "jrS", "jalrD1S", "movzDST", "movnDST", + "syscallY", "breakY", false, "sync", + "mfhiD", "mthiS", "mfloD", "mtloS", + "dsllvDST", false, "dsrlvDST", "dsravDST", + "multST", "multuST", "divST", "divuST", + "dmultST", "dmultuST", "ddivST", "ddivuST", + "addDST", "addu|moveDST0", "subDST", "subu|neguDS0T", + "andDST", "or|moveDST0", "xorDST", "nor|notDST0", + false, false, "sltDST", "sltuDST", + "daddDST", "dadduDST", "dsubDST", "dsubuDST", + "tgeSTZ", "tgeuSTZ", "tltSTZ", "tltuSTZ", + "teqSTZ", false, "tneSTZ", false, + "dsllDTA", false, "dsrlDTA", "dsraDTA", + "dsll32DTA", false, "dsrl32DTA", "dsra32DTA", +} + +local map_special2 = { + shift = 0, mask = 63, + [0] = "maddST", "madduST", "mulDST", false, + "msubST", "msubuST", + [32] = "clzDS", [33] = "cloDS", + [63] = "sdbbpY", +} + +local map_bshfl = { + shift = 6, mask = 31, + [2] = "wsbhDT", + [16] = "sebDT", + [24] = "sehDT", +} + +local map_dbshfl = { + shift = 6, mask = 31, + [2] = "dsbhDT", + [5] = "dshdDT", +} + +local map_special3 = { + shift = 0, mask = 63, + [0] = "extTSAK", [1] = "dextmTSAP", [3] = "dextTSAK", + [4] = "insTSAL", [6] = "dinsuTSEQ", [7] = "dinsTSAL", + [32] = map_bshfl, [36] = map_dbshfl, [59] = "rdhwrTD", +} + +local map_regimm = { + shift = 16, mask = 31, + [0] = "bltzSB", "bgezSB", "bltzlSB", "bgezlSB", + false, false, false, false, + "tgeiSI", "tgeiuSI", "tltiSI", "tltiuSI", + "teqiSI", false, "tneiSI", false, + "bltzalSB", "bgezalSB", "bltzallSB", "bgezallSB", + false, false, false, false, + false, false, false, false, + false, false, false, "synciSO", +} + +local map_cop1s = { + shift = 0, mask = 63, + [0] = "add.sFGH", "sub.sFGH", "mul.sFGH", "div.sFGH", + "sqrt.sFG", "abs.sFG", "mov.sFG", "neg.sFG", + "round.l.sFG", "trunc.l.sFG", "ceil.l.sFG", "floor.l.sFG", + "round.w.sFG", "trunc.w.sFG", "ceil.w.sFG", "floor.w.sFG", + false, + { shift = 16, mask = 1, [0] = "movf.sFGC", "movt.sFGC" }, + "movz.sFGT", "movn.sFGT", + false, "recip.sFG", "rsqrt.sFG", false, + false, false, false, false, + false, false, false, false, + false, "cvt.d.sFG", false, false, + "cvt.w.sFG", "cvt.l.sFG", "cvt.ps.sFGH", false, + false, false, false, false, + false, false, false, false, + "c.f.sVGH", "c.un.sVGH", "c.eq.sVGH", "c.ueq.sVGH", + "c.olt.sVGH", "c.ult.sVGH", "c.ole.sVGH", "c.ule.sVGH", + "c.sf.sVGH", "c.ngle.sVGH", "c.seq.sVGH", "c.ngl.sVGH", + "c.lt.sVGH", "c.nge.sVGH", "c.le.sVGH", "c.ngt.sVGH", +} + +local map_cop1d = { + shift = 0, mask = 63, + [0] = "add.dFGH", "sub.dFGH", "mul.dFGH", "div.dFGH", + "sqrt.dFG", "abs.dFG", "mov.dFG", "neg.dFG", + "round.l.dFG", "trunc.l.dFG", "ceil.l.dFG", "floor.l.dFG", + "round.w.dFG", "trunc.w.dFG", "ceil.w.dFG", "floor.w.dFG", + false, + { shift = 16, mask = 1, [0] = "movf.dFGC", "movt.dFGC" }, + "movz.dFGT", "movn.dFGT", + false, "recip.dFG", "rsqrt.dFG", false, + false, false, false, false, + false, false, false, false, + "cvt.s.dFG", false, false, false, + "cvt.w.dFG", "cvt.l.dFG", false, false, + false, false, false, false, + false, false, false, false, + "c.f.dVGH", "c.un.dVGH", "c.eq.dVGH", "c.ueq.dVGH", + "c.olt.dVGH", "c.ult.dVGH", "c.ole.dVGH", "c.ule.dVGH", + "c.df.dVGH", "c.ngle.dVGH", "c.deq.dVGH", "c.ngl.dVGH", + "c.lt.dVGH", "c.nge.dVGH", "c.le.dVGH", "c.ngt.dVGH", +} + +local map_cop1ps = { + shift = 0, mask = 63, + [0] = "add.psFGH", "sub.psFGH", "mul.psFGH", false, + false, "abs.psFG", "mov.psFG", "neg.psFG", + false, false, false, false, + false, false, false, false, + false, + { shift = 16, mask = 1, [0] = "movf.psFGC", "movt.psFGC" }, + "movz.psFGT", "movn.psFGT", + false, false, false, false, + false, false, false, false, + false, false, false, false, + "cvt.s.puFG", false, false, false, + false, false, false, false, + "cvt.s.plFG", false, false, false, + "pll.psFGH", "plu.psFGH", "pul.psFGH", "puu.psFGH", + "c.f.psVGH", "c.un.psVGH", "c.eq.psVGH", "c.ueq.psVGH", + "c.olt.psVGH", "c.ult.psVGH", "c.ole.psVGH", "c.ule.psVGH", + "c.psf.psVGH", "c.ngle.psVGH", "c.pseq.psVGH", "c.ngl.psVGH", + "c.lt.psVGH", "c.nge.psVGH", "c.le.psVGH", "c.ngt.psVGH", +} + +local map_cop1w = { + shift = 0, mask = 63, + [32] = "cvt.s.wFG", [33] = "cvt.d.wFG", +} + +local map_cop1l = { + shift = 0, mask = 63, + [32] = "cvt.s.lFG", [33] = "cvt.d.lFG", +} + +local map_cop1bc = { + shift = 16, mask = 3, + [0] = "bc1fCB", "bc1tCB", "bc1flCB", "bc1tlCB", +} + +local map_cop1 = { + shift = 21, mask = 31, + [0] = "mfc1TG", "dmfc1TG", "cfc1TG", "mfhc1TG", + "mtc1TG", "dmtc1TG", "ctc1TG", "mthc1TG", + map_cop1bc, false, false, false, + false, false, false, false, + map_cop1s, map_cop1d, false, false, + map_cop1w, map_cop1l, map_cop1ps, +} + +local map_cop1x = { + shift = 0, mask = 63, + [0] = "lwxc1FSX", "ldxc1FSX", false, false, + false, "luxc1FSX", false, false, + "swxc1FSX", "sdxc1FSX", false, false, + false, "suxc1FSX", false, "prefxMSX", + false, false, false, false, + false, false, false, false, + false, false, false, false, + false, false, "alnv.psFGHS", false, + "madd.sFRGH", "madd.dFRGH", false, false, + false, false, "madd.psFRGH", false, + "msub.sFRGH", "msub.dFRGH", false, false, + false, false, "msub.psFRGH", false, + "nmadd.sFRGH", "nmadd.dFRGH", false, false, + false, false, "nmadd.psFRGH", false, + "nmsub.sFRGH", "nmsub.dFRGH", false, false, + false, false, "nmsub.psFRGH", false, +} + +local map_pri = { + [0] = map_special, map_regimm, "jJ", "jalJ", + "beq|beqz|bST00B", "bne|bnezST0B", "blezSB", "bgtzSB", + "addiTSI", "addiu|liTS0I", "sltiTSI", "sltiuTSI", + "andiTSU", "ori|liTS0U", "xoriTSU", "luiTU", + map_cop0, map_cop1, false, map_cop1x, + "beql|beqzlST0B", "bnel|bnezlST0B", "blezlSB", "bgtzlSB", + "daddiTSI", "daddiuTSI", false, false, + map_special2, "jalxJ", false, map_special3, + "lbTSO", "lhTSO", "lwlTSO", "lwTSO", + "lbuTSO", "lhuTSO", "lwrTSO", false, + "sbTSO", "shTSO", "swlTSO", "swTSO", + false, false, "swrTSO", "cacheNSO", + "llTSO", "lwc1HSO", "lwc2TSO", "prefNSO", + false, "ldc1HSO", "ldc2TSO", "ldTSO", + "scTSO", "swc1HSO", "swc2TSO", false, + false, "sdc1HSO", "sdc2TSO", "sdTSO", +} + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps for MIPS R6 +------------------------------------------------------------------------------ + +local map_mul_r6 = { shift = 6, mask = 3, [2] = "mulDST", [3] = "muhDST" } +local map_mulu_r6 = { shift = 6, mask = 3, [2] = "muluDST", [3] = "muhuDST" } +local map_div_r6 = { shift = 6, mask = 3, [2] = "divDST", [3] = "modDST" } +local map_divu_r6 = { shift = 6, mask = 3, [2] = "divuDST", [3] = "moduDST" } +local map_dmul_r6 = { shift = 6, mask = 3, [2] = "dmulDST", [3] = "dmuhDST" } +local map_dmulu_r6 = { shift = 6, mask = 3, [2] = "dmuluDST", [3] = "dmuhuDST" } +local map_ddiv_r6 = { shift = 6, mask = 3, [2] = "ddivDST", [3] = "dmodDST" } +local map_ddivu_r6 = { shift = 6, mask = 3, [2] = "ddivuDST", [3] = "dmoduDST" } + +local map_special_r6 = { + shift = 0, mask = 63, + [0] = { shift = 0, mask = -1, [0] = "nop", _ = "sllDTA" }, + false, map_srl, "sraDTA", + "sllvDTS", false, map_srlv, "sravDTS", + "jrS", "jalrD1S", false, false, + "syscallY", "breakY", false, "sync", + "clzDS", "cloDS", "dclzDS", "dcloDS", + "dsllvDST", "dlsaDSTA", "dsrlvDST", "dsravDST", + map_mul_r6, map_mulu_r6, map_div_r6, map_divu_r6, + map_dmul_r6, map_dmulu_r6, map_ddiv_r6, map_ddivu_r6, + "addDST", "addu|moveDST0", "subDST", "subu|neguDS0T", + "andDST", "or|moveDST0", "xorDST", "nor|notDST0", + false, false, "sltDST", "sltuDST", + "daddDST", "dadduDST", "dsubDST", "dsubuDST", + "tgeSTZ", "tgeuSTZ", "tltSTZ", "tltuSTZ", + "teqSTZ", "seleqzDST", "tneSTZ", "selnezDST", + "dsllDTA", false, "dsrlDTA", "dsraDTA", + "dsll32DTA", false, "dsrl32DTA", "dsra32DTA", +} + +local map_bshfl_r6 = { + shift = 9, mask = 3, + [1] = "alignDSTa", + _ = { + shift = 6, mask = 31, + [0] = "bitswapDT", + [2] = "wsbhDT", + [16] = "sebDT", + [24] = "sehDT", + } +} + +local map_dbshfl_r6 = { + shift = 9, mask = 3, + [1] = "dalignDSTa", + _ = { + shift = 6, mask = 31, + [0] = "dbitswapDT", + [2] = "dsbhDT", + [5] = "dshdDT", + } +} + +local map_special3_r6 = { + shift = 0, mask = 63, + [0] = "extTSAK", [1] = "dextmTSAP", [3] = "dextTSAK", + [4] = "insTSAL", [6] = "dinsuTSEQ", [7] = "dinsTSAL", + [32] = map_bshfl_r6, [36] = map_dbshfl_r6, [59] = "rdhwrTD", +} + +local map_regimm_r6 = { + shift = 16, mask = 31, + [0] = "bltzSB", [1] = "bgezSB", + [6] = "dahiSI", [30] = "datiSI", + [23] = "sigrieI", [31] = "synciSO", +} + +local map_pcrel_r6 = { + shift = 19, mask = 3, + [0] = "addiupcS2", "lwpcS2", "lwupcS2", { + shift = 18, mask = 1, + [0] = "ldpcS3", { shift = 16, mask = 3, [2] = "auipcSI", [3] = "aluipcSI" } + } +} + +local map_cop1s_r6 = { + shift = 0, mask = 63, + [0] = "add.sFGH", "sub.sFGH", "mul.sFGH", "div.sFGH", + "sqrt.sFG", "abs.sFG", "mov.sFG", "neg.sFG", + "round.l.sFG", "trunc.l.sFG", "ceil.l.sFG", "floor.l.sFG", + "round.w.sFG", "trunc.w.sFG", "ceil.w.sFG", "floor.w.sFG", + "sel.sFGH", false, false, false, + "seleqz.sFGH", "recip.sFG", "rsqrt.sFG", "selnez.sFGH", + "maddf.sFGH", "msubf.sFGH", "rint.sFG", "class.sFG", + "min.sFGH", "mina.sFGH", "max.sFGH", "maxa.sFGH", + false, "cvt.d.sFG", false, false, + "cvt.w.sFG", "cvt.l.sFG", +} + +local map_cop1d_r6 = { + shift = 0, mask = 63, + [0] = "add.dFGH", "sub.dFGH", "mul.dFGH", "div.dFGH", + "sqrt.dFG", "abs.dFG", "mov.dFG", "neg.dFG", + "round.l.dFG", "trunc.l.dFG", "ceil.l.dFG", "floor.l.dFG", + "round.w.dFG", "trunc.w.dFG", "ceil.w.dFG", "floor.w.dFG", + "sel.dFGH", false, false, false, + "seleqz.dFGH", "recip.dFG", "rsqrt.dFG", "selnez.dFGH", + "maddf.dFGH", "msubf.dFGH", "rint.dFG", "class.dFG", + "min.dFGH", "mina.dFGH", "max.dFGH", "maxa.dFGH", + "cvt.s.dFG", false, false, false, + "cvt.w.dFG", "cvt.l.dFG", +} + +local map_cop1w_r6 = { + shift = 0, mask = 63, + [0] = "cmp.af.sFGH", "cmp.un.sFGH", "cmp.eq.sFGH", "cmp.ueq.sFGH", + "cmp.lt.sFGH", "cmp.ult.sFGH", "cmp.le.sFGH", "cmp.ule.sFGH", + "cmp.saf.sFGH", "cmp.sun.sFGH", "cmp.seq.sFGH", "cmp.sueq.sFGH", + "cmp.slt.sFGH", "cmp.sult.sFGH", "cmp.sle.sFGH", "cmp.sule.sFGH", + false, "cmp.or.sFGH", "cmp.une.sFGH", "cmp.ne.sFGH", + false, false, false, false, + false, "cmp.sor.sFGH", "cmp.sune.sFGH", "cmp.sne.sFGH", + false, false, false, false, + "cvt.s.wFG", "cvt.d.wFG", +} + +local map_cop1l_r6 = { + shift = 0, mask = 63, + [0] = "cmp.af.dFGH", "cmp.un.dFGH", "cmp.eq.dFGH", "cmp.ueq.dFGH", + "cmp.lt.dFGH", "cmp.ult.dFGH", "cmp.le.dFGH", "cmp.ule.dFGH", + "cmp.saf.dFGH", "cmp.sun.dFGH", "cmp.seq.dFGH", "cmp.sueq.dFGH", + "cmp.slt.dFGH", "cmp.sult.dFGH", "cmp.sle.dFGH", "cmp.sule.dFGH", + false, "cmp.or.dFGH", "cmp.une.dFGH", "cmp.ne.dFGH", + false, false, false, false, + false, "cmp.sor.dFGH", "cmp.sune.dFGH", "cmp.sne.dFGH", + false, false, false, false, + "cvt.s.lFG", "cvt.d.lFG", +} + +local map_cop1_r6 = { + shift = 21, mask = 31, + [0] = "mfc1TG", "dmfc1TG", "cfc1TG", "mfhc1TG", + "mtc1TG", "dmtc1TG", "ctc1TG", "mthc1TG", + false, "bc1eqzHB", false, false, + false, "bc1nezHB", false, false, + map_cop1s_r6, map_cop1d_r6, false, false, + map_cop1w_r6, map_cop1l_r6, +} + +local function maprs_popTS(rs, rt) + if rt == 0 then return 0 elseif rs == 0 then return 1 + elseif rs == rt then return 2 else return 3 end +end + +local map_pop06_r6 = { + maprs = maprs_popTS, [0] = "blezSB", "blezalcTB", "bgezalcTB", "bgeucSTB" +} +local map_pop07_r6 = { + maprs = maprs_popTS, [0] = "bgtzSB", "bgtzalcTB", "bltzalcTB", "bltucSTB" +} +local map_pop26_r6 = { + maprs = maprs_popTS, "blezcTB", "bgezcTB", "bgecSTB" +} +local map_pop27_r6 = { + maprs = maprs_popTS, "bgtzcTB", "bltzcTB", "bltcSTB" +} + +local function maprs_popS(rs, rt) + if rs == 0 then return 0 else return 1 end +end + +local map_pop66_r6 = { + maprs = maprs_popS, [0] = "jicTI", "beqzcSb" +} +local map_pop76_r6 = { + maprs = maprs_popS, [0] = "jialcTI", "bnezcSb" +} + +local function maprs_popST(rs, rt) + if rs >= rt then return 0 elseif rs == 0 then return 1 else return 2 end +end + +local map_pop10_r6 = { + maprs = maprs_popST, [0] = "bovcSTB", "beqzalcTB", "beqcSTB" +} +local map_pop30_r6 = { + maprs = maprs_popST, [0] = "bnvcSTB", "bnezalcTB", "bnecSTB" +} + +local map_pri_r6 = { + [0] = map_special_r6, map_regimm_r6, "jJ", "jalJ", + "beq|beqz|bST00B", "bne|bnezST0B", map_pop06_r6, map_pop07_r6, + map_pop10_r6, "addiu|liTS0I", "sltiTSI", "sltiuTSI", + "andiTSU", "ori|liTS0U", "xoriTSU", "aui|luiTS0U", + map_cop0, map_cop1_r6, false, false, + false, false, map_pop26_r6, map_pop27_r6, + map_pop30_r6, "daddiuTSI", false, false, + false, "dauiTSI", false, map_special3_r6, + "lbTSO", "lhTSO", false, "lwTSO", + "lbuTSO", "lhuTSO", false, false, + "sbTSO", "shTSO", false, "swTSO", + false, false, false, false, + false, "lwc1HSO", "bc#", false, + false, "ldc1HSO", map_pop66_r6, "ldTSO", + false, "swc1HSO", "balc#", map_pcrel_r6, + false, "sdc1HSO", map_pop76_r6, "sdTSO", +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15", + "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23", + "r24", "r25", "r26", "r27", "r28", "sp", "r30", "ra", +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then extra = "\t->"..sym end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-7s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-7s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +local function get_be(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + return bor(lshift(b0, 24), lshift(b1, 16), lshift(b2, 8), b3) +end + +local function get_le(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + return bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local op = ctx:get() + local operands = {} + local last = nil + ctx.op = op + ctx.rel = nil + + local opat = ctx.map_pri[rshift(op, 26)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + if opat.maprs then + opat = opat[opat.maprs(band(rshift(op,21),31), band(rshift(op,16),31))] + else + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + end + local name, pat = match(opat, "^([a-z0-9_.]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.|]*)(.*)") + if altname then pat = pat2 end + + for p in gmatch(pat, ".") do + local x = nil + if p == "S" then + x = map_gpr[band(rshift(op, 21), 31)] + elseif p == "T" then + x = map_gpr[band(rshift(op, 16), 31)] + elseif p == "D" then + x = map_gpr[band(rshift(op, 11), 31)] + elseif p == "F" then + x = "f"..band(rshift(op, 6), 31) + elseif p == "G" then + x = "f"..band(rshift(op, 11), 31) + elseif p == "H" then + x = "f"..band(rshift(op, 16), 31) + elseif p == "R" then + x = "f"..band(rshift(op, 21), 31) + elseif p == "A" then + x = band(rshift(op, 6), 31) + elseif p == "a" then + x = band(rshift(op, 6), 7) + elseif p == "E" then + x = band(rshift(op, 6), 31) + 32 + elseif p == "M" then + x = band(rshift(op, 11), 31) + elseif p == "N" then + x = band(rshift(op, 16), 31) + elseif p == "C" then + x = band(rshift(op, 18), 7) + if x == 0 then x = nil end + elseif p == "K" then + x = band(rshift(op, 11), 31) + 1 + elseif p == "P" then + x = band(rshift(op, 11), 31) + 33 + elseif p == "L" then + x = band(rshift(op, 11), 31) - last + 1 + elseif p == "Q" then + x = band(rshift(op, 11), 31) - last + 33 + elseif p == "I" then + x = arshift(lshift(op, 16), 16) + elseif p == "2" then + x = arshift(lshift(op, 13), 11) + elseif p == "3" then + x = arshift(lshift(op, 14), 11) + elseif p == "U" then + x = band(op, 0xffff) + elseif p == "O" then + local disp = arshift(lshift(op, 16), 16) + operands[#operands] = format("%d(%s)", disp, last) + elseif p == "X" then + local index = map_gpr[band(rshift(op, 16), 31)] + operands[#operands] = format("%s(%s)", index, last) + elseif p == "B" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 16), 14) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "b" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 11), 9) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "#" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 6), 4) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "J" then + local a = ctx.addr + ctx.pos + x = a - band(a, 0x0fffffff) + band(op, 0x03ffffff)*4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "V" then + x = band(rshift(op, 8), 7) + if x == 0 then x = nil end + elseif p == "W" then + x = band(op, 7) + if x == 0 then x = nil end + elseif p == "Y" then + x = band(rshift(op, 6), 0x000fffff) + if x == 0 then x = nil end + elseif p == "Z" then + x = band(rshift(op, 6), 1023) + if x == 0 then x = nil end + elseif p == "0" then + if last == "r0" or last == 0 then + local n = #operands + operands[n] = nil + last = operands[n-1] + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if a1 then name, altname = a1, a2 + else name = altname end + end + end + elseif p == "1" then + if last == "ra" then + operands[#operands] = nil + end + else + assert(false) + end + if x then operands[#operands+1] = x; last = x end + end + + return putop(ctx, name, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + stop = stop - stop % 4 + ctx.pos = ofs - ofs % 4 + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + ctx.get = get_be + ctx.map_pri = map_pri + return ctx +end + +local function create_el(code, addr, out) + local ctx = create(code, addr, out) + ctx.get = get_le + return ctx +end + +local function create_r6(code, addr, out) + local ctx = create(code, addr, out) + ctx.map_pri = map_pri_r6 + return ctx +end + +local function create_r6_el(code, addr, out) + local ctx = create(code, addr, out) + ctx.get = get_le + ctx.map_pri = map_pri_r6 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +local function disass_el(code, addr, out) + create_el(code, addr, out):disass() +end + +local function disass_r6(code, addr, out) + create_r6(code, addr, out):disass() +end + +local function disass_r6_el(code, addr, out) + create_r6_el(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_gpr[r] end + return "f"..(r-32) +end + +-- Public module functions. +return { + create = create, + create_el = create_el, + create_r6 = create_r6, + create_r6_el = create_r6_el, + disass = disass, + disass_el = disass_el, + disass_r6 = disass_r6, + disass_r6_el = disass_r6_el, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64.lua new file mode 100644 index 00000000..5ad48f8f --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the big-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create, + disass = dis_mips.disass, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64el.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64el.lua new file mode 100644 index 00000000..d50e3a18 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64el.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64EL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_el, + disass = dis_mips.disass_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64r6.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64r6.lua new file mode 100644 index 00000000..921b3cbe --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64r6.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64R6 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the r6 big-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_r6, + disass = dis_mips.disass_r6, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64r6el.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64r6el.lua new file mode 100644 index 00000000..aadef9f3 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mips64r6el.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64R6EL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the r6 little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_r6_el, + disass = dis_mips.disass_r6_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mipsel.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mipsel.lua new file mode 100644 index 00000000..52cebefb --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_mipsel.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPSEL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_el, + disass = dis_mips.disass_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_ppc.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_ppc.lua new file mode 100644 index 00000000..08d742f1 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_ppc.lua @@ -0,0 +1,591 @@ +---------------------------------------------------------------------------- +-- LuaJIT PPC disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT/X license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles all common, non-privileged 32/64 bit PowerPC instructions +-- plus the e500 SPE instructions and some Cell/Xenon extensions. +-- +-- NYI: VMX, VMX128 +------------------------------------------------------------------------------ + +local type = type +local byte, format = string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local concat = table.concat +local bit = require("bit") +local band, bor, tohex = bit.band, bit.bor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps +------------------------------------------------------------------------------ + +local map_crops = { + shift = 1, mask = 1023, + [0] = "mcrfXX", + [33] = "crnor|crnotCCC=", [129] = "crandcCCC", + [193] = "crxor|crclrCCC%", [225] = "crnandCCC", + [257] = "crandCCC", [289] = "creqv|crsetCCC%", + [417] = "crorcCCC", [449] = "cror|crmoveCCC=", + [16] = "b_lrKB", [528] = "b_ctrKB", + [150] = "isync", +} + +local map_rlwinm = setmetatable({ + shift = 0, mask = -1, +}, +{ __index = function(t, x) + local rot = band(rshift(x, 11), 31) + local mb = band(rshift(x, 6), 31) + local me = band(rshift(x, 1), 31) + if mb == 0 and me == 31-rot then + return "slwiRR~A." + elseif me == 31 and mb == 32-rot then + return "srwiRR~-A." + else + return "rlwinmRR~AAA." + end + end +}) + +local map_rld = { + shift = 2, mask = 7, + [0] = "rldiclRR~HM.", "rldicrRR~HM.", "rldicRR~HM.", "rldimiRR~HM.", + { + shift = 1, mask = 1, + [0] = "rldclRR~RM.", "rldcrRR~RM.", + }, +} + +local map_ext = setmetatable({ + shift = 1, mask = 1023, + + [0] = "cmp_YLRR", [32] = "cmpl_YLRR", + [4] = "twARR", [68] = "tdARR", + + [8] = "subfcRRR.", [40] = "subfRRR.", + [104] = "negRR.", [136] = "subfeRRR.", + [200] = "subfzeRR.", [232] = "subfmeRR.", + [520] = "subfcoRRR.", [552] = "subfoRRR.", + [616] = "negoRR.", [648] = "subfeoRRR.", + [712] = "subfzeoRR.", [744] = "subfmeoRR.", + + [9] = "mulhduRRR.", [73] = "mulhdRRR.", [233] = "mulldRRR.", + [457] = "divduRRR.", [489] = "divdRRR.", + [745] = "mulldoRRR.", + [969] = "divduoRRR.", [1001] = "divdoRRR.", + + [10] = "addcRRR.", [138] = "addeRRR.", + [202] = "addzeRR.", [234] = "addmeRR.", [266] = "addRRR.", + [522] = "addcoRRR.", [650] = "addeoRRR.", + [714] = "addzeoRR.", [746] = "addmeoRR.", [778] = "addoRRR.", + + [11] = "mulhwuRRR.", [75] = "mulhwRRR.", [235] = "mullwRRR.", + [459] = "divwuRRR.", [491] = "divwRRR.", + [747] = "mullwoRRR.", + [971] = "divwouRRR.", [1003] = "divwoRRR.", + + [15] = "iselltRRR", [47] = "iselgtRRR", [79] = "iseleqRRR", + + [144] = { shift = 20, mask = 1, [0] = "mtcrfRZ~", "mtocrfRZ~", }, + [19] = { shift = 20, mask = 1, [0] = "mfcrR", "mfocrfRZ", }, + [371] = { shift = 11, mask = 1023, [392] = "mftbR", [424] = "mftbuR", }, + [339] = { + shift = 11, mask = 1023, + [32] = "mferR", [256] = "mflrR", [288] = "mfctrR", [16] = "mfspefscrR", + }, + [467] = { + shift = 11, mask = 1023, + [32] = "mtxerR", [256] = "mtlrR", [288] = "mtctrR", [16] = "mtspefscrR", + }, + + [20] = "lwarxRR0R", [84] = "ldarxRR0R", + + [21] = "ldxRR0R", [53] = "lduxRRR", + [149] = "stdxRR0R", [181] = "stduxRRR", + [341] = "lwaxRR0R", [373] = "lwauxRRR", + + [23] = "lwzxRR0R", [55] = "lwzuxRRR", + [87] = "lbzxRR0R", [119] = "lbzuxRRR", + [151] = "stwxRR0R", [183] = "stwuxRRR", + [215] = "stbxRR0R", [247] = "stbuxRRR", + [279] = "lhzxRR0R", [311] = "lhzuxRRR", + [343] = "lhaxRR0R", [375] = "lhauxRRR", + [407] = "sthxRR0R", [439] = "sthuxRRR", + + [54] = "dcbst-R0R", [86] = "dcbf-R0R", + [150] = "stwcxRR0R.", [214] = "stdcxRR0R.", + [246] = "dcbtst-R0R", [278] = "dcbt-R0R", + [310] = "eciwxRR0R", [438] = "ecowxRR0R", + [470] = "dcbi-RR", + + [598] = { + shift = 21, mask = 3, + [0] = "sync", "lwsync", "ptesync", + }, + [758] = "dcba-RR", + [854] = "eieio", [982] = "icbi-R0R", [1014] = "dcbz-R0R", + + [26] = "cntlzwRR~", [58] = "cntlzdRR~", + [122] = "popcntbRR~", + [154] = "prtywRR~", [186] = "prtydRR~", + + [28] = "andRR~R.", [60] = "andcRR~R.", [124] = "nor|notRR~R=.", + [284] = "eqvRR~R.", [316] = "xorRR~R.", + [412] = "orcRR~R.", [444] = "or|mrRR~R=.", [476] = "nandRR~R.", + [508] = "cmpbRR~R", + + [512] = "mcrxrX", + + [532] = "ldbrxRR0R", [660] = "stdbrxRR0R", + + [533] = "lswxRR0R", [597] = "lswiRR0A", + [661] = "stswxRR0R", [725] = "stswiRR0A", + + [534] = "lwbrxRR0R", [662] = "stwbrxRR0R", + [790] = "lhbrxRR0R", [918] = "sthbrxRR0R", + + [535] = "lfsxFR0R", [567] = "lfsuxFRR", + [599] = "lfdxFR0R", [631] = "lfduxFRR", + [663] = "stfsxFR0R", [695] = "stfsuxFRR", + [727] = "stfdxFR0R", [759] = "stfduxFR0R", + [855] = "lfiwaxFR0R", + [983] = "stfiwxFR0R", + + [24] = "slwRR~R.", + + [27] = "sldRR~R.", [536] = "srwRR~R.", + [792] = "srawRR~R.", [824] = "srawiRR~A.", + + [794] = "sradRR~R.", [826] = "sradiRR~H.", [827] = "sradiRR~H.", + [922] = "extshRR~.", [954] = "extsbRR~.", [986] = "extswRR~.", + + [539] = "srdRR~R.", +}, +{ __index = function(t, x) + if band(x, 31) == 15 then return "iselRRRC" end + end +}) + +local map_ld = { + shift = 0, mask = 3, + [0] = "ldRRE", "lduRRE", "lwaRRE", +} + +local map_std = { + shift = 0, mask = 3, + [0] = "stdRRE", "stduRRE", +} + +local map_fps = { + shift = 5, mask = 1, + { + shift = 1, mask = 15, + [0] = false, false, "fdivsFFF.", false, + "fsubsFFF.", "faddsFFF.", "fsqrtsF-F.", false, + "fresF-F.", "fmulsFF-F.", "frsqrtesF-F.", false, + "fmsubsFFFF~.", "fmaddsFFFF~.", "fnmsubsFFFF~.", "fnmaddsFFFF~.", + } +} + +local map_fpd = { + shift = 5, mask = 1, + [0] = { + shift = 1, mask = 1023, + [0] = "fcmpuXFF", [32] = "fcmpoXFF", [64] = "mcrfsXX", + [38] = "mtfsb1A.", [70] = "mtfsb0A.", [134] = "mtfsfiA>>-A>", + [8] = "fcpsgnFFF.", [40] = "fnegF-F.", [72] = "fmrF-F.", + [136] = "fnabsF-F.", [264] = "fabsF-F.", + [12] = "frspF-F.", + [14] = "fctiwF-F.", [15] = "fctiwzF-F.", + [583] = "mffsF.", [711] = "mtfsfZF.", + [392] = "frinF-F.", [424] = "frizF-F.", + [456] = "fripF-F.", [488] = "frimF-F.", + [814] = "fctidF-F.", [815] = "fctidzF-F.", [846] = "fcfidF-F.", + }, + { + shift = 1, mask = 15, + [0] = false, false, "fdivFFF.", false, + "fsubFFF.", "faddFFF.", "fsqrtF-F.", "fselFFFF~.", + "freF-F.", "fmulFF-F.", "frsqrteF-F.", false, + "fmsubFFFF~.", "fmaddFFFF~.", "fnmsubFFFF~.", "fnmaddFFFF~.", + } +} + +local map_spe = { + shift = 0, mask = 2047, + + [512] = "evaddwRRR", [514] = "evaddiwRAR~", + [516] = "evsubwRRR~", [518] = "evsubiwRAR~", + [520] = "evabsRR", [521] = "evnegRR", + [522] = "evextsbRR", [523] = "evextshRR", [524] = "evrndwRR", + [525] = "evcntlzwRR", [526] = "evcntlswRR", + + [527] = "brincRRR", + + [529] = "evandRRR", [530] = "evandcRRR", [534] = "evxorRRR", + [535] = "evor|evmrRRR=", [536] = "evnor|evnotRRR=", + [537] = "eveqvRRR", [539] = "evorcRRR", [542] = "evnandRRR", + + [544] = "evsrwuRRR", [545] = "evsrwsRRR", + [546] = "evsrwiuRRA", [547] = "evsrwisRRA", + [548] = "evslwRRR", [550] = "evslwiRRA", + [552] = "evrlwRRR", [553] = "evsplatiRS", + [554] = "evrlwiRRA", [555] = "evsplatfiRS", + [556] = "evmergehiRRR", [557] = "evmergeloRRR", + [558] = "evmergehiloRRR", [559] = "evmergelohiRRR", + + [560] = "evcmpgtuYRR", [561] = "evcmpgtsYRR", + [562] = "evcmpltuYRR", [563] = "evcmpltsYRR", + [564] = "evcmpeqYRR", + + [632] = "evselRRR", [633] = "evselRRRW", + [634] = "evselRRRW", [635] = "evselRRRW", + [636] = "evselRRRW", [637] = "evselRRRW", + [638] = "evselRRRW", [639] = "evselRRRW", + + [640] = "evfsaddRRR", [641] = "evfssubRRR", + [644] = "evfsabsRR", [645] = "evfsnabsRR", [646] = "evfsnegRR", + [648] = "evfsmulRRR", [649] = "evfsdivRRR", + [652] = "evfscmpgtYRR", [653] = "evfscmpltYRR", [654] = "evfscmpeqYRR", + [656] = "evfscfuiR-R", [657] = "evfscfsiR-R", + [658] = "evfscfufR-R", [659] = "evfscfsfR-R", + [660] = "evfsctuiR-R", [661] = "evfsctsiR-R", + [662] = "evfsctufR-R", [663] = "evfsctsfR-R", + [664] = "evfsctuizR-R", [666] = "evfsctsizR-R", + [668] = "evfststgtYRR", [669] = "evfststltYRR", [670] = "evfststeqYRR", + + [704] = "efsaddRRR", [705] = "efssubRRR", + [708] = "efsabsRR", [709] = "efsnabsRR", [710] = "efsnegRR", + [712] = "efsmulRRR", [713] = "efsdivRRR", + [716] = "efscmpgtYRR", [717] = "efscmpltYRR", [718] = "efscmpeqYRR", + [719] = "efscfdR-R", + [720] = "efscfuiR-R", [721] = "efscfsiR-R", + [722] = "efscfufR-R", [723] = "efscfsfR-R", + [724] = "efsctuiR-R", [725] = "efsctsiR-R", + [726] = "efsctufR-R", [727] = "efsctsfR-R", + [728] = "efsctuizR-R", [730] = "efsctsizR-R", + [732] = "efststgtYRR", [733] = "efststltYRR", [734] = "efststeqYRR", + + [736] = "efdaddRRR", [737] = "efdsubRRR", + [738] = "efdcfuidR-R", [739] = "efdcfsidR-R", + [740] = "efdabsRR", [741] = "efdnabsRR", [742] = "efdnegRR", + [744] = "efdmulRRR", [745] = "efddivRRR", + [746] = "efdctuidzR-R", [747] = "efdctsidzR-R", + [748] = "efdcmpgtYRR", [749] = "efdcmpltYRR", [750] = "efdcmpeqYRR", + [751] = "efdcfsR-R", + [752] = "efdcfuiR-R", [753] = "efdcfsiR-R", + [754] = "efdcfufR-R", [755] = "efdcfsfR-R", + [756] = "efdctuiR-R", [757] = "efdctsiR-R", + [758] = "efdctufR-R", [759] = "efdctsfR-R", + [760] = "efdctuizR-R", [762] = "efdctsizR-R", + [764] = "efdtstgtYRR", [765] = "efdtstltYRR", [766] = "efdtsteqYRR", + + [768] = "evlddxRR0R", [769] = "evlddRR8", + [770] = "evldwxRR0R", [771] = "evldwRR8", + [772] = "evldhxRR0R", [773] = "evldhRR8", + [776] = "evlhhesplatxRR0R", [777] = "evlhhesplatRR2", + [780] = "evlhhousplatxRR0R", [781] = "evlhhousplatRR2", + [782] = "evlhhossplatxRR0R", [783] = "evlhhossplatRR2", + [784] = "evlwhexRR0R", [785] = "evlwheRR4", + [788] = "evlwhouxRR0R", [789] = "evlwhouRR4", + [790] = "evlwhosxRR0R", [791] = "evlwhosRR4", + [792] = "evlwwsplatxRR0R", [793] = "evlwwsplatRR4", + [796] = "evlwhsplatxRR0R", [797] = "evlwhsplatRR4", + + [800] = "evstddxRR0R", [801] = "evstddRR8", + [802] = "evstdwxRR0R", [803] = "evstdwRR8", + [804] = "evstdhxRR0R", [805] = "evstdhRR8", + [816] = "evstwhexRR0R", [817] = "evstwheRR4", + [820] = "evstwhoxRR0R", [821] = "evstwhoRR4", + [824] = "evstwwexRR0R", [825] = "evstwweRR4", + [828] = "evstwwoxRR0R", [829] = "evstwwoRR4", + + [1027] = "evmhessfRRR", [1031] = "evmhossfRRR", [1032] = "evmheumiRRR", + [1033] = "evmhesmiRRR", [1035] = "evmhesmfRRR", [1036] = "evmhoumiRRR", + [1037] = "evmhosmiRRR", [1039] = "evmhosmfRRR", [1059] = "evmhessfaRRR", + [1063] = "evmhossfaRRR", [1064] = "evmheumiaRRR", [1065] = "evmhesmiaRRR", + [1067] = "evmhesmfaRRR", [1068] = "evmhoumiaRRR", [1069] = "evmhosmiaRRR", + [1071] = "evmhosmfaRRR", [1095] = "evmwhssfRRR", [1096] = "evmwlumiRRR", + [1100] = "evmwhumiRRR", [1101] = "evmwhsmiRRR", [1103] = "evmwhsmfRRR", + [1107] = "evmwssfRRR", [1112] = "evmwumiRRR", [1113] = "evmwsmiRRR", + [1115] = "evmwsmfRRR", [1127] = "evmwhssfaRRR", [1128] = "evmwlumiaRRR", + [1132] = "evmwhumiaRRR", [1133] = "evmwhsmiaRRR", [1135] = "evmwhsmfaRRR", + [1139] = "evmwssfaRRR", [1144] = "evmwumiaRRR", [1145] = "evmwsmiaRRR", + [1147] = "evmwsmfaRRR", + + [1216] = "evaddusiaawRR", [1217] = "evaddssiaawRR", + [1218] = "evsubfusiaawRR", [1219] = "evsubfssiaawRR", + [1220] = "evmraRR", + [1222] = "evdivwsRRR", [1223] = "evdivwuRRR", + [1224] = "evaddumiaawRR", [1225] = "evaddsmiaawRR", + [1226] = "evsubfumiaawRR", [1227] = "evsubfsmiaawRR", + + [1280] = "evmheusiaawRRR", [1281] = "evmhessiaawRRR", + [1283] = "evmhessfaawRRR", [1284] = "evmhousiaawRRR", + [1285] = "evmhossiaawRRR", [1287] = "evmhossfaawRRR", + [1288] = "evmheumiaawRRR", [1289] = "evmhesmiaawRRR", + [1291] = "evmhesmfaawRRR", [1292] = "evmhoumiaawRRR", + [1293] = "evmhosmiaawRRR", [1295] = "evmhosmfaawRRR", + [1320] = "evmhegumiaaRRR", [1321] = "evmhegsmiaaRRR", + [1323] = "evmhegsmfaaRRR", [1324] = "evmhogumiaaRRR", + [1325] = "evmhogsmiaaRRR", [1327] = "evmhogsmfaaRRR", + [1344] = "evmwlusiaawRRR", [1345] = "evmwlssiaawRRR", + [1352] = "evmwlumiaawRRR", [1353] = "evmwlsmiaawRRR", + [1363] = "evmwssfaaRRR", [1368] = "evmwumiaaRRR", + [1369] = "evmwsmiaaRRR", [1371] = "evmwsmfaaRRR", + [1408] = "evmheusianwRRR", [1409] = "evmhessianwRRR", + [1411] = "evmhessfanwRRR", [1412] = "evmhousianwRRR", + [1413] = "evmhossianwRRR", [1415] = "evmhossfanwRRR", + [1416] = "evmheumianwRRR", [1417] = "evmhesmianwRRR", + [1419] = "evmhesmfanwRRR", [1420] = "evmhoumianwRRR", + [1421] = "evmhosmianwRRR", [1423] = "evmhosmfanwRRR", + [1448] = "evmhegumianRRR", [1449] = "evmhegsmianRRR", + [1451] = "evmhegsmfanRRR", [1452] = "evmhogumianRRR", + [1453] = "evmhogsmianRRR", [1455] = "evmhogsmfanRRR", + [1472] = "evmwlusianwRRR", [1473] = "evmwlssianwRRR", + [1480] = "evmwlumianwRRR", [1481] = "evmwlsmianwRRR", + [1491] = "evmwssfanRRR", [1496] = "evmwumianRRR", + [1497] = "evmwsmianRRR", [1499] = "evmwsmfanRRR", +} + +local map_pri = { + [0] = false, false, "tdiARI", "twiARI", + map_spe, false, false, "mulliRRI", + "subficRRI", false, "cmpl_iYLRU", "cmp_iYLRI", + "addicRRI", "addic.RRI", "addi|liRR0I", "addis|lisRR0I", + "b_KBJ", "sc", "bKJ", map_crops, + "rlwimiRR~AAA.", map_rlwinm, false, "rlwnmRR~RAA.", + "oriNRR~U", "orisRR~U", "xoriRR~U", "xorisRR~U", + "andi.RR~U", "andis.RR~U", map_rld, map_ext, + "lwzRRD", "lwzuRRD", "lbzRRD", "lbzuRRD", + "stwRRD", "stwuRRD", "stbRRD", "stbuRRD", + "lhzRRD", "lhzuRRD", "lhaRRD", "lhauRRD", + "sthRRD", "sthuRRD", "lmwRRD", "stmwRRD", + "lfsFRD", "lfsuFRD", "lfdFRD", "lfduFRD", + "stfsFRD", "stfsuFRD", "stfdFRD", "stfduFRD", + false, false, map_ld, map_fps, + false, false, map_std, map_fpd, +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "sp", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15", + "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23", + "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31", +} + +local map_cond = { [0] = "lt", "gt", "eq", "so", "ge", "le", "ne", "ns", } + +-- Format a condition bit. +local function condfmt(cond) + if cond <= 3 then + return map_cond[band(cond, 3)] + else + return format("4*cr%d+%s", rshift(cond, 2), map_cond[band(cond, 3)]) + end +end + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then extra = "\t->"..sym end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-7s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-7s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b0, 24), lshift(b1, 16), lshift(b2, 8), b3) + local operands = {} + local last = nil + local rs = 21 + ctx.op = op + ctx.rel = nil + + local opat = map_pri[rshift(b0, 2)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] + end + local name, pat = match(opat, "^([a-z0-9_.]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.]*)(.*)") + if altname then pat = pat2 end + + for p in gmatch(pat, ".") do + local x = nil + if p == "R" then + x = map_gpr[band(rshift(op, rs), 31)] + rs = rs - 5 + elseif p == "F" then + x = "f"..band(rshift(op, rs), 31) + rs = rs - 5 + elseif p == "A" then + x = band(rshift(op, rs), 31) + rs = rs - 5 + elseif p == "S" then + x = arshift(lshift(op, 27-rs), 27) + rs = rs - 5 + elseif p == "I" then + x = arshift(lshift(op, 16), 16) + elseif p == "U" then + x = band(op, 0xffff) + elseif p == "D" or p == "E" then + local disp = arshift(lshift(op, 16), 16) + if p == "E" then disp = band(disp, -4) end + if last == "r0" then last = "0" end + operands[#operands] = format("%d(%s)", disp, last) + elseif p >= "2" and p <= "8" then + local disp = band(rshift(op, rs), 31) * p + if last == "r0" then last = "0" end + operands[#operands] = format("%d(%s)", disp, last) + elseif p == "H" then + x = band(rshift(op, rs), 31) + lshift(band(op, 2), 4) + rs = rs - 5 + elseif p == "M" then + x = band(rshift(op, rs), 31) + band(op, 0x20) + elseif p == "C" then + x = condfmt(band(rshift(op, rs), 31)) + rs = rs - 5 + elseif p == "B" then + local bo = rshift(op, 21) + local cond = band(rshift(op, 16), 31) + local cn = "" + rs = rs - 10 + if band(bo, 4) == 0 then + cn = band(bo, 2) == 0 and "dnz" or "dz" + if band(bo, 0x10) == 0 then + cn = cn..(band(bo, 8) == 0 and "f" or "t") + end + if band(bo, 0x10) == 0 then x = condfmt(cond) end + name = name..(band(bo, 1) == band(rshift(op, 15), 1) and "-" or "+") + elseif band(bo, 0x10) == 0 then + cn = map_cond[band(cond, 3) + (band(bo, 8) == 0 and 4 or 0)] + if cond > 3 then x = "cr"..rshift(cond, 2) end + name = name..(band(bo, 1) == band(rshift(op, 15), 1) and "-" or "+") + end + name = gsub(name, "_", cn) + elseif p == "J" then + x = arshift(lshift(op, 27-rs), 29-rs)*4 + if band(op, 2) == 0 then x = ctx.addr + pos + x end + ctx.rel = x + x = "0x"..tohex(x) + elseif p == "K" then + if band(op, 1) ~= 0 then name = name.."l" end + if band(op, 2) ~= 0 then name = name.."a" end + elseif p == "X" or p == "Y" then + x = band(rshift(op, rs+2), 7) + if x == 0 and p == "Y" then x = nil else x = "cr"..x end + rs = rs - 5 + elseif p == "W" then + x = "cr"..band(op, 7) + elseif p == "Z" then + x = band(rshift(op, rs-4), 255) + rs = rs - 10 + elseif p == ">" then + operands[#operands] = rshift(operands[#operands], 1) + elseif p == "0" then + if last == "r0" then + operands[#operands] = nil + if altname then name = altname end + end + elseif p == "L" then + name = gsub(name, "_", band(op, 0x00200000) ~= 0 and "d" or "w") + elseif p == "." then + if band(op, 1) == 1 then name = name.."." end + elseif p == "N" then + if op == 0x60000000 then name = "nop"; break end + elseif p == "~" then + local n = #operands + operands[n-1], operands[n] = operands[n], operands[n-1] + elseif p == "=" then + local n = #operands + if last == operands[n-1] then + operands[n] = nil + name = altname + end + elseif p == "%" then + local n = #operands + if last == operands[n-1] and last == operands[n-2] then + operands[n] = nil + operands[n-1] = nil + name = altname + end + elseif p == "-" then + rs = rs - 5 + else + assert(false) + end + if x then operands[#operands+1] = x; last = x end + end + + return putop(ctx, name, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + stop = stop - stop % 4 + ctx.pos = ofs - ofs % 4 + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_gpr[r] end + return "f"..(r-32) +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_x64.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_x64.lua new file mode 100644 index 00000000..2d37423e --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_x64.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT x64 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the 64 bit functions from the combined +-- x86/x64 disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_x86 = require((string.match(..., ".*%.") or "").."dis_x86") +return { + create = dis_x86.create64, + disass = dis_x86.disass64, + regname = dis_x86.regname64 +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_x86.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_x86.lua new file mode 100644 index 00000000..5480854c --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dis_x86.lua @@ -0,0 +1,953 @@ +---------------------------------------------------------------------------- +-- LuaJIT x86/x64 disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- Sending small code snippets to an external disassembler and mixing the +-- output with our own stuff was too fragile. So I had to bite the bullet +-- and write yet another x86 disassembler. Oh well ... +-- +-- The output format is very similar to what ndisasm generates. But it has +-- been developed independently by looking at the opcode tables from the +-- Intel and AMD manuals. The supported instruction set is quite extensive +-- and reflects what a current generation Intel or AMD CPU implements in +-- 32 bit and 64 bit mode. Yes, this includes MMX, SSE, SSE2, SSE3, SSSE3, +-- SSE4.1, SSE4.2, SSE4a, AVX, AVX2 and even privileged and hypervisor +-- (VMX/SVM) instructions. +-- +-- Notes: +-- * The (useless) a16 prefix, 3DNow and pre-586 opcodes are unsupported. +-- * No attempt at optimization has been made -- it's fast enough for my needs. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local lower, rep = string.lower, string.rep +local bit = require("bit") +local tohex = bit.tohex + +-- Map for 1st opcode byte in 32 bit mode. Ugly? Well ... read on. +local map_opc1_32 = { +--0x +[0]="addBmr","addVmr","addBrm","addVrm","addBai","addVai","push es","pop es", +"orBmr","orVmr","orBrm","orVrm","orBai","orVai","push cs","opc2*", +--1x +"adcBmr","adcVmr","adcBrm","adcVrm","adcBai","adcVai","push ss","pop ss", +"sbbBmr","sbbVmr","sbbBrm","sbbVrm","sbbBai","sbbVai","push ds","pop ds", +--2x +"andBmr","andVmr","andBrm","andVrm","andBai","andVai","es:seg","daa", +"subBmr","subVmr","subBrm","subVrm","subBai","subVai","cs:seg","das", +--3x +"xorBmr","xorVmr","xorBrm","xorVrm","xorBai","xorVai","ss:seg","aaa", +"cmpBmr","cmpVmr","cmpBrm","cmpVrm","cmpBai","cmpVai","ds:seg","aas", +--4x +"incVR","incVR","incVR","incVR","incVR","incVR","incVR","incVR", +"decVR","decVR","decVR","decVR","decVR","decVR","decVR","decVR", +--5x +"pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR", +"popUR","popUR","popUR","popUR","popUR","popUR","popUR","popUR", +--6x +"sz*pushaw,pusha","sz*popaw,popa","boundVrm","arplWmr", +"fs:seg","gs:seg","o16:","a16", +"pushUi","imulVrmi","pushBs","imulVrms", +"insb","insVS","outsb","outsVS", +--7x +"joBj","jnoBj","jbBj","jnbBj","jzBj","jnzBj","jbeBj","jaBj", +"jsBj","jnsBj","jpeBj","jpoBj","jlBj","jgeBj","jleBj","jgBj", +--8x +"arith!Bmi","arith!Vmi","arith!Bmi","arith!Vms", +"testBmr","testVmr","xchgBrm","xchgVrm", +"movBmr","movVmr","movBrm","movVrm", +"movVmg","leaVrm","movWgm","popUm", +--9x +"nop*xchgVaR|pause|xchgWaR|repne nop","xchgVaR","xchgVaR","xchgVaR", +"xchgVaR","xchgVaR","xchgVaR","xchgVaR", +"sz*cbw,cwde,cdqe","sz*cwd,cdq,cqo","call farViw","wait", +"sz*pushfw,pushf","sz*popfw,popf","sahf","lahf", +--Ax +"movBao","movVao","movBoa","movVoa", +"movsb","movsVS","cmpsb","cmpsVS", +"testBai","testVai","stosb","stosVS", +"lodsb","lodsVS","scasb","scasVS", +--Bx +"movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi", +"movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI", +--Cx +"shift!Bmu","shift!Vmu","retBw","ret","vex*3$lesVrm","vex*2$ldsVrm","movBmi","movVmi", +"enterBwu","leave","retfBw","retf","int3","intBu","into","iretVS", +--Dx +"shift!Bm1","shift!Vm1","shift!Bmc","shift!Vmc","aamBu","aadBu","salc","xlatb", +"fp*0","fp*1","fp*2","fp*3","fp*4","fp*5","fp*6","fp*7", +--Ex +"loopneBj","loopeBj","loopBj","sz*jcxzBj,jecxzBj,jrcxzBj", +"inBau","inVau","outBua","outVua", +"callVj","jmpVj","jmp farViw","jmpBj","inBad","inVad","outBda","outVda", +--Fx +"lock:","int1","repne:rep","rep:","hlt","cmc","testb!Bm","testv!Vm", +"clc","stc","cli","sti","cld","std","incb!Bm","incd!Vm", +} +assert(#map_opc1_32 == 255) + +-- Map for 1st opcode byte in 64 bit mode (overrides only). +local map_opc1_64 = setmetatable({ + [0x06]=false, [0x07]=false, [0x0e]=false, + [0x16]=false, [0x17]=false, [0x1e]=false, [0x1f]=false, + [0x27]=false, [0x2f]=false, [0x37]=false, [0x3f]=false, + [0x60]=false, [0x61]=false, [0x62]=false, [0x63]="movsxdVrDmt", [0x67]="a32:", + [0x40]="rex*", [0x41]="rex*b", [0x42]="rex*x", [0x43]="rex*xb", + [0x44]="rex*r", [0x45]="rex*rb", [0x46]="rex*rx", [0x47]="rex*rxb", + [0x48]="rex*w", [0x49]="rex*wb", [0x4a]="rex*wx", [0x4b]="rex*wxb", + [0x4c]="rex*wr", [0x4d]="rex*wrb", [0x4e]="rex*wrx", [0x4f]="rex*wrxb", + [0x82]=false, [0x9a]=false, [0xc4]="vex*3", [0xc5]="vex*2", [0xce]=false, + [0xd4]=false, [0xd5]=false, [0xd6]=false, [0xea]=false, +}, { __index = map_opc1_32 }) + +-- Map for 2nd opcode byte (0F xx). True CISC hell. Hey, I told you. +-- Prefix dependent MMX/SSE opcodes: (none)|rep|o16|repne, -|F3|66|F2 +local map_opc2 = { +--0x +[0]="sldt!Dmp","sgdt!Ump","larVrm","lslVrm",nil,"syscall","clts","sysret", +"invd","wbinvd",nil,"ud1",nil,"$prefetch!Bm","femms","3dnowMrmu", +--1x +"movupsXrm|movssXrvm|movupdXrm|movsdXrvm", +"movupsXmr|movssXmvr|movupdXmr|movsdXmvr", +"movhlpsXrm$movlpsXrm|movsldupXrm|movlpdXrm|movddupXrm", +"movlpsXmr||movlpdXmr", +"unpcklpsXrvm||unpcklpdXrvm", +"unpckhpsXrvm||unpckhpdXrvm", +"movlhpsXrm$movhpsXrm|movshdupXrm|movhpdXrm", +"movhpsXmr||movhpdXmr", +"$prefetcht!Bm","hintnopVm","hintnopVm","hintnopVm", +"hintnopVm","hintnopVm","hintnopVm","hintnopVm", +--2x +"movUmx$","movUmy$","movUxm$","movUym$","movUmz$",nil,"movUzm$",nil, +"movapsXrm||movapdXrm", +"movapsXmr||movapdXmr", +"cvtpi2psXrMm|cvtsi2ssXrvVmt|cvtpi2pdXrMm|cvtsi2sdXrvVmt", +"movntpsXmr|movntssXmr|movntpdXmr|movntsdXmr", +"cvttps2piMrXm|cvttss2siVrXm|cvttpd2piMrXm|cvttsd2siVrXm", +"cvtps2piMrXm|cvtss2siVrXm|cvtpd2piMrXm|cvtsd2siVrXm", +"ucomissXrm||ucomisdXrm", +"comissXrm||comisdXrm", +--3x +"wrmsr","rdtsc","rdmsr","rdpmc","sysenter","sysexit",nil,"getsec", +"opc3*38",nil,"opc3*3a",nil,nil,nil,nil,nil, +--4x +"cmovoVrm","cmovnoVrm","cmovbVrm","cmovnbVrm", +"cmovzVrm","cmovnzVrm","cmovbeVrm","cmovaVrm", +"cmovsVrm","cmovnsVrm","cmovpeVrm","cmovpoVrm", +"cmovlVrm","cmovgeVrm","cmovleVrm","cmovgVrm", +--5x +"movmskpsVrXm$||movmskpdVrXm$","sqrtpsXrm|sqrtssXrm|sqrtpdXrm|sqrtsdXrm", +"rsqrtpsXrm|rsqrtssXrvm","rcppsXrm|rcpssXrvm", +"andpsXrvm||andpdXrvm","andnpsXrvm||andnpdXrvm", +"orpsXrvm||orpdXrvm","xorpsXrvm||xorpdXrvm", +"addpsXrvm|addssXrvm|addpdXrvm|addsdXrvm","mulpsXrvm|mulssXrvm|mulpdXrvm|mulsdXrvm", +"cvtps2pdXrm|cvtss2sdXrvm|cvtpd2psXrm|cvtsd2ssXrvm", +"cvtdq2psXrm|cvttps2dqXrm|cvtps2dqXrm", +"subpsXrvm|subssXrvm|subpdXrvm|subsdXrvm","minpsXrvm|minssXrvm|minpdXrvm|minsdXrvm", +"divpsXrvm|divssXrvm|divpdXrvm|divsdXrvm","maxpsXrvm|maxssXrvm|maxpdXrvm|maxsdXrvm", +--6x +"punpcklbwPrvm","punpcklwdPrvm","punpckldqPrvm","packsswbPrvm", +"pcmpgtbPrvm","pcmpgtwPrvm","pcmpgtdPrvm","packuswbPrvm", +"punpckhbwPrvm","punpckhwdPrvm","punpckhdqPrvm","packssdwPrvm", +"||punpcklqdqXrvm","||punpckhqdqXrvm", +"movPrVSm","movqMrm|movdquXrm|movdqaXrm", +--7x +"pshufwMrmu|pshufhwXrmu|pshufdXrmu|pshuflwXrmu","pshiftw!Pvmu", +"pshiftd!Pvmu","pshiftq!Mvmu||pshiftdq!Xvmu", +"pcmpeqbPrvm","pcmpeqwPrvm","pcmpeqdPrvm","emms*|", +"vmreadUmr||extrqXmuu$|insertqXrmuu$","vmwriteUrm||extrqXrm$|insertqXrm$", +nil,nil, +"||haddpdXrvm|haddpsXrvm","||hsubpdXrvm|hsubpsXrvm", +"movVSmMr|movqXrm|movVSmXr","movqMmr|movdquXmr|movdqaXmr", +--8x +"joVj","jnoVj","jbVj","jnbVj","jzVj","jnzVj","jbeVj","jaVj", +"jsVj","jnsVj","jpeVj","jpoVj","jlVj","jgeVj","jleVj","jgVj", +--9x +"setoBm","setnoBm","setbBm","setnbBm","setzBm","setnzBm","setbeBm","setaBm", +"setsBm","setnsBm","setpeBm","setpoBm","setlBm","setgeBm","setleBm","setgBm", +--Ax +"push fs","pop fs","cpuid","btVmr","shldVmru","shldVmrc",nil,nil, +"push gs","pop gs","rsm","btsVmr","shrdVmru","shrdVmrc","fxsave!Dmp","imulVrm", +--Bx +"cmpxchgBmr","cmpxchgVmr","$lssVrm","btrVmr", +"$lfsVrm","$lgsVrm","movzxVrBmt","movzxVrWmt", +"|popcntVrm","ud2Dp","bt!Vmu","btcVmr", +"bsfVrm","bsrVrm|lzcntVrm|bsrWrm","movsxVrBmt","movsxVrWmt", +--Cx +"xaddBmr","xaddVmr", +"cmppsXrvmu|cmpssXrvmu|cmppdXrvmu|cmpsdXrvmu","$movntiVmr|", +"pinsrwPrvWmu","pextrwDrPmu", +"shufpsXrvmu||shufpdXrvmu","$cmpxchg!Qmp", +"bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR", +--Dx +"||addsubpdXrvm|addsubpsXrvm","psrlwPrvm","psrldPrvm","psrlqPrvm", +"paddqPrvm","pmullwPrvm", +"|movq2dqXrMm|movqXmr|movdq2qMrXm$","pmovmskbVrMm||pmovmskbVrXm", +"psubusbPrvm","psubuswPrvm","pminubPrvm","pandPrvm", +"paddusbPrvm","padduswPrvm","pmaxubPrvm","pandnPrvm", +--Ex +"pavgbPrvm","psrawPrvm","psradPrvm","pavgwPrvm", +"pmulhuwPrvm","pmulhwPrvm", +"|cvtdq2pdXrm|cvttpd2dqXrm|cvtpd2dqXrm","$movntqMmr||$movntdqXmr", +"psubsbPrvm","psubswPrvm","pminswPrvm","porPrvm", +"paddsbPrvm","paddswPrvm","pmaxswPrvm","pxorPrvm", +--Fx +"|||lddquXrm","psllwPrvm","pslldPrvm","psllqPrvm", +"pmuludqPrvm","pmaddwdPrvm","psadbwPrvm","maskmovqMrm||maskmovdquXrm$", +"psubbPrvm","psubwPrvm","psubdPrvm","psubqPrvm", +"paddbPrvm","paddwPrvm","padddPrvm","ud", +} +assert(map_opc2[255] == "ud") + +-- Map for three-byte opcodes. Can't wait for their next invention. +local map_opc3 = { +["38"] = { -- [66] 0f 38 xx +--0x +[0]="pshufbPrvm","phaddwPrvm","phadddPrvm","phaddswPrvm", +"pmaddubswPrvm","phsubwPrvm","phsubdPrvm","phsubswPrvm", +"psignbPrvm","psignwPrvm","psigndPrvm","pmulhrswPrvm", +"||permilpsXrvm","||permilpdXrvm",nil,nil, +--1x +"||pblendvbXrma",nil,nil,nil, +"||blendvpsXrma","||blendvpdXrma","||permpsXrvm","||ptestXrm", +"||broadcastssXrm","||broadcastsdXrm","||broadcastf128XrlXm",nil, +"pabsbPrm","pabswPrm","pabsdPrm",nil, +--2x +"||pmovsxbwXrm","||pmovsxbdXrm","||pmovsxbqXrm","||pmovsxwdXrm", +"||pmovsxwqXrm","||pmovsxdqXrm",nil,nil, +"||pmuldqXrvm","||pcmpeqqXrvm","||$movntdqaXrm","||packusdwXrvm", +"||maskmovpsXrvm","||maskmovpdXrvm","||maskmovpsXmvr","||maskmovpdXmvr", +--3x +"||pmovzxbwXrm","||pmovzxbdXrm","||pmovzxbqXrm","||pmovzxwdXrm", +"||pmovzxwqXrm","||pmovzxdqXrm","||permdXrvm","||pcmpgtqXrvm", +"||pminsbXrvm","||pminsdXrvm","||pminuwXrvm","||pminudXrvm", +"||pmaxsbXrvm","||pmaxsdXrvm","||pmaxuwXrvm","||pmaxudXrvm", +--4x +"||pmulddXrvm","||phminposuwXrm",nil,nil, +nil,"||psrlvVSXrvm","||psravdXrvm","||psllvVSXrvm", +--5x +[0x58] = "||pbroadcastdXrlXm",[0x59] = "||pbroadcastqXrlXm", +[0x5a] = "||broadcasti128XrlXm", +--7x +[0x78] = "||pbroadcastbXrlXm",[0x79] = "||pbroadcastwXrlXm", +--8x +[0x8c] = "||pmaskmovXrvVSm", +[0x8e] = "||pmaskmovVSmXvr", +--9x +[0x96] = "||fmaddsub132pHXrvm",[0x97] = "||fmsubadd132pHXrvm", +[0x98] = "||fmadd132pHXrvm",[0x99] = "||fmadd132sHXrvm", +[0x9a] = "||fmsub132pHXrvm",[0x9b] = "||fmsub132sHXrvm", +[0x9c] = "||fnmadd132pHXrvm",[0x9d] = "||fnmadd132sHXrvm", +[0x9e] = "||fnmsub132pHXrvm",[0x9f] = "||fnmsub132sHXrvm", +--Ax +[0xa6] = "||fmaddsub213pHXrvm",[0xa7] = "||fmsubadd213pHXrvm", +[0xa8] = "||fmadd213pHXrvm",[0xa9] = "||fmadd213sHXrvm", +[0xaa] = "||fmsub213pHXrvm",[0xab] = "||fmsub213sHXrvm", +[0xac] = "||fnmadd213pHXrvm",[0xad] = "||fnmadd213sHXrvm", +[0xae] = "||fnmsub213pHXrvm",[0xaf] = "||fnmsub213sHXrvm", +--Bx +[0xb6] = "||fmaddsub231pHXrvm",[0xb7] = "||fmsubadd231pHXrvm", +[0xb8] = "||fmadd231pHXrvm",[0xb9] = "||fmadd231sHXrvm", +[0xba] = "||fmsub231pHXrvm",[0xbb] = "||fmsub231sHXrvm", +[0xbc] = "||fnmadd231pHXrvm",[0xbd] = "||fnmadd231sHXrvm", +[0xbe] = "||fnmsub231pHXrvm",[0xbf] = "||fnmsub231sHXrvm", +--Dx +[0xdc] = "||aesencXrvm", [0xdd] = "||aesenclastXrvm", +[0xde] = "||aesdecXrvm", [0xdf] = "||aesdeclastXrvm", +--Fx +[0xf0] = "|||crc32TrBmt",[0xf1] = "|||crc32TrVmt", +[0xf7] = "| sarxVrmv| shlxVrmv| shrxVrmv", +}, + +["3a"] = { -- [66] 0f 3a xx +--0x +[0x00]="||permqXrmu","||permpdXrmu","||pblenddXrvmu",nil, +"||permilpsXrmu","||permilpdXrmu","||perm2f128Xrvmu",nil, +"||roundpsXrmu","||roundpdXrmu","||roundssXrvmu","||roundsdXrvmu", +"||blendpsXrvmu","||blendpdXrvmu","||pblendwXrvmu","palignrPrvmu", +--1x +nil,nil,nil,nil, +"||pextrbVmXru","||pextrwVmXru","||pextrVmSXru","||extractpsVmXru", +"||insertf128XrvlXmu","||extractf128XlXmYru",nil,nil, +nil,nil,nil,nil, +--2x +"||pinsrbXrvVmu","||insertpsXrvmu","||pinsrXrvVmuS",nil, +--3x +[0x38] = "||inserti128Xrvmu",[0x39] = "||extracti128XlXmYru", +--4x +[0x40] = "||dppsXrvmu", +[0x41] = "||dppdXrvmu", +[0x42] = "||mpsadbwXrvmu", +[0x44] = "||pclmulqdqXrvmu", +[0x46] = "||perm2i128Xrvmu", +[0x4a] = "||blendvpsXrvmb",[0x4b] = "||blendvpdXrvmb", +[0x4c] = "||pblendvbXrvmb", +--6x +[0x60] = "||pcmpestrmXrmu",[0x61] = "||pcmpestriXrmu", +[0x62] = "||pcmpistrmXrmu",[0x63] = "||pcmpistriXrmu", +[0xdf] = "||aeskeygenassistXrmu", +--Fx +[0xf0] = "||| rorxVrmu", +}, +} + +-- Map for VMX/SVM opcodes 0F 01 C0-FF (sgdt group with register operands). +local map_opcvm = { +[0xc1]="vmcall",[0xc2]="vmlaunch",[0xc3]="vmresume",[0xc4]="vmxoff", +[0xc8]="monitor",[0xc9]="mwait", +[0xd8]="vmrun",[0xd9]="vmmcall",[0xda]="vmload",[0xdb]="vmsave", +[0xdc]="stgi",[0xdd]="clgi",[0xde]="skinit",[0xdf]="invlpga", +[0xf8]="swapgs",[0xf9]="rdtscp", +} + +-- Map for FP opcodes. And you thought stack machines are simple? +local map_opcfp = { +-- D8-DF 00-BF: opcodes with a memory operand. +-- D8 +[0]="faddFm","fmulFm","fcomFm","fcompFm","fsubFm","fsubrFm","fdivFm","fdivrFm", +"fldFm",nil,"fstFm","fstpFm","fldenvVm","fldcwWm","fnstenvVm","fnstcwWm", +-- DA +"fiaddDm","fimulDm","ficomDm","ficompDm", +"fisubDm","fisubrDm","fidivDm","fidivrDm", +-- DB +"fildDm","fisttpDm","fistDm","fistpDm",nil,"fld twordFmp",nil,"fstp twordFmp", +-- DC +"faddGm","fmulGm","fcomGm","fcompGm","fsubGm","fsubrGm","fdivGm","fdivrGm", +-- DD +"fldGm","fisttpQm","fstGm","fstpGm","frstorDmp",nil,"fnsaveDmp","fnstswWm", +-- DE +"fiaddWm","fimulWm","ficomWm","ficompWm", +"fisubWm","fisubrWm","fidivWm","fidivrWm", +-- DF +"fildWm","fisttpWm","fistWm","fistpWm", +"fbld twordFmp","fildQm","fbstp twordFmp","fistpQm", +-- xx C0-FF: opcodes with a pseudo-register operand. +-- D8 +"faddFf","fmulFf","fcomFf","fcompFf","fsubFf","fsubrFf","fdivFf","fdivrFf", +-- D9 +"fldFf","fxchFf",{"fnop"},nil, +{"fchs","fabs",nil,nil,"ftst","fxam"}, +{"fld1","fldl2t","fldl2e","fldpi","fldlg2","fldln2","fldz"}, +{"f2xm1","fyl2x","fptan","fpatan","fxtract","fprem1","fdecstp","fincstp"}, +{"fprem","fyl2xp1","fsqrt","fsincos","frndint","fscale","fsin","fcos"}, +-- DA +"fcmovbFf","fcmoveFf","fcmovbeFf","fcmovuFf",nil,{nil,"fucompp"},nil,nil, +-- DB +"fcmovnbFf","fcmovneFf","fcmovnbeFf","fcmovnuFf", +{nil,nil,"fnclex","fninit"},"fucomiFf","fcomiFf",nil, +-- DC +"fadd toFf","fmul toFf",nil,nil, +"fsub toFf","fsubr toFf","fdivr toFf","fdiv toFf", +-- DD +"ffreeFf",nil,"fstFf","fstpFf","fucomFf","fucompFf",nil,nil, +-- DE +"faddpFf","fmulpFf",nil,{nil,"fcompp"}, +"fsubrpFf","fsubpFf","fdivrpFf","fdivpFf", +-- DF +nil,nil,nil,nil,{"fnstsw ax"},"fucomipFf","fcomipFf",nil, +} +assert(map_opcfp[126] == "fcomipFf") + +-- Map for opcode groups. The subkey is sp from the ModRM byte. +local map_opcgroup = { + arith = { "add", "or", "adc", "sbb", "and", "sub", "xor", "cmp" }, + shift = { "rol", "ror", "rcl", "rcr", "shl", "shr", "sal", "sar" }, + testb = { "testBmi", "testBmi", "not", "neg", "mul", "imul", "div", "idiv" }, + testv = { "testVmi", "testVmi", "not", "neg", "mul", "imul", "div", "idiv" }, + incb = { "inc", "dec" }, + incd = { "inc", "dec", "callUmp", "$call farDmp", + "jmpUmp", "$jmp farDmp", "pushUm" }, + sldt = { "sldt", "str", "lldt", "ltr", "verr", "verw" }, + sgdt = { "vm*$sgdt", "vm*$sidt", "$lgdt", "vm*$lidt", + "smsw", nil, "lmsw", "vm*$invlpg" }, + bt = { nil, nil, nil, nil, "bt", "bts", "btr", "btc" }, + cmpxchg = { nil, "sz*,cmpxchg8bQmp,cmpxchg16bXmp", nil, nil, + nil, nil, "vmptrld|vmxon|vmclear", "vmptrst" }, + pshiftw = { nil, nil, "psrlw", nil, "psraw", nil, "psllw" }, + pshiftd = { nil, nil, "psrld", nil, "psrad", nil, "pslld" }, + pshiftq = { nil, nil, "psrlq", nil, nil, nil, "psllq" }, + pshiftdq = { nil, nil, "psrlq", "psrldq", nil, nil, "psllq", "pslldq" }, + fxsave = { "$fxsave", "$fxrstor", "$ldmxcsr", "$stmxcsr", + nil, "lfenceDp$", "mfenceDp$", "sfenceDp$clflush" }, + prefetch = { "prefetch", "prefetchw" }, + prefetcht = { "prefetchnta", "prefetcht0", "prefetcht1", "prefetcht2" }, +} + +------------------------------------------------------------------------------ + +-- Maps for register names. +local map_regs = { + B = { "al", "cl", "dl", "bl", "ah", "ch", "dh", "bh", + "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, + B64 = { "al", "cl", "dl", "bl", "spl", "bpl", "sil", "dil", + "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, + W = { "ax", "cx", "dx", "bx", "sp", "bp", "si", "di", + "r8w", "r9w", "r10w", "r11w", "r12w", "r13w", "r14w", "r15w" }, + D = { "eax", "ecx", "edx", "ebx", "esp", "ebp", "esi", "edi", + "r8d", "r9d", "r10d", "r11d", "r12d", "r13d", "r14d", "r15d" }, + Q = { "rax", "rcx", "rdx", "rbx", "rsp", "rbp", "rsi", "rdi", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15" }, + M = { "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7", + "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7" }, -- No x64 ext! + X = { "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7", + "xmm8", "xmm9", "xmm10", "xmm11", "xmm12", "xmm13", "xmm14", "xmm15" }, + Y = { "ymm0", "ymm1", "ymm2", "ymm3", "ymm4", "ymm5", "ymm6", "ymm7", + "ymm8", "ymm9", "ymm10", "ymm11", "ymm12", "ymm13", "ymm14", "ymm15" }, +} +local map_segregs = { "es", "cs", "ss", "ds", "fs", "gs", "segr6", "segr7" } + +-- Maps for size names. +local map_sz2n = { + B = 1, W = 2, D = 4, Q = 8, M = 8, X = 16, Y = 32, +} +local map_sz2prefix = { + B = "byte", W = "word", D = "dword", + Q = "qword", + M = "qword", X = "xword", Y = "yword", + F = "dword", G = "qword", -- No need for sizes/register names for these two. +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local code, pos, hex = ctx.code, ctx.pos, "" + local hmax = ctx.hexdump + if hmax > 0 then + for i=ctx.start,pos-1 do + hex = hex..format("%02X", byte(code, i, i)) + end + if #hex > hmax then hex = sub(hex, 1, hmax)..". " + else hex = hex..rep(" ", hmax-#hex+2) end + end + if operands then text = text.." "..operands end + if ctx.o16 then text = "o16 "..text; ctx.o16 = false end + if ctx.a32 then text = "a32 "..text; ctx.a32 = false end + if ctx.rep then text = ctx.rep.." "..text; ctx.rep = false end + if ctx.rex then + local t = (ctx.rexw and "w" or "")..(ctx.rexr and "r" or "").. + (ctx.rexx and "x" or "")..(ctx.rexb and "b" or "").. + (ctx.vexl and "l" or "") + if ctx.vexv and ctx.vexv ~= 0 then t = t.."v"..ctx.vexv end + if t ~= "" then text = ctx.rex.."."..t.." "..gsub(text, "^ ", "") + elseif ctx.rex == "vex" then text = gsub("v"..text, "^v ", "") end + ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false + ctx.rex = false; ctx.vexl = false; ctx.vexv = false + end + if ctx.seg then + local text2, n = gsub(text, "%[", "["..ctx.seg..":") + if n == 0 then text = ctx.seg.." "..text else text = text2 end + ctx.seg = false + end + if ctx.lock then text = "lock "..text; ctx.lock = false end + local imm = ctx.imm + if imm then + local sym = ctx.symtab[imm] + if sym then text = text.."\t->"..sym end + end + ctx.out(format("%08x %s%s\n", ctx.addr+ctx.start, hex, text)) + ctx.mrm = false + ctx.vexv = false + ctx.start = pos + ctx.imm = nil +end + +-- Clear all prefix flags. +local function clearprefixes(ctx) + ctx.o16 = false; ctx.seg = false; ctx.lock = false; ctx.rep = false + ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false + ctx.rex = false; ctx.a32 = false; ctx.vexl = false +end + +-- Fallback for incomplete opcodes at the end. +local function incomplete(ctx) + ctx.pos = ctx.stop+1 + clearprefixes(ctx) + return putop(ctx, "(incomplete)") +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + clearprefixes(ctx) + return putop(ctx, "(unknown)") +end + +-- Return an immediate of the specified size. +local function getimm(ctx, pos, n) + if pos+n-1 > ctx.stop then return incomplete(ctx) end + local code = ctx.code + if n == 1 then + local b1 = byte(code, pos, pos) + return b1 + elseif n == 2 then + local b1, b2 = byte(code, pos, pos+1) + return b1+b2*256 + else + local b1, b2, b3, b4 = byte(code, pos, pos+3) + local imm = b1+b2*256+b3*65536+b4*16777216 + ctx.imm = imm + return imm + end +end + +-- Process pattern string and generate the operands. +local function putpat(ctx, name, pat) + local operands, regs, sz, mode, sp, rm, sc, rx, sdisp + local code, pos, stop, vexl = ctx.code, ctx.pos, ctx.stop, ctx.vexl + + -- Chars used: 1DFGHIMPQRSTUVWXYabcdfgijlmoprstuvwxyz + for p in gmatch(pat, ".") do + local x = nil + if p == "V" or p == "U" then + if ctx.rexw then sz = "Q"; ctx.rexw = false + elseif ctx.o16 then sz = "W"; ctx.o16 = false + elseif p == "U" and ctx.x64 then sz = "Q" + else sz = "D" end + regs = map_regs[sz] + elseif p == "T" then + if ctx.rexw then sz = "Q"; ctx.rexw = false else sz = "D" end + regs = map_regs[sz] + elseif p == "B" then + sz = "B" + regs = ctx.rex and map_regs.B64 or map_regs.B + elseif match(p, "[WDQMXYFG]") then + sz = p + if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end + regs = map_regs[sz] + elseif p == "P" then + sz = ctx.o16 and "X" or "M"; ctx.o16 = false + if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end + regs = map_regs[sz] + elseif p == "H" then + name = name..(ctx.rexw and "d" or "s") + ctx.rexw = false + elseif p == "S" then + name = name..lower(sz) + elseif p == "s" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = imm <= 127 and format("+0x%02x", imm) + or format("-0x%02x", 256-imm) + pos = pos+1 + elseif p == "u" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = format("0x%02x", imm) + pos = pos+1 + elseif p == "b" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = regs[imm/16+1] + pos = pos+1 + elseif p == "w" then + local imm = getimm(ctx, pos, 2); if not imm then return end + x = format("0x%x", imm) + pos = pos+2 + elseif p == "o" then -- [offset] + if ctx.x64 then + local imm1 = getimm(ctx, pos, 4); if not imm1 then return end + local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end + x = format("[0x%08x%08x]", imm2, imm1) + pos = pos+8 + else + local imm = getimm(ctx, pos, 4); if not imm then return end + x = format("[0x%08x]", imm) + pos = pos+4 + end + elseif p == "i" or p == "I" then + local n = map_sz2n[sz] + if n == 8 and ctx.x64 and p == "I" then + local imm1 = getimm(ctx, pos, 4); if not imm1 then return end + local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end + x = format("0x%08x%08x", imm2, imm1) + else + if n == 8 then n = 4 end + local imm = getimm(ctx, pos, n); if not imm then return end + if sz == "Q" and (imm < 0 or imm > 0x7fffffff) then + imm = (0xffffffff+1)-imm + x = format(imm > 65535 and "-0x%08x" or "-0x%x", imm) + else + x = format(imm > 65535 and "0x%08x" or "0x%x", imm) + end + end + pos = pos+n + elseif p == "j" then + local n = map_sz2n[sz] + if n == 8 then n = 4 end + local imm = getimm(ctx, pos, n); if not imm then return end + if sz == "B" and imm > 127 then imm = imm-256 + elseif imm > 2147483647 then imm = imm-4294967296 end + pos = pos+n + imm = imm + pos + ctx.addr + if imm > 4294967295 and not ctx.x64 then imm = imm-4294967296 end + ctx.imm = imm + if sz == "W" then + x = format("word 0x%04x", imm%65536) + elseif ctx.x64 then + local lo = imm % 0x1000000 + x = format("0x%02x%06x", (imm-lo) / 0x1000000, lo) + else + x = "0x"..tohex(imm) + end + elseif p == "R" then + local r = byte(code, pos-1, pos-1)%8 + if ctx.rexb then r = r + 8; ctx.rexb = false end + x = regs[r+1] + elseif p == "a" then x = regs[1] + elseif p == "c" then x = "cl" + elseif p == "d" then x = "dx" + elseif p == "1" then x = "1" + else + if not mode then + mode = ctx.mrm + if not mode then + if pos > stop then return incomplete(ctx) end + mode = byte(code, pos, pos) + pos = pos+1 + end + rm = mode%8; mode = (mode-rm)/8 + sp = mode%8; mode = (mode-sp)/8 + sdisp = "" + if mode < 3 then + if rm == 4 then + if pos > stop then return incomplete(ctx) end + sc = byte(code, pos, pos) + pos = pos+1 + rm = sc%8; sc = (sc-rm)/8 + rx = sc%8; sc = (sc-rx)/8 + if ctx.rexx then rx = rx + 8; ctx.rexx = false end + if rx == 4 then rx = nil end + end + if mode > 0 or rm == 5 then + local dsz = mode + if dsz ~= 1 then dsz = 4 end + local disp = getimm(ctx, pos, dsz); if not disp then return end + if mode == 0 then rm = nil end + if rm or rx or (not sc and ctx.x64 and not ctx.a32) then + if dsz == 1 and disp > 127 then + sdisp = format("-0x%x", 256-disp) + elseif disp >= 0 and disp <= 0x7fffffff then + sdisp = format("+0x%x", disp) + else + sdisp = format("-0x%x", (0xffffffff+1)-disp) + end + else + sdisp = format(ctx.x64 and not ctx.a32 and + not (disp >= 0 and disp <= 0x7fffffff) + and "0xffffffff%08x" or "0x%08x", disp) + end + pos = pos+dsz + end + end + if rm and ctx.rexb then rm = rm + 8; ctx.rexb = false end + if ctx.rexr then sp = sp + 8; ctx.rexr = false end + end + if p == "m" then + if mode == 3 then x = regs[rm+1] + else + local aregs = ctx.a32 and map_regs.D or ctx.aregs + local srm, srx = "", "" + if rm then srm = aregs[rm+1] + elseif not sc and ctx.x64 and not ctx.a32 then srm = "rip" end + ctx.a32 = false + if rx then + if rm then srm = srm.."+" end + srx = aregs[rx+1] + if sc > 0 then srx = srx.."*"..(2^sc) end + end + x = format("[%s%s%s]", srm, srx, sdisp) + end + if mode < 3 and + (not match(pat, "[aRrgp]") or match(pat, "t")) then -- Yuck. + x = map_sz2prefix[sz].." "..x + end + elseif p == "r" then x = regs[sp+1] + elseif p == "g" then x = map_segregs[sp+1] + elseif p == "p" then -- Suppress prefix. + elseif p == "f" then x = "st"..rm + elseif p == "x" then + if sp == 0 and ctx.lock and not ctx.x64 then + x = "CR8"; ctx.lock = false + else + x = "CR"..sp + end + elseif p == "v" then + if ctx.vexv then + x = regs[ctx.vexv+1]; ctx.vexv = false + end + elseif p == "y" then x = "DR"..sp + elseif p == "z" then x = "TR"..sp + elseif p == "l" then vexl = false + elseif p == "t" then + else + error("bad pattern `"..pat.."'") + end + end + if x then operands = operands and operands..", "..x or x end + end + ctx.pos = pos + return putop(ctx, name, operands) +end + +-- Forward declaration. +local map_act + +-- Fetch and cache MRM byte. +local function getmrm(ctx) + local mrm = ctx.mrm + if not mrm then + local pos = ctx.pos + if pos > ctx.stop then return nil end + mrm = byte(ctx.code, pos, pos) + ctx.pos = pos+1 + ctx.mrm = mrm + end + return mrm +end + +-- Dispatch to handler depending on pattern. +local function dispatch(ctx, opat, patgrp) + if not opat then return unknown(ctx) end + if match(opat, "%|") then -- MMX/SSE variants depending on prefix. + local p + if ctx.rep then + p = ctx.rep=="rep" and "%|([^%|]*)" or "%|[^%|]*%|[^%|]*%|([^%|]*)" + ctx.rep = false + elseif ctx.o16 then p = "%|[^%|]*%|([^%|]*)"; ctx.o16 = false + else p = "^[^%|]*" end + opat = match(opat, p) + if not opat then return unknown(ctx) end +-- ctx.rep = false; ctx.o16 = false + --XXX fails for 66 f2 0f 38 f1 06 crc32 eax,WORD PTR [esi] + --XXX remove in branches? + end + if match(opat, "%$") then -- reg$mem variants. + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + opat = match(opat, mrm >= 192 and "^[^%$]*" or "%$(.*)") + if opat == "" then return unknown(ctx) end + end + if opat == "" then return unknown(ctx) end + local name, pat = match(opat, "^([a-z0-9 ]*)(.*)") + if pat == "" and patgrp then pat = patgrp end + return map_act[sub(pat, 1, 1)](ctx, name, pat) +end + +-- Get a pattern from an opcode map and dispatch to handler. +local function dispatchmap(ctx, opcmap) + local pos = ctx.pos + local opat = opcmap[byte(ctx.code, pos, pos)] + pos = pos + 1 + ctx.pos = pos + return dispatch(ctx, opat) +end + +-- Map for action codes. The key is the first char after the name. +map_act = { + -- Simple opcodes without operands. + [""] = function(ctx, name, pat) + return putop(ctx, name) + end, + + -- Operand size chars fall right through. + B = putpat, W = putpat, D = putpat, Q = putpat, + V = putpat, U = putpat, T = putpat, + M = putpat, X = putpat, P = putpat, + F = putpat, G = putpat, Y = putpat, + H = putpat, + + -- Collect prefixes. + [":"] = function(ctx, name, pat) + ctx[pat == ":" and name or sub(pat, 2)] = name + if ctx.pos - ctx.start > 5 then return unknown(ctx) end -- Limit #prefixes. + end, + + -- Chain to special handler specified by name. + ["*"] = function(ctx, name, pat) + return map_act[name](ctx, name, sub(pat, 2)) + end, + + -- Use named subtable for opcode group. + ["!"] = function(ctx, name, pat) + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + return dispatch(ctx, map_opcgroup[name][((mrm-(mrm%8))/8)%8+1], sub(pat, 2)) + end, + + -- o16,o32[,o64] variants. + sz = function(ctx, name, pat) + if ctx.o16 then ctx.o16 = false + else + pat = match(pat, ",(.*)") + if ctx.rexw then + local p = match(pat, ",(.*)") + if p then pat = p; ctx.rexw = false end + end + end + pat = match(pat, "^[^,]*") + return dispatch(ctx, pat) + end, + + -- Two-byte opcode dispatch. + opc2 = function(ctx, name, pat) + return dispatchmap(ctx, map_opc2) + end, + + -- Three-byte opcode dispatch. + opc3 = function(ctx, name, pat) + return dispatchmap(ctx, map_opc3[pat]) + end, + + -- VMX/SVM dispatch. + vm = function(ctx, name, pat) + return dispatch(ctx, map_opcvm[ctx.mrm]) + end, + + -- Floating point opcode dispatch. + fp = function(ctx, name, pat) + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + local rm = mrm%8 + local idx = pat*8 + ((mrm-rm)/8)%8 + if mrm >= 192 then idx = idx + 64 end + local opat = map_opcfp[idx] + if type(opat) == "table" then opat = opat[rm+1] end + return dispatch(ctx, opat) + end, + + -- REX prefix. + rex = function(ctx, name, pat) + if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. + for p in gmatch(pat, ".") do ctx["rex"..p] = true end + ctx.rex = "rex" + end, + + -- VEX prefix. + vex = function(ctx, name, pat) + if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. + ctx.rex = "vex" + local pos = ctx.pos + if ctx.mrm then + ctx.mrm = nil + pos = pos-1 + end + local b = byte(ctx.code, pos, pos) + if not b then return incomplete(ctx) end + pos = pos+1 + if b < 128 then ctx.rexr = true end + local m = 1 + if pat == "3" then + m = b%32; b = (b-m)/32 + local nb = b%2; b = (b-nb)/2 + if nb == 0 then ctx.rexb = true end + local nx = b%2 + if nx == 0 then ctx.rexx = true end + b = byte(ctx.code, pos, pos) + if not b then return incomplete(ctx) end + pos = pos+1 + if b >= 128 then ctx.rexw = true end + end + ctx.pos = pos + local map + if m == 1 then map = map_opc2 + elseif m == 2 then map = map_opc3["38"] + elseif m == 3 then map = map_opc3["3a"] + else return unknown(ctx) end + local p = b%4; b = (b-p)/4 + if p == 1 then ctx.o16 = "o16" + elseif p == 2 then ctx.rep = "rep" + elseif p == 3 then ctx.rep = "repne" end + local l = b%2; b = (b-l)/2 + if l ~= 0 then ctx.vexl = true end + ctx.vexv = (-1-b)%16 + return dispatchmap(ctx, map) + end, + + -- Special case for nop with REX prefix. + nop = function(ctx, name, pat) + return dispatch(ctx, ctx.rex and pat or "nop") + end, + + -- Special case for 0F 77. + emms = function(ctx, name, pat) + if ctx.rex ~= "vex" then + return putop(ctx, "emms") + elseif ctx.vexl then + ctx.vexl = false + return putop(ctx, "zeroall") + else + return putop(ctx, "zeroupper") + end + end, +} + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ofs = ofs + 1 + ctx.start = ofs + ctx.pos = ofs + ctx.stop = stop + ctx.imm = nil + ctx.mrm = false + clearprefixes(ctx) + while ctx.pos <= stop do dispatchmap(ctx, ctx.map1) end + if ctx.pos ~= ctx.start then incomplete(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = (addr or 0) - 1 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 16 + ctx.x64 = false + ctx.map1 = map_opc1_32 + ctx.aregs = map_regs.D + return ctx +end + +local function create64(code, addr, out) + local ctx = create(code, addr, out) + ctx.x64 = true + ctx.map1 = map_opc1_64 + ctx.aregs = map_regs.Q + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +local function disass64(code, addr, out) + create64(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 8 then return map_regs.D[r+1] end + return map_regs.X[r-7] +end + +local function regname64(r) + if r < 16 then return map_regs.Q[r+1] end + return map_regs.X[r-15] +end + +-- Public module functions. +return { + create = create, + create64 = create64, + disass = disass, + disass64 = disass64, + regname = regname, + regname64 = regname64 +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dump.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dump.lua new file mode 100644 index 00000000..9eda08c4 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/dump.lua @@ -0,0 +1,726 @@ +---------------------------------------------------------------------------- +-- LuaJIT compiler dump module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module can be used to debug the JIT compiler itself. It dumps the +-- code representations and structures used in various compiler stages. +-- +-- Example usage: +-- +-- luajit -jdump -e "local x=0; for i=1,1e6 do x=x+i end; print(x)" +-- luajit -jdump=im -e "for i=1,1000 do for j=1,1000 do end end" | less -R +-- luajit -jdump=is myapp.lua | less -R +-- luajit -jdump=-b myapp.lua +-- luajit -jdump=+aH,myapp.html myapp.lua +-- luajit -jdump=ixT,myapp.dump myapp.lua +-- +-- The first argument specifies the dump mode. The second argument gives +-- the output file name. Default output is to stdout, unless the environment +-- variable LUAJIT_DUMPFILE is set. The file is overwritten every time the +-- module is started. +-- +-- Different features can be turned on or off with the dump mode. If the +-- mode starts with a '+', the following features are added to the default +-- set of features; a '-' removes them. Otherwise the features are replaced. +-- +-- The following dump features are available (* marks the default): +-- +-- * t Print a line for each started, ended or aborted trace (see also -jv). +-- * b Dump the traced bytecode. +-- * i Dump the IR (intermediate representation). +-- r Augment the IR with register/stack slots. +-- s Dump the snapshot map. +-- * m Dump the generated machine code. +-- x Print each taken trace exit. +-- X Print each taken trace exit and the contents of all registers. +-- a Print the IR of aborted traces, too. +-- +-- The output format can be set with the following characters: +-- +-- T Plain text output. +-- A ANSI-colored text output +-- H Colorized HTML + CSS output. +-- +-- The default output format is plain text. It's set to ANSI-colored text +-- if the COLORTERM variable is set. Note: this is independent of any output +-- redirection, which is actually considered a feature. +-- +-- You probably want to use less -R to enjoy viewing ANSI-colored text from +-- a pipe or a file. Add this to your ~/.bashrc: export LESS="-R" +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local funcinfo, funcbc = jutil.funcinfo, jutil.funcbc +local traceinfo, traceir, tracek = jutil.traceinfo, jutil.traceir, jutil.tracek +local tracemc, tracesnap = jutil.tracemc, jutil.tracesnap +local traceexitstub, ircalladdr = jutil.traceexitstub, jutil.ircalladdr +local bit = require("bit") +local band, shr, tohex = bit.band, bit.rshift, bit.tohex +local sub, gsub, format = string.sub, string.gsub, string.format +local byte, rep = string.byte, string.rep +local type, tostring = type, tostring +local stdout, stderr = io.stdout, io.stderr + +-- Load other modules on-demand. +local bcline, disass + +-- Active flag, output file handle and dump mode. +local active, out, dumpmode + +------------------------------------------------------------------------------ + +local symtabmt = { __index = false } +local symtab = {} +local nexitsym = 0 + +-- Fill nested symbol table with per-trace exit stub addresses. +local function fillsymtab_tr(tr, nexit) + local t = {} + symtabmt.__index = t + if jit.arch:sub(1, 4) == "mips" then + t[traceexitstub(tr, 0)] = "exit" + return + end + for i=0,nexit-1 do + local addr = traceexitstub(tr, i) + if addr < 0 then addr = addr + 2^32 end + t[addr] = tostring(i) + end + local addr = traceexitstub(tr, nexit) + if addr then t[addr] = "stack_check" end +end + +-- Fill symbol table with trace exit stub addresses. +local function fillsymtab(tr, nexit) + local t = symtab + if nexitsym == 0 then + local maskaddr = jit.arch == "arm" and -2 + local ircall = vmdef.ircall + for i=0,#ircall do + local addr = ircalladdr(i) + if addr ~= 0 then + if maskaddr then addr = band(addr, maskaddr) end + if addr < 0 then addr = addr + 2^32 end + t[addr] = ircall[i] + end + end + end + if nexitsym == 1000000 then -- Per-trace exit stubs. + fillsymtab_tr(tr, nexit) + elseif nexit > nexitsym then -- Shared exit stubs. + for i=nexitsym,nexit-1 do + local addr = traceexitstub(i) + if addr == nil then -- Fall back to per-trace exit stubs. + fillsymtab_tr(tr, nexit) + setmetatable(symtab, symtabmt) + nexit = 1000000 + break + end + if addr < 0 then addr = addr + 2^32 end + t[addr] = tostring(i) + end + nexitsym = nexit + end + return t +end + +local function dumpwrite(s) + out:write(s) +end + +-- Disassemble machine code. +local function dump_mcode(tr) + local info = traceinfo(tr) + if not info then return end + local mcode, addr, loop = tracemc(tr) + if not mcode then return end + if not disass then disass = require("jit.dis_"..jit.arch) end + if addr < 0 then addr = addr + 2^32 end + out:write("---- TRACE ", tr, " mcode ", #mcode, "\n") + local ctx = disass.create(mcode, addr, dumpwrite) + ctx.hexdump = 0 + ctx.symtab = fillsymtab(tr, info.nexit) + if loop ~= 0 then + symtab[addr+loop] = "LOOP" + ctx:disass(0, loop) + out:write("->LOOP:\n") + ctx:disass(loop, #mcode-loop) + symtab[addr+loop] = nil + else + ctx:disass(0, #mcode) + end +end + +------------------------------------------------------------------------------ + +local irtype_text = { + [0] = "nil", + "fal", + "tru", + "lud", + "str", + "p32", + "thr", + "pro", + "fun", + "p64", + "cdt", + "tab", + "udt", + "flt", + "num", + "i8 ", + "u8 ", + "i16", + "u16", + "int", + "u32", + "i64", + "u64", + "sfp", +} + +local colortype_ansi = { + [0] = "%s", + "%s", + "%s", + "\027[36m%s\027[m", + "\027[32m%s\027[m", + "%s", + "\027[1m%s\027[m", + "%s", + "\027[1m%s\027[m", + "%s", + "\027[33m%s\027[m", + "\027[31m%s\027[m", + "\027[36m%s\027[m", + "\027[34m%s\027[m", + "\027[34m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", +} + +local function colorize_text(s) + return s +end + +local function colorize_ansi(s, t, extra) + local out = format(colortype_ansi[t], s) + if extra then out = "\027[3m"..out end + return out +end + +local irtype_ansi = setmetatable({}, + { __index = function(tab, t) + local s = colorize_ansi(irtype_text[t], t); tab[t] = s; return s; end }) + +local html_escape = { ["<"] = "<", [">"] = ">", ["&"] = "&", } + +local function colorize_html(s, t, extra) + s = gsub(s, "[<>&]", html_escape) + return format('%s', + irtype_text[t], extra and " irt_extra" or "", s) +end + +local irtype_html = setmetatable({}, + { __index = function(tab, t) + local s = colorize_html(irtype_text[t], t); tab[t] = s; return s; end }) + +local header_html = [[ + +]] + +local colorize, irtype + +-- Lookup tables to convert some literals into names. +local litname = { + ["SLOAD "] = setmetatable({}, { __index = function(t, mode) + local s = "" + if band(mode, 1) ~= 0 then s = s.."P" end + if band(mode, 2) ~= 0 then s = s.."F" end + if band(mode, 4) ~= 0 then s = s.."T" end + if band(mode, 8) ~= 0 then s = s.."C" end + if band(mode, 16) ~= 0 then s = s.."R" end + if band(mode, 32) ~= 0 then s = s.."I" end + if band(mode, 64) ~= 0 then s = s.."K" end + t[mode] = s + return s + end}), + ["XLOAD "] = { [0] = "", "R", "V", "RV", "U", "RU", "VU", "RVU", }, + ["CONV "] = setmetatable({}, { __index = function(t, mode) + local s = irtype[band(mode, 31)] + s = irtype[band(shr(mode, 5), 31)].."."..s + if band(mode, 0x800) ~= 0 then s = s.." sext" end + local c = shr(mode, 12) + if c == 1 then s = s.." none" + elseif c == 2 then s = s.." index" + elseif c == 3 then s = s.." check" end + t[mode] = s + return s + end}), + ["FLOAD "] = vmdef.irfield, + ["FREF "] = vmdef.irfield, + ["FPMATH"] = vmdef.irfpm, + ["TMPREF"] = { [0] = "", "IN", "OUT", "INOUT", "", "", "OUT2", "INOUT2" }, + ["BUFHDR"] = { [0] = "RESET", "APPEND", "WRITE" }, + ["TOSTR "] = { [0] = "INT", "NUM", "CHAR" }, +} + +local function ctlsub(c) + if c == "\n" then return "\\n" + elseif c == "\r" then return "\\r" + elseif c == "\t" then return "\\t" + else return format("\\%03d", byte(c)) + end +end + +local function fmtfunc(func, pc) + local fi = funcinfo(func, pc) + if fi.loc then + return fi.loc + elseif fi.ffid then + return vmdef.ffnames[fi.ffid] + elseif fi.addr then + return format("C:%x", fi.addr) + else + return "(?)" + end +end + +local function formatk(tr, idx, sn) + local k, t, slot = tracek(tr, idx) + local tn = type(k) + local s + if tn == "number" then + if t < 12 then + s = k == 0 and "NULL" or format("[0x%08x]", k) + elseif band(sn or 0, 0x30000) ~= 0 then + s = band(sn, 0x20000) ~= 0 and "contpc" or "ftsz" + elseif k == 2^52+2^51 then + s = "bias" + else + s = format(0 < k and k < 0x1p-1026 and "%+a" or "%+.14g", k) + end + elseif tn == "string" then + s = format(#k > 20 and '"%.20s"~' or '"%s"', gsub(k, "%c", ctlsub)) + elseif tn == "function" then + s = fmtfunc(k) + elseif tn == "table" then + s = format("{%p}", k) + elseif tn == "userdata" then + if t == 12 then + s = format("userdata:%p", k) + else + s = format("[%p]", k) + if s == "[NULL]" then s = "NULL" end + end + elseif t == 21 then -- int64_t + s = sub(tostring(k), 1, -3) + if sub(s, 1, 1) ~= "-" then s = "+"..s end + elseif sn == 0x1057fff then -- SNAP(1, SNAP_FRAME | SNAP_NORESTORE, REF_NIL) + return "----" -- Special case for LJ_FR2 slot 1. + else + s = tostring(k) -- For primitives. + end + s = colorize(format("%-4s", s), t, band(sn or 0, 0x100000) ~= 0) + if slot then + s = format("%s @%d", s, slot) + end + return s +end + +local function printsnap(tr, snap) + local n = 2 + for s=0,snap[1]-1 do + local sn = snap[n] + if shr(sn, 24) == s then + n = n + 1 + local ref = band(sn, 0xffff) - 0x8000 -- REF_BIAS + if ref < 0 then + out:write(formatk(tr, ref, sn)) + elseif band(sn, 0x80000) ~= 0 then -- SNAP_SOFTFPNUM + out:write(colorize(format("%04d/%04d", ref, ref+1), 14)) + else + local m, ot, op1, op2 = traceir(tr, ref) + out:write(colorize(format("%04d", ref), band(ot, 31), band(sn, 0x100000) ~= 0)) + end + out:write(band(sn, 0x10000) == 0 and " " or "|") -- SNAP_FRAME + else + out:write("---- ") + end + end + out:write("]\n") +end + +-- Dump snapshots (not interleaved with IR). +local function dump_snap(tr) + out:write("---- TRACE ", tr, " snapshots\n") + for i=0,1000000000 do + local snap = tracesnap(tr, i) + if not snap then break end + out:write(format("#%-3d %04d [ ", i, snap[0])) + printsnap(tr, snap) + end +end + +-- Return a register name or stack slot for a rid/sp location. +local function ridsp_name(ridsp, ins) + if not disass then disass = require("jit.dis_"..jit.arch) end + local rid, slot = band(ridsp, 0xff), shr(ridsp, 8) + if rid == 253 or rid == 254 then + return (slot == 0 or slot == 255) and " {sink" or format(" {%04d", ins-slot) + end + if ridsp > 255 then return format("[%x]", slot*4) end + if rid < 128 then return disass.regname(rid) end + return "" +end + +-- Dump CALL* function ref and return optional ctype. +local function dumpcallfunc(tr, ins) + local ctype + if ins > 0 then + local m, ot, op1, op2 = traceir(tr, ins) + if band(ot, 31) == 0 then -- nil type means CARG(func, ctype). + ins = op1 + ctype = formatk(tr, op2) + end + end + if ins < 0 then + out:write(format("[0x%x](", tonumber((tracek(tr, ins))))) + else + out:write(format("%04d (", ins)) + end + return ctype +end + +-- Recursively gather CALL* args and dump them. +local function dumpcallargs(tr, ins) + if ins < 0 then + out:write(formatk(tr, ins)) + else + local m, ot, op1, op2 = traceir(tr, ins) + local oidx = 6*shr(ot, 8) + local op = sub(vmdef.irnames, oidx+1, oidx+6) + if op == "CARG " then + dumpcallargs(tr, op1) + if op2 < 0 then + out:write(" ", formatk(tr, op2)) + else + out:write(" ", format("%04d", op2)) + end + else + out:write(format("%04d", ins)) + end + end +end + +-- Dump IR and interleaved snapshots. +local function dump_ir(tr, dumpsnap, dumpreg) + local info = traceinfo(tr) + if not info then return end + local nins = info.nins + out:write("---- TRACE ", tr, " IR\n") + local irnames = vmdef.irnames + local snapref = 65536 + local snap, snapno + if dumpsnap then + snap = tracesnap(tr, 0) + snapref = snap[0] + snapno = 0 + end + for ins=1,nins do + if ins >= snapref then + if dumpreg then + out:write(format(".... SNAP #%-3d [ ", snapno)) + else + out:write(format(".... SNAP #%-3d [ ", snapno)) + end + printsnap(tr, snap) + snapno = snapno + 1 + snap = tracesnap(tr, snapno) + snapref = snap and snap[0] or 65536 + end + local m, ot, op1, op2, ridsp = traceir(tr, ins) + local oidx, t = 6*shr(ot, 8), band(ot, 31) + local op = sub(irnames, oidx+1, oidx+6) + if op == "LOOP " then + if dumpreg then + out:write(format("%04d ------------ LOOP ------------\n", ins)) + else + out:write(format("%04d ------ LOOP ------------\n", ins)) + end + elseif op ~= "NOP " and op ~= "CARG " and + (dumpreg or op ~= "RENAME") then + local rid = band(ridsp, 255) + if dumpreg then + out:write(format("%04d %-6s", ins, ridsp_name(ridsp, ins))) + else + out:write(format("%04d ", ins)) + end + out:write(format("%s%s %s %s ", + (rid == 254 or rid == 253) and "}" or + (band(ot, 128) == 0 and " " or ">"), + band(ot, 64) == 0 and " " or "+", + irtype[t], op)) + local m1, m2 = band(m, 3), band(m, 3*4) + if sub(op, 1, 4) == "CALL" then + local ctype + if m2 == 1*4 then -- op2 == IRMlit + out:write(format("%-10s (", vmdef.ircall[op2])) + else + ctype = dumpcallfunc(tr, op2) + end + if op1 ~= -1 then dumpcallargs(tr, op1) end + out:write(")") + if ctype then out:write(" ctype ", ctype) end + elseif op == "CNEW " and op2 == -1 then + out:write(formatk(tr, op1)) + elseif m1 ~= 3 then -- op1 != IRMnone + if op1 < 0 then + out:write(formatk(tr, op1)) + else + out:write(format(m1 == 0 and "%04d" or "#%-3d", op1)) + end + if m2 ~= 3*4 then -- op2 != IRMnone + if m2 == 1*4 then -- op2 == IRMlit + local litn = litname[op] + if litn and litn[op2] then + out:write(" ", litn[op2]) + elseif op == "UREFO " or op == "UREFC " then + out:write(format(" #%-3d", shr(op2, 8))) + else + out:write(format(" #%-3d", op2)) + end + elseif op2 < 0 then + out:write(" ", formatk(tr, op2)) + else + out:write(format(" %04d", op2)) + end + end + end + out:write("\n") + end + end + if snap then + if dumpreg then + out:write(format(".... SNAP #%-3d [ ", snapno)) + else + out:write(format(".... SNAP #%-3d [ ", snapno)) + end + printsnap(tr, snap) + end +end + +------------------------------------------------------------------------------ + +local recprefix = "" +local recdepth = 0 + +-- Format trace error message. +local function fmterr(err, info) + if type(err) == "number" then + if type(info) == "function" then info = fmtfunc(info) end + err = format(vmdef.traceerr[err], info) + end + return err +end + +-- Dump trace states. +local function dump_trace(what, tr, func, pc, otr, oex) + if what == "stop" or (what == "abort" and dumpmode.a) then + if dumpmode.i then dump_ir(tr, dumpmode.s, dumpmode.r and what == "stop") + elseif dumpmode.s then dump_snap(tr) end + if dumpmode.m then dump_mcode(tr) end + end + if what == "start" then + if dumpmode.H then out:write('
\n') end
+    out:write("---- TRACE ", tr, " ", what)
+    if otr then out:write(" ", otr, "/", oex == -1 and "stitch" or oex) end
+    out:write(" ", fmtfunc(func, pc), "\n")
+  elseif what == "stop" or what == "abort" then
+    out:write("---- TRACE ", tr, " ", what)
+    if what == "abort" then
+      out:write(" ", fmtfunc(func, pc), " -- ", fmterr(otr, oex), "\n")
+    else
+      local info = traceinfo(tr)
+      local link, ltype = info.link, info.linktype
+      if link == tr or link == 0 then
+	out:write(" -> ", ltype, "\n")
+      elseif ltype == "root" then
+	out:write(" -> ", link, "\n")
+      else
+	out:write(" -> ", link, " ", ltype, "\n")
+      end
+    end
+    if dumpmode.H then out:write("
\n\n") else out:write("\n") end + else + if what == "flush" then symtab, nexitsym = {}, 0 end + out:write("---- TRACE ", what, "\n\n") + end + out:flush() +end + +-- Dump recorded bytecode. +local function dump_record(tr, func, pc, depth) + if depth ~= recdepth then + recdepth = depth + recprefix = rep(" .", depth) + end + local line + if pc >= 0 then + line = bcline(func, pc, recprefix) + if dumpmode.H then line = gsub(line, "[<>&]", html_escape) end + else + line = "0000 "..recprefix.." FUNCC \n" + end + if pc <= 0 then + out:write(sub(line, 1, -2), " ; ", fmtfunc(func), "\n") + else + out:write(line) + end + if pc >= 0 and band(funcbc(func, pc), 0xff) < 16 then -- ORDER BC + out:write(bcline(func, pc+1, recprefix)) -- Write JMP for cond. + end +end + +------------------------------------------------------------------------------ + +local gpr64 = jit.arch:match("64") +local fprmips32 = jit.arch == "mips" or jit.arch == "mipsel" + +-- Dump taken trace exits. +local function dump_texit(tr, ex, ngpr, nfpr, ...) + out:write("---- TRACE ", tr, " exit ", ex, "\n") + if dumpmode.X then + local regs = {...} + if gpr64 then + for i=1,ngpr do + out:write(format(" %016x", regs[i])) + if i % 4 == 0 then out:write("\n") end + end + else + for i=1,ngpr do + out:write(" ", tohex(regs[i])) + if i % 8 == 0 then out:write("\n") end + end + end + if fprmips32 then + for i=1,nfpr,2 do + out:write(format(" %+17.14g", regs[ngpr+i])) + if i % 8 == 7 then out:write("\n") end + end + else + for i=1,nfpr do + out:write(format(" %+17.14g", regs[ngpr+i])) + if i % 4 == 0 then out:write("\n") end + end + end + end +end + +------------------------------------------------------------------------------ + +-- Detach dump handlers. +local function dumpoff() + if active then + active = false + jit.attach(dump_texit) + jit.attach(dump_record) + jit.attach(dump_trace) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach dump handlers. +local function dumpon(opt, outfile) + if active then dumpoff() end + + local term = os.getenv("TERM") + local colormode = (term and term:match("color") or os.getenv("COLORTERM")) and "A" or "T" + if opt then + opt = gsub(opt, "[TAH]", function(mode) colormode = mode; return ""; end) + end + + local m = { t=true, b=true, i=true, m=true, } + if opt and opt ~= "" then + local o = sub(opt, 1, 1) + if o ~= "+" and o ~= "-" then m = {} end + for i=1,#opt do m[sub(opt, i, i)] = (o ~= "-") end + end + dumpmode = m + + if m.t or m.b or m.i or m.s or m.m then + jit.attach(dump_trace, "trace") + end + if m.b then + jit.attach(dump_record, "record") + if not bcline then bcline = require("jit.bc").line end + end + if m.x or m.X then + jit.attach(dump_texit, "texit") + end + + if not outfile then outfile = os.getenv("LUAJIT_DUMPFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stdout + end + + m[colormode] = true + if colormode == "A" then + colorize = colorize_ansi + irtype = irtype_ansi + elseif colormode == "H" then + colorize = colorize_html + irtype = irtype_html + out:write(header_html) + else + colorize = colorize_text + irtype = irtype_text + end + + active = true +end + +-- Public module functions. +return { + on = dumpon, + off = dumpoff, + start = dumpon -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/p.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/p.lua new file mode 100644 index 00000000..c9ec1d8b --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/p.lua @@ -0,0 +1,312 @@ +---------------------------------------------------------------------------- +-- LuaJIT profiler. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module is a simple command line interface to the built-in +-- low-overhead profiler of LuaJIT. +-- +-- The lower-level API of the profiler is accessible via the "jit.profile" +-- module or the luaJIT_profile_* C API. +-- +-- Example usage: +-- +-- luajit -jp myapp.lua +-- luajit -jp=s myapp.lua +-- luajit -jp=-s myapp.lua +-- luajit -jp=vl myapp.lua +-- luajit -jp=G,profile.txt myapp.lua +-- +-- The following dump features are available: +-- +-- f Stack dump: function name, otherwise module:line. Default mode. +-- F Stack dump: ditto, but always prepend module. +-- l Stack dump: module:line. +-- stack dump depth (callee < caller). Default: 1. +-- - Inverse stack dump depth (caller > callee). +-- s Split stack dump after first stack level. Implies abs(depth) >= 2. +-- p Show full path for module names. +-- v Show VM states. Can be combined with stack dumps, e.g. vf or fv. +-- z Show zones. Can be combined with stack dumps, e.g. zf or fz. +-- r Show raw sample counts. Default: show percentages. +-- a Annotate excerpts from source code files. +-- A Annotate complete source code files. +-- G Produce raw output suitable for graphical tools (e.g. flame graphs). +-- m Minimum sample percentage to be shown. Default: 3. +-- i Sampling interval in milliseconds. Default: 10. +-- +---------------------------------------------------------------------------- + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local profile = require("jit.profile") +local vmdef = require("jit.vmdef") +local math = math +local pairs, ipairs, tonumber, floor = pairs, ipairs, tonumber, math.floor +local sort, format = table.sort, string.format +local stdout = io.stdout +local zone -- Load jit.zone module on demand. + +-- Output file handle. +local out + +------------------------------------------------------------------------------ + +local prof_ud +local prof_states, prof_split, prof_min, prof_raw, prof_fmt, prof_depth +local prof_ann, prof_count1, prof_count2, prof_samples + +local map_vmmode = { + N = "Compiled", + I = "Interpreted", + C = "C code", + G = "Garbage Collector", + J = "JIT Compiler", +} + +-- Profiler callback. +local function prof_cb(th, samples, vmmode) + prof_samples = prof_samples + samples + local key_stack, key_stack2, key_state + -- Collect keys for sample. + if prof_states then + if prof_states == "v" then + key_state = map_vmmode[vmmode] or vmmode + else + key_state = zone:get() or "(none)" + end + end + if prof_fmt then + key_stack = profile.dumpstack(th, prof_fmt, prof_depth) + key_stack = key_stack:gsub("%[builtin#(%d+)%]", function(x) + return vmdef.ffnames[tonumber(x)] + end) + if prof_split == 2 then + local k1, k2 = key_stack:match("(.-) [<>] (.*)") + if k2 then key_stack, key_stack2 = k1, k2 end + elseif prof_split == 3 then + key_stack2 = profile.dumpstack(th, "l", 1) + end + end + -- Order keys. + local k1, k2 + if prof_split == 1 then + if key_state then + k1 = key_state + if key_stack then k2 = key_stack end + end + elseif key_stack then + k1 = key_stack + if key_stack2 then k2 = key_stack2 elseif key_state then k2 = key_state end + end + -- Coalesce samples in one or two levels. + if k1 then + local t1 = prof_count1 + t1[k1] = (t1[k1] or 0) + samples + if k2 then + local t2 = prof_count2 + local t3 = t2[k1] + if not t3 then t3 = {}; t2[k1] = t3 end + t3[k2] = (t3[k2] or 0) + samples + end + end +end + +------------------------------------------------------------------------------ + +-- Show top N list. +local function prof_top(count1, count2, samples, indent) + local t, n = {}, 0 + for k in pairs(count1) do + n = n + 1 + t[n] = k + end + sort(t, function(a, b) return count1[a] > count1[b] end) + for i=1,n do + local k = t[i] + local v = count1[k] + local pct = floor(v*100/samples + 0.5) + if pct < prof_min then break end + if not prof_raw then + out:write(format("%s%2d%% %s\n", indent, pct, k)) + elseif prof_raw == "r" then + out:write(format("%s%5d %s\n", indent, v, k)) + else + out:write(format("%s %d\n", k, v)) + end + if count2 then + local r = count2[k] + if r then + prof_top(r, nil, v, (prof_split == 3 or prof_split == 1) and " -- " or + (prof_depth < 0 and " -> " or " <- ")) + end + end + end +end + +-- Annotate source code +local function prof_annotate(count1, samples) + local files = {} + local ms = 0 + for k, v in pairs(count1) do + local pct = floor(v*100/samples + 0.5) + ms = math.max(ms, v) + if pct >= prof_min then + local file, line = k:match("^(.*):(%d+)$") + if not file then file = k; line = 0 end + local fl = files[file] + if not fl then fl = {}; files[file] = fl; files[#files+1] = file end + line = tonumber(line) + fl[line] = prof_raw and v or pct + end + end + sort(files) + local fmtv, fmtn = " %3d%% | %s\n", " | %s\n" + if prof_raw then + local n = math.max(5, math.ceil(math.log10(ms))) + fmtv = "%"..n.."d | %s\n" + fmtn = (" "):rep(n).." | %s\n" + end + local ann = prof_ann + for _, file in ipairs(files) do + local f0 = file:byte() + if f0 == 40 or f0 == 91 then + out:write(format("\n====== %s ======\n[Cannot annotate non-file]\n", file)) + break + end + local fp, err = io.open(file) + if not fp then + out:write(format("====== ERROR: %s: %s\n", file, err)) + break + end + out:write(format("\n====== %s ======\n", file)) + local fl = files[file] + local n, show = 1, false + if ann ~= 0 then + for i=1,ann do + if fl[i] then show = true; out:write("@@ 1 @@\n"); break end + end + end + for line in fp:lines() do + if line:byte() == 27 then + out:write("[Cannot annotate bytecode file]\n") + break + end + local v = fl[n] + if ann ~= 0 then + local v2 = fl[n+ann] + if show then + if v2 then show = n+ann elseif v then show = n + elseif show+ann < n then show = false end + elseif v2 then + show = n+ann + out:write(format("@@ %d @@\n", n)) + end + if not show then goto next end + end + if v then + out:write(format(fmtv, v, line)) + else + out:write(format(fmtn, line)) + end + ::next:: + n = n + 1 + end + fp:close() + end +end + +------------------------------------------------------------------------------ + +-- Finish profiling and dump result. +local function prof_finish() + if prof_ud then + profile.stop() + local samples = prof_samples + if samples == 0 then + if prof_raw ~= true then out:write("[No samples collected]\n") end + return + end + if prof_ann then + prof_annotate(prof_count1, samples) + else + prof_top(prof_count1, prof_count2, samples, "") + end + prof_count1 = nil + prof_count2 = nil + prof_ud = nil + if out ~= stdout then out:close() end + end +end + +-- Start profiling. +local function prof_start(mode) + local interval = "" + mode = mode:gsub("i%d*", function(s) interval = s; return "" end) + prof_min = 3 + mode = mode:gsub("m(%d+)", function(s) prof_min = tonumber(s); return "" end) + prof_depth = 1 + mode = mode:gsub("%-?%d+", function(s) prof_depth = tonumber(s); return "" end) + local m = {} + for c in mode:gmatch(".") do m[c] = c end + prof_states = m.z or m.v + if prof_states == "z" then zone = require("jit.zone") end + local scope = m.l or m.f or m.F or (prof_states and "" or "f") + local flags = (m.p or "") + prof_raw = m.r + if m.s then + prof_split = 2 + if prof_depth == -1 or m["-"] then prof_depth = -2 + elseif prof_depth == 1 then prof_depth = 2 end + elseif mode:find("[fF].*l") then + scope = "l" + prof_split = 3 + else + prof_split = (scope == "" or mode:find("[zv].*[lfF]")) and 1 or 0 + end + prof_ann = m.A and 0 or (m.a and 3) + if prof_ann then + scope = "l" + prof_fmt = "pl" + prof_split = 0 + prof_depth = 1 + elseif m.G and scope ~= "" then + prof_fmt = flags..scope.."Z;" + prof_depth = -100 + prof_raw = true + prof_min = 0 + elseif scope == "" then + prof_fmt = false + else + local sc = prof_split == 3 and m.f or m.F or scope + prof_fmt = flags..sc..(prof_depth >= 0 and "Z < " or "Z > ") + end + prof_count1 = {} + prof_count2 = {} + prof_samples = 0 + profile.start(scope:lower()..interval, prof_cb) + prof_ud = newproxy(true) + getmetatable(prof_ud).__gc = prof_finish +end + +------------------------------------------------------------------------------ + +local function start(mode, outfile) + if not outfile then outfile = os.getenv("LUAJIT_PROFILEFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stdout + end + prof_start(mode or "f") +end + +-- Public module functions. +return { + start = start, -- For -j command line option. + stop = prof_finish +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/v.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/v.lua new file mode 100644 index 00000000..83589143 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/v.lua @@ -0,0 +1,170 @@ +---------------------------------------------------------------------------- +-- Verbose mode of the LuaJIT compiler. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module shows verbose information about the progress of the +-- JIT compiler. It prints one line for each generated trace. This module +-- is useful to see which code has been compiled or where the compiler +-- punts and falls back to the interpreter. +-- +-- Example usage: +-- +-- luajit -jv -e "for i=1,1000 do for j=1,1000 do end end" +-- luajit -jv=myapp.out myapp.lua +-- +-- Default output is to stderr. To redirect the output to a file, pass a +-- filename as an argument (use '-' for stdout) or set the environment +-- variable LUAJIT_VERBOSEFILE. The file is overwritten every time the +-- module is started. +-- +-- The output from the first example should look like this: +-- +-- [TRACE 1 (command line):1 loop] +-- [TRACE 2 (1/3) (command line):1 -> 1] +-- +-- The first number in each line is the internal trace number. Next are +-- the file name ('(command line)') and the line number (':1') where the +-- trace has started. Side traces also show the parent trace number and +-- the exit number where they are attached to in parentheses ('(1/3)'). +-- An arrow at the end shows where the trace links to ('-> 1'), unless +-- it loops to itself. +-- +-- In this case the inner loop gets hot and is traced first, generating +-- a root trace. Then the last exit from the 1st trace gets hot, too, +-- and triggers generation of the 2nd trace. The side trace follows the +-- path along the outer loop and *around* the inner loop, back to its +-- start, and then links to the 1st trace. Yes, this may seem unusual, +-- if you know how traditional compilers work. Trace compilers are full +-- of surprises like this -- have fun! :-) +-- +-- Aborted traces are shown like this: +-- +-- [TRACE --- foo.lua:44 -- leaving loop in root trace at foo:lua:50] +-- +-- Don't worry -- trace aborts are quite common, even in programs which +-- can be fully compiled. The compiler may retry several times until it +-- finds a suitable trace. +-- +-- Of course this doesn't work with features that are not-yet-implemented +-- (NYI error messages). The VM simply falls back to the interpreter. This +-- may not matter at all if the particular trace is not very high up in +-- the CPU usage profile. Oh, and the interpreter is quite fast, too. +-- +-- Also check out the -jdump module, which prints all the gory details. +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local funcinfo, traceinfo = jutil.funcinfo, jutil.traceinfo +local type, format = type, string.format +local stdout, stderr = io.stdout, io.stderr + +-- Active flag and output file handle. +local active, out + +------------------------------------------------------------------------------ + +local startloc, startex + +local function fmtfunc(func, pc) + local fi = funcinfo(func, pc) + if fi.loc then + return fi.loc + elseif fi.ffid then + return vmdef.ffnames[fi.ffid] + elseif fi.addr then + return format("C:%x", fi.addr) + else + return "(?)" + end +end + +-- Format trace error message. +local function fmterr(err, info) + if type(err) == "number" then + if type(info) == "function" then info = fmtfunc(info) end + err = format(vmdef.traceerr[err], info) + end + return err +end + +-- Dump trace states. +local function dump_trace(what, tr, func, pc, otr, oex) + if what == "start" then + startloc = fmtfunc(func, pc) + startex = otr and "("..otr.."/"..(oex == -1 and "stitch" or oex)..") " or "" + else + if what == "abort" then + local loc = fmtfunc(func, pc) + if loc ~= startloc then + out:write(format("[TRACE --- %s%s -- %s at %s]\n", + startex, startloc, fmterr(otr, oex), loc)) + else + out:write(format("[TRACE --- %s%s -- %s]\n", + startex, startloc, fmterr(otr, oex))) + end + elseif what == "stop" then + local info = traceinfo(tr) + local link, ltype = info.link, info.linktype + if ltype == "interpreter" then + out:write(format("[TRACE %3s %s%s -- fallback to interpreter]\n", + tr, startex, startloc)) + elseif ltype == "stitch" then + out:write(format("[TRACE %3s %s%s %s %s]\n", + tr, startex, startloc, ltype, fmtfunc(func, pc))) + elseif link == tr or link == 0 then + out:write(format("[TRACE %3s %s%s %s]\n", + tr, startex, startloc, ltype)) + elseif ltype == "root" then + out:write(format("[TRACE %3s %s%s -> %d]\n", + tr, startex, startloc, link)) + else + out:write(format("[TRACE %3s %s%s -> %d %s]\n", + tr, startex, startloc, link, ltype)) + end + else + out:write(format("[TRACE %s]\n", what)) + end + out:flush() + end +end + +------------------------------------------------------------------------------ + +-- Detach dump handlers. +local function dumpoff() + if active then + active = false + jit.attach(dump_trace) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach dump handlers. +local function dumpon(outfile) + if active then dumpoff() end + if not outfile then outfile = os.getenv("LUAJIT_VERBOSEFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stderr + end + jit.attach(dump_trace, "trace") + active = true +end + +-- Public module functions. +return { + on = dumpon, + off = dumpoff, + start = dumpon -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/vmdef.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/vmdef.lua new file mode 100644 index 00000000..68ff7ba2 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/vmdef.lua @@ -0,0 +1,395 @@ +-- This is a generated file. DO NOT EDIT! + +return { + +bcnames = "ISLT ISGE ISLE ISGT ISEQV ISNEV ISEQS ISNES ISEQN ISNEN ISEQP ISNEP ISTC ISFC IST ISF ISTYPEISNUM MOV NOT UNM LEN ADDVN SUBVN MULVN DIVVN MODVN ADDNV SUBNV MULNV DIVNV MODNV ADDVV SUBVV MULVV DIVVV MODVV POW CAT KSTR KCDATAKSHORTKNUM KPRI KNIL UGET USETV USETS USETN USETP UCLO FNEW TNEW TDUP GGET GSET TGETV TGETS TGETB TGETR TSETV TSETS TSETB TSETM TSETR CALLM CALL CALLMTCALLT ITERC ITERN VARG ISNEXTRETM RET RET0 RET1 FORI JFORI FORL IFORL JFORL ITERL IITERLJITERLLOOP ILOOP JLOOP JMP FUNCF IFUNCFJFUNCFFUNCV IFUNCVJFUNCVFUNCC FUNCCW", + +irnames = "LT GE LE GT ULT UGE ULE UGT EQ NE ABC RETF NOP BASE PVAL GCSTEPHIOP LOOP USE PHI RENAMEPROF KPRI KINT KGC KPTR KKPTR KNULL KNUM KINT64KSLOT BNOT BSWAP BAND BOR BXOR BSHL BSHR BSAR BROL BROR ADD SUB MUL DIV MOD POW NEG ABS LDEXP MIN MAX FPMATHADDOV SUBOV MULOV AREF HREFK HREF NEWREFUREFO UREFC FREF TMPREFSTRREFLREF ALOAD HLOAD ULOAD FLOAD XLOAD SLOAD VLOAD ALEN ASTOREHSTOREUSTOREFSTOREXSTORESNEW XSNEW TNEW TDUP CNEW CNEWI BUFHDRBUFPUTBUFSTRTBAR OBAR XBAR CONV TOBIT TOSTR STRTO CALLN CALLA CALLL CALLS CALLXSCARG ", + +irfpm = { [0]="floor", "ceil", "trunc", "sqrt", "log", "log2", "other", }, + +irfield = { [0]="str.len", "func.env", "func.pc", "func.ffid", "thread.env", "tab.meta", "tab.array", "tab.node", "tab.asize", "tab.hmask", "tab.nomm", "udata.meta", "udata.udtype", "udata.file", "sbuf.w", "sbuf.e", "sbuf.b", "sbuf.l", "sbuf.ref", "sbuf.r", "cdata.ctypeid", "cdata.ptr", "cdata.int", "cdata.int64", "cdata.int64_4", }, + +ircall = { +[0]="lj_str_cmp", +"lj_str_find", +"lj_str_new", +"lj_strscan_num", +"lj_strfmt_int", +"lj_strfmt_num", +"lj_strfmt_char", +"lj_strfmt_putint", +"lj_strfmt_putnum", +"lj_strfmt_putquoted", +"lj_strfmt_putfxint", +"lj_strfmt_putfnum_int", +"lj_strfmt_putfnum_uint", +"lj_strfmt_putfnum", +"lj_strfmt_putfstr", +"lj_strfmt_putfchar", +"lj_buf_putmem", +"lj_buf_putstr", +"lj_buf_putchar", +"lj_buf_putstr_reverse", +"lj_buf_putstr_lower", +"lj_buf_putstr_upper", +"lj_buf_putstr_rep", +"lj_buf_puttab", +"lj_bufx_set", +"lj_bufx_more", +"lj_serialize_put", +"lj_serialize_get", +"lj_serialize_encode", +"lj_serialize_decode", +"lj_buf_tostr", +"lj_tab_new_ah", +"lj_tab_new1", +"lj_tab_dup", +"lj_tab_clear", +"lj_tab_newkey", +"lj_tab_keyindex", +"lj_vm_next", +"lj_tab_len", +"lj_tab_len_hint", +"lj_gc_step_jit", +"lj_gc_barrieruv", +"lj_mem_newgco", +"lj_prng_u64d", +"lj_vm_modi", +"log10", +"exp", +"sin", +"cos", +"tan", +"asin", +"acos", +"atan", +"sinh", +"cosh", +"tanh", +"fputc", +"fwrite", +"fflush", +"lj_vm_floor", +"lj_vm_ceil", +"lj_vm_trunc", +"sqrt", +"log", +"lj_vm_log2", +"lj_vm_powi", +"pow", +"atan2", +"ldexp", +"lj_vm_tobit", +"softfp_add", +"softfp_sub", +"softfp_mul", +"softfp_div", +"softfp_cmp", +"softfp_i2d", +"softfp_d2i", +"lj_vm_sfmin", +"lj_vm_sfmax", +"lj_vm_tointg", +"softfp_ui2d", +"softfp_f2d", +"softfp_d2ui", +"softfp_d2f", +"softfp_i2f", +"softfp_ui2f", +"softfp_f2i", +"softfp_f2ui", +"fp64_l2d", +"fp64_ul2d", +"fp64_l2f", +"fp64_ul2f", +"fp64_d2l", +"fp64_d2ul", +"fp64_f2l", +"fp64_f2ul", +"lj_carith_divi64", +"lj_carith_divu64", +"lj_carith_modi64", +"lj_carith_modu64", +"lj_carith_powi64", +"lj_carith_powu64", +"lj_cdata_newv", +"lj_cdata_setfin", +"strlen", +"memcpy", +"memset", +"lj_vm_errno", +"lj_carith_mul64", +"lj_carith_shl64", +"lj_carith_shr64", +"lj_carith_sar64", +"lj_carith_rol64", +"lj_carith_ror64", +}, + +traceerr = { +[0]="error thrown or hook called during recording", +"trace too short", +"trace too long", +"trace too deep", +"too many snapshots", +"blacklisted", +"retry recording", +"NYI: bytecode %d", +"leaving loop in root trace", +"inner loop in root trace", +"loop unroll limit reached", +"bad argument type", +"JIT compilation disabled for function", +"call unroll limit reached", +"down-recursion, restarting", +"NYI: unsupported variant of FastFunc %s", +"NYI: return to lower frame", +"store with nil or NaN key", +"missing metamethod", +"looping index lookup", +"NYI: mixed sparse/dense table", +"symbol not in cache", +"NYI: unsupported C type conversion", +"NYI: unsupported C function type", +"guard would always fail", +"too many PHIs", +"persistent type instability", +"failed to allocate mcode memory", +"machine code too long", +"hit mcode limit (retrying)", +"too many spill slots", +"inconsistent register allocation", +"NYI: cannot assemble IR instruction %d", +"NYI: PHI shuffling too complex", +"NYI: register coalescing too complex", +}, + +ffnames = { +[0]="Lua", +"C", +"assert", +"type", +"next", +"pairs", +"ipairs_aux", +"ipairs", +"getmetatable", +"setmetatable", +"getfenv", +"setfenv", +"rawget", +"rawset", +"rawequal", +"unpack", +"select", +"tonumber", +"tostring", +"error", +"pcall", +"xpcall", +"loadfile", +"load", +"loadstring", +"dofile", +"gcinfo", +"collectgarbage", +"newproxy", +"print", +"coroutine.status", +"coroutine.running", +"coroutine.isyieldable", +"coroutine.create", +"coroutine.yield", +"coroutine.resume", +"coroutine.wrap_aux", +"coroutine.wrap", +"math.abs", +"math.floor", +"math.ceil", +"math.sqrt", +"math.log10", +"math.exp", +"math.sin", +"math.cos", +"math.tan", +"math.asin", +"math.acos", +"math.atan", +"math.sinh", +"math.cosh", +"math.tanh", +"math.frexp", +"math.modf", +"math.log", +"math.atan2", +"math.pow", +"math.fmod", +"math.ldexp", +"math.min", +"math.max", +"math.random", +"math.randomseed", +"bit.tobit", +"bit.bnot", +"bit.bswap", +"bit.lshift", +"bit.rshift", +"bit.arshift", +"bit.rol", +"bit.ror", +"bit.band", +"bit.bor", +"bit.bxor", +"bit.tohex", +"string.byte", +"string.char", +"string.sub", +"string.rep", +"string.reverse", +"string.lower", +"string.upper", +"string.dump", +"string.find", +"string.match", +"string.gmatch_aux", +"string.gmatch", +"string.gsub", +"string.format", +"table.maxn", +"table.insert", +"table.concat", +"table.sort", +"table.new", +"table.clear", +"io.method.close", +"io.method.read", +"io.method.write", +"io.method.flush", +"io.method.seek", +"io.method.setvbuf", +"io.method.lines", +"io.method.__gc", +"io.method.__tostring", +"io.open", +"io.popen", +"io.tmpfile", +"io.close", +"io.read", +"io.write", +"io.flush", +"io.input", +"io.output", +"io.lines", +"io.type", +"os.execute", +"os.remove", +"os.rename", +"os.tmpname", +"os.getenv", +"os.exit", +"os.clock", +"os.date", +"os.time", +"os.difftime", +"os.setlocale", +"debug.getregistry", +"debug.getmetatable", +"debug.setmetatable", +"debug.getfenv", +"debug.setfenv", +"debug.getinfo", +"debug.getlocal", +"debug.setlocal", +"debug.getupvalue", +"debug.setupvalue", +"debug.upvalueid", +"debug.upvaluejoin", +"debug.sethook", +"debug.gethook", +"debug.debug", +"debug.traceback", +"jit.on", +"jit.off", +"jit.flush", +"jit.status", +"jit.security", +"jit.attach", +"jit.util.funcinfo", +"jit.util.funcbc", +"jit.util.funck", +"jit.util.funcuvname", +"jit.util.traceinfo", +"jit.util.traceir", +"jit.util.tracek", +"jit.util.tracesnap", +"jit.util.tracemc", +"jit.util.traceexitstub", +"jit.util.ircalladdr", +"jit.opt.start", +"jit.profile.start", +"jit.profile.stop", +"jit.profile.dumpstack", +"ffi.meta.__index", +"ffi.meta.__newindex", +"ffi.meta.__eq", +"ffi.meta.__len", +"ffi.meta.__lt", +"ffi.meta.__le", +"ffi.meta.__concat", +"ffi.meta.__call", +"ffi.meta.__add", +"ffi.meta.__sub", +"ffi.meta.__mul", +"ffi.meta.__div", +"ffi.meta.__mod", +"ffi.meta.__pow", +"ffi.meta.__unm", +"ffi.meta.__tostring", +"ffi.meta.__pairs", +"ffi.meta.__ipairs", +"ffi.clib.__index", +"ffi.clib.__newindex", +"ffi.clib.__gc", +"ffi.callback.free", +"ffi.callback.set", +"ffi.cdef", +"ffi.new", +"ffi.cast", +"ffi.typeof", +"ffi.typeinfo", +"ffi.istype", +"ffi.sizeof", +"ffi.alignof", +"ffi.offsetof", +"ffi.errno", +"ffi.string", +"ffi.copy", +"ffi.fill", +"ffi.abi", +"ffi.metatype", +"ffi.gc", +"ffi.load", +"buffer.method.free", +"buffer.method.reset", +"buffer.method.skip", +"buffer.method.set", +"buffer.method.put", +"buffer.method.putf", +"buffer.method.get", +"buffer.method.putcdata", +"buffer.method.reserve", +"buffer.method.commit", +"buffer.method.ref", +"buffer.method.encode", +"buffer.method.decode", +"buffer.method.__gc", +"buffer.method.__tostring", +"buffer.method.__len", +"buffer.new", +"buffer.encode", +"buffer.decode", +}, + +} + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/zone.lua b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/zone.lua new file mode 100644 index 00000000..94357854 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/jit/zone.lua @@ -0,0 +1,45 @@ +---------------------------------------------------------------------------- +-- LuaJIT profiler zones. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module implements a simple hierarchical zone model. +-- +-- Example usage: +-- +-- local zone = require("jit.zone") +-- zone("AI") +-- ... +-- zone("A*") +-- ... +-- print(zone:get()) --> "A*" +-- ... +-- zone() +-- ... +-- print(zone:get()) --> "AI" +-- ... +-- zone() +-- +---------------------------------------------------------------------------- + +local remove = table.remove + +return setmetatable({ + flush = function(t) + for i=#t,1,-1 do t[i] = nil end + end, + get = function(t) + return t[#t] + end +}, { + __call = function(t, zone) + if zone then + t[#t+1] = zone + else + return (assert(remove(t), "empty zone stack")) + end + end +}) + diff --git a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/libluajit.a b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/libluajit.a index 87646f43..d9dd4fef 100644 Binary files a/love/src/jni/LuaJIT-2.1/android/arm64-v8a/libluajit.a and b/love/src/jni/LuaJIT-2.1/android/arm64-v8a/libluajit.a differ diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/bc.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/bc.lua new file mode 100644 index 00000000..e58a3fef --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/bc.lua @@ -0,0 +1,190 @@ +---------------------------------------------------------------------------- +-- LuaJIT bytecode listing module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module lists the bytecode of a Lua function. If it's loaded by -jbc +-- it hooks into the parser and lists all functions of a chunk as they +-- are parsed. +-- +-- Example usage: +-- +-- luajit -jbc -e 'local x=0; for i=1,1e6 do x=x+i end; print(x)' +-- luajit -jbc=- foo.lua +-- luajit -jbc=foo.list foo.lua +-- +-- Default output is to stderr. To redirect the output to a file, pass a +-- filename as an argument (use '-' for stdout) or set the environment +-- variable LUAJIT_LISTFILE. The file is overwritten every time the module +-- is started. +-- +-- This module can also be used programmatically: +-- +-- local bc = require("jit.bc") +-- +-- local function foo() print("hello") end +-- +-- bc.dump(foo) --> -- BYTECODE -- [...] +-- print(bc.line(foo, 2)) --> 0002 KSTR 1 1 ; "hello" +-- +-- local out = { +-- -- Do something with each line: +-- write = function(t, ...) io.write(...) end, +-- close = function(t) end, +-- flush = function(t) end, +-- } +-- bc.dump(foo, out) +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local bit = require("bit") +local sub, gsub, format = string.sub, string.gsub, string.format +local byte, band, shr = string.byte, bit.band, bit.rshift +local funcinfo, funcbc, funck = jutil.funcinfo, jutil.funcbc, jutil.funck +local funcuvname = jutil.funcuvname +local bcnames = vmdef.bcnames +local stdout, stderr = io.stdout, io.stderr + +------------------------------------------------------------------------------ + +local function ctlsub(c) + if c == "\n" then return "\\n" + elseif c == "\r" then return "\\r" + elseif c == "\t" then return "\\t" + else return format("\\%03d", byte(c)) + end +end + +-- Return one bytecode line. +local function bcline(func, pc, prefix) + local ins, m = funcbc(func, pc) + if not ins then return end + local ma, mb, mc = band(m, 7), band(m, 15*8), band(m, 15*128) + local a = band(shr(ins, 8), 0xff) + local oidx = 6*band(ins, 0xff) + local op = sub(bcnames, oidx+1, oidx+6) + local s = format("%04d %s %-6s %3s ", + pc, prefix or " ", op, ma == 0 and "" or a) + local d = shr(ins, 16) + if mc == 13*128 then -- BCMjump + return format("%s=> %04d\n", s, pc+d-0x7fff) + end + if mb ~= 0 then + d = band(d, 0xff) + elseif mc == 0 then + return s.."\n" + end + local kc + if mc == 10*128 then -- BCMstr + kc = funck(func, -d-1) + kc = format(#kc > 40 and '"%.40s"~' or '"%s"', gsub(kc, "%c", ctlsub)) + elseif mc == 9*128 then -- BCMnum + kc = funck(func, d) + if op == "TSETM " then kc = kc - 2^52 end + elseif mc == 12*128 then -- BCMfunc + local fi = funcinfo(funck(func, -d-1)) + if fi.ffid then + kc = vmdef.ffnames[fi.ffid] + else + kc = fi.loc + end + elseif mc == 5*128 then -- BCMuv + kc = funcuvname(func, d) + end + if ma == 5 then -- BCMuv + local ka = funcuvname(func, a) + if kc then kc = ka.." ; "..kc else kc = ka end + end + if mb ~= 0 then + local b = shr(ins, 24) + if kc then return format("%s%3d %3d ; %s\n", s, b, d, kc) end + return format("%s%3d %3d\n", s, b, d) + end + if kc then return format("%s%3d ; %s\n", s, d, kc) end + if mc == 7*128 and d > 32767 then d = d - 65536 end -- BCMlits + return format("%s%3d\n", s, d) +end + +-- Collect branch targets of a function. +local function bctargets(func) + local target = {} + for pc=1,1000000000 do + local ins, m = funcbc(func, pc) + if not ins then break end + if band(m, 15*128) == 13*128 then target[pc+shr(ins, 16)-0x7fff] = true end + end + return target +end + +-- Dump bytecode instructions of a function. +local function bcdump(func, out, all) + if not out then out = stdout end + local fi = funcinfo(func) + if all and fi.children then + for n=-1,-1000000000,-1 do + local k = funck(func, n) + if not k then break end + if type(k) == "proto" then bcdump(k, out, true) end + end + end + out:write(format("-- BYTECODE -- %s-%d\n", fi.loc, fi.lastlinedefined)) + local target = bctargets(func) + for pc=1,1000000000 do + local s = bcline(func, pc, target[pc] and "=>") + if not s then break end + out:write(s) + end + out:write("\n") + out:flush() +end + +------------------------------------------------------------------------------ + +-- Active flag and output file handle. +local active, out + +-- List handler. +local function h_list(func) + return bcdump(func, out) +end + +-- Detach list handler. +local function bclistoff() + if active then + active = false + jit.attach(h_list) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach list handler. +local function bcliston(outfile) + if active then bclistoff() end + if not outfile then outfile = os.getenv("LUAJIT_LISTFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stderr + end + jit.attach(h_list, "bc") + active = true +end + +-- Public module functions. +return { + line = bcline, + dump = bcdump, + targets = bctargets, + on = bcliston, + off = bclistoff, + start = bcliston -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/bcsave.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/bcsave.lua new file mode 100644 index 00000000..ab13667a --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/bcsave.lua @@ -0,0 +1,665 @@ +---------------------------------------------------------------------------- +-- LuaJIT module to save/list bytecode. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module saves or lists the bytecode for an input file. +-- It's run by the -b command line option. +-- +------------------------------------------------------------------------------ + +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local bit = require("bit") + +-- Symbol name prefix for LuaJIT bytecode. +local LJBC_PREFIX = "luaJIT_BC_" + +local type, assert = type, assert +local format = string.format +local tremove, tconcat = table.remove, table.concat + +------------------------------------------------------------------------------ + +local function usage() + io.stderr:write[[ +Save LuaJIT bytecode: luajit -b[options] input output + -l Only list bytecode. + -s Strip debug info (default). + -g Keep debug info. + -n name Set module name (default: auto-detect from input name). + -t type Set output file type (default: auto-detect from output name). + -a arch Override architecture for object files (default: native). + -o os Override OS for object files (default: native). + -e chunk Use chunk string as input. + -- Stop handling options. + - Use stdin as input and/or stdout as output. + +File types: c h obj o raw (default) +]] + os.exit(1) +end + +local function check(ok, ...) + if ok then return ok, ... end + io.stderr:write("luajit: ", ...) + io.stderr:write("\n") + os.exit(1) +end + +local function readfile(input) + if type(input) == "function" then return input end + if input == "-" then input = nil end + return check(loadfile(input)) +end + +local function savefile(name, mode) + if name == "-" then return io.stdout end + return check(io.open(name, mode)) +end + +------------------------------------------------------------------------------ + +local map_type = { + raw = "raw", c = "c", h = "h", o = "obj", obj = "obj", +} + +local map_arch = { + x86 = { e = "le", b = 32, m = 3, p = 0x14c, }, + x64 = { e = "le", b = 64, m = 62, p = 0x8664, }, + arm = { e = "le", b = 32, m = 40, p = 0x1c0, }, + arm64 = { e = "le", b = 64, m = 183, p = 0xaa64, }, + arm64be = { e = "be", b = 64, m = 183, }, + ppc = { e = "be", b = 32, m = 20, }, + mips = { e = "be", b = 32, m = 8, f = 0x50001006, }, + mipsel = { e = "le", b = 32, m = 8, f = 0x50001006, }, + mips64 = { e = "be", b = 64, m = 8, f = 0x80000007, }, + mips64el = { e = "le", b = 64, m = 8, f = 0x80000007, }, + mips64r6 = { e = "be", b = 64, m = 8, f = 0xa0000407, }, + mips64r6el = { e = "le", b = 64, m = 8, f = 0xa0000407, }, +} + +local map_os = { + linux = true, windows = true, osx = true, freebsd = true, netbsd = true, + openbsd = true, dragonfly = true, solaris = true, +} + +local function checkarg(str, map, err) + str = str:lower() + local s = check(map[str], "unknown ", err) + return type(s) == "string" and s or str +end + +local function detecttype(str) + local ext = str:lower():match("%.(%a+)$") + return map_type[ext] or "raw" +end + +local function checkmodname(str) + check(str:match("^[%w_.%-]+$"), "bad module name") + return str:gsub("[%.%-]", "_") +end + +local function detectmodname(str) + if type(str) == "string" then + local tail = str:match("[^/\\]+$") + if tail then str = tail end + local head = str:match("^(.*)%.[^.]*$") + if head then str = head end + str = str:match("^[%w_.%-]+") + else + str = nil + end + check(str, "cannot derive module name, use -n name") + return str:gsub("[%.%-]", "_") +end + +------------------------------------------------------------------------------ + +local function bcsave_tail(fp, output, s) + local ok, err = fp:write(s) + if ok and output ~= "-" then ok, err = fp:close() end + check(ok, "cannot write ", output, ": ", err) +end + +local function bcsave_raw(output, s) + local fp = savefile(output, "wb") + bcsave_tail(fp, output, s) +end + +local function bcsave_c(ctx, output, s) + local fp = savefile(output, "w") + if ctx.type == "c" then + fp:write(format([[ +#ifdef __cplusplus +extern "C" +#endif +#ifdef _WIN32 +__declspec(dllexport) +#endif +const unsigned char %s%s[] = { +]], LJBC_PREFIX, ctx.modname)) + else + fp:write(format([[ +#define %s%s_SIZE %d +static const unsigned char %s%s[] = { +]], LJBC_PREFIX, ctx.modname, #s, LJBC_PREFIX, ctx.modname)) + end + local t, n, m = {}, 0, 0 + for i=1,#s do + local b = tostring(string.byte(s, i)) + m = m + #b + 1 + if m > 78 then + fp:write(tconcat(t, ",", 1, n), ",\n") + n, m = 0, #b + 1 + end + n = n + 1 + t[n] = b + end + bcsave_tail(fp, output, tconcat(t, ",", 1, n).."\n};\n") +end + +local function bcsave_elfobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct { + uint8_t emagic[4], eclass, eendian, eversion, eosabi, eabiversion, epad[7]; + uint16_t type, machine; + uint32_t version; + uint32_t entry, phofs, shofs; + uint32_t flags; + uint16_t ehsize, phentsize, phnum, shentsize, shnum, shstridx; +} ELF32header; +typedef struct { + uint8_t emagic[4], eclass, eendian, eversion, eosabi, eabiversion, epad[7]; + uint16_t type, machine; + uint32_t version; + uint64_t entry, phofs, shofs; + uint32_t flags; + uint16_t ehsize, phentsize, phnum, shentsize, shnum, shstridx; +} ELF64header; +typedef struct { + uint32_t name, type, flags, addr, ofs, size, link, info, align, entsize; +} ELF32sectheader; +typedef struct { + uint32_t name, type; + uint64_t flags, addr, ofs, size; + uint32_t link, info; + uint64_t align, entsize; +} ELF64sectheader; +typedef struct { + uint32_t name, value, size; + uint8_t info, other; + uint16_t sectidx; +} ELF32symbol; +typedef struct { + uint32_t name; + uint8_t info, other; + uint16_t sectidx; + uint64_t value, size; +} ELF64symbol; +typedef struct { + ELF32header hdr; + ELF32sectheader sect[6]; + ELF32symbol sym[2]; + uint8_t space[4096]; +} ELF32obj; +typedef struct { + ELF64header hdr; + ELF64sectheader sect[6]; + ELF64symbol sym[2]; + uint8_t space[4096]; +} ELF64obj; +]] + local symname = LJBC_PREFIX..ctx.modname + local ai = assert(map_arch[ctx.arch]) + local is64, isbe = ai.b == 64, ai.e == "be" + + -- Handle different host/target endianess. + local function f32(x) return x end + local f16, fofs = f32, f32 + if ffi.abi("be") ~= isbe then + f32 = bit.bswap + function f16(x) return bit.rshift(bit.bswap(x), 16) end + if is64 then + local two32 = ffi.cast("int64_t", 2^32) + function fofs(x) return bit.bswap(x)*two32 end + else + fofs = f32 + end + end + + -- Create ELF object and fill in header. + local o = ffi.new(is64 and "ELF64obj" or "ELF32obj") + local hdr = o.hdr + if ctx.os == "bsd" or ctx.os == "other" then -- Determine native hdr.eosabi. + local bf = assert(io.open("/bin/ls", "rb")) + local bs = bf:read(9) + bf:close() + ffi.copy(o, bs, 9) + check(hdr.emagic[0] == 127, "no support for writing native object files") + else + hdr.emagic = "\127ELF" + hdr.eosabi = ({ freebsd=9, netbsd=2, openbsd=12, solaris=6 })[ctx.os] or 0 + end + hdr.eclass = is64 and 2 or 1 + hdr.eendian = isbe and 2 or 1 + hdr.eversion = 1 + hdr.type = f16(1) + hdr.machine = f16(ai.m) + hdr.flags = f32(ai.f or 0) + hdr.version = f32(1) + hdr.shofs = fofs(ffi.offsetof(o, "sect")) + hdr.ehsize = f16(ffi.sizeof(hdr)) + hdr.shentsize = f16(ffi.sizeof(o.sect[0])) + hdr.shnum = f16(6) + hdr.shstridx = f16(2) + + -- Fill in sections and symbols. + local sofs, ofs = ffi.offsetof(o, "space"), 1 + for i,name in ipairs{ + ".symtab", ".shstrtab", ".strtab", ".rodata", ".note.GNU-stack", + } do + local sect = o.sect[i] + sect.align = fofs(1) + sect.name = f32(ofs) + ffi.copy(o.space+ofs, name) + ofs = ofs + #name+1 + end + o.sect[1].type = f32(2) -- .symtab + o.sect[1].link = f32(3) + o.sect[1].info = f32(1) + o.sect[1].align = fofs(8) + o.sect[1].ofs = fofs(ffi.offsetof(o, "sym")) + o.sect[1].entsize = fofs(ffi.sizeof(o.sym[0])) + o.sect[1].size = fofs(ffi.sizeof(o.sym)) + o.sym[1].name = f32(1) + o.sym[1].sectidx = f16(4) + o.sym[1].size = fofs(#s) + o.sym[1].info = 17 + o.sect[2].type = f32(3) -- .shstrtab + o.sect[2].ofs = fofs(sofs) + o.sect[2].size = fofs(ofs) + o.sect[3].type = f32(3) -- .strtab + o.sect[3].ofs = fofs(sofs + ofs) + o.sect[3].size = fofs(#symname+2) + ffi.copy(o.space+ofs+1, symname) + ofs = ofs + #symname + 2 + o.sect[4].type = f32(1) -- .rodata + o.sect[4].flags = fofs(2) + o.sect[4].ofs = fofs(sofs + ofs) + o.sect[4].size = fofs(#s) + o.sect[5].type = f32(1) -- .note.GNU-stack + o.sect[5].ofs = fofs(sofs + ofs + #s) + + -- Write ELF object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, ffi.sizeof(o)-4096+ofs)) + bcsave_tail(fp, output, s) +end + +local function bcsave_peobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct { + uint16_t arch, nsects; + uint32_t time, symtabofs, nsyms; + uint16_t opthdrsz, flags; +} PEheader; +typedef struct { + char name[8]; + uint32_t vsize, vaddr, size, ofs, relocofs, lineofs; + uint16_t nreloc, nline; + uint32_t flags; +} PEsection; +typedef struct __attribute((packed)) { + union { + char name[8]; + uint32_t nameref[2]; + }; + uint32_t value; + int16_t sect; + uint16_t type; + uint8_t scl, naux; +} PEsym; +typedef struct __attribute((packed)) { + uint32_t size; + uint16_t nreloc, nline; + uint32_t cksum; + uint16_t assoc; + uint8_t comdatsel, unused[3]; +} PEsymaux; +typedef struct { + PEheader hdr; + PEsection sect[2]; + // Must be an even number of symbol structs. + PEsym sym0; + PEsymaux sym0aux; + PEsym sym1; + PEsymaux sym1aux; + PEsym sym2; + PEsym sym3; + uint32_t strtabsize; + uint8_t space[4096]; +} PEobj; +]] + local symname = LJBC_PREFIX..ctx.modname + local ai = assert(map_arch[ctx.arch]) + local is64 = ai.b == 64 + local symexport = " /EXPORT:"..symname..",DATA " + + -- The file format is always little-endian. Swap if the host is big-endian. + local function f32(x) return x end + local f16 = f32 + if ffi.abi("be") then + f32 = bit.bswap + function f16(x) return bit.rshift(bit.bswap(x), 16) end + end + + -- Create PE object and fill in header. + local o = ffi.new("PEobj") + local hdr = o.hdr + hdr.arch = f16(assert(ai.p)) + hdr.nsects = f16(2) + hdr.symtabofs = f32(ffi.offsetof(o, "sym0")) + hdr.nsyms = f32(6) + + -- Fill in sections and symbols. + o.sect[0].name = ".drectve" + o.sect[0].size = f32(#symexport) + o.sect[0].flags = f32(0x00100a00) + o.sym0.sect = f16(1) + o.sym0.scl = 3 + o.sym0.name = ".drectve" + o.sym0.naux = 1 + o.sym0aux.size = f32(#symexport) + o.sect[1].name = ".rdata" + o.sect[1].size = f32(#s) + o.sect[1].flags = f32(0x40300040) + o.sym1.sect = f16(2) + o.sym1.scl = 3 + o.sym1.name = ".rdata" + o.sym1.naux = 1 + o.sym1aux.size = f32(#s) + o.sym2.sect = f16(2) + o.sym2.scl = 2 + o.sym2.nameref[1] = f32(4) + o.sym3.sect = f16(-1) + o.sym3.scl = 2 + o.sym3.value = f32(1) + o.sym3.name = "@feat.00" -- Mark as SafeSEH compliant. + ffi.copy(o.space, symname) + local ofs = #symname + 1 + o.strtabsize = f32(ofs + 4) + o.sect[0].ofs = f32(ffi.offsetof(o, "space") + ofs) + ffi.copy(o.space + ofs, symexport) + ofs = ofs + #symexport + o.sect[1].ofs = f32(ffi.offsetof(o, "space") + ofs) + + -- Write PE object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, ffi.sizeof(o)-4096+ofs)) + bcsave_tail(fp, output, s) +end + +local function bcsave_machobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct +{ + uint32_t magic, cputype, cpusubtype, filetype, ncmds, sizeofcmds, flags; +} mach_header; +typedef struct +{ + mach_header; uint32_t reserved; +} mach_header_64; +typedef struct { + uint32_t cmd, cmdsize; + char segname[16]; + uint32_t vmaddr, vmsize, fileoff, filesize; + uint32_t maxprot, initprot, nsects, flags; +} mach_segment_command; +typedef struct { + uint32_t cmd, cmdsize; + char segname[16]; + uint64_t vmaddr, vmsize, fileoff, filesize; + uint32_t maxprot, initprot, nsects, flags; +} mach_segment_command_64; +typedef struct { + char sectname[16], segname[16]; + uint32_t addr, size; + uint32_t offset, align, reloff, nreloc, flags; + uint32_t reserved1, reserved2; +} mach_section; +typedef struct { + char sectname[16], segname[16]; + uint64_t addr, size; + uint32_t offset, align, reloff, nreloc, flags; + uint32_t reserved1, reserved2, reserved3; +} mach_section_64; +typedef struct { + uint32_t cmd, cmdsize, symoff, nsyms, stroff, strsize; +} mach_symtab_command; +typedef struct { + int32_t strx; + uint8_t type, sect; + int16_t desc; + uint32_t value; +} mach_nlist; +typedef struct { + uint32_t strx; + uint8_t type, sect; + uint16_t desc; + uint64_t value; +} mach_nlist_64; +typedef struct +{ + uint32_t magic, nfat_arch; +} mach_fat_header; +typedef struct +{ + uint32_t cputype, cpusubtype, offset, size, align; +} mach_fat_arch; +typedef struct { + struct { + mach_header hdr; + mach_segment_command seg; + mach_section sec; + mach_symtab_command sym; + } arch[1]; + mach_nlist sym_entry; + uint8_t space[4096]; +} mach_obj; +typedef struct { + struct { + mach_header_64 hdr; + mach_segment_command_64 seg; + mach_section_64 sec; + mach_symtab_command sym; + } arch[1]; + mach_nlist_64 sym_entry; + uint8_t space[4096]; +} mach_obj_64; +typedef struct { + mach_fat_header fat; + mach_fat_arch fat_arch[2]; + struct { + mach_header hdr; + mach_segment_command seg; + mach_section sec; + mach_symtab_command sym; + } arch[2]; + mach_nlist sym_entry; + uint8_t space[4096]; +} mach_fat_obj; +]] + local symname = '_'..LJBC_PREFIX..ctx.modname + local isfat, is64, align, mobj = false, false, 4, "mach_obj" + if ctx.arch == "x64" then + is64, align, mobj = true, 8, "mach_obj_64" + elseif ctx.arch == "arm" then + isfat, mobj = true, "mach_fat_obj" + elseif ctx.arch == "arm64" then + is64, align, isfat, mobj = true, 8, true, "mach_fat_obj" + else + check(ctx.arch == "x86", "unsupported architecture for OSX") + end + local function aligned(v, a) return bit.band(v+a-1, -a) end + local be32 = bit.bswap -- Mach-O FAT is BE, supported archs are LE. + + -- Create Mach-O object and fill in header. + local o = ffi.new(mobj) + local mach_size = aligned(ffi.offsetof(o, "space")+#symname+2, align) + local cputype = ({ x86={7}, x64={0x01000007}, arm={7,12}, arm64={0x01000007,0x0100000c} })[ctx.arch] + local cpusubtype = ({ x86={3}, x64={3}, arm={3,9}, arm64={3,0} })[ctx.arch] + if isfat then + o.fat.magic = be32(0xcafebabe) + o.fat.nfat_arch = be32(#cpusubtype) + end + + -- Fill in sections and symbols. + for i=0,#cpusubtype-1 do + local ofs = 0 + if isfat then + local a = o.fat_arch[i] + a.cputype = be32(cputype[i+1]) + a.cpusubtype = be32(cpusubtype[i+1]) + -- Subsequent slices overlap each other to share data. + ofs = ffi.offsetof(o, "arch") + i*ffi.sizeof(o.arch[0]) + a.offset = be32(ofs) + a.size = be32(mach_size-ofs+#s) + end + local a = o.arch[i] + a.hdr.magic = is64 and 0xfeedfacf or 0xfeedface + a.hdr.cputype = cputype[i+1] + a.hdr.cpusubtype = cpusubtype[i+1] + a.hdr.filetype = 1 + a.hdr.ncmds = 2 + a.hdr.sizeofcmds = ffi.sizeof(a.seg)+ffi.sizeof(a.sec)+ffi.sizeof(a.sym) + a.seg.cmd = is64 and 0x19 or 0x1 + a.seg.cmdsize = ffi.sizeof(a.seg)+ffi.sizeof(a.sec) + a.seg.vmsize = #s + a.seg.fileoff = mach_size-ofs + a.seg.filesize = #s + a.seg.maxprot = 1 + a.seg.initprot = 1 + a.seg.nsects = 1 + ffi.copy(a.sec.sectname, "__data") + ffi.copy(a.sec.segname, "__DATA") + a.sec.size = #s + a.sec.offset = mach_size-ofs + a.sym.cmd = 2 + a.sym.cmdsize = ffi.sizeof(a.sym) + a.sym.symoff = ffi.offsetof(o, "sym_entry")-ofs + a.sym.nsyms = 1 + a.sym.stroff = ffi.offsetof(o, "sym_entry")+ffi.sizeof(o.sym_entry)-ofs + a.sym.strsize = aligned(#symname+2, align) + end + o.sym_entry.type = 0xf + o.sym_entry.sect = 1 + o.sym_entry.strx = 1 + ffi.copy(o.space+1, symname) + + -- Write Macho-O object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, mach_size)) + bcsave_tail(fp, output, s) +end + +local function bcsave_obj(ctx, output, s) + local ok, ffi = pcall(require, "ffi") + check(ok, "FFI library required to write this file type") + if ctx.os == "windows" then + return bcsave_peobj(ctx, output, s, ffi) + elseif ctx.os == "osx" then + return bcsave_machobj(ctx, output, s, ffi) + else + return bcsave_elfobj(ctx, output, s, ffi) + end +end + +------------------------------------------------------------------------------ + +local function bclist(input, output) + local f = readfile(input) + require("jit.bc").dump(f, savefile(output, "w"), true) +end + +local function bcsave(ctx, input, output) + local f = readfile(input) + local s = string.dump(f, ctx.strip) + local t = ctx.type + if not t then + t = detecttype(output) + ctx.type = t + end + if t == "raw" then + bcsave_raw(output, s) + else + if not ctx.modname then ctx.modname = detectmodname(input) end + if t == "obj" then + bcsave_obj(ctx, output, s) + else + bcsave_c(ctx, output, s) + end + end +end + +local function docmd(...) + local arg = {...} + local n = 1 + local list = false + local ctx = { + strip = true, arch = jit.arch, os = jit.os:lower(), + type = false, modname = false, + } + while n <= #arg do + local a = arg[n] + if type(a) == "string" and a:sub(1, 1) == "-" and a ~= "-" then + tremove(arg, n) + if a == "--" then break end + for m=2,#a do + local opt = a:sub(m, m) + if opt == "l" then + list = true + elseif opt == "s" then + ctx.strip = true + elseif opt == "g" then + ctx.strip = false + else + if arg[n] == nil or m ~= #a then usage() end + if opt == "e" then + if n ~= 1 then usage() end + arg[1] = check(loadstring(arg[1])) + elseif opt == "n" then + ctx.modname = checkmodname(tremove(arg, n)) + elseif opt == "t" then + ctx.type = checkarg(tremove(arg, n), map_type, "file type") + elseif opt == "a" then + ctx.arch = checkarg(tremove(arg, n), map_arch, "architecture") + elseif opt == "o" then + ctx.os = checkarg(tremove(arg, n), map_os, "OS name") + else + usage() + end + end + end + else + n = n + 1 + end + end + if list then + if #arg == 0 or #arg > 2 then usage() end + bclist(arg[1], arg[2] or "-") + else + if #arg ~= 2 then usage() end + bcsave(ctx, arg[1], arg[2]) + end +end + +------------------------------------------------------------------------------ + +-- Public module functions. +return { + start = docmd -- Process -b command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm.lua new file mode 100644 index 00000000..ba79c47e --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm.lua @@ -0,0 +1,689 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles most user-mode ARMv7 instructions +-- NYI: Advanced SIMD and VFP instructions. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch = string.match, string.gmatch +local concat = table.concat +local bit = require("bit") +local band, bor, ror, tohex = bit.band, bit.bor, bit.ror, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Opcode maps +------------------------------------------------------------------------------ + +local map_loadc = { + shift = 8, mask = 15, + [10] = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "vmovFmDN", "vstmFNdr", + _ = { + shift = 21, mask = 1, + [0] = "vstrFdl", + { shift = 16, mask = 15, [13] = "vpushFdr", _ = "vstmdbFNdr", } + }, + }, + { + shift = 23, mask = 3, + [0] = "vmovFDNm", + { shift = 16, mask = 15, [13] = "vpopFdr", _ = "vldmFNdr", }, + _ = { + shift = 21, mask = 1, + [0] = "vldrFdl", "vldmdbFNdr", + }, + }, + }, + [11] = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "vmovGmDN", "vstmGNdr", + _ = { + shift = 21, mask = 1, + [0] = "vstrGdl", + { shift = 16, mask = 15, [13] = "vpushGdr", _ = "vstmdbGNdr", } + }, + }, + { + shift = 23, mask = 3, + [0] = "vmovGDNm", + { shift = 16, mask = 15, [13] = "vpopGdr", _ = "vldmGNdr", }, + _ = { + shift = 21, mask = 1, + [0] = "vldrGdl", "vldmdbGNdr", + }, + }, + }, + _ = { + shift = 0, mask = 0 -- NYI ldc, mcrr, mrrc. + }, +} + +local map_vfps = { + shift = 6, mask = 0x2c001, + [0] = "vmlaF.dnm", "vmlsF.dnm", + [0x04000] = "vnmlsF.dnm", [0x04001] = "vnmlaF.dnm", + [0x08000] = "vmulF.dnm", [0x08001] = "vnmulF.dnm", + [0x0c000] = "vaddF.dnm", [0x0c001] = "vsubF.dnm", + [0x20000] = "vdivF.dnm", + [0x24000] = "vfnmsF.dnm", [0x24001] = "vfnmaF.dnm", + [0x28000] = "vfmaF.dnm", [0x28001] = "vfmsF.dnm", + [0x2c000] = "vmovF.dY", + [0x2c001] = { + shift = 7, mask = 0x1e01, + [0] = "vmovF.dm", "vabsF.dm", + [0x0200] = "vnegF.dm", [0x0201] = "vsqrtF.dm", + [0x0800] = "vcmpF.dm", [0x0801] = "vcmpeF.dm", + [0x0a00] = "vcmpzF.d", [0x0a01] = "vcmpzeF.d", + [0x0e01] = "vcvtG.dF.m", + [0x1000] = "vcvt.f32.u32Fdm", [0x1001] = "vcvt.f32.s32Fdm", + [0x1800] = "vcvtr.u32F.dm", [0x1801] = "vcvt.u32F.dm", + [0x1a00] = "vcvtr.s32F.dm", [0x1a01] = "vcvt.s32F.dm", + }, +} + +local map_vfpd = { + shift = 6, mask = 0x2c001, + [0] = "vmlaG.dnm", "vmlsG.dnm", + [0x04000] = "vnmlsG.dnm", [0x04001] = "vnmlaG.dnm", + [0x08000] = "vmulG.dnm", [0x08001] = "vnmulG.dnm", + [0x0c000] = "vaddG.dnm", [0x0c001] = "vsubG.dnm", + [0x20000] = "vdivG.dnm", + [0x24000] = "vfnmsG.dnm", [0x24001] = "vfnmaG.dnm", + [0x28000] = "vfmaG.dnm", [0x28001] = "vfmsG.dnm", + [0x2c000] = "vmovG.dY", + [0x2c001] = { + shift = 7, mask = 0x1e01, + [0] = "vmovG.dm", "vabsG.dm", + [0x0200] = "vnegG.dm", [0x0201] = "vsqrtG.dm", + [0x0800] = "vcmpG.dm", [0x0801] = "vcmpeG.dm", + [0x0a00] = "vcmpzG.d", [0x0a01] = "vcmpzeG.d", + [0x0e01] = "vcvtF.dG.m", + [0x1000] = "vcvt.f64.u32GdFm", [0x1001] = "vcvt.f64.s32GdFm", + [0x1800] = "vcvtr.u32FdG.m", [0x1801] = "vcvt.u32FdG.m", + [0x1a00] = "vcvtr.s32FdG.m", [0x1a01] = "vcvt.s32FdG.m", + }, +} + +local map_datac = { + shift = 24, mask = 1, + [0] = { + shift = 4, mask = 1, + [0] = { + shift = 8, mask = 15, + [10] = map_vfps, + [11] = map_vfpd, + -- NYI cdp, mcr, mrc. + }, + { + shift = 8, mask = 15, + [10] = { + shift = 20, mask = 15, + [0] = "vmovFnD", "vmovFDn", + [14] = "vmsrD", + [15] = { shift = 12, mask = 15, [15] = "vmrs", _ = "vmrsD", }, + }, + }, + }, + "svcT", +} + +local map_loadcu = { + shift = 0, mask = 0, -- NYI unconditional CP load/store. +} + +local map_datacu = { + shift = 0, mask = 0, -- NYI unconditional CP data. +} + +local map_simddata = { + shift = 0, mask = 0, -- NYI SIMD data. +} + +local map_simdload = { + shift = 0, mask = 0, -- NYI SIMD load/store, preload. +} + +local map_preload = { + shift = 0, mask = 0, -- NYI preload. +} + +local map_media = { + shift = 20, mask = 31, + [0] = false, + { --01 + shift = 5, mask = 7, + [0] = "sadd16DNM", "sasxDNM", "ssaxDNM", "ssub16DNM", + "sadd8DNM", false, false, "ssub8DNM", + }, + { --02 + shift = 5, mask = 7, + [0] = "qadd16DNM", "qasxDNM", "qsaxDNM", "qsub16DNM", + "qadd8DNM", false, false, "qsub8DNM", + }, + { --03 + shift = 5, mask = 7, + [0] = "shadd16DNM", "shasxDNM", "shsaxDNM", "shsub16DNM", + "shadd8DNM", false, false, "shsub8DNM", + }, + false, + { --05 + shift = 5, mask = 7, + [0] = "uadd16DNM", "uasxDNM", "usaxDNM", "usub16DNM", + "uadd8DNM", false, false, "usub8DNM", + }, + { --06 + shift = 5, mask = 7, + [0] = "uqadd16DNM", "uqasxDNM", "uqsaxDNM", "uqsub16DNM", + "uqadd8DNM", false, false, "uqsub8DNM", + }, + { --07 + shift = 5, mask = 7, + [0] = "uhadd16DNM", "uhasxDNM", "uhsaxDNM", "uhsub16DNM", + "uhadd8DNM", false, false, "uhsub8DNM", + }, + { --08 + shift = 5, mask = 7, + [0] = "pkhbtDNMU", false, "pkhtbDNMU", + { shift = 16, mask = 15, [15] = "sxtb16DMU", _ = "sxtab16DNMU", }, + "pkhbtDNMU", "selDNM", "pkhtbDNMU", + }, + false, + { --0a + shift = 5, mask = 7, + [0] = "ssatDxMu", "ssat16DxM", "ssatDxMu", + { shift = 16, mask = 15, [15] = "sxtbDMU", _ = "sxtabDNMU", }, + "ssatDxMu", false, "ssatDxMu", + }, + { --0b + shift = 5, mask = 7, + [0] = "ssatDxMu", "revDM", "ssatDxMu", + { shift = 16, mask = 15, [15] = "sxthDMU", _ = "sxtahDNMU", }, + "ssatDxMu", "rev16DM", "ssatDxMu", + }, + { --0c + shift = 5, mask = 7, + [3] = { shift = 16, mask = 15, [15] = "uxtb16DMU", _ = "uxtab16DNMU", }, + }, + false, + { --0e + shift = 5, mask = 7, + [0] = "usatDwMu", "usat16DwM", "usatDwMu", + { shift = 16, mask = 15, [15] = "uxtbDMU", _ = "uxtabDNMU", }, + "usatDwMu", false, "usatDwMu", + }, + { --0f + shift = 5, mask = 7, + [0] = "usatDwMu", "rbitDM", "usatDwMu", + { shift = 16, mask = 15, [15] = "uxthDMU", _ = "uxtahDNMU", }, + "usatDwMu", "revshDM", "usatDwMu", + }, + { --10 + shift = 12, mask = 15, + [15] = { + shift = 5, mask = 7, + "smuadNMS", "smuadxNMS", "smusdNMS", "smusdxNMS", + }, + _ = { + shift = 5, mask = 7, + [0] = "smladNMSD", "smladxNMSD", "smlsdNMSD", "smlsdxNMSD", + }, + }, + false, false, false, + { --14 + shift = 5, mask = 7, + [0] = "smlaldDNMS", "smlaldxDNMS", "smlsldDNMS", "smlsldxDNMS", + }, + { --15 + shift = 5, mask = 7, + [0] = { shift = 12, mask = 15, [15] = "smmulNMS", _ = "smmlaNMSD", }, + { shift = 12, mask = 15, [15] = "smmulrNMS", _ = "smmlarNMSD", }, + false, false, false, false, + "smmlsNMSD", "smmlsrNMSD", + }, + false, false, + { --18 + shift = 5, mask = 7, + [0] = { shift = 12, mask = 15, [15] = "usad8NMS", _ = "usada8NMSD", }, + }, + false, + { --1a + shift = 5, mask = 3, [2] = "sbfxDMvw", + }, + { --1b + shift = 5, mask = 3, [2] = "sbfxDMvw", + }, + { --1c + shift = 5, mask = 3, + [0] = { shift = 0, mask = 15, [15] = "bfcDvX", _ = "bfiDMvX", }, + }, + { --1d + shift = 5, mask = 3, + [0] = { shift = 0, mask = 15, [15] = "bfcDvX", _ = "bfiDMvX", }, + }, + { --1e + shift = 5, mask = 3, [2] = "ubfxDMvw", + }, + { --1f + shift = 5, mask = 3, [2] = "ubfxDMvw", + }, +} + +local map_load = { + shift = 21, mask = 9, + { + shift = 20, mask = 5, + [0] = "strtDL", "ldrtDL", [4] = "strbtDL", [5] = "ldrbtDL", + }, + _ = { + shift = 20, mask = 5, + [0] = "strDL", "ldrDL", [4] = "strbDL", [5] = "ldrbDL", + } +} + +local map_load1 = { + shift = 4, mask = 1, + [0] = map_load, map_media, +} + +local map_loadm = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "stmdaNR", "stmNR", + { shift = 16, mask = 63, [45] = "pushR", _ = "stmdbNR", }, "stmibNR", + }, + { + shift = 23, mask = 3, + [0] = "ldmdaNR", { shift = 16, mask = 63, [61] = "popR", _ = "ldmNR", }, + "ldmdbNR", "ldmibNR", + }, +} + +local map_data = { + shift = 21, mask = 15, + [0] = "andDNPs", "eorDNPs", "subDNPs", "rsbDNPs", + "addDNPs", "adcDNPs", "sbcDNPs", "rscDNPs", + "tstNP", "teqNP", "cmpNP", "cmnNP", + "orrDNPs", "movDPs", "bicDNPs", "mvnDPs", +} + +local map_mul = { + shift = 21, mask = 7, + [0] = "mulNMSs", "mlaNMSDs", "umaalDNMS", "mlsDNMS", + "umullDNMSs", "umlalDNMSs", "smullDNMSs", "smlalDNMSs", +} + +local map_sync = { + shift = 20, mask = 15, -- NYI: brackets around N. R(D+1) for ldrexd/strexd. + [0] = "swpDMN", false, false, false, + "swpbDMN", false, false, false, + "strexDMN", "ldrexDN", "strexdDN", "ldrexdDN", + "strexbDMN", "ldrexbDN", "strexhDN", "ldrexhDN", +} + +local map_mulh = { + shift = 21, mask = 3, + [0] = { shift = 5, mask = 3, + [0] = "smlabbNMSD", "smlatbNMSD", "smlabtNMSD", "smlattNMSD", }, + { shift = 5, mask = 3, + [0] = "smlawbNMSD", "smulwbNMS", "smlawtNMSD", "smulwtNMS", }, + { shift = 5, mask = 3, + [0] = "smlalbbDNMS", "smlaltbDNMS", "smlalbtDNMS", "smlalttDNMS", }, + { shift = 5, mask = 3, + [0] = "smulbbNMS", "smultbNMS", "smulbtNMS", "smulttNMS", }, +} + +local map_misc = { + shift = 4, mask = 7, + -- NYI: decode PSR bits of msr. + [0] = { shift = 21, mask = 1, [0] = "mrsD", "msrM", }, + { shift = 21, mask = 3, "bxM", false, "clzDM", }, + { shift = 21, mask = 3, "bxjM", }, + { shift = 21, mask = 3, "blxM", }, + false, + { shift = 21, mask = 3, [0] = "qaddDMN", "qsubDMN", "qdaddDMN", "qdsubDMN", }, + false, + { shift = 21, mask = 3, "bkptK", }, +} + +local map_datar = { + shift = 4, mask = 9, + [9] = { + shift = 5, mask = 3, + [0] = { shift = 24, mask = 1, [0] = map_mul, map_sync, }, + { shift = 20, mask = 1, [0] = "strhDL", "ldrhDL", }, + { shift = 20, mask = 1, [0] = "ldrdDL", "ldrsbDL", }, + { shift = 20, mask = 1, [0] = "strdDL", "ldrshDL", }, + }, + _ = { + shift = 20, mask = 25, + [16] = { shift = 7, mask = 1, [0] = map_misc, map_mulh, }, + _ = { + shift = 0, mask = 0xffffffff, + [bor(0xe1a00000)] = "nop", + _ = map_data, + } + }, +} + +local map_datai = { + shift = 20, mask = 31, -- NYI: decode PSR bits of msr. Decode imm12. + [16] = "movwDW", [20] = "movtDW", + [18] = { shift = 0, mask = 0xf00ff, [0] = "nopv6", _ = "msrNW", }, + [22] = "msrNW", + _ = map_data, +} + +local map_branch = { + shift = 24, mask = 1, + [0] = "bB", "blB" +} + +local map_condins = { + [0] = map_datar, map_datai, map_load, map_load1, + map_loadm, map_branch, map_loadc, map_datac +} + +-- NYI: setend. +local map_uncondins = { + [0] = false, map_simddata, map_simdload, map_preload, + false, "blxB", map_loadcu, map_datacu, +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "sp", "lr", "pc", +} + +local map_cond = { + [0] = "eq", "ne", "hs", "lo", "mi", "pl", "vs", "vc", + "hi", "ls", "ge", "lt", "gt", "le", "al", +} + +local map_shift = { [0] = "lsl", "lsr", "asr", "ror", } + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then + extra = "\t->"..sym + elseif band(ctx.op, 0x0e000000) ~= 0x0a000000 then + extra = "\t; 0x"..tohex(ctx.rel) + end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-5s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-5s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +-- Format operand 2 of load/store opcodes. +local function fmtload(ctx, op, pos) + local base = map_gpr[band(rshift(op, 16), 15)] + local x, ofs + local ext = (band(op, 0x04000000) == 0) + if not ext and band(op, 0x02000000) == 0 then + ofs = band(op, 4095) + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + ofs = "#"..ofs + elseif ext and band(op, 0x00400000) ~= 0 then + ofs = band(op, 15) + band(rshift(op, 4), 0xf0) + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + ofs = "#"..ofs + else + ofs = map_gpr[band(op, 15)] + if ext or band(op, 0xfe0) == 0 then + elseif band(op, 0xfe0) == 0x60 then + ofs = format("%s, rrx", ofs) + else + local sh = band(rshift(op, 7), 31) + if sh == 0 then sh = 32 end + ofs = format("%s, %s #%d", ofs, map_shift[band(rshift(op, 5), 3)], sh) + end + if band(op, 0x00800000) == 0 then ofs = "-"..ofs end + end + if ofs == "#0" then + x = format("[%s]", base) + elseif band(op, 0x01000000) == 0 then + x = format("[%s], %s", base, ofs) + else + x = format("[%s, %s]", base, ofs) + end + if band(op, 0x01200000) == 0x01200000 then x = x.."!" end + return x +end + +-- Format operand 2 of vector load/store opcodes. +local function fmtvload(ctx, op, pos) + local base = map_gpr[band(rshift(op, 16), 15)] + local ofs = band(op, 255)*4 + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + if ofs == 0 then + return format("[%s]", base) + else + return format("[%s, #%d]", base, ofs) + end +end + +local function fmtvr(op, vr, sh0, sh1) + if vr == "s" then + return format("s%d", 2*band(rshift(op, sh0), 15)+band(rshift(op, sh1), 1)) + else + return format("d%d", band(rshift(op, sh0), 15)+band(rshift(op, sh1-4), 16)) + end +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) + local operands = {} + local suffix = "" + local last, name, pat + local vr + ctx.op = op + ctx.rel = nil + + local cond = rshift(op, 28) + local opat + if cond == 15 then + opat = map_uncondins[band(rshift(op, 25), 7)] + else + if cond ~= 14 then suffix = map_cond[cond] end + opat = map_condins[band(rshift(op, 25), 7)] + end + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + name, pat = match(opat, "^([a-z0-9]*)(.*)") + if sub(pat, 1, 1) == "." then + local s2, p2 = match(pat, "^([a-z0-9.]*)(.*)") + suffix = suffix..s2 + pat = p2 + end + + for p in gmatch(pat, ".") do + local x = nil + if p == "D" then + x = map_gpr[band(rshift(op, 12), 15)] + elseif p == "N" then + x = map_gpr[band(rshift(op, 16), 15)] + elseif p == "S" then + x = map_gpr[band(rshift(op, 8), 15)] + elseif p == "M" then + x = map_gpr[band(op, 15)] + elseif p == "d" then + x = fmtvr(op, vr, 12, 22) + elseif p == "n" then + x = fmtvr(op, vr, 16, 7) + elseif p == "m" then + x = fmtvr(op, vr, 0, 5) + elseif p == "P" then + if band(op, 0x02000000) ~= 0 then + x = ror(band(op, 255), 2*band(rshift(op, 8), 15)) + else + x = map_gpr[band(op, 15)] + if band(op, 0xff0) ~= 0 then + operands[#operands+1] = x + local s = map_shift[band(rshift(op, 5), 3)] + local r = nil + if band(op, 0xf90) == 0 then + if s == "ror" then s = "rrx" else r = "#32" end + elseif band(op, 0x10) == 0 then + r = "#"..band(rshift(op, 7), 31) + else + r = map_gpr[band(rshift(op, 8), 15)] + end + if name == "mov" then name = s; x = r + elseif r then x = format("%s %s", s, r) + else x = s end + end + end + elseif p == "L" then + x = fmtload(ctx, op, pos) + elseif p == "l" then + x = fmtvload(ctx, op, pos) + elseif p == "B" then + local addr = ctx.addr + pos + 8 + arshift(lshift(op, 8), 6) + if cond == 15 then addr = addr + band(rshift(op, 23), 2) end + ctx.rel = addr + x = "0x"..tohex(addr) + elseif p == "F" then + vr = "s" + elseif p == "G" then + vr = "d" + elseif p == "." then + suffix = suffix..(vr == "s" and ".f32" or ".f64") + elseif p == "R" then + if band(op, 0x00200000) ~= 0 and #operands == 1 then + operands[1] = operands[1].."!" + end + local t = {} + for i=0,15 do + if band(rshift(op, i), 1) == 1 then t[#t+1] = map_gpr[i] end + end + x = "{"..concat(t, ", ").."}" + elseif p == "r" then + if band(op, 0x00200000) ~= 0 and #operands == 2 then + operands[1] = operands[1].."!" + end + local s = tonumber(sub(last, 2)) + local n = band(op, 255) + if vr == "d" then n = rshift(n, 1) end + operands[#operands] = format("{%s-%s%d}", last, vr, s+n-1) + elseif p == "W" then + x = band(op, 0x0fff) + band(rshift(op, 4), 0xf000) + elseif p == "T" then + x = "#0x"..tohex(band(op, 0x00ffffff), 6) + elseif p == "U" then + x = band(rshift(op, 7), 31) + if x == 0 then x = nil end + elseif p == "u" then + x = band(rshift(op, 7), 31) + if band(op, 0x40) == 0 then + if x == 0 then x = nil else x = "lsl #"..x end + else + if x == 0 then x = "asr #32" else x = "asr #"..x end + end + elseif p == "v" then + x = band(rshift(op, 7), 31) + elseif p == "w" then + x = band(rshift(op, 16), 31) + elseif p == "x" then + x = band(rshift(op, 16), 31) + 1 + elseif p == "X" then + x = band(rshift(op, 16), 31) - last + 1 + elseif p == "Y" then + x = band(rshift(op, 12), 0xf0) + band(op, 0x0f) + elseif p == "K" then + x = "#0x"..tohex(band(rshift(op, 4), 0x0000fff0) + band(op, 15), 4) + elseif p == "s" then + if band(op, 0x00100000) ~= 0 then suffix = "s"..suffix end + else + assert(false) + end + if x then + last = x + if type(x) == "number" then x = "#"..x end + operands[#operands+1] = x + end + end + + return putop(ctx, name..suffix, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ctx.pos = ofs + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 16 then return map_gpr[r] end + return "d"..(r-16) +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm64.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm64.lua new file mode 100644 index 00000000..ad909fbd --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm64.lua @@ -0,0 +1,1216 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM64 disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +-- +-- Contributed by Djordje Kovacevic and Stefan Pejic from RT-RK.com. +-- Sponsored by Cisco Systems, Inc. +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles most user-mode AArch64 instructions. +-- NYI: Advanced SIMD and VFP instructions. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local concat = table.concat +local bit = require("bit") +local band, bor, bxor, tohex = bit.band, bit.bor, bit.bxor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift +local ror = bit.ror + +------------------------------------------------------------------------------ +-- Opcode maps +------------------------------------------------------------------------------ + +local map_adr = { -- PC-relative addressing. + shift = 31, mask = 1, + [0] = "adrDBx", "adrpDBx" +} + +local map_addsubi = { -- Add/subtract immediate. + shift = 29, mask = 3, + [0] = "add|movDNIg", "adds|cmnD0NIg", "subDNIg", "subs|cmpD0NIg", +} + +local map_logi = { -- Logical immediate. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "andDNig", "orr|movDN0ig", "eorDNig", "ands|tstD0Nig" + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = "andDNig", "orr|movDN0ig", "eorDNig", "ands|tstD0Nig" + } +} + +local map_movwi = { -- Move wide immediate. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "movnDWRg", false, "movz|movDYRg", "movkDWRg" + }, false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = "movnDWRg", false, "movz|movDYRg", "movkDWRg" + }, +} + +local map_bitf = { -- Bitfield. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "sbfm|sbfiz|sbfx|asr|sxtw|sxth|sxtbDN12w", + "bfm|bfi|bfxilDN13w", + "ubfm|ubfiz|ubfx|lsr|lsl|uxth|uxtbDN12w" + } + }, + { + shift = 22, mask = 1, + { + shift = 29, mask = 3, + [0] = "sbfm|sbfiz|sbfx|asr|sxtw|sxth|sxtbDN12x", + "bfm|bfi|bfxilDN13x", + "ubfm|ubfiz|ubfx|lsr|lsl|uxth|uxtbDN12x" + } + } +} + +local map_datai = { -- Data processing - immediate. + shift = 23, mask = 7, + [0] = map_adr, map_adr, map_addsubi, false, + map_logi, map_movwi, map_bitf, + { + shift = 15, mask = 0x1c0c1, + [0] = "extr|rorDNM4w", [0x10080] = "extr|rorDNM4x", + [0x10081] = "extr|rorDNM4x" + } +} + +local map_logsr = { -- Logical, shifted register. + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = { + shift = 21, mask = 7, + [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg", + "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg" + }, + { + shift = 21, mask = 7, + [0] ="orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg", + "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg" + }, + { + shift = 21, mask = 7, + [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg", + "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg" + }, + { + shift = 21, mask = 7, + [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg", + "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg" + } + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = { + shift = 21, mask = 7, + [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg", + "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg" + }, + { + shift = 21, mask = 7, + [0] = "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg", + "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg" + }, + { + shift = 21, mask = 7, + [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg", + "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg" + }, + { + shift = 21, mask = 7, + [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg", + "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg" + } + } +} + +local map_assh = { + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "addDNMSg", "addDNMSg", "addDNMSg", "addDNMg" + }, + { + shift = 22, mask = 3, + [0] = "adds|cmnD0NMSg", "adds|cmnD0NMSg", + "adds|cmnD0NMSg", "adds|cmnD0NMg" + }, + { + shift = 22, mask = 3, + [0] = "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0Mg" + }, + { + shift = 22, mask = 3, + [0] = "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0MzSg", + "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0Mzg" + }, + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "addDNMSg", "addDNMSg", "addDNMSg", "addDNMg" + }, + { + shift = 22, mask = 3, + [0] = "adds|cmnD0NMSg", "adds|cmnD0NMSg", "adds|cmnD0NMSg", + "adds|cmnD0NMg" + }, + { + shift = 22, mask = 3, + [0] = "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0Mg" + }, + { + shift = 22, mask = 3, + [0] = "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0MzSg", + "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0Mzg" + } + } +} + +local map_addsubsh = { -- Add/subtract, shifted register. + shift = 22, mask = 3, + [0] = map_assh, map_assh, map_assh +} + +local map_addsubex = { -- Add/subtract, extended register. + shift = 22, mask = 3, + [0] = { + shift = 29, mask = 3, + [0] = "addDNMXg", "adds|cmnD0NMXg", "subDNMXg", "subs|cmpD0NMzXg", + } +} + +local map_addsubc = { -- Add/subtract, with carry. + shift = 10, mask = 63, + [0] = { + shift = 29, mask = 3, + [0] = "adcDNMg", "adcsDNMg", "sbc|ngcDN0Mg", "sbcs|ngcsDN0Mg", + } +} + +local map_ccomp = { + shift = 4, mask = 1, + [0] = { + shift = 10, mask = 3, + [0] = { -- Conditional compare register. + shift = 29, mask = 3, + "ccmnNMVCg", false, "ccmpNMVCg", + }, + [2] = { -- Conditional compare immediate. + shift = 29, mask = 3, + "ccmnN5VCg", false, "ccmpN5VCg", + } + } +} + +local map_csel = { -- Conditional select. + shift = 11, mask = 1, + [0] = { + shift = 10, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "cselDNMzCg", false, "csinv|cinv|csetmDNMcg", false, + }, + { + shift = 29, mask = 3, + [0] = "csinc|cinc|csetDNMcg", false, "csneg|cnegDNMcg", false, + } + } +} + +local map_data1s = { -- Data processing, 1 source. + shift = 29, mask = 1, + [0] = { + shift = 31, mask = 1, + [0] = { + shift = 10, mask = 0x7ff, + [0] = "rbitDNg", "rev16DNg", "revDNw", false, "clzDNg", "clsDNg" + }, + { + shift = 10, mask = 0x7ff, + [0] = "rbitDNg", "rev16DNg", "rev32DNx", "revDNx", "clzDNg", "clsDNg" + } + } +} + +local map_data2s = { -- Data processing, 2 sources. + shift = 29, mask = 1, + [0] = { + shift = 10, mask = 63, + false, "udivDNMg", "sdivDNMg", false, false, false, false, "lslDNMg", + "lsrDNMg", "asrDNMg", "rorDNMg" + } +} + +local map_data3s = { -- Data processing, 3 sources. + shift = 29, mask = 7, + [0] = { + shift = 21, mask = 7, + [0] = { + shift = 15, mask = 1, + [0] = "madd|mulDNMA0g", "msub|mnegDNMA0g" + } + }, false, false, false, + { + shift = 15, mask = 1, + [0] = { + shift = 21, mask = 7, + [0] = "madd|mulDNMA0g", "smaddl|smullDxNMwA0x", "smulhDNMx", false, + false, "umaddl|umullDxNMwA0x", "umulhDNMx" + }, + { + shift = 21, mask = 7, + [0] = "msub|mnegDNMA0g", "smsubl|smneglDxNMwA0x", false, false, + false, "umsubl|umneglDxNMwA0x" + } + } +} + +local map_datar = { -- Data processing, register. + shift = 28, mask = 1, + [0] = { + shift = 24, mask = 1, + [0] = map_logsr, + { + shift = 21, mask = 1, + [0] = map_addsubsh, map_addsubex + } + }, + { + shift = 21, mask = 15, + [0] = map_addsubc, false, map_ccomp, false, map_csel, false, + { + shift = 30, mask = 1, + [0] = map_data2s, map_data1s + }, + false, map_data3s, map_data3s, map_data3s, map_data3s, map_data3s, + map_data3s, map_data3s, map_data3s + } +} + +local map_lrl = { -- Load register, literal. + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = "ldrDwB", "ldrDxB", "ldrswDxB" + }, + { + shift = 30, mask = 3, + [0] = "ldrDsB", "ldrDdB" + } +} + +local map_lsriind = { -- Load/store register, immediate pre/post-indexed. + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwzL", "ldrbDwzL", "ldrsbDxzL", "ldrsbDwzL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strhDwzL", "ldrhDwzL", "ldrshDxzL", "ldrshDwzL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strDwzL", "ldrDwzL", "ldrswDxzL" + }, + { + shift = 22, mask = 3, + [0] = "strDszL", "ldrDszL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strDxzL", "ldrDxzL" + }, + { + shift = 22, mask = 3, + [0] = "strDdzL", "ldrDdzL" + } + } +} + +local map_lsriro = { + shift = 21, mask = 1, + [0] = { -- Load/store register immediate. + shift = 10, mask = 3, + [0] = { -- Unscaled immediate. + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "sturbDwK", "ldurbDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturhDwK", "ldurhDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturDwK", "ldurDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturDxK", "ldurDxK" + } + } + }, map_lsriind, false, map_lsriind + }, + { -- Load/store register, register offset. + shift = 10, mask = 3, + [2] = { + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwO", "ldrbDwO", "ldrsbDxO", "ldrsbDwO" + }, + { + shift = 22, mask = 3, + [0] = "strhDwO", "ldrhDwO", "ldrshDxO", "ldrshDwO" + }, + { + shift = 22, mask = 3, + [0] = "strDwO", "ldrDwO", "ldrswDxO" + }, + { + shift = 22, mask = 3, + [0] = "strDxO", "ldrDxO" + } + }, + { + shift = 30, mask = 3, + [2] = { + shift = 22, mask = 3, + [0] = "strDsO", "ldrDsO" + }, + [3] = { + shift = 22, mask = 3, + [0] = "strDdO", "ldrDdO" + } + } + } + } +} + +local map_lsp = { -- Load/store register pair, offset. + shift = 22, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = "stpDzAzwP", "stpDzAzsP", + }, + { + shift = 26, mask = 1, + "stpDzAzdP" + }, + { + shift = 26, mask = 1, + [0] = "stpDzAzxP" + } + }, + { + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = "ldpDzAzwP", "ldpDzAzsP", + }, + { + shift = 26, mask = 1, + [0] = "ldpswDAxP", "ldpDzAzdP" + }, + { + shift = 26, mask = 1, + [0] = "ldpDzAzxP" + } + } +} + +local map_ls = { -- Loads and stores. + shift = 24, mask = 0x31, + [0x10] = map_lrl, [0x30] = map_lsriro, + [0x20] = { + shift = 23, mask = 3, + map_lsp, map_lsp, map_lsp + }, + [0x21] = { + shift = 23, mask = 3, + map_lsp, map_lsp, map_lsp + }, + [0x31] = { + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwzU", "ldrbDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strhDwzU", "ldrhDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strDwzU", "ldrDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strDxzU", "ldrDxzU" + } + }, + { + shift = 30, mask = 3, + [2] = { + shift = 22, mask = 3, + [0] = "strDszU", "ldrDszU" + }, + [3] = { + shift = 22, mask = 3, + [0] = "strDdzU", "ldrDdzU" + } + } + }, +} + +local map_datafp = { -- Data processing, SIMD and FP. + shift = 28, mask = 7, + { -- 001 + shift = 24, mask = 1, + [0] = { + shift = 21, mask = 1, + { + shift = 10, mask = 3, + [0] = { + shift = 12, mask = 1, + [0] = { + shift = 13, mask = 1, + [0] = { + shift = 14, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { -- FP/int conversion. + shift = 31, mask = 1, + [0] = { + shift = 16, mask = 0xff, + [0x20] = "fcvtnsDwNs", [0x21] = "fcvtnuDwNs", + [0x22] = "scvtfDsNw", [0x23] = "ucvtfDsNw", + [0x24] = "fcvtasDwNs", [0x25] = "fcvtauDwNs", + [0x26] = "fmovDwNs", [0x27] = "fmovDsNw", + [0x28] = "fcvtpsDwNs", [0x29] = "fcvtpuDwNs", + [0x30] = "fcvtmsDwNs", [0x31] = "fcvtmuDwNs", + [0x38] = "fcvtzsDwNs", [0x39] = "fcvtzuDwNs", + [0x60] = "fcvtnsDwNd", [0x61] = "fcvtnuDwNd", + [0x62] = "scvtfDdNw", [0x63] = "ucvtfDdNw", + [0x64] = "fcvtasDwNd", [0x65] = "fcvtauDwNd", + [0x68] = "fcvtpsDwNd", [0x69] = "fcvtpuDwNd", + [0x70] = "fcvtmsDwNd", [0x71] = "fcvtmuDwNd", + [0x78] = "fcvtzsDwNd", [0x79] = "fcvtzuDwNd" + }, + { + shift = 16, mask = 0xff, + [0x20] = "fcvtnsDxNs", [0x21] = "fcvtnuDxNs", + [0x22] = "scvtfDsNx", [0x23] = "ucvtfDsNx", + [0x24] = "fcvtasDxNs", [0x25] = "fcvtauDxNs", + [0x28] = "fcvtpsDxNs", [0x29] = "fcvtpuDxNs", + [0x30] = "fcvtmsDxNs", [0x31] = "fcvtmuDxNs", + [0x38] = "fcvtzsDxNs", [0x39] = "fcvtzuDxNs", + [0x60] = "fcvtnsDxNd", [0x61] = "fcvtnuDxNd", + [0x62] = "scvtfDdNx", [0x63] = "ucvtfDdNx", + [0x64] = "fcvtasDxNd", [0x65] = "fcvtauDxNd", + [0x66] = "fmovDxNd", [0x67] = "fmovDdNx", + [0x68] = "fcvtpsDxNd", [0x69] = "fcvtpuDxNd", + [0x70] = "fcvtmsDxNd", [0x71] = "fcvtmuDxNd", + [0x78] = "fcvtzsDxNd", [0x79] = "fcvtzuDxNd" + } + } + }, + { -- FP data-processing, 1 source. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = { + shift = 15, mask = 63, + [0] = "fmovDNf", "fabsDNf", "fnegDNf", + "fsqrtDNf", false, "fcvtDdNs", false, false, + "frintnDNf", "frintpDNf", "frintmDNf", "frintzDNf", + "frintaDNf", false, "frintxDNf", "frintiDNf", + }, + { + shift = 15, mask = 63, + [0] = "fmovDNf", "fabsDNf", "fnegDNf", + "fsqrtDNf", "fcvtDsNd", false, false, false, + "frintnDNf", "frintpDNf", "frintmDNf", "frintzDNf", + "frintaDNf", false, "frintxDNf", "frintiDNf", + } + } + } + }, + { -- FP compare. + shift = 31, mask = 1, + [0] = { + shift = 14, mask = 3, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 0, mask = 31, + [0] = "fcmpNMf", [8] = "fcmpNZf", + [16] = "fcmpeNMf", [24] = "fcmpeNZf", + } + } + } + } + }, + { -- FP immediate. + shift = 31, mask = 1, + [0] = { + shift = 5, mask = 31, + [0] = { + shift = 23, mask = 1, + [0] = "fmovDFf" + } + } + } + }, + { -- FP conditional compare. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 4, mask = 1, + [0] = "fccmpNMVCf", "fccmpeNMVCf" + } + } + }, + { -- FP data-processing, 2 sources. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 12, mask = 15, + [0] = "fmulDNMf", "fdivDNMf", "faddDNMf", "fsubDNMf", + "fmaxDNMf", "fminDNMf", "fmaxnmDNMf", "fminnmDNMf", + "fnmulDNMf" + } + } + }, + { -- FP conditional select. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = "fcselDNMCf" + } + } + } + }, + { -- FP data-processing, 3 sources. + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 21, mask = 5, + [0] = "fmaddDNMAf", "fnmaddDNMAf" + }, + { + shift = 21, mask = 5, + [0] = "fmsubDNMAf", "fnmsubDNMAf" + } + } + } + } +} + +local map_br = { -- Branches, exception generating and system instructions. + shift = 29, mask = 7, + [0] = "bB", + { -- Compare & branch, immediate. + shift = 24, mask = 3, + [0] = "cbzDBg", "cbnzDBg", "tbzDTBw", "tbnzDTBw" + }, + { -- Conditional branch, immediate. + shift = 24, mask = 3, + [0] = { + shift = 4, mask = 1, + [0] = { + shift = 0, mask = 15, + [0] = "beqB", "bneB", "bhsB", "bloB", "bmiB", "bplB", "bvsB", "bvcB", + "bhiB", "blsB", "bgeB", "bltB", "bgtB", "bleB", "balB" + } + } + }, false, "blB", + { -- Compare & branch, immediate. + shift = 24, mask = 3, + [0] = "cbzDBg", "cbnzDBg", "tbzDTBx", "tbnzDTBx" + }, + { + shift = 24, mask = 3, + [0] = { -- Exception generation. + shift = 0, mask = 0xe0001f, + [0x200000] = "brkW" + }, + { -- System instructions. + shift = 0, mask = 0x3fffff, + [0x03201f] = "nop" + }, + { -- Unconditional branch, register. + shift = 0, mask = 0xfffc1f, + [0x1f0000] = "brNx", [0x3f0000] = "blrNx", + [0x5f0000] = "retNx" + }, + } +} + +local map_init = { + shift = 25, mask = 15, + [0] = false, false, false, false, map_ls, map_datar, map_ls, map_datafp, + map_datai, map_datai, map_br, map_br, map_ls, map_datar, map_ls, map_datafp +} + +------------------------------------------------------------------------------ + +local map_regs = { x = {}, w = {}, d = {}, s = {} } + +for i=0,30 do + map_regs.x[i] = "x"..i + map_regs.w[i] = "w"..i + map_regs.d[i] = "d"..i + map_regs.s[i] = "s"..i +end +map_regs.x[31] = "sp" +map_regs.w[31] = "wsp" +map_regs.d[31] = "d31" +map_regs.s[31] = "s31" + +local map_cond = { + [0] = "eq", "ne", "cs", "cc", "mi", "pl", "vs", "vc", + "hi", "ls", "ge", "lt", "gt", "le", "al", +} + +local map_shift = { [0] = "lsl", "lsr", "asr", } + +local map_extend = { + [0] = "uxtb", "uxth", "uxtw", "uxtx", "sxtb", "sxth", "sxtw", "sxtx", +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then + extra = "\t->"..sym + end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-5s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-5s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +local function match_reg(p, pat, regnum) + return map_regs[match(pat, p.."%w-([xwds])")][regnum] +end + +local function fmt_hex32(x) + if x < 0 then + return tohex(x) + else + return format("%x", x) + end +end + +local imm13_rep = { 0x55555555, 0x11111111, 0x01010101, 0x00010001, 0x00000001 } + +local function decode_imm13(op) + local imms = band(rshift(op, 10), 63) + local immr = band(rshift(op, 16), 63) + if band(op, 0x00400000) == 0 then + local len = 5 + if imms >= 56 then + if imms >= 60 then len = 1 else len = 2 end + elseif imms >= 48 then len = 3 elseif imms >= 32 then len = 4 end + local l = lshift(1, len)-1 + local s = band(imms, l) + local r = band(immr, l) + local imm = ror(rshift(-1, 31-s), r) + if len ~= 5 then imm = band(imm, lshift(1, l)-1) + rshift(imm, 31-l) end + imm = imm * imm13_rep[len] + local ix = fmt_hex32(imm) + if rshift(op, 31) ~= 0 then + return ix..tohex(imm) + else + return ix + end + else + local lo, hi = -1, 0 + if imms < 32 then lo = rshift(-1, 31-imms) else hi = rshift(-1, 63-imms) end + if immr ~= 0 then + lo, hi = ror(lo, immr), ror(hi, immr) + local x = immr == 32 and 0 or band(bxor(lo, hi), lshift(-1, 32-immr)) + lo, hi = bxor(lo, x), bxor(hi, x) + if immr >= 32 then lo, hi = hi, lo end + end + if hi ~= 0 then + return fmt_hex32(hi)..tohex(lo) + else + return fmt_hex32(lo) + end + end +end + +local function parse_immpc(op, name) + if name == "b" or name == "bl" then + return arshift(lshift(op, 6), 4) + elseif name == "adr" or name == "adrp" then + local immlo = band(rshift(op, 29), 3) + local immhi = lshift(arshift(lshift(op, 8), 13), 2) + return bor(immhi, immlo) + elseif name == "tbz" or name == "tbnz" then + return lshift(arshift(lshift(op, 13), 18), 2) + else + return lshift(arshift(lshift(op, 8), 13), 2) + end +end + +local function parse_fpimm8(op) + local sign = band(op, 0x100000) == 0 and 1 or -1 + local exp = bxor(rshift(arshift(lshift(op, 12), 5), 24), 0x80) - 131 + local frac = 16+band(rshift(op, 13), 15) + return sign * frac * 2^exp +end + +local function prefer_bfx(sf, uns, imms, immr) + if imms < immr or imms == 31 or imms == 63 then + return false + end + if immr == 0 then + if sf == 0 and (imms == 7 or imms == 15) then + return false + end + if sf ~= 0 and uns == 0 and (imms == 7 or imms == 15 or imms == 31) then + return false + end + end + return true +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) + local operands = {} + local suffix = "" + local last, name, pat + local map_reg + ctx.op = op + ctx.rel = nil + last = nil + local opat + opat = map_init[band(rshift(op, 25), 15)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + name, pat = match(opat, "^([a-z0-9]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.|]*)(.*)") + if altname then pat = pat2 end + if sub(pat, 1, 1) == "." then + local s2, p2 = match(pat, "^([a-z0-9.]*)(.*)") + suffix = suffix..s2 + pat = p2 + end + + local rt = match(pat, "[gf]") + if rt then + if rt == "g" then + map_reg = band(op, 0x80000000) ~= 0 and map_regs.x or map_regs.w + else + map_reg = band(op, 0x400000) ~= 0 and map_regs.d or map_regs.s + end + end + + local second0, immr + + for p in gmatch(pat, ".") do + local x = nil + if p == "D" then + local regnum = band(op, 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "N" then + local regnum = band(rshift(op, 5), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "M" then + local regnum = band(rshift(op, 16), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "A" then + local regnum = band(rshift(op, 10), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "B" then + local addr = ctx.addr + pos + parse_immpc(op, name) + ctx.rel = addr + x = "0x"..tohex(addr) + elseif p == "T" then + x = bor(band(rshift(op, 26), 32), band(rshift(op, 19), 31)) + elseif p == "V" then + x = band(op, 15) + elseif p == "C" then + x = map_cond[band(rshift(op, 12), 15)] + elseif p == "c" then + local rn = band(rshift(op, 5), 31) + local rm = band(rshift(op, 16), 31) + local cond = band(rshift(op, 12), 15) + local invc = bxor(cond, 1) + x = map_cond[cond] + if altname and cond ~= 14 and cond ~= 15 then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if rn == rm then + local n = #operands + operands[n] = nil + x = map_cond[invc] + if rn ~= 31 then + if a1 then name = a1 else name = altname end + else + operands[n-1] = nil + name = a2 + end + end + end + elseif p == "W" then + x = band(rshift(op, 5), 0xffff) + elseif p == "Y" then + x = band(rshift(op, 5), 0xffff) + local hw = band(rshift(op, 21), 3) + if altname and (hw == 0 or x ~= 0) then + name = altname + end + elseif p == "L" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local imm9 = arshift(lshift(op, 11), 23) + if band(op, 0x800) ~= 0 then + x = "["..rn..", #"..imm9.."]!" + else + x = "["..rn.."], #"..imm9 + end + elseif p == "U" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local sz = band(rshift(op, 30), 3) + local imm12 = lshift(arshift(lshift(op, 10), 20), sz) + if imm12 ~= 0 then + x = "["..rn..", #"..imm12.."]" + else + x = "["..rn.."]" + end + elseif p == "K" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local imm9 = arshift(lshift(op, 11), 23) + if imm9 ~= 0 then + x = "["..rn..", #"..imm9.."]" + else + x = "["..rn.."]" + end + elseif p == "O" then + local rn, rm = map_regs.x[band(rshift(op, 5), 31)] + local m = band(rshift(op, 13), 1) + if m == 0 then + rm = map_regs.w[band(rshift(op, 16), 31)] + else + rm = map_regs.x[band(rshift(op, 16), 31)] + end + x = "["..rn..", "..rm + local opt = band(rshift(op, 13), 7) + local s = band(rshift(op, 12), 1) + local sz = band(rshift(op, 30), 3) + -- extension to be applied + if opt == 3 then + if s == 0 then x = x.."]" + else x = x..", lsl #"..sz.."]" end + elseif opt == 2 or opt == 6 or opt == 7 then + if s == 0 then x = x..", "..map_extend[opt].."]" + else x = x..", "..map_extend[opt].." #"..sz.."]" end + else + x = x.."]" + end + elseif p == "P" then + local opcv, sh = rshift(op, 26), 2 + if opcv >= 0x2a then sh = 4 elseif opcv >= 0x1b then sh = 3 end + local imm7 = lshift(arshift(lshift(op, 10), 25), sh) + local rn = map_regs.x[band(rshift(op, 5), 31)] + local ind = band(rshift(op, 23), 3) + if ind == 1 then + x = "["..rn.."], #"..imm7 + elseif ind == 2 then + if imm7 == 0 then + x = "["..rn.."]" + else + x = "["..rn..", #"..imm7.."]" + end + elseif ind == 3 then + x = "["..rn..", #"..imm7.."]!" + end + elseif p == "I" then + local shf = band(rshift(op, 22), 3) + local imm12 = band(rshift(op, 10), 0x0fff) + local rn, rd = band(rshift(op, 5), 31), band(op, 31) + if altname == "mov" and shf == 0 and imm12 == 0 and (rn == 31 or rd == 31) then + name = altname + x = nil + elseif shf == 0 then + x = imm12 + elseif shf == 1 then + x = imm12..", lsl #12" + end + elseif p == "i" then + x = "#0x"..decode_imm13(op) + elseif p == "1" then + immr = band(rshift(op, 16), 63) + x = immr + elseif p == "2" then + x = band(rshift(op, 10), 63) + if altname then + local a1, a2, a3, a4, a5, a6 = + match(altname, "([^|]*)|([^|]*)|([^|]*)|([^|]*)|([^|]*)|(.*)") + local sf = band(rshift(op, 26), 32) + local uns = band(rshift(op, 30), 1) + if prefer_bfx(sf, uns, x, immr) then + name = a2 + x = x - immr + 1 + elseif immr == 0 and x == 7 then + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + name = a6 + x = nil + elseif immr == 0 and x == 15 then + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + name = a5 + x = nil + elseif x == 31 or x == 63 then + if x == 31 and immr == 0 and name == "sbfm" then + name = a4 + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + else + name = a3 + end + x = nil + elseif band(x, 31) ~= 31 and immr == x+1 and name == "ubfm" then + name = a4 + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = nil + elseif x < immr then + name = a1 + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = x + 1 + end + end + elseif p == "3" then + x = band(rshift(op, 10), 63) + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if x < immr then + name = a1 + local sf = band(rshift(op, 26), 32) + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = x + 1 + else + name = a2 + x = x - immr + 1 + end + end + elseif p == "4" then + x = band(rshift(op, 10), 63) + local rn = band(rshift(op, 5), 31) + local rm = band(rshift(op, 16), 31) + if altname and rn == rm then + local n = #operands + operands[n] = nil + last = operands[n-1] + name = altname + end + elseif p == "5" then + x = band(rshift(op, 16), 31) + elseif p == "S" then + x = band(rshift(op, 10), 63) + if x == 0 then x = nil + else x = map_shift[band(rshift(op, 22), 3)].." #"..x end + elseif p == "X" then + local opt = band(rshift(op, 13), 7) + -- Width specifier . + if opt ~= 3 and opt ~= 7 then + last = map_regs.w[band(rshift(op, 16), 31)] + operands[#operands] = last + end + x = band(rshift(op, 10), 7) + -- Extension. + if opt == 2 + band(rshift(op, 31), 1) and + band(rshift(op, second0 and 5 or 0), 31) == 31 then + if x == 0 then x = nil + else x = "lsl #"..x end + else + if x == 0 then x = map_extend[band(rshift(op, 13), 7)] + else x = map_extend[band(rshift(op, 13), 7)].." #"..x end + end + elseif p == "R" then + x = band(rshift(op,21), 3) + if x == 0 then x = nil + else x = "lsl #"..x*16 end + elseif p == "z" then + local n = #operands + if operands[n] == "sp" then operands[n] = "xzr" + elseif operands[n] == "wsp" then operands[n] = "wzr" + end + elseif p == "Z" then + x = 0 + elseif p == "F" then + x = parse_fpimm8(op) + elseif p == "g" or p == "f" or p == "x" or p == "w" or + p == "d" or p == "s" then + -- These are handled in D/N/M/A. + elseif p == "0" then + if last == "sp" or last == "wsp" then + local n = #operands + operands[n] = nil + last = operands[n-1] + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if not a1 then + name = altname + elseif second0 then + name, altname = a2, a1 + else + name, altname = a1, a2 + end + end + end + second0 = true + else + assert(false) + end + if x then + last = x + if type(x) == "number" then x = "#"..x end + operands[#operands+1] = x + end + end + + return putop(ctx, name..suffix, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ctx.pos = ofs + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_regs.x[r] end + return map_regs.d[r-32] +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm64be.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm64be.lua new file mode 100644 index 00000000..edcbffa8 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_arm64be.lua @@ -0,0 +1,12 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM64BE disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- ARM64 instructions are always little-endian. So just forward to the +-- common ARM64 disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +return require((string.match(..., ".*%.") or "").."dis_arm64") + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips.lua new file mode 100644 index 00000000..6ad17f54 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips.lua @@ -0,0 +1,694 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT/X license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles all standard MIPS32R1/R2 instructions. +-- Default mode is big-endian, but see: dis_mipsel.lua +------------------------------------------------------------------------------ + +local type = type +local byte, format = string.byte, string.format +local match, gmatch = string.match, string.gmatch +local concat = table.concat +local bit = require("bit") +local band, bor, tohex = bit.band, bit.bor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Extended opcode maps common to all MIPS releases +------------------------------------------------------------------------------ + +local map_srl = { shift = 21, mask = 1, [0] = "srlDTA", "rotrDTA", } +local map_srlv = { shift = 6, mask = 1, [0] = "srlvDTS", "rotrvDTS", } + +local map_cop0 = { + shift = 25, mask = 1, + [0] = { + shift = 21, mask = 15, + [0] = "mfc0TDW", [4] = "mtc0TDW", + [10] = "rdpgprDT", + [11] = { shift = 5, mask = 1, [0] = "diT0", "eiT0", }, + [14] = "wrpgprDT", + }, { + shift = 0, mask = 63, + [1] = "tlbr", [2] = "tlbwi", [6] = "tlbwr", [8] = "tlbp", + [24] = "eret", [31] = "deret", + [32] = "wait", + }, +} + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps for MIPS R1-R5 +------------------------------------------------------------------------------ + +local map_movci = { shift = 16, mask = 1, [0] = "movfDSC", "movtDSC", } + +local map_special = { + shift = 0, mask = 63, + [0] = { shift = 0, mask = -1, [0] = "nop", _ = "sllDTA" }, + map_movci, map_srl, "sraDTA", + "sllvDTS", false, map_srlv, "sravDTS", + "jrS", "jalrD1S", "movzDST", "movnDST", + "syscallY", "breakY", false, "sync", + "mfhiD", "mthiS", "mfloD", "mtloS", + "dsllvDST", false, "dsrlvDST", "dsravDST", + "multST", "multuST", "divST", "divuST", + "dmultST", "dmultuST", "ddivST", "ddivuST", + "addDST", "addu|moveDST0", "subDST", "subu|neguDS0T", + "andDST", "or|moveDST0", "xorDST", "nor|notDST0", + false, false, "sltDST", "sltuDST", + "daddDST", "dadduDST", "dsubDST", "dsubuDST", + "tgeSTZ", "tgeuSTZ", "tltSTZ", "tltuSTZ", + "teqSTZ", false, "tneSTZ", false, + "dsllDTA", false, "dsrlDTA", "dsraDTA", + "dsll32DTA", false, "dsrl32DTA", "dsra32DTA", +} + +local map_special2 = { + shift = 0, mask = 63, + [0] = "maddST", "madduST", "mulDST", false, + "msubST", "msubuST", + [32] = "clzDS", [33] = "cloDS", + [63] = "sdbbpY", +} + +local map_bshfl = { + shift = 6, mask = 31, + [2] = "wsbhDT", + [16] = "sebDT", + [24] = "sehDT", +} + +local map_dbshfl = { + shift = 6, mask = 31, + [2] = "dsbhDT", + [5] = "dshdDT", +} + +local map_special3 = { + shift = 0, mask = 63, + [0] = "extTSAK", [1] = "dextmTSAP", [3] = "dextTSAK", + [4] = "insTSAL", [6] = "dinsuTSEQ", [7] = "dinsTSAL", + [32] = map_bshfl, [36] = map_dbshfl, [59] = "rdhwrTD", +} + +local map_regimm = { + shift = 16, mask = 31, + [0] = "bltzSB", "bgezSB", "bltzlSB", "bgezlSB", + false, false, false, false, + "tgeiSI", "tgeiuSI", "tltiSI", "tltiuSI", + "teqiSI", false, "tneiSI", false, + "bltzalSB", "bgezalSB", "bltzallSB", "bgezallSB", + false, false, false, false, + false, false, false, false, + false, false, false, "synciSO", +} + +local map_cop1s = { + shift = 0, mask = 63, + [0] = "add.sFGH", "sub.sFGH", "mul.sFGH", "div.sFGH", + "sqrt.sFG", "abs.sFG", "mov.sFG", "neg.sFG", + "round.l.sFG", "trunc.l.sFG", "ceil.l.sFG", "floor.l.sFG", + "round.w.sFG", "trunc.w.sFG", "ceil.w.sFG", "floor.w.sFG", + false, + { shift = 16, mask = 1, [0] = "movf.sFGC", "movt.sFGC" }, + "movz.sFGT", "movn.sFGT", + false, "recip.sFG", "rsqrt.sFG", false, + false, false, false, false, + false, false, false, false, + false, "cvt.d.sFG", false, false, + "cvt.w.sFG", "cvt.l.sFG", "cvt.ps.sFGH", false, + false, false, false, false, + false, false, false, false, + "c.f.sVGH", "c.un.sVGH", "c.eq.sVGH", "c.ueq.sVGH", + "c.olt.sVGH", "c.ult.sVGH", "c.ole.sVGH", "c.ule.sVGH", + "c.sf.sVGH", "c.ngle.sVGH", "c.seq.sVGH", "c.ngl.sVGH", + "c.lt.sVGH", "c.nge.sVGH", "c.le.sVGH", "c.ngt.sVGH", +} + +local map_cop1d = { + shift = 0, mask = 63, + [0] = "add.dFGH", "sub.dFGH", "mul.dFGH", "div.dFGH", + "sqrt.dFG", "abs.dFG", "mov.dFG", "neg.dFG", + "round.l.dFG", "trunc.l.dFG", "ceil.l.dFG", "floor.l.dFG", + "round.w.dFG", "trunc.w.dFG", "ceil.w.dFG", "floor.w.dFG", + false, + { shift = 16, mask = 1, [0] = "movf.dFGC", "movt.dFGC" }, + "movz.dFGT", "movn.dFGT", + false, "recip.dFG", "rsqrt.dFG", false, + false, false, false, false, + false, false, false, false, + "cvt.s.dFG", false, false, false, + "cvt.w.dFG", "cvt.l.dFG", false, false, + false, false, false, false, + false, false, false, false, + "c.f.dVGH", "c.un.dVGH", "c.eq.dVGH", "c.ueq.dVGH", + "c.olt.dVGH", "c.ult.dVGH", "c.ole.dVGH", "c.ule.dVGH", + "c.df.dVGH", "c.ngle.dVGH", "c.deq.dVGH", "c.ngl.dVGH", + "c.lt.dVGH", "c.nge.dVGH", "c.le.dVGH", "c.ngt.dVGH", +} + +local map_cop1ps = { + shift = 0, mask = 63, + [0] = "add.psFGH", "sub.psFGH", "mul.psFGH", false, + false, "abs.psFG", "mov.psFG", "neg.psFG", + false, false, false, false, + false, false, false, false, + false, + { shift = 16, mask = 1, [0] = "movf.psFGC", "movt.psFGC" }, + "movz.psFGT", "movn.psFGT", + false, false, false, false, + false, false, false, false, + false, false, false, false, + "cvt.s.puFG", false, false, false, + false, false, false, false, + "cvt.s.plFG", false, false, false, + "pll.psFGH", "plu.psFGH", "pul.psFGH", "puu.psFGH", + "c.f.psVGH", "c.un.psVGH", "c.eq.psVGH", "c.ueq.psVGH", + "c.olt.psVGH", "c.ult.psVGH", "c.ole.psVGH", "c.ule.psVGH", + "c.psf.psVGH", "c.ngle.psVGH", "c.pseq.psVGH", "c.ngl.psVGH", + "c.lt.psVGH", "c.nge.psVGH", "c.le.psVGH", "c.ngt.psVGH", +} + +local map_cop1w = { + shift = 0, mask = 63, + [32] = "cvt.s.wFG", [33] = "cvt.d.wFG", +} + +local map_cop1l = { + shift = 0, mask = 63, + [32] = "cvt.s.lFG", [33] = "cvt.d.lFG", +} + +local map_cop1bc = { + shift = 16, mask = 3, + [0] = "bc1fCB", "bc1tCB", "bc1flCB", "bc1tlCB", +} + +local map_cop1 = { + shift = 21, mask = 31, + [0] = "mfc1TG", "dmfc1TG", "cfc1TG", "mfhc1TG", + "mtc1TG", "dmtc1TG", "ctc1TG", "mthc1TG", + map_cop1bc, false, false, false, + false, false, false, false, + map_cop1s, map_cop1d, false, false, + map_cop1w, map_cop1l, map_cop1ps, +} + +local map_cop1x = { + shift = 0, mask = 63, + [0] = "lwxc1FSX", "ldxc1FSX", false, false, + false, "luxc1FSX", false, false, + "swxc1FSX", "sdxc1FSX", false, false, + false, "suxc1FSX", false, "prefxMSX", + false, false, false, false, + false, false, false, false, + false, false, false, false, + false, false, "alnv.psFGHS", false, + "madd.sFRGH", "madd.dFRGH", false, false, + false, false, "madd.psFRGH", false, + "msub.sFRGH", "msub.dFRGH", false, false, + false, false, "msub.psFRGH", false, + "nmadd.sFRGH", "nmadd.dFRGH", false, false, + false, false, "nmadd.psFRGH", false, + "nmsub.sFRGH", "nmsub.dFRGH", false, false, + false, false, "nmsub.psFRGH", false, +} + +local map_pri = { + [0] = map_special, map_regimm, "jJ", "jalJ", + "beq|beqz|bST00B", "bne|bnezST0B", "blezSB", "bgtzSB", + "addiTSI", "addiu|liTS0I", "sltiTSI", "sltiuTSI", + "andiTSU", "ori|liTS0U", "xoriTSU", "luiTU", + map_cop0, map_cop1, false, map_cop1x, + "beql|beqzlST0B", "bnel|bnezlST0B", "blezlSB", "bgtzlSB", + "daddiTSI", "daddiuTSI", false, false, + map_special2, "jalxJ", false, map_special3, + "lbTSO", "lhTSO", "lwlTSO", "lwTSO", + "lbuTSO", "lhuTSO", "lwrTSO", false, + "sbTSO", "shTSO", "swlTSO", "swTSO", + false, false, "swrTSO", "cacheNSO", + "llTSO", "lwc1HSO", "lwc2TSO", "prefNSO", + false, "ldc1HSO", "ldc2TSO", "ldTSO", + "scTSO", "swc1HSO", "swc2TSO", false, + false, "sdc1HSO", "sdc2TSO", "sdTSO", +} + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps for MIPS R6 +------------------------------------------------------------------------------ + +local map_mul_r6 = { shift = 6, mask = 3, [2] = "mulDST", [3] = "muhDST" } +local map_mulu_r6 = { shift = 6, mask = 3, [2] = "muluDST", [3] = "muhuDST" } +local map_div_r6 = { shift = 6, mask = 3, [2] = "divDST", [3] = "modDST" } +local map_divu_r6 = { shift = 6, mask = 3, [2] = "divuDST", [3] = "moduDST" } +local map_dmul_r6 = { shift = 6, mask = 3, [2] = "dmulDST", [3] = "dmuhDST" } +local map_dmulu_r6 = { shift = 6, mask = 3, [2] = "dmuluDST", [3] = "dmuhuDST" } +local map_ddiv_r6 = { shift = 6, mask = 3, [2] = "ddivDST", [3] = "dmodDST" } +local map_ddivu_r6 = { shift = 6, mask = 3, [2] = "ddivuDST", [3] = "dmoduDST" } + +local map_special_r6 = { + shift = 0, mask = 63, + [0] = { shift = 0, mask = -1, [0] = "nop", _ = "sllDTA" }, + false, map_srl, "sraDTA", + "sllvDTS", false, map_srlv, "sravDTS", + "jrS", "jalrD1S", false, false, + "syscallY", "breakY", false, "sync", + "clzDS", "cloDS", "dclzDS", "dcloDS", + "dsllvDST", "dlsaDSTA", "dsrlvDST", "dsravDST", + map_mul_r6, map_mulu_r6, map_div_r6, map_divu_r6, + map_dmul_r6, map_dmulu_r6, map_ddiv_r6, map_ddivu_r6, + "addDST", "addu|moveDST0", "subDST", "subu|neguDS0T", + "andDST", "or|moveDST0", "xorDST", "nor|notDST0", + false, false, "sltDST", "sltuDST", + "daddDST", "dadduDST", "dsubDST", "dsubuDST", + "tgeSTZ", "tgeuSTZ", "tltSTZ", "tltuSTZ", + "teqSTZ", "seleqzDST", "tneSTZ", "selnezDST", + "dsllDTA", false, "dsrlDTA", "dsraDTA", + "dsll32DTA", false, "dsrl32DTA", "dsra32DTA", +} + +local map_bshfl_r6 = { + shift = 9, mask = 3, + [1] = "alignDSTa", + _ = { + shift = 6, mask = 31, + [0] = "bitswapDT", + [2] = "wsbhDT", + [16] = "sebDT", + [24] = "sehDT", + } +} + +local map_dbshfl_r6 = { + shift = 9, mask = 3, + [1] = "dalignDSTa", + _ = { + shift = 6, mask = 31, + [0] = "dbitswapDT", + [2] = "dsbhDT", + [5] = "dshdDT", + } +} + +local map_special3_r6 = { + shift = 0, mask = 63, + [0] = "extTSAK", [1] = "dextmTSAP", [3] = "dextTSAK", + [4] = "insTSAL", [6] = "dinsuTSEQ", [7] = "dinsTSAL", + [32] = map_bshfl_r6, [36] = map_dbshfl_r6, [59] = "rdhwrTD", +} + +local map_regimm_r6 = { + shift = 16, mask = 31, + [0] = "bltzSB", [1] = "bgezSB", + [6] = "dahiSI", [30] = "datiSI", + [23] = "sigrieI", [31] = "synciSO", +} + +local map_pcrel_r6 = { + shift = 19, mask = 3, + [0] = "addiupcS2", "lwpcS2", "lwupcS2", { + shift = 18, mask = 1, + [0] = "ldpcS3", { shift = 16, mask = 3, [2] = "auipcSI", [3] = "aluipcSI" } + } +} + +local map_cop1s_r6 = { + shift = 0, mask = 63, + [0] = "add.sFGH", "sub.sFGH", "mul.sFGH", "div.sFGH", + "sqrt.sFG", "abs.sFG", "mov.sFG", "neg.sFG", + "round.l.sFG", "trunc.l.sFG", "ceil.l.sFG", "floor.l.sFG", + "round.w.sFG", "trunc.w.sFG", "ceil.w.sFG", "floor.w.sFG", + "sel.sFGH", false, false, false, + "seleqz.sFGH", "recip.sFG", "rsqrt.sFG", "selnez.sFGH", + "maddf.sFGH", "msubf.sFGH", "rint.sFG", "class.sFG", + "min.sFGH", "mina.sFGH", "max.sFGH", "maxa.sFGH", + false, "cvt.d.sFG", false, false, + "cvt.w.sFG", "cvt.l.sFG", +} + +local map_cop1d_r6 = { + shift = 0, mask = 63, + [0] = "add.dFGH", "sub.dFGH", "mul.dFGH", "div.dFGH", + "sqrt.dFG", "abs.dFG", "mov.dFG", "neg.dFG", + "round.l.dFG", "trunc.l.dFG", "ceil.l.dFG", "floor.l.dFG", + "round.w.dFG", "trunc.w.dFG", "ceil.w.dFG", "floor.w.dFG", + "sel.dFGH", false, false, false, + "seleqz.dFGH", "recip.dFG", "rsqrt.dFG", "selnez.dFGH", + "maddf.dFGH", "msubf.dFGH", "rint.dFG", "class.dFG", + "min.dFGH", "mina.dFGH", "max.dFGH", "maxa.dFGH", + "cvt.s.dFG", false, false, false, + "cvt.w.dFG", "cvt.l.dFG", +} + +local map_cop1w_r6 = { + shift = 0, mask = 63, + [0] = "cmp.af.sFGH", "cmp.un.sFGH", "cmp.eq.sFGH", "cmp.ueq.sFGH", + "cmp.lt.sFGH", "cmp.ult.sFGH", "cmp.le.sFGH", "cmp.ule.sFGH", + "cmp.saf.sFGH", "cmp.sun.sFGH", "cmp.seq.sFGH", "cmp.sueq.sFGH", + "cmp.slt.sFGH", "cmp.sult.sFGH", "cmp.sle.sFGH", "cmp.sule.sFGH", + false, "cmp.or.sFGH", "cmp.une.sFGH", "cmp.ne.sFGH", + false, false, false, false, + false, "cmp.sor.sFGH", "cmp.sune.sFGH", "cmp.sne.sFGH", + false, false, false, false, + "cvt.s.wFG", "cvt.d.wFG", +} + +local map_cop1l_r6 = { + shift = 0, mask = 63, + [0] = "cmp.af.dFGH", "cmp.un.dFGH", "cmp.eq.dFGH", "cmp.ueq.dFGH", + "cmp.lt.dFGH", "cmp.ult.dFGH", "cmp.le.dFGH", "cmp.ule.dFGH", + "cmp.saf.dFGH", "cmp.sun.dFGH", "cmp.seq.dFGH", "cmp.sueq.dFGH", + "cmp.slt.dFGH", "cmp.sult.dFGH", "cmp.sle.dFGH", "cmp.sule.dFGH", + false, "cmp.or.dFGH", "cmp.une.dFGH", "cmp.ne.dFGH", + false, false, false, false, + false, "cmp.sor.dFGH", "cmp.sune.dFGH", "cmp.sne.dFGH", + false, false, false, false, + "cvt.s.lFG", "cvt.d.lFG", +} + +local map_cop1_r6 = { + shift = 21, mask = 31, + [0] = "mfc1TG", "dmfc1TG", "cfc1TG", "mfhc1TG", + "mtc1TG", "dmtc1TG", "ctc1TG", "mthc1TG", + false, "bc1eqzHB", false, false, + false, "bc1nezHB", false, false, + map_cop1s_r6, map_cop1d_r6, false, false, + map_cop1w_r6, map_cop1l_r6, +} + +local function maprs_popTS(rs, rt) + if rt == 0 then return 0 elseif rs == 0 then return 1 + elseif rs == rt then return 2 else return 3 end +end + +local map_pop06_r6 = { + maprs = maprs_popTS, [0] = "blezSB", "blezalcTB", "bgezalcTB", "bgeucSTB" +} +local map_pop07_r6 = { + maprs = maprs_popTS, [0] = "bgtzSB", "bgtzalcTB", "bltzalcTB", "bltucSTB" +} +local map_pop26_r6 = { + maprs = maprs_popTS, "blezcTB", "bgezcTB", "bgecSTB" +} +local map_pop27_r6 = { + maprs = maprs_popTS, "bgtzcTB", "bltzcTB", "bltcSTB" +} + +local function maprs_popS(rs, rt) + if rs == 0 then return 0 else return 1 end +end + +local map_pop66_r6 = { + maprs = maprs_popS, [0] = "jicTI", "beqzcSb" +} +local map_pop76_r6 = { + maprs = maprs_popS, [0] = "jialcTI", "bnezcSb" +} + +local function maprs_popST(rs, rt) + if rs >= rt then return 0 elseif rs == 0 then return 1 else return 2 end +end + +local map_pop10_r6 = { + maprs = maprs_popST, [0] = "bovcSTB", "beqzalcTB", "beqcSTB" +} +local map_pop30_r6 = { + maprs = maprs_popST, [0] = "bnvcSTB", "bnezalcTB", "bnecSTB" +} + +local map_pri_r6 = { + [0] = map_special_r6, map_regimm_r6, "jJ", "jalJ", + "beq|beqz|bST00B", "bne|bnezST0B", map_pop06_r6, map_pop07_r6, + map_pop10_r6, "addiu|liTS0I", "sltiTSI", "sltiuTSI", + "andiTSU", "ori|liTS0U", "xoriTSU", "aui|luiTS0U", + map_cop0, map_cop1_r6, false, false, + false, false, map_pop26_r6, map_pop27_r6, + map_pop30_r6, "daddiuTSI", false, false, + false, "dauiTSI", false, map_special3_r6, + "lbTSO", "lhTSO", false, "lwTSO", + "lbuTSO", "lhuTSO", false, false, + "sbTSO", "shTSO", false, "swTSO", + false, false, false, false, + false, "lwc1HSO", "bc#", false, + false, "ldc1HSO", map_pop66_r6, "ldTSO", + false, "swc1HSO", "balc#", map_pcrel_r6, + false, "sdc1HSO", map_pop76_r6, "sdTSO", +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15", + "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23", + "r24", "r25", "r26", "r27", "r28", "sp", "r30", "ra", +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then extra = "\t->"..sym end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-7s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-7s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +local function get_be(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + return bor(lshift(b0, 24), lshift(b1, 16), lshift(b2, 8), b3) +end + +local function get_le(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + return bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local op = ctx:get() + local operands = {} + local last = nil + ctx.op = op + ctx.rel = nil + + local opat = ctx.map_pri[rshift(op, 26)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + if opat.maprs then + opat = opat[opat.maprs(band(rshift(op,21),31), band(rshift(op,16),31))] + else + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + end + local name, pat = match(opat, "^([a-z0-9_.]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.|]*)(.*)") + if altname then pat = pat2 end + + for p in gmatch(pat, ".") do + local x = nil + if p == "S" then + x = map_gpr[band(rshift(op, 21), 31)] + elseif p == "T" then + x = map_gpr[band(rshift(op, 16), 31)] + elseif p == "D" then + x = map_gpr[band(rshift(op, 11), 31)] + elseif p == "F" then + x = "f"..band(rshift(op, 6), 31) + elseif p == "G" then + x = "f"..band(rshift(op, 11), 31) + elseif p == "H" then + x = "f"..band(rshift(op, 16), 31) + elseif p == "R" then + x = "f"..band(rshift(op, 21), 31) + elseif p == "A" then + x = band(rshift(op, 6), 31) + elseif p == "a" then + x = band(rshift(op, 6), 7) + elseif p == "E" then + x = band(rshift(op, 6), 31) + 32 + elseif p == "M" then + x = band(rshift(op, 11), 31) + elseif p == "N" then + x = band(rshift(op, 16), 31) + elseif p == "C" then + x = band(rshift(op, 18), 7) + if x == 0 then x = nil end + elseif p == "K" then + x = band(rshift(op, 11), 31) + 1 + elseif p == "P" then + x = band(rshift(op, 11), 31) + 33 + elseif p == "L" then + x = band(rshift(op, 11), 31) - last + 1 + elseif p == "Q" then + x = band(rshift(op, 11), 31) - last + 33 + elseif p == "I" then + x = arshift(lshift(op, 16), 16) + elseif p == "2" then + x = arshift(lshift(op, 13), 11) + elseif p == "3" then + x = arshift(lshift(op, 14), 11) + elseif p == "U" then + x = band(op, 0xffff) + elseif p == "O" then + local disp = arshift(lshift(op, 16), 16) + operands[#operands] = format("%d(%s)", disp, last) + elseif p == "X" then + local index = map_gpr[band(rshift(op, 16), 31)] + operands[#operands] = format("%s(%s)", index, last) + elseif p == "B" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 16), 14) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "b" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 11), 9) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "#" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 6), 4) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "J" then + local a = ctx.addr + ctx.pos + x = a - band(a, 0x0fffffff) + band(op, 0x03ffffff)*4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "V" then + x = band(rshift(op, 8), 7) + if x == 0 then x = nil end + elseif p == "W" then + x = band(op, 7) + if x == 0 then x = nil end + elseif p == "Y" then + x = band(rshift(op, 6), 0x000fffff) + if x == 0 then x = nil end + elseif p == "Z" then + x = band(rshift(op, 6), 1023) + if x == 0 then x = nil end + elseif p == "0" then + if last == "r0" or last == 0 then + local n = #operands + operands[n] = nil + last = operands[n-1] + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if a1 then name, altname = a1, a2 + else name = altname end + end + end + elseif p == "1" then + if last == "ra" then + operands[#operands] = nil + end + else + assert(false) + end + if x then operands[#operands+1] = x; last = x end + end + + return putop(ctx, name, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + stop = stop - stop % 4 + ctx.pos = ofs - ofs % 4 + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + ctx.get = get_be + ctx.map_pri = map_pri + return ctx +end + +local function create_el(code, addr, out) + local ctx = create(code, addr, out) + ctx.get = get_le + return ctx +end + +local function create_r6(code, addr, out) + local ctx = create(code, addr, out) + ctx.map_pri = map_pri_r6 + return ctx +end + +local function create_r6_el(code, addr, out) + local ctx = create(code, addr, out) + ctx.get = get_le + ctx.map_pri = map_pri_r6 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +local function disass_el(code, addr, out) + create_el(code, addr, out):disass() +end + +local function disass_r6(code, addr, out) + create_r6(code, addr, out):disass() +end + +local function disass_r6_el(code, addr, out) + create_r6_el(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_gpr[r] end + return "f"..(r-32) +end + +-- Public module functions. +return { + create = create, + create_el = create_el, + create_r6 = create_r6, + create_r6_el = create_r6_el, + disass = disass, + disass_el = disass_el, + disass_r6 = disass_r6, + disass_r6_el = disass_r6_el, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64.lua new file mode 100644 index 00000000..5ad48f8f --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the big-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create, + disass = dis_mips.disass, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64el.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64el.lua new file mode 100644 index 00000000..d50e3a18 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64el.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64EL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_el, + disass = dis_mips.disass_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64r6.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64r6.lua new file mode 100644 index 00000000..921b3cbe --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64r6.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64R6 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the r6 big-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_r6, + disass = dis_mips.disass_r6, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64r6el.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64r6el.lua new file mode 100644 index 00000000..aadef9f3 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mips64r6el.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64R6EL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the r6 little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_r6_el, + disass = dis_mips.disass_r6_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mipsel.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mipsel.lua new file mode 100644 index 00000000..52cebefb --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_mipsel.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPSEL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_el, + disass = dis_mips.disass_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_ppc.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_ppc.lua new file mode 100644 index 00000000..08d742f1 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_ppc.lua @@ -0,0 +1,591 @@ +---------------------------------------------------------------------------- +-- LuaJIT PPC disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT/X license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles all common, non-privileged 32/64 bit PowerPC instructions +-- plus the e500 SPE instructions and some Cell/Xenon extensions. +-- +-- NYI: VMX, VMX128 +------------------------------------------------------------------------------ + +local type = type +local byte, format = string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local concat = table.concat +local bit = require("bit") +local band, bor, tohex = bit.band, bit.bor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps +------------------------------------------------------------------------------ + +local map_crops = { + shift = 1, mask = 1023, + [0] = "mcrfXX", + [33] = "crnor|crnotCCC=", [129] = "crandcCCC", + [193] = "crxor|crclrCCC%", [225] = "crnandCCC", + [257] = "crandCCC", [289] = "creqv|crsetCCC%", + [417] = "crorcCCC", [449] = "cror|crmoveCCC=", + [16] = "b_lrKB", [528] = "b_ctrKB", + [150] = "isync", +} + +local map_rlwinm = setmetatable({ + shift = 0, mask = -1, +}, +{ __index = function(t, x) + local rot = band(rshift(x, 11), 31) + local mb = band(rshift(x, 6), 31) + local me = band(rshift(x, 1), 31) + if mb == 0 and me == 31-rot then + return "slwiRR~A." + elseif me == 31 and mb == 32-rot then + return "srwiRR~-A." + else + return "rlwinmRR~AAA." + end + end +}) + +local map_rld = { + shift = 2, mask = 7, + [0] = "rldiclRR~HM.", "rldicrRR~HM.", "rldicRR~HM.", "rldimiRR~HM.", + { + shift = 1, mask = 1, + [0] = "rldclRR~RM.", "rldcrRR~RM.", + }, +} + +local map_ext = setmetatable({ + shift = 1, mask = 1023, + + [0] = "cmp_YLRR", [32] = "cmpl_YLRR", + [4] = "twARR", [68] = "tdARR", + + [8] = "subfcRRR.", [40] = "subfRRR.", + [104] = "negRR.", [136] = "subfeRRR.", + [200] = "subfzeRR.", [232] = "subfmeRR.", + [520] = "subfcoRRR.", [552] = "subfoRRR.", + [616] = "negoRR.", [648] = "subfeoRRR.", + [712] = "subfzeoRR.", [744] = "subfmeoRR.", + + [9] = "mulhduRRR.", [73] = "mulhdRRR.", [233] = "mulldRRR.", + [457] = "divduRRR.", [489] = "divdRRR.", + [745] = "mulldoRRR.", + [969] = "divduoRRR.", [1001] = "divdoRRR.", + + [10] = "addcRRR.", [138] = "addeRRR.", + [202] = "addzeRR.", [234] = "addmeRR.", [266] = "addRRR.", + [522] = "addcoRRR.", [650] = "addeoRRR.", + [714] = "addzeoRR.", [746] = "addmeoRR.", [778] = "addoRRR.", + + [11] = "mulhwuRRR.", [75] = "mulhwRRR.", [235] = "mullwRRR.", + [459] = "divwuRRR.", [491] = "divwRRR.", + [747] = "mullwoRRR.", + [971] = "divwouRRR.", [1003] = "divwoRRR.", + + [15] = "iselltRRR", [47] = "iselgtRRR", [79] = "iseleqRRR", + + [144] = { shift = 20, mask = 1, [0] = "mtcrfRZ~", "mtocrfRZ~", }, + [19] = { shift = 20, mask = 1, [0] = "mfcrR", "mfocrfRZ", }, + [371] = { shift = 11, mask = 1023, [392] = "mftbR", [424] = "mftbuR", }, + [339] = { + shift = 11, mask = 1023, + [32] = "mferR", [256] = "mflrR", [288] = "mfctrR", [16] = "mfspefscrR", + }, + [467] = { + shift = 11, mask = 1023, + [32] = "mtxerR", [256] = "mtlrR", [288] = "mtctrR", [16] = "mtspefscrR", + }, + + [20] = "lwarxRR0R", [84] = "ldarxRR0R", + + [21] = "ldxRR0R", [53] = "lduxRRR", + [149] = "stdxRR0R", [181] = "stduxRRR", + [341] = "lwaxRR0R", [373] = "lwauxRRR", + + [23] = "lwzxRR0R", [55] = "lwzuxRRR", + [87] = "lbzxRR0R", [119] = "lbzuxRRR", + [151] = "stwxRR0R", [183] = "stwuxRRR", + [215] = "stbxRR0R", [247] = "stbuxRRR", + [279] = "lhzxRR0R", [311] = "lhzuxRRR", + [343] = "lhaxRR0R", [375] = "lhauxRRR", + [407] = "sthxRR0R", [439] = "sthuxRRR", + + [54] = "dcbst-R0R", [86] = "dcbf-R0R", + [150] = "stwcxRR0R.", [214] = "stdcxRR0R.", + [246] = "dcbtst-R0R", [278] = "dcbt-R0R", + [310] = "eciwxRR0R", [438] = "ecowxRR0R", + [470] = "dcbi-RR", + + [598] = { + shift = 21, mask = 3, + [0] = "sync", "lwsync", "ptesync", + }, + [758] = "dcba-RR", + [854] = "eieio", [982] = "icbi-R0R", [1014] = "dcbz-R0R", + + [26] = "cntlzwRR~", [58] = "cntlzdRR~", + [122] = "popcntbRR~", + [154] = "prtywRR~", [186] = "prtydRR~", + + [28] = "andRR~R.", [60] = "andcRR~R.", [124] = "nor|notRR~R=.", + [284] = "eqvRR~R.", [316] = "xorRR~R.", + [412] = "orcRR~R.", [444] = "or|mrRR~R=.", [476] = "nandRR~R.", + [508] = "cmpbRR~R", + + [512] = "mcrxrX", + + [532] = "ldbrxRR0R", [660] = "stdbrxRR0R", + + [533] = "lswxRR0R", [597] = "lswiRR0A", + [661] = "stswxRR0R", [725] = "stswiRR0A", + + [534] = "lwbrxRR0R", [662] = "stwbrxRR0R", + [790] = "lhbrxRR0R", [918] = "sthbrxRR0R", + + [535] = "lfsxFR0R", [567] = "lfsuxFRR", + [599] = "lfdxFR0R", [631] = "lfduxFRR", + [663] = "stfsxFR0R", [695] = "stfsuxFRR", + [727] = "stfdxFR0R", [759] = "stfduxFR0R", + [855] = "lfiwaxFR0R", + [983] = "stfiwxFR0R", + + [24] = "slwRR~R.", + + [27] = "sldRR~R.", [536] = "srwRR~R.", + [792] = "srawRR~R.", [824] = "srawiRR~A.", + + [794] = "sradRR~R.", [826] = "sradiRR~H.", [827] = "sradiRR~H.", + [922] = "extshRR~.", [954] = "extsbRR~.", [986] = "extswRR~.", + + [539] = "srdRR~R.", +}, +{ __index = function(t, x) + if band(x, 31) == 15 then return "iselRRRC" end + end +}) + +local map_ld = { + shift = 0, mask = 3, + [0] = "ldRRE", "lduRRE", "lwaRRE", +} + +local map_std = { + shift = 0, mask = 3, + [0] = "stdRRE", "stduRRE", +} + +local map_fps = { + shift = 5, mask = 1, + { + shift = 1, mask = 15, + [0] = false, false, "fdivsFFF.", false, + "fsubsFFF.", "faddsFFF.", "fsqrtsF-F.", false, + "fresF-F.", "fmulsFF-F.", "frsqrtesF-F.", false, + "fmsubsFFFF~.", "fmaddsFFFF~.", "fnmsubsFFFF~.", "fnmaddsFFFF~.", + } +} + +local map_fpd = { + shift = 5, mask = 1, + [0] = { + shift = 1, mask = 1023, + [0] = "fcmpuXFF", [32] = "fcmpoXFF", [64] = "mcrfsXX", + [38] = "mtfsb1A.", [70] = "mtfsb0A.", [134] = "mtfsfiA>>-A>", + [8] = "fcpsgnFFF.", [40] = "fnegF-F.", [72] = "fmrF-F.", + [136] = "fnabsF-F.", [264] = "fabsF-F.", + [12] = "frspF-F.", + [14] = "fctiwF-F.", [15] = "fctiwzF-F.", + [583] = "mffsF.", [711] = "mtfsfZF.", + [392] = "frinF-F.", [424] = "frizF-F.", + [456] = "fripF-F.", [488] = "frimF-F.", + [814] = "fctidF-F.", [815] = "fctidzF-F.", [846] = "fcfidF-F.", + }, + { + shift = 1, mask = 15, + [0] = false, false, "fdivFFF.", false, + "fsubFFF.", "faddFFF.", "fsqrtF-F.", "fselFFFF~.", + "freF-F.", "fmulFF-F.", "frsqrteF-F.", false, + "fmsubFFFF~.", "fmaddFFFF~.", "fnmsubFFFF~.", "fnmaddFFFF~.", + } +} + +local map_spe = { + shift = 0, mask = 2047, + + [512] = "evaddwRRR", [514] = "evaddiwRAR~", + [516] = "evsubwRRR~", [518] = "evsubiwRAR~", + [520] = "evabsRR", [521] = "evnegRR", + [522] = "evextsbRR", [523] = "evextshRR", [524] = "evrndwRR", + [525] = "evcntlzwRR", [526] = "evcntlswRR", + + [527] = "brincRRR", + + [529] = "evandRRR", [530] = "evandcRRR", [534] = "evxorRRR", + [535] = "evor|evmrRRR=", [536] = "evnor|evnotRRR=", + [537] = "eveqvRRR", [539] = "evorcRRR", [542] = "evnandRRR", + + [544] = "evsrwuRRR", [545] = "evsrwsRRR", + [546] = "evsrwiuRRA", [547] = "evsrwisRRA", + [548] = "evslwRRR", [550] = "evslwiRRA", + [552] = "evrlwRRR", [553] = "evsplatiRS", + [554] = "evrlwiRRA", [555] = "evsplatfiRS", + [556] = "evmergehiRRR", [557] = "evmergeloRRR", + [558] = "evmergehiloRRR", [559] = "evmergelohiRRR", + + [560] = "evcmpgtuYRR", [561] = "evcmpgtsYRR", + [562] = "evcmpltuYRR", [563] = "evcmpltsYRR", + [564] = "evcmpeqYRR", + + [632] = "evselRRR", [633] = "evselRRRW", + [634] = "evselRRRW", [635] = "evselRRRW", + [636] = "evselRRRW", [637] = "evselRRRW", + [638] = "evselRRRW", [639] = "evselRRRW", + + [640] = "evfsaddRRR", [641] = "evfssubRRR", + [644] = "evfsabsRR", [645] = "evfsnabsRR", [646] = "evfsnegRR", + [648] = "evfsmulRRR", [649] = "evfsdivRRR", + [652] = "evfscmpgtYRR", [653] = "evfscmpltYRR", [654] = "evfscmpeqYRR", + [656] = "evfscfuiR-R", [657] = "evfscfsiR-R", + [658] = "evfscfufR-R", [659] = "evfscfsfR-R", + [660] = "evfsctuiR-R", [661] = "evfsctsiR-R", + [662] = "evfsctufR-R", [663] = "evfsctsfR-R", + [664] = "evfsctuizR-R", [666] = "evfsctsizR-R", + [668] = "evfststgtYRR", [669] = "evfststltYRR", [670] = "evfststeqYRR", + + [704] = "efsaddRRR", [705] = "efssubRRR", + [708] = "efsabsRR", [709] = "efsnabsRR", [710] = "efsnegRR", + [712] = "efsmulRRR", [713] = "efsdivRRR", + [716] = "efscmpgtYRR", [717] = "efscmpltYRR", [718] = "efscmpeqYRR", + [719] = "efscfdR-R", + [720] = "efscfuiR-R", [721] = "efscfsiR-R", + [722] = "efscfufR-R", [723] = "efscfsfR-R", + [724] = "efsctuiR-R", [725] = "efsctsiR-R", + [726] = "efsctufR-R", [727] = "efsctsfR-R", + [728] = "efsctuizR-R", [730] = "efsctsizR-R", + [732] = "efststgtYRR", [733] = "efststltYRR", [734] = "efststeqYRR", + + [736] = "efdaddRRR", [737] = "efdsubRRR", + [738] = "efdcfuidR-R", [739] = "efdcfsidR-R", + [740] = "efdabsRR", [741] = "efdnabsRR", [742] = "efdnegRR", + [744] = "efdmulRRR", [745] = "efddivRRR", + [746] = "efdctuidzR-R", [747] = "efdctsidzR-R", + [748] = "efdcmpgtYRR", [749] = "efdcmpltYRR", [750] = "efdcmpeqYRR", + [751] = "efdcfsR-R", + [752] = "efdcfuiR-R", [753] = "efdcfsiR-R", + [754] = "efdcfufR-R", [755] = "efdcfsfR-R", + [756] = "efdctuiR-R", [757] = "efdctsiR-R", + [758] = "efdctufR-R", [759] = "efdctsfR-R", + [760] = "efdctuizR-R", [762] = "efdctsizR-R", + [764] = "efdtstgtYRR", [765] = "efdtstltYRR", [766] = "efdtsteqYRR", + + [768] = "evlddxRR0R", [769] = "evlddRR8", + [770] = "evldwxRR0R", [771] = "evldwRR8", + [772] = "evldhxRR0R", [773] = "evldhRR8", + [776] = "evlhhesplatxRR0R", [777] = "evlhhesplatRR2", + [780] = "evlhhousplatxRR0R", [781] = "evlhhousplatRR2", + [782] = "evlhhossplatxRR0R", [783] = "evlhhossplatRR2", + [784] = "evlwhexRR0R", [785] = "evlwheRR4", + [788] = "evlwhouxRR0R", [789] = "evlwhouRR4", + [790] = "evlwhosxRR0R", [791] = "evlwhosRR4", + [792] = "evlwwsplatxRR0R", [793] = "evlwwsplatRR4", + [796] = "evlwhsplatxRR0R", [797] = "evlwhsplatRR4", + + [800] = "evstddxRR0R", [801] = "evstddRR8", + [802] = "evstdwxRR0R", [803] = "evstdwRR8", + [804] = "evstdhxRR0R", [805] = "evstdhRR8", + [816] = "evstwhexRR0R", [817] = "evstwheRR4", + [820] = "evstwhoxRR0R", [821] = "evstwhoRR4", + [824] = "evstwwexRR0R", [825] = "evstwweRR4", + [828] = "evstwwoxRR0R", [829] = "evstwwoRR4", + + [1027] = "evmhessfRRR", [1031] = "evmhossfRRR", [1032] = "evmheumiRRR", + [1033] = "evmhesmiRRR", [1035] = "evmhesmfRRR", [1036] = "evmhoumiRRR", + [1037] = "evmhosmiRRR", [1039] = "evmhosmfRRR", [1059] = "evmhessfaRRR", + [1063] = "evmhossfaRRR", [1064] = "evmheumiaRRR", [1065] = "evmhesmiaRRR", + [1067] = "evmhesmfaRRR", [1068] = "evmhoumiaRRR", [1069] = "evmhosmiaRRR", + [1071] = "evmhosmfaRRR", [1095] = "evmwhssfRRR", [1096] = "evmwlumiRRR", + [1100] = "evmwhumiRRR", [1101] = "evmwhsmiRRR", [1103] = "evmwhsmfRRR", + [1107] = "evmwssfRRR", [1112] = "evmwumiRRR", [1113] = "evmwsmiRRR", + [1115] = "evmwsmfRRR", [1127] = "evmwhssfaRRR", [1128] = "evmwlumiaRRR", + [1132] = "evmwhumiaRRR", [1133] = "evmwhsmiaRRR", [1135] = "evmwhsmfaRRR", + [1139] = "evmwssfaRRR", [1144] = "evmwumiaRRR", [1145] = "evmwsmiaRRR", + [1147] = "evmwsmfaRRR", + + [1216] = "evaddusiaawRR", [1217] = "evaddssiaawRR", + [1218] = "evsubfusiaawRR", [1219] = "evsubfssiaawRR", + [1220] = "evmraRR", + [1222] = "evdivwsRRR", [1223] = "evdivwuRRR", + [1224] = "evaddumiaawRR", [1225] = "evaddsmiaawRR", + [1226] = "evsubfumiaawRR", [1227] = "evsubfsmiaawRR", + + [1280] = "evmheusiaawRRR", [1281] = "evmhessiaawRRR", + [1283] = "evmhessfaawRRR", [1284] = "evmhousiaawRRR", + [1285] = "evmhossiaawRRR", [1287] = "evmhossfaawRRR", + [1288] = "evmheumiaawRRR", [1289] = "evmhesmiaawRRR", + [1291] = "evmhesmfaawRRR", [1292] = "evmhoumiaawRRR", + [1293] = "evmhosmiaawRRR", [1295] = "evmhosmfaawRRR", + [1320] = "evmhegumiaaRRR", [1321] = "evmhegsmiaaRRR", + [1323] = "evmhegsmfaaRRR", [1324] = "evmhogumiaaRRR", + [1325] = "evmhogsmiaaRRR", [1327] = "evmhogsmfaaRRR", + [1344] = "evmwlusiaawRRR", [1345] = "evmwlssiaawRRR", + [1352] = "evmwlumiaawRRR", [1353] = "evmwlsmiaawRRR", + [1363] = "evmwssfaaRRR", [1368] = "evmwumiaaRRR", + [1369] = "evmwsmiaaRRR", [1371] = "evmwsmfaaRRR", + [1408] = "evmheusianwRRR", [1409] = "evmhessianwRRR", + [1411] = "evmhessfanwRRR", [1412] = "evmhousianwRRR", + [1413] = "evmhossianwRRR", [1415] = "evmhossfanwRRR", + [1416] = "evmheumianwRRR", [1417] = "evmhesmianwRRR", + [1419] = "evmhesmfanwRRR", [1420] = "evmhoumianwRRR", + [1421] = "evmhosmianwRRR", [1423] = "evmhosmfanwRRR", + [1448] = "evmhegumianRRR", [1449] = "evmhegsmianRRR", + [1451] = "evmhegsmfanRRR", [1452] = "evmhogumianRRR", + [1453] = "evmhogsmianRRR", [1455] = "evmhogsmfanRRR", + [1472] = "evmwlusianwRRR", [1473] = "evmwlssianwRRR", + [1480] = "evmwlumianwRRR", [1481] = "evmwlsmianwRRR", + [1491] = "evmwssfanRRR", [1496] = "evmwumianRRR", + [1497] = "evmwsmianRRR", [1499] = "evmwsmfanRRR", +} + +local map_pri = { + [0] = false, false, "tdiARI", "twiARI", + map_spe, false, false, "mulliRRI", + "subficRRI", false, "cmpl_iYLRU", "cmp_iYLRI", + "addicRRI", "addic.RRI", "addi|liRR0I", "addis|lisRR0I", + "b_KBJ", "sc", "bKJ", map_crops, + "rlwimiRR~AAA.", map_rlwinm, false, "rlwnmRR~RAA.", + "oriNRR~U", "orisRR~U", "xoriRR~U", "xorisRR~U", + "andi.RR~U", "andis.RR~U", map_rld, map_ext, + "lwzRRD", "lwzuRRD", "lbzRRD", "lbzuRRD", + "stwRRD", "stwuRRD", "stbRRD", "stbuRRD", + "lhzRRD", "lhzuRRD", "lhaRRD", "lhauRRD", + "sthRRD", "sthuRRD", "lmwRRD", "stmwRRD", + "lfsFRD", "lfsuFRD", "lfdFRD", "lfduFRD", + "stfsFRD", "stfsuFRD", "stfdFRD", "stfduFRD", + false, false, map_ld, map_fps, + false, false, map_std, map_fpd, +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "sp", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15", + "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23", + "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31", +} + +local map_cond = { [0] = "lt", "gt", "eq", "so", "ge", "le", "ne", "ns", } + +-- Format a condition bit. +local function condfmt(cond) + if cond <= 3 then + return map_cond[band(cond, 3)] + else + return format("4*cr%d+%s", rshift(cond, 2), map_cond[band(cond, 3)]) + end +end + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then extra = "\t->"..sym end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-7s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-7s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b0, 24), lshift(b1, 16), lshift(b2, 8), b3) + local operands = {} + local last = nil + local rs = 21 + ctx.op = op + ctx.rel = nil + + local opat = map_pri[rshift(b0, 2)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] + end + local name, pat = match(opat, "^([a-z0-9_.]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.]*)(.*)") + if altname then pat = pat2 end + + for p in gmatch(pat, ".") do + local x = nil + if p == "R" then + x = map_gpr[band(rshift(op, rs), 31)] + rs = rs - 5 + elseif p == "F" then + x = "f"..band(rshift(op, rs), 31) + rs = rs - 5 + elseif p == "A" then + x = band(rshift(op, rs), 31) + rs = rs - 5 + elseif p == "S" then + x = arshift(lshift(op, 27-rs), 27) + rs = rs - 5 + elseif p == "I" then + x = arshift(lshift(op, 16), 16) + elseif p == "U" then + x = band(op, 0xffff) + elseif p == "D" or p == "E" then + local disp = arshift(lshift(op, 16), 16) + if p == "E" then disp = band(disp, -4) end + if last == "r0" then last = "0" end + operands[#operands] = format("%d(%s)", disp, last) + elseif p >= "2" and p <= "8" then + local disp = band(rshift(op, rs), 31) * p + if last == "r0" then last = "0" end + operands[#operands] = format("%d(%s)", disp, last) + elseif p == "H" then + x = band(rshift(op, rs), 31) + lshift(band(op, 2), 4) + rs = rs - 5 + elseif p == "M" then + x = band(rshift(op, rs), 31) + band(op, 0x20) + elseif p == "C" then + x = condfmt(band(rshift(op, rs), 31)) + rs = rs - 5 + elseif p == "B" then + local bo = rshift(op, 21) + local cond = band(rshift(op, 16), 31) + local cn = "" + rs = rs - 10 + if band(bo, 4) == 0 then + cn = band(bo, 2) == 0 and "dnz" or "dz" + if band(bo, 0x10) == 0 then + cn = cn..(band(bo, 8) == 0 and "f" or "t") + end + if band(bo, 0x10) == 0 then x = condfmt(cond) end + name = name..(band(bo, 1) == band(rshift(op, 15), 1) and "-" or "+") + elseif band(bo, 0x10) == 0 then + cn = map_cond[band(cond, 3) + (band(bo, 8) == 0 and 4 or 0)] + if cond > 3 then x = "cr"..rshift(cond, 2) end + name = name..(band(bo, 1) == band(rshift(op, 15), 1) and "-" or "+") + end + name = gsub(name, "_", cn) + elseif p == "J" then + x = arshift(lshift(op, 27-rs), 29-rs)*4 + if band(op, 2) == 0 then x = ctx.addr + pos + x end + ctx.rel = x + x = "0x"..tohex(x) + elseif p == "K" then + if band(op, 1) ~= 0 then name = name.."l" end + if band(op, 2) ~= 0 then name = name.."a" end + elseif p == "X" or p == "Y" then + x = band(rshift(op, rs+2), 7) + if x == 0 and p == "Y" then x = nil else x = "cr"..x end + rs = rs - 5 + elseif p == "W" then + x = "cr"..band(op, 7) + elseif p == "Z" then + x = band(rshift(op, rs-4), 255) + rs = rs - 10 + elseif p == ">" then + operands[#operands] = rshift(operands[#operands], 1) + elseif p == "0" then + if last == "r0" then + operands[#operands] = nil + if altname then name = altname end + end + elseif p == "L" then + name = gsub(name, "_", band(op, 0x00200000) ~= 0 and "d" or "w") + elseif p == "." then + if band(op, 1) == 1 then name = name.."." end + elseif p == "N" then + if op == 0x60000000 then name = "nop"; break end + elseif p == "~" then + local n = #operands + operands[n-1], operands[n] = operands[n], operands[n-1] + elseif p == "=" then + local n = #operands + if last == operands[n-1] then + operands[n] = nil + name = altname + end + elseif p == "%" then + local n = #operands + if last == operands[n-1] and last == operands[n-2] then + operands[n] = nil + operands[n-1] = nil + name = altname + end + elseif p == "-" then + rs = rs - 5 + else + assert(false) + end + if x then operands[#operands+1] = x; last = x end + end + + return putop(ctx, name, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + stop = stop - stop % 4 + ctx.pos = ofs - ofs % 4 + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_gpr[r] end + return "f"..(r-32) +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_x64.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_x64.lua new file mode 100644 index 00000000..2d37423e --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_x64.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT x64 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the 64 bit functions from the combined +-- x86/x64 disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_x86 = require((string.match(..., ".*%.") or "").."dis_x86") +return { + create = dis_x86.create64, + disass = dis_x86.disass64, + regname = dis_x86.regname64 +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_x86.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_x86.lua new file mode 100644 index 00000000..5480854c --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dis_x86.lua @@ -0,0 +1,953 @@ +---------------------------------------------------------------------------- +-- LuaJIT x86/x64 disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- Sending small code snippets to an external disassembler and mixing the +-- output with our own stuff was too fragile. So I had to bite the bullet +-- and write yet another x86 disassembler. Oh well ... +-- +-- The output format is very similar to what ndisasm generates. But it has +-- been developed independently by looking at the opcode tables from the +-- Intel and AMD manuals. The supported instruction set is quite extensive +-- and reflects what a current generation Intel or AMD CPU implements in +-- 32 bit and 64 bit mode. Yes, this includes MMX, SSE, SSE2, SSE3, SSSE3, +-- SSE4.1, SSE4.2, SSE4a, AVX, AVX2 and even privileged and hypervisor +-- (VMX/SVM) instructions. +-- +-- Notes: +-- * The (useless) a16 prefix, 3DNow and pre-586 opcodes are unsupported. +-- * No attempt at optimization has been made -- it's fast enough for my needs. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local lower, rep = string.lower, string.rep +local bit = require("bit") +local tohex = bit.tohex + +-- Map for 1st opcode byte in 32 bit mode. Ugly? Well ... read on. +local map_opc1_32 = { +--0x +[0]="addBmr","addVmr","addBrm","addVrm","addBai","addVai","push es","pop es", +"orBmr","orVmr","orBrm","orVrm","orBai","orVai","push cs","opc2*", +--1x +"adcBmr","adcVmr","adcBrm","adcVrm","adcBai","adcVai","push ss","pop ss", +"sbbBmr","sbbVmr","sbbBrm","sbbVrm","sbbBai","sbbVai","push ds","pop ds", +--2x +"andBmr","andVmr","andBrm","andVrm","andBai","andVai","es:seg","daa", +"subBmr","subVmr","subBrm","subVrm","subBai","subVai","cs:seg","das", +--3x +"xorBmr","xorVmr","xorBrm","xorVrm","xorBai","xorVai","ss:seg","aaa", +"cmpBmr","cmpVmr","cmpBrm","cmpVrm","cmpBai","cmpVai","ds:seg","aas", +--4x +"incVR","incVR","incVR","incVR","incVR","incVR","incVR","incVR", +"decVR","decVR","decVR","decVR","decVR","decVR","decVR","decVR", +--5x +"pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR", +"popUR","popUR","popUR","popUR","popUR","popUR","popUR","popUR", +--6x +"sz*pushaw,pusha","sz*popaw,popa","boundVrm","arplWmr", +"fs:seg","gs:seg","o16:","a16", +"pushUi","imulVrmi","pushBs","imulVrms", +"insb","insVS","outsb","outsVS", +--7x +"joBj","jnoBj","jbBj","jnbBj","jzBj","jnzBj","jbeBj","jaBj", +"jsBj","jnsBj","jpeBj","jpoBj","jlBj","jgeBj","jleBj","jgBj", +--8x +"arith!Bmi","arith!Vmi","arith!Bmi","arith!Vms", +"testBmr","testVmr","xchgBrm","xchgVrm", +"movBmr","movVmr","movBrm","movVrm", +"movVmg","leaVrm","movWgm","popUm", +--9x +"nop*xchgVaR|pause|xchgWaR|repne nop","xchgVaR","xchgVaR","xchgVaR", +"xchgVaR","xchgVaR","xchgVaR","xchgVaR", +"sz*cbw,cwde,cdqe","sz*cwd,cdq,cqo","call farViw","wait", +"sz*pushfw,pushf","sz*popfw,popf","sahf","lahf", +--Ax +"movBao","movVao","movBoa","movVoa", +"movsb","movsVS","cmpsb","cmpsVS", +"testBai","testVai","stosb","stosVS", +"lodsb","lodsVS","scasb","scasVS", +--Bx +"movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi", +"movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI", +--Cx +"shift!Bmu","shift!Vmu","retBw","ret","vex*3$lesVrm","vex*2$ldsVrm","movBmi","movVmi", +"enterBwu","leave","retfBw","retf","int3","intBu","into","iretVS", +--Dx +"shift!Bm1","shift!Vm1","shift!Bmc","shift!Vmc","aamBu","aadBu","salc","xlatb", +"fp*0","fp*1","fp*2","fp*3","fp*4","fp*5","fp*6","fp*7", +--Ex +"loopneBj","loopeBj","loopBj","sz*jcxzBj,jecxzBj,jrcxzBj", +"inBau","inVau","outBua","outVua", +"callVj","jmpVj","jmp farViw","jmpBj","inBad","inVad","outBda","outVda", +--Fx +"lock:","int1","repne:rep","rep:","hlt","cmc","testb!Bm","testv!Vm", +"clc","stc","cli","sti","cld","std","incb!Bm","incd!Vm", +} +assert(#map_opc1_32 == 255) + +-- Map for 1st opcode byte in 64 bit mode (overrides only). +local map_opc1_64 = setmetatable({ + [0x06]=false, [0x07]=false, [0x0e]=false, + [0x16]=false, [0x17]=false, [0x1e]=false, [0x1f]=false, + [0x27]=false, [0x2f]=false, [0x37]=false, [0x3f]=false, + [0x60]=false, [0x61]=false, [0x62]=false, [0x63]="movsxdVrDmt", [0x67]="a32:", + [0x40]="rex*", [0x41]="rex*b", [0x42]="rex*x", [0x43]="rex*xb", + [0x44]="rex*r", [0x45]="rex*rb", [0x46]="rex*rx", [0x47]="rex*rxb", + [0x48]="rex*w", [0x49]="rex*wb", [0x4a]="rex*wx", [0x4b]="rex*wxb", + [0x4c]="rex*wr", [0x4d]="rex*wrb", [0x4e]="rex*wrx", [0x4f]="rex*wrxb", + [0x82]=false, [0x9a]=false, [0xc4]="vex*3", [0xc5]="vex*2", [0xce]=false, + [0xd4]=false, [0xd5]=false, [0xd6]=false, [0xea]=false, +}, { __index = map_opc1_32 }) + +-- Map for 2nd opcode byte (0F xx). True CISC hell. Hey, I told you. +-- Prefix dependent MMX/SSE opcodes: (none)|rep|o16|repne, -|F3|66|F2 +local map_opc2 = { +--0x +[0]="sldt!Dmp","sgdt!Ump","larVrm","lslVrm",nil,"syscall","clts","sysret", +"invd","wbinvd",nil,"ud1",nil,"$prefetch!Bm","femms","3dnowMrmu", +--1x +"movupsXrm|movssXrvm|movupdXrm|movsdXrvm", +"movupsXmr|movssXmvr|movupdXmr|movsdXmvr", +"movhlpsXrm$movlpsXrm|movsldupXrm|movlpdXrm|movddupXrm", +"movlpsXmr||movlpdXmr", +"unpcklpsXrvm||unpcklpdXrvm", +"unpckhpsXrvm||unpckhpdXrvm", +"movlhpsXrm$movhpsXrm|movshdupXrm|movhpdXrm", +"movhpsXmr||movhpdXmr", +"$prefetcht!Bm","hintnopVm","hintnopVm","hintnopVm", +"hintnopVm","hintnopVm","hintnopVm","hintnopVm", +--2x +"movUmx$","movUmy$","movUxm$","movUym$","movUmz$",nil,"movUzm$",nil, +"movapsXrm||movapdXrm", +"movapsXmr||movapdXmr", +"cvtpi2psXrMm|cvtsi2ssXrvVmt|cvtpi2pdXrMm|cvtsi2sdXrvVmt", +"movntpsXmr|movntssXmr|movntpdXmr|movntsdXmr", +"cvttps2piMrXm|cvttss2siVrXm|cvttpd2piMrXm|cvttsd2siVrXm", +"cvtps2piMrXm|cvtss2siVrXm|cvtpd2piMrXm|cvtsd2siVrXm", +"ucomissXrm||ucomisdXrm", +"comissXrm||comisdXrm", +--3x +"wrmsr","rdtsc","rdmsr","rdpmc","sysenter","sysexit",nil,"getsec", +"opc3*38",nil,"opc3*3a",nil,nil,nil,nil,nil, +--4x +"cmovoVrm","cmovnoVrm","cmovbVrm","cmovnbVrm", +"cmovzVrm","cmovnzVrm","cmovbeVrm","cmovaVrm", +"cmovsVrm","cmovnsVrm","cmovpeVrm","cmovpoVrm", +"cmovlVrm","cmovgeVrm","cmovleVrm","cmovgVrm", +--5x +"movmskpsVrXm$||movmskpdVrXm$","sqrtpsXrm|sqrtssXrm|sqrtpdXrm|sqrtsdXrm", +"rsqrtpsXrm|rsqrtssXrvm","rcppsXrm|rcpssXrvm", +"andpsXrvm||andpdXrvm","andnpsXrvm||andnpdXrvm", +"orpsXrvm||orpdXrvm","xorpsXrvm||xorpdXrvm", +"addpsXrvm|addssXrvm|addpdXrvm|addsdXrvm","mulpsXrvm|mulssXrvm|mulpdXrvm|mulsdXrvm", +"cvtps2pdXrm|cvtss2sdXrvm|cvtpd2psXrm|cvtsd2ssXrvm", +"cvtdq2psXrm|cvttps2dqXrm|cvtps2dqXrm", +"subpsXrvm|subssXrvm|subpdXrvm|subsdXrvm","minpsXrvm|minssXrvm|minpdXrvm|minsdXrvm", +"divpsXrvm|divssXrvm|divpdXrvm|divsdXrvm","maxpsXrvm|maxssXrvm|maxpdXrvm|maxsdXrvm", +--6x +"punpcklbwPrvm","punpcklwdPrvm","punpckldqPrvm","packsswbPrvm", +"pcmpgtbPrvm","pcmpgtwPrvm","pcmpgtdPrvm","packuswbPrvm", +"punpckhbwPrvm","punpckhwdPrvm","punpckhdqPrvm","packssdwPrvm", +"||punpcklqdqXrvm","||punpckhqdqXrvm", +"movPrVSm","movqMrm|movdquXrm|movdqaXrm", +--7x +"pshufwMrmu|pshufhwXrmu|pshufdXrmu|pshuflwXrmu","pshiftw!Pvmu", +"pshiftd!Pvmu","pshiftq!Mvmu||pshiftdq!Xvmu", +"pcmpeqbPrvm","pcmpeqwPrvm","pcmpeqdPrvm","emms*|", +"vmreadUmr||extrqXmuu$|insertqXrmuu$","vmwriteUrm||extrqXrm$|insertqXrm$", +nil,nil, +"||haddpdXrvm|haddpsXrvm","||hsubpdXrvm|hsubpsXrvm", +"movVSmMr|movqXrm|movVSmXr","movqMmr|movdquXmr|movdqaXmr", +--8x +"joVj","jnoVj","jbVj","jnbVj","jzVj","jnzVj","jbeVj","jaVj", +"jsVj","jnsVj","jpeVj","jpoVj","jlVj","jgeVj","jleVj","jgVj", +--9x +"setoBm","setnoBm","setbBm","setnbBm","setzBm","setnzBm","setbeBm","setaBm", +"setsBm","setnsBm","setpeBm","setpoBm","setlBm","setgeBm","setleBm","setgBm", +--Ax +"push fs","pop fs","cpuid","btVmr","shldVmru","shldVmrc",nil,nil, +"push gs","pop gs","rsm","btsVmr","shrdVmru","shrdVmrc","fxsave!Dmp","imulVrm", +--Bx +"cmpxchgBmr","cmpxchgVmr","$lssVrm","btrVmr", +"$lfsVrm","$lgsVrm","movzxVrBmt","movzxVrWmt", +"|popcntVrm","ud2Dp","bt!Vmu","btcVmr", +"bsfVrm","bsrVrm|lzcntVrm|bsrWrm","movsxVrBmt","movsxVrWmt", +--Cx +"xaddBmr","xaddVmr", +"cmppsXrvmu|cmpssXrvmu|cmppdXrvmu|cmpsdXrvmu","$movntiVmr|", +"pinsrwPrvWmu","pextrwDrPmu", +"shufpsXrvmu||shufpdXrvmu","$cmpxchg!Qmp", +"bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR", +--Dx +"||addsubpdXrvm|addsubpsXrvm","psrlwPrvm","psrldPrvm","psrlqPrvm", +"paddqPrvm","pmullwPrvm", +"|movq2dqXrMm|movqXmr|movdq2qMrXm$","pmovmskbVrMm||pmovmskbVrXm", +"psubusbPrvm","psubuswPrvm","pminubPrvm","pandPrvm", +"paddusbPrvm","padduswPrvm","pmaxubPrvm","pandnPrvm", +--Ex +"pavgbPrvm","psrawPrvm","psradPrvm","pavgwPrvm", +"pmulhuwPrvm","pmulhwPrvm", +"|cvtdq2pdXrm|cvttpd2dqXrm|cvtpd2dqXrm","$movntqMmr||$movntdqXmr", +"psubsbPrvm","psubswPrvm","pminswPrvm","porPrvm", +"paddsbPrvm","paddswPrvm","pmaxswPrvm","pxorPrvm", +--Fx +"|||lddquXrm","psllwPrvm","pslldPrvm","psllqPrvm", +"pmuludqPrvm","pmaddwdPrvm","psadbwPrvm","maskmovqMrm||maskmovdquXrm$", +"psubbPrvm","psubwPrvm","psubdPrvm","psubqPrvm", +"paddbPrvm","paddwPrvm","padddPrvm","ud", +} +assert(map_opc2[255] == "ud") + +-- Map for three-byte opcodes. Can't wait for their next invention. +local map_opc3 = { +["38"] = { -- [66] 0f 38 xx +--0x +[0]="pshufbPrvm","phaddwPrvm","phadddPrvm","phaddswPrvm", +"pmaddubswPrvm","phsubwPrvm","phsubdPrvm","phsubswPrvm", +"psignbPrvm","psignwPrvm","psigndPrvm","pmulhrswPrvm", +"||permilpsXrvm","||permilpdXrvm",nil,nil, +--1x +"||pblendvbXrma",nil,nil,nil, +"||blendvpsXrma","||blendvpdXrma","||permpsXrvm","||ptestXrm", +"||broadcastssXrm","||broadcastsdXrm","||broadcastf128XrlXm",nil, +"pabsbPrm","pabswPrm","pabsdPrm",nil, +--2x +"||pmovsxbwXrm","||pmovsxbdXrm","||pmovsxbqXrm","||pmovsxwdXrm", +"||pmovsxwqXrm","||pmovsxdqXrm",nil,nil, +"||pmuldqXrvm","||pcmpeqqXrvm","||$movntdqaXrm","||packusdwXrvm", +"||maskmovpsXrvm","||maskmovpdXrvm","||maskmovpsXmvr","||maskmovpdXmvr", +--3x +"||pmovzxbwXrm","||pmovzxbdXrm","||pmovzxbqXrm","||pmovzxwdXrm", +"||pmovzxwqXrm","||pmovzxdqXrm","||permdXrvm","||pcmpgtqXrvm", +"||pminsbXrvm","||pminsdXrvm","||pminuwXrvm","||pminudXrvm", +"||pmaxsbXrvm","||pmaxsdXrvm","||pmaxuwXrvm","||pmaxudXrvm", +--4x +"||pmulddXrvm","||phminposuwXrm",nil,nil, +nil,"||psrlvVSXrvm","||psravdXrvm","||psllvVSXrvm", +--5x +[0x58] = "||pbroadcastdXrlXm",[0x59] = "||pbroadcastqXrlXm", +[0x5a] = "||broadcasti128XrlXm", +--7x +[0x78] = "||pbroadcastbXrlXm",[0x79] = "||pbroadcastwXrlXm", +--8x +[0x8c] = "||pmaskmovXrvVSm", +[0x8e] = "||pmaskmovVSmXvr", +--9x +[0x96] = "||fmaddsub132pHXrvm",[0x97] = "||fmsubadd132pHXrvm", +[0x98] = "||fmadd132pHXrvm",[0x99] = "||fmadd132sHXrvm", +[0x9a] = "||fmsub132pHXrvm",[0x9b] = "||fmsub132sHXrvm", +[0x9c] = "||fnmadd132pHXrvm",[0x9d] = "||fnmadd132sHXrvm", +[0x9e] = "||fnmsub132pHXrvm",[0x9f] = "||fnmsub132sHXrvm", +--Ax +[0xa6] = "||fmaddsub213pHXrvm",[0xa7] = "||fmsubadd213pHXrvm", +[0xa8] = "||fmadd213pHXrvm",[0xa9] = "||fmadd213sHXrvm", +[0xaa] = "||fmsub213pHXrvm",[0xab] = "||fmsub213sHXrvm", +[0xac] = "||fnmadd213pHXrvm",[0xad] = "||fnmadd213sHXrvm", +[0xae] = "||fnmsub213pHXrvm",[0xaf] = "||fnmsub213sHXrvm", +--Bx +[0xb6] = "||fmaddsub231pHXrvm",[0xb7] = "||fmsubadd231pHXrvm", +[0xb8] = "||fmadd231pHXrvm",[0xb9] = "||fmadd231sHXrvm", +[0xba] = "||fmsub231pHXrvm",[0xbb] = "||fmsub231sHXrvm", +[0xbc] = "||fnmadd231pHXrvm",[0xbd] = "||fnmadd231sHXrvm", +[0xbe] = "||fnmsub231pHXrvm",[0xbf] = "||fnmsub231sHXrvm", +--Dx +[0xdc] = "||aesencXrvm", [0xdd] = "||aesenclastXrvm", +[0xde] = "||aesdecXrvm", [0xdf] = "||aesdeclastXrvm", +--Fx +[0xf0] = "|||crc32TrBmt",[0xf1] = "|||crc32TrVmt", +[0xf7] = "| sarxVrmv| shlxVrmv| shrxVrmv", +}, + +["3a"] = { -- [66] 0f 3a xx +--0x +[0x00]="||permqXrmu","||permpdXrmu","||pblenddXrvmu",nil, +"||permilpsXrmu","||permilpdXrmu","||perm2f128Xrvmu",nil, +"||roundpsXrmu","||roundpdXrmu","||roundssXrvmu","||roundsdXrvmu", +"||blendpsXrvmu","||blendpdXrvmu","||pblendwXrvmu","palignrPrvmu", +--1x +nil,nil,nil,nil, +"||pextrbVmXru","||pextrwVmXru","||pextrVmSXru","||extractpsVmXru", +"||insertf128XrvlXmu","||extractf128XlXmYru",nil,nil, +nil,nil,nil,nil, +--2x +"||pinsrbXrvVmu","||insertpsXrvmu","||pinsrXrvVmuS",nil, +--3x +[0x38] = "||inserti128Xrvmu",[0x39] = "||extracti128XlXmYru", +--4x +[0x40] = "||dppsXrvmu", +[0x41] = "||dppdXrvmu", +[0x42] = "||mpsadbwXrvmu", +[0x44] = "||pclmulqdqXrvmu", +[0x46] = "||perm2i128Xrvmu", +[0x4a] = "||blendvpsXrvmb",[0x4b] = "||blendvpdXrvmb", +[0x4c] = "||pblendvbXrvmb", +--6x +[0x60] = "||pcmpestrmXrmu",[0x61] = "||pcmpestriXrmu", +[0x62] = "||pcmpistrmXrmu",[0x63] = "||pcmpistriXrmu", +[0xdf] = "||aeskeygenassistXrmu", +--Fx +[0xf0] = "||| rorxVrmu", +}, +} + +-- Map for VMX/SVM opcodes 0F 01 C0-FF (sgdt group with register operands). +local map_opcvm = { +[0xc1]="vmcall",[0xc2]="vmlaunch",[0xc3]="vmresume",[0xc4]="vmxoff", +[0xc8]="monitor",[0xc9]="mwait", +[0xd8]="vmrun",[0xd9]="vmmcall",[0xda]="vmload",[0xdb]="vmsave", +[0xdc]="stgi",[0xdd]="clgi",[0xde]="skinit",[0xdf]="invlpga", +[0xf8]="swapgs",[0xf9]="rdtscp", +} + +-- Map for FP opcodes. And you thought stack machines are simple? +local map_opcfp = { +-- D8-DF 00-BF: opcodes with a memory operand. +-- D8 +[0]="faddFm","fmulFm","fcomFm","fcompFm","fsubFm","fsubrFm","fdivFm","fdivrFm", +"fldFm",nil,"fstFm","fstpFm","fldenvVm","fldcwWm","fnstenvVm","fnstcwWm", +-- DA +"fiaddDm","fimulDm","ficomDm","ficompDm", +"fisubDm","fisubrDm","fidivDm","fidivrDm", +-- DB +"fildDm","fisttpDm","fistDm","fistpDm",nil,"fld twordFmp",nil,"fstp twordFmp", +-- DC +"faddGm","fmulGm","fcomGm","fcompGm","fsubGm","fsubrGm","fdivGm","fdivrGm", +-- DD +"fldGm","fisttpQm","fstGm","fstpGm","frstorDmp",nil,"fnsaveDmp","fnstswWm", +-- DE +"fiaddWm","fimulWm","ficomWm","ficompWm", +"fisubWm","fisubrWm","fidivWm","fidivrWm", +-- DF +"fildWm","fisttpWm","fistWm","fistpWm", +"fbld twordFmp","fildQm","fbstp twordFmp","fistpQm", +-- xx C0-FF: opcodes with a pseudo-register operand. +-- D8 +"faddFf","fmulFf","fcomFf","fcompFf","fsubFf","fsubrFf","fdivFf","fdivrFf", +-- D9 +"fldFf","fxchFf",{"fnop"},nil, +{"fchs","fabs",nil,nil,"ftst","fxam"}, +{"fld1","fldl2t","fldl2e","fldpi","fldlg2","fldln2","fldz"}, +{"f2xm1","fyl2x","fptan","fpatan","fxtract","fprem1","fdecstp","fincstp"}, +{"fprem","fyl2xp1","fsqrt","fsincos","frndint","fscale","fsin","fcos"}, +-- DA +"fcmovbFf","fcmoveFf","fcmovbeFf","fcmovuFf",nil,{nil,"fucompp"},nil,nil, +-- DB +"fcmovnbFf","fcmovneFf","fcmovnbeFf","fcmovnuFf", +{nil,nil,"fnclex","fninit"},"fucomiFf","fcomiFf",nil, +-- DC +"fadd toFf","fmul toFf",nil,nil, +"fsub toFf","fsubr toFf","fdivr toFf","fdiv toFf", +-- DD +"ffreeFf",nil,"fstFf","fstpFf","fucomFf","fucompFf",nil,nil, +-- DE +"faddpFf","fmulpFf",nil,{nil,"fcompp"}, +"fsubrpFf","fsubpFf","fdivrpFf","fdivpFf", +-- DF +nil,nil,nil,nil,{"fnstsw ax"},"fucomipFf","fcomipFf",nil, +} +assert(map_opcfp[126] == "fcomipFf") + +-- Map for opcode groups. The subkey is sp from the ModRM byte. +local map_opcgroup = { + arith = { "add", "or", "adc", "sbb", "and", "sub", "xor", "cmp" }, + shift = { "rol", "ror", "rcl", "rcr", "shl", "shr", "sal", "sar" }, + testb = { "testBmi", "testBmi", "not", "neg", "mul", "imul", "div", "idiv" }, + testv = { "testVmi", "testVmi", "not", "neg", "mul", "imul", "div", "idiv" }, + incb = { "inc", "dec" }, + incd = { "inc", "dec", "callUmp", "$call farDmp", + "jmpUmp", "$jmp farDmp", "pushUm" }, + sldt = { "sldt", "str", "lldt", "ltr", "verr", "verw" }, + sgdt = { "vm*$sgdt", "vm*$sidt", "$lgdt", "vm*$lidt", + "smsw", nil, "lmsw", "vm*$invlpg" }, + bt = { nil, nil, nil, nil, "bt", "bts", "btr", "btc" }, + cmpxchg = { nil, "sz*,cmpxchg8bQmp,cmpxchg16bXmp", nil, nil, + nil, nil, "vmptrld|vmxon|vmclear", "vmptrst" }, + pshiftw = { nil, nil, "psrlw", nil, "psraw", nil, "psllw" }, + pshiftd = { nil, nil, "psrld", nil, "psrad", nil, "pslld" }, + pshiftq = { nil, nil, "psrlq", nil, nil, nil, "psllq" }, + pshiftdq = { nil, nil, "psrlq", "psrldq", nil, nil, "psllq", "pslldq" }, + fxsave = { "$fxsave", "$fxrstor", "$ldmxcsr", "$stmxcsr", + nil, "lfenceDp$", "mfenceDp$", "sfenceDp$clflush" }, + prefetch = { "prefetch", "prefetchw" }, + prefetcht = { "prefetchnta", "prefetcht0", "prefetcht1", "prefetcht2" }, +} + +------------------------------------------------------------------------------ + +-- Maps for register names. +local map_regs = { + B = { "al", "cl", "dl", "bl", "ah", "ch", "dh", "bh", + "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, + B64 = { "al", "cl", "dl", "bl", "spl", "bpl", "sil", "dil", + "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, + W = { "ax", "cx", "dx", "bx", "sp", "bp", "si", "di", + "r8w", "r9w", "r10w", "r11w", "r12w", "r13w", "r14w", "r15w" }, + D = { "eax", "ecx", "edx", "ebx", "esp", "ebp", "esi", "edi", + "r8d", "r9d", "r10d", "r11d", "r12d", "r13d", "r14d", "r15d" }, + Q = { "rax", "rcx", "rdx", "rbx", "rsp", "rbp", "rsi", "rdi", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15" }, + M = { "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7", + "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7" }, -- No x64 ext! + X = { "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7", + "xmm8", "xmm9", "xmm10", "xmm11", "xmm12", "xmm13", "xmm14", "xmm15" }, + Y = { "ymm0", "ymm1", "ymm2", "ymm3", "ymm4", "ymm5", "ymm6", "ymm7", + "ymm8", "ymm9", "ymm10", "ymm11", "ymm12", "ymm13", "ymm14", "ymm15" }, +} +local map_segregs = { "es", "cs", "ss", "ds", "fs", "gs", "segr6", "segr7" } + +-- Maps for size names. +local map_sz2n = { + B = 1, W = 2, D = 4, Q = 8, M = 8, X = 16, Y = 32, +} +local map_sz2prefix = { + B = "byte", W = "word", D = "dword", + Q = "qword", + M = "qword", X = "xword", Y = "yword", + F = "dword", G = "qword", -- No need for sizes/register names for these two. +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local code, pos, hex = ctx.code, ctx.pos, "" + local hmax = ctx.hexdump + if hmax > 0 then + for i=ctx.start,pos-1 do + hex = hex..format("%02X", byte(code, i, i)) + end + if #hex > hmax then hex = sub(hex, 1, hmax)..". " + else hex = hex..rep(" ", hmax-#hex+2) end + end + if operands then text = text.." "..operands end + if ctx.o16 then text = "o16 "..text; ctx.o16 = false end + if ctx.a32 then text = "a32 "..text; ctx.a32 = false end + if ctx.rep then text = ctx.rep.." "..text; ctx.rep = false end + if ctx.rex then + local t = (ctx.rexw and "w" or "")..(ctx.rexr and "r" or "").. + (ctx.rexx and "x" or "")..(ctx.rexb and "b" or "").. + (ctx.vexl and "l" or "") + if ctx.vexv and ctx.vexv ~= 0 then t = t.."v"..ctx.vexv end + if t ~= "" then text = ctx.rex.."."..t.." "..gsub(text, "^ ", "") + elseif ctx.rex == "vex" then text = gsub("v"..text, "^v ", "") end + ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false + ctx.rex = false; ctx.vexl = false; ctx.vexv = false + end + if ctx.seg then + local text2, n = gsub(text, "%[", "["..ctx.seg..":") + if n == 0 then text = ctx.seg.." "..text else text = text2 end + ctx.seg = false + end + if ctx.lock then text = "lock "..text; ctx.lock = false end + local imm = ctx.imm + if imm then + local sym = ctx.symtab[imm] + if sym then text = text.."\t->"..sym end + end + ctx.out(format("%08x %s%s\n", ctx.addr+ctx.start, hex, text)) + ctx.mrm = false + ctx.vexv = false + ctx.start = pos + ctx.imm = nil +end + +-- Clear all prefix flags. +local function clearprefixes(ctx) + ctx.o16 = false; ctx.seg = false; ctx.lock = false; ctx.rep = false + ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false + ctx.rex = false; ctx.a32 = false; ctx.vexl = false +end + +-- Fallback for incomplete opcodes at the end. +local function incomplete(ctx) + ctx.pos = ctx.stop+1 + clearprefixes(ctx) + return putop(ctx, "(incomplete)") +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + clearprefixes(ctx) + return putop(ctx, "(unknown)") +end + +-- Return an immediate of the specified size. +local function getimm(ctx, pos, n) + if pos+n-1 > ctx.stop then return incomplete(ctx) end + local code = ctx.code + if n == 1 then + local b1 = byte(code, pos, pos) + return b1 + elseif n == 2 then + local b1, b2 = byte(code, pos, pos+1) + return b1+b2*256 + else + local b1, b2, b3, b4 = byte(code, pos, pos+3) + local imm = b1+b2*256+b3*65536+b4*16777216 + ctx.imm = imm + return imm + end +end + +-- Process pattern string and generate the operands. +local function putpat(ctx, name, pat) + local operands, regs, sz, mode, sp, rm, sc, rx, sdisp + local code, pos, stop, vexl = ctx.code, ctx.pos, ctx.stop, ctx.vexl + + -- Chars used: 1DFGHIMPQRSTUVWXYabcdfgijlmoprstuvwxyz + for p in gmatch(pat, ".") do + local x = nil + if p == "V" or p == "U" then + if ctx.rexw then sz = "Q"; ctx.rexw = false + elseif ctx.o16 then sz = "W"; ctx.o16 = false + elseif p == "U" and ctx.x64 then sz = "Q" + else sz = "D" end + regs = map_regs[sz] + elseif p == "T" then + if ctx.rexw then sz = "Q"; ctx.rexw = false else sz = "D" end + regs = map_regs[sz] + elseif p == "B" then + sz = "B" + regs = ctx.rex and map_regs.B64 or map_regs.B + elseif match(p, "[WDQMXYFG]") then + sz = p + if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end + regs = map_regs[sz] + elseif p == "P" then + sz = ctx.o16 and "X" or "M"; ctx.o16 = false + if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end + regs = map_regs[sz] + elseif p == "H" then + name = name..(ctx.rexw and "d" or "s") + ctx.rexw = false + elseif p == "S" then + name = name..lower(sz) + elseif p == "s" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = imm <= 127 and format("+0x%02x", imm) + or format("-0x%02x", 256-imm) + pos = pos+1 + elseif p == "u" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = format("0x%02x", imm) + pos = pos+1 + elseif p == "b" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = regs[imm/16+1] + pos = pos+1 + elseif p == "w" then + local imm = getimm(ctx, pos, 2); if not imm then return end + x = format("0x%x", imm) + pos = pos+2 + elseif p == "o" then -- [offset] + if ctx.x64 then + local imm1 = getimm(ctx, pos, 4); if not imm1 then return end + local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end + x = format("[0x%08x%08x]", imm2, imm1) + pos = pos+8 + else + local imm = getimm(ctx, pos, 4); if not imm then return end + x = format("[0x%08x]", imm) + pos = pos+4 + end + elseif p == "i" or p == "I" then + local n = map_sz2n[sz] + if n == 8 and ctx.x64 and p == "I" then + local imm1 = getimm(ctx, pos, 4); if not imm1 then return end + local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end + x = format("0x%08x%08x", imm2, imm1) + else + if n == 8 then n = 4 end + local imm = getimm(ctx, pos, n); if not imm then return end + if sz == "Q" and (imm < 0 or imm > 0x7fffffff) then + imm = (0xffffffff+1)-imm + x = format(imm > 65535 and "-0x%08x" or "-0x%x", imm) + else + x = format(imm > 65535 and "0x%08x" or "0x%x", imm) + end + end + pos = pos+n + elseif p == "j" then + local n = map_sz2n[sz] + if n == 8 then n = 4 end + local imm = getimm(ctx, pos, n); if not imm then return end + if sz == "B" and imm > 127 then imm = imm-256 + elseif imm > 2147483647 then imm = imm-4294967296 end + pos = pos+n + imm = imm + pos + ctx.addr + if imm > 4294967295 and not ctx.x64 then imm = imm-4294967296 end + ctx.imm = imm + if sz == "W" then + x = format("word 0x%04x", imm%65536) + elseif ctx.x64 then + local lo = imm % 0x1000000 + x = format("0x%02x%06x", (imm-lo) / 0x1000000, lo) + else + x = "0x"..tohex(imm) + end + elseif p == "R" then + local r = byte(code, pos-1, pos-1)%8 + if ctx.rexb then r = r + 8; ctx.rexb = false end + x = regs[r+1] + elseif p == "a" then x = regs[1] + elseif p == "c" then x = "cl" + elseif p == "d" then x = "dx" + elseif p == "1" then x = "1" + else + if not mode then + mode = ctx.mrm + if not mode then + if pos > stop then return incomplete(ctx) end + mode = byte(code, pos, pos) + pos = pos+1 + end + rm = mode%8; mode = (mode-rm)/8 + sp = mode%8; mode = (mode-sp)/8 + sdisp = "" + if mode < 3 then + if rm == 4 then + if pos > stop then return incomplete(ctx) end + sc = byte(code, pos, pos) + pos = pos+1 + rm = sc%8; sc = (sc-rm)/8 + rx = sc%8; sc = (sc-rx)/8 + if ctx.rexx then rx = rx + 8; ctx.rexx = false end + if rx == 4 then rx = nil end + end + if mode > 0 or rm == 5 then + local dsz = mode + if dsz ~= 1 then dsz = 4 end + local disp = getimm(ctx, pos, dsz); if not disp then return end + if mode == 0 then rm = nil end + if rm or rx or (not sc and ctx.x64 and not ctx.a32) then + if dsz == 1 and disp > 127 then + sdisp = format("-0x%x", 256-disp) + elseif disp >= 0 and disp <= 0x7fffffff then + sdisp = format("+0x%x", disp) + else + sdisp = format("-0x%x", (0xffffffff+1)-disp) + end + else + sdisp = format(ctx.x64 and not ctx.a32 and + not (disp >= 0 and disp <= 0x7fffffff) + and "0xffffffff%08x" or "0x%08x", disp) + end + pos = pos+dsz + end + end + if rm and ctx.rexb then rm = rm + 8; ctx.rexb = false end + if ctx.rexr then sp = sp + 8; ctx.rexr = false end + end + if p == "m" then + if mode == 3 then x = regs[rm+1] + else + local aregs = ctx.a32 and map_regs.D or ctx.aregs + local srm, srx = "", "" + if rm then srm = aregs[rm+1] + elseif not sc and ctx.x64 and not ctx.a32 then srm = "rip" end + ctx.a32 = false + if rx then + if rm then srm = srm.."+" end + srx = aregs[rx+1] + if sc > 0 then srx = srx.."*"..(2^sc) end + end + x = format("[%s%s%s]", srm, srx, sdisp) + end + if mode < 3 and + (not match(pat, "[aRrgp]") or match(pat, "t")) then -- Yuck. + x = map_sz2prefix[sz].." "..x + end + elseif p == "r" then x = regs[sp+1] + elseif p == "g" then x = map_segregs[sp+1] + elseif p == "p" then -- Suppress prefix. + elseif p == "f" then x = "st"..rm + elseif p == "x" then + if sp == 0 and ctx.lock and not ctx.x64 then + x = "CR8"; ctx.lock = false + else + x = "CR"..sp + end + elseif p == "v" then + if ctx.vexv then + x = regs[ctx.vexv+1]; ctx.vexv = false + end + elseif p == "y" then x = "DR"..sp + elseif p == "z" then x = "TR"..sp + elseif p == "l" then vexl = false + elseif p == "t" then + else + error("bad pattern `"..pat.."'") + end + end + if x then operands = operands and operands..", "..x or x end + end + ctx.pos = pos + return putop(ctx, name, operands) +end + +-- Forward declaration. +local map_act + +-- Fetch and cache MRM byte. +local function getmrm(ctx) + local mrm = ctx.mrm + if not mrm then + local pos = ctx.pos + if pos > ctx.stop then return nil end + mrm = byte(ctx.code, pos, pos) + ctx.pos = pos+1 + ctx.mrm = mrm + end + return mrm +end + +-- Dispatch to handler depending on pattern. +local function dispatch(ctx, opat, patgrp) + if not opat then return unknown(ctx) end + if match(opat, "%|") then -- MMX/SSE variants depending on prefix. + local p + if ctx.rep then + p = ctx.rep=="rep" and "%|([^%|]*)" or "%|[^%|]*%|[^%|]*%|([^%|]*)" + ctx.rep = false + elseif ctx.o16 then p = "%|[^%|]*%|([^%|]*)"; ctx.o16 = false + else p = "^[^%|]*" end + opat = match(opat, p) + if not opat then return unknown(ctx) end +-- ctx.rep = false; ctx.o16 = false + --XXX fails for 66 f2 0f 38 f1 06 crc32 eax,WORD PTR [esi] + --XXX remove in branches? + end + if match(opat, "%$") then -- reg$mem variants. + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + opat = match(opat, mrm >= 192 and "^[^%$]*" or "%$(.*)") + if opat == "" then return unknown(ctx) end + end + if opat == "" then return unknown(ctx) end + local name, pat = match(opat, "^([a-z0-9 ]*)(.*)") + if pat == "" and patgrp then pat = patgrp end + return map_act[sub(pat, 1, 1)](ctx, name, pat) +end + +-- Get a pattern from an opcode map and dispatch to handler. +local function dispatchmap(ctx, opcmap) + local pos = ctx.pos + local opat = opcmap[byte(ctx.code, pos, pos)] + pos = pos + 1 + ctx.pos = pos + return dispatch(ctx, opat) +end + +-- Map for action codes. The key is the first char after the name. +map_act = { + -- Simple opcodes without operands. + [""] = function(ctx, name, pat) + return putop(ctx, name) + end, + + -- Operand size chars fall right through. + B = putpat, W = putpat, D = putpat, Q = putpat, + V = putpat, U = putpat, T = putpat, + M = putpat, X = putpat, P = putpat, + F = putpat, G = putpat, Y = putpat, + H = putpat, + + -- Collect prefixes. + [":"] = function(ctx, name, pat) + ctx[pat == ":" and name or sub(pat, 2)] = name + if ctx.pos - ctx.start > 5 then return unknown(ctx) end -- Limit #prefixes. + end, + + -- Chain to special handler specified by name. + ["*"] = function(ctx, name, pat) + return map_act[name](ctx, name, sub(pat, 2)) + end, + + -- Use named subtable for opcode group. + ["!"] = function(ctx, name, pat) + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + return dispatch(ctx, map_opcgroup[name][((mrm-(mrm%8))/8)%8+1], sub(pat, 2)) + end, + + -- o16,o32[,o64] variants. + sz = function(ctx, name, pat) + if ctx.o16 then ctx.o16 = false + else + pat = match(pat, ",(.*)") + if ctx.rexw then + local p = match(pat, ",(.*)") + if p then pat = p; ctx.rexw = false end + end + end + pat = match(pat, "^[^,]*") + return dispatch(ctx, pat) + end, + + -- Two-byte opcode dispatch. + opc2 = function(ctx, name, pat) + return dispatchmap(ctx, map_opc2) + end, + + -- Three-byte opcode dispatch. + opc3 = function(ctx, name, pat) + return dispatchmap(ctx, map_opc3[pat]) + end, + + -- VMX/SVM dispatch. + vm = function(ctx, name, pat) + return dispatch(ctx, map_opcvm[ctx.mrm]) + end, + + -- Floating point opcode dispatch. + fp = function(ctx, name, pat) + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + local rm = mrm%8 + local idx = pat*8 + ((mrm-rm)/8)%8 + if mrm >= 192 then idx = idx + 64 end + local opat = map_opcfp[idx] + if type(opat) == "table" then opat = opat[rm+1] end + return dispatch(ctx, opat) + end, + + -- REX prefix. + rex = function(ctx, name, pat) + if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. + for p in gmatch(pat, ".") do ctx["rex"..p] = true end + ctx.rex = "rex" + end, + + -- VEX prefix. + vex = function(ctx, name, pat) + if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. + ctx.rex = "vex" + local pos = ctx.pos + if ctx.mrm then + ctx.mrm = nil + pos = pos-1 + end + local b = byte(ctx.code, pos, pos) + if not b then return incomplete(ctx) end + pos = pos+1 + if b < 128 then ctx.rexr = true end + local m = 1 + if pat == "3" then + m = b%32; b = (b-m)/32 + local nb = b%2; b = (b-nb)/2 + if nb == 0 then ctx.rexb = true end + local nx = b%2 + if nx == 0 then ctx.rexx = true end + b = byte(ctx.code, pos, pos) + if not b then return incomplete(ctx) end + pos = pos+1 + if b >= 128 then ctx.rexw = true end + end + ctx.pos = pos + local map + if m == 1 then map = map_opc2 + elseif m == 2 then map = map_opc3["38"] + elseif m == 3 then map = map_opc3["3a"] + else return unknown(ctx) end + local p = b%4; b = (b-p)/4 + if p == 1 then ctx.o16 = "o16" + elseif p == 2 then ctx.rep = "rep" + elseif p == 3 then ctx.rep = "repne" end + local l = b%2; b = (b-l)/2 + if l ~= 0 then ctx.vexl = true end + ctx.vexv = (-1-b)%16 + return dispatchmap(ctx, map) + end, + + -- Special case for nop with REX prefix. + nop = function(ctx, name, pat) + return dispatch(ctx, ctx.rex and pat or "nop") + end, + + -- Special case for 0F 77. + emms = function(ctx, name, pat) + if ctx.rex ~= "vex" then + return putop(ctx, "emms") + elseif ctx.vexl then + ctx.vexl = false + return putop(ctx, "zeroall") + else + return putop(ctx, "zeroupper") + end + end, +} + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ofs = ofs + 1 + ctx.start = ofs + ctx.pos = ofs + ctx.stop = stop + ctx.imm = nil + ctx.mrm = false + clearprefixes(ctx) + while ctx.pos <= stop do dispatchmap(ctx, ctx.map1) end + if ctx.pos ~= ctx.start then incomplete(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = (addr or 0) - 1 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 16 + ctx.x64 = false + ctx.map1 = map_opc1_32 + ctx.aregs = map_regs.D + return ctx +end + +local function create64(code, addr, out) + local ctx = create(code, addr, out) + ctx.x64 = true + ctx.map1 = map_opc1_64 + ctx.aregs = map_regs.Q + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +local function disass64(code, addr, out) + create64(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 8 then return map_regs.D[r+1] end + return map_regs.X[r-7] +end + +local function regname64(r) + if r < 16 then return map_regs.Q[r+1] end + return map_regs.X[r-15] +end + +-- Public module functions. +return { + create = create, + create64 = create64, + disass = disass, + disass64 = disass64, + regname = regname, + regname64 = regname64 +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dump.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dump.lua new file mode 100644 index 00000000..9eda08c4 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/dump.lua @@ -0,0 +1,726 @@ +---------------------------------------------------------------------------- +-- LuaJIT compiler dump module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module can be used to debug the JIT compiler itself. It dumps the +-- code representations and structures used in various compiler stages. +-- +-- Example usage: +-- +-- luajit -jdump -e "local x=0; for i=1,1e6 do x=x+i end; print(x)" +-- luajit -jdump=im -e "for i=1,1000 do for j=1,1000 do end end" | less -R +-- luajit -jdump=is myapp.lua | less -R +-- luajit -jdump=-b myapp.lua +-- luajit -jdump=+aH,myapp.html myapp.lua +-- luajit -jdump=ixT,myapp.dump myapp.lua +-- +-- The first argument specifies the dump mode. The second argument gives +-- the output file name. Default output is to stdout, unless the environment +-- variable LUAJIT_DUMPFILE is set. The file is overwritten every time the +-- module is started. +-- +-- Different features can be turned on or off with the dump mode. If the +-- mode starts with a '+', the following features are added to the default +-- set of features; a '-' removes them. Otherwise the features are replaced. +-- +-- The following dump features are available (* marks the default): +-- +-- * t Print a line for each started, ended or aborted trace (see also -jv). +-- * b Dump the traced bytecode. +-- * i Dump the IR (intermediate representation). +-- r Augment the IR with register/stack slots. +-- s Dump the snapshot map. +-- * m Dump the generated machine code. +-- x Print each taken trace exit. +-- X Print each taken trace exit and the contents of all registers. +-- a Print the IR of aborted traces, too. +-- +-- The output format can be set with the following characters: +-- +-- T Plain text output. +-- A ANSI-colored text output +-- H Colorized HTML + CSS output. +-- +-- The default output format is plain text. It's set to ANSI-colored text +-- if the COLORTERM variable is set. Note: this is independent of any output +-- redirection, which is actually considered a feature. +-- +-- You probably want to use less -R to enjoy viewing ANSI-colored text from +-- a pipe or a file. Add this to your ~/.bashrc: export LESS="-R" +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local funcinfo, funcbc = jutil.funcinfo, jutil.funcbc +local traceinfo, traceir, tracek = jutil.traceinfo, jutil.traceir, jutil.tracek +local tracemc, tracesnap = jutil.tracemc, jutil.tracesnap +local traceexitstub, ircalladdr = jutil.traceexitstub, jutil.ircalladdr +local bit = require("bit") +local band, shr, tohex = bit.band, bit.rshift, bit.tohex +local sub, gsub, format = string.sub, string.gsub, string.format +local byte, rep = string.byte, string.rep +local type, tostring = type, tostring +local stdout, stderr = io.stdout, io.stderr + +-- Load other modules on-demand. +local bcline, disass + +-- Active flag, output file handle and dump mode. +local active, out, dumpmode + +------------------------------------------------------------------------------ + +local symtabmt = { __index = false } +local symtab = {} +local nexitsym = 0 + +-- Fill nested symbol table with per-trace exit stub addresses. +local function fillsymtab_tr(tr, nexit) + local t = {} + symtabmt.__index = t + if jit.arch:sub(1, 4) == "mips" then + t[traceexitstub(tr, 0)] = "exit" + return + end + for i=0,nexit-1 do + local addr = traceexitstub(tr, i) + if addr < 0 then addr = addr + 2^32 end + t[addr] = tostring(i) + end + local addr = traceexitstub(tr, nexit) + if addr then t[addr] = "stack_check" end +end + +-- Fill symbol table with trace exit stub addresses. +local function fillsymtab(tr, nexit) + local t = symtab + if nexitsym == 0 then + local maskaddr = jit.arch == "arm" and -2 + local ircall = vmdef.ircall + for i=0,#ircall do + local addr = ircalladdr(i) + if addr ~= 0 then + if maskaddr then addr = band(addr, maskaddr) end + if addr < 0 then addr = addr + 2^32 end + t[addr] = ircall[i] + end + end + end + if nexitsym == 1000000 then -- Per-trace exit stubs. + fillsymtab_tr(tr, nexit) + elseif nexit > nexitsym then -- Shared exit stubs. + for i=nexitsym,nexit-1 do + local addr = traceexitstub(i) + if addr == nil then -- Fall back to per-trace exit stubs. + fillsymtab_tr(tr, nexit) + setmetatable(symtab, symtabmt) + nexit = 1000000 + break + end + if addr < 0 then addr = addr + 2^32 end + t[addr] = tostring(i) + end + nexitsym = nexit + end + return t +end + +local function dumpwrite(s) + out:write(s) +end + +-- Disassemble machine code. +local function dump_mcode(tr) + local info = traceinfo(tr) + if not info then return end + local mcode, addr, loop = tracemc(tr) + if not mcode then return end + if not disass then disass = require("jit.dis_"..jit.arch) end + if addr < 0 then addr = addr + 2^32 end + out:write("---- TRACE ", tr, " mcode ", #mcode, "\n") + local ctx = disass.create(mcode, addr, dumpwrite) + ctx.hexdump = 0 + ctx.symtab = fillsymtab(tr, info.nexit) + if loop ~= 0 then + symtab[addr+loop] = "LOOP" + ctx:disass(0, loop) + out:write("->LOOP:\n") + ctx:disass(loop, #mcode-loop) + symtab[addr+loop] = nil + else + ctx:disass(0, #mcode) + end +end + +------------------------------------------------------------------------------ + +local irtype_text = { + [0] = "nil", + "fal", + "tru", + "lud", + "str", + "p32", + "thr", + "pro", + "fun", + "p64", + "cdt", + "tab", + "udt", + "flt", + "num", + "i8 ", + "u8 ", + "i16", + "u16", + "int", + "u32", + "i64", + "u64", + "sfp", +} + +local colortype_ansi = { + [0] = "%s", + "%s", + "%s", + "\027[36m%s\027[m", + "\027[32m%s\027[m", + "%s", + "\027[1m%s\027[m", + "%s", + "\027[1m%s\027[m", + "%s", + "\027[33m%s\027[m", + "\027[31m%s\027[m", + "\027[36m%s\027[m", + "\027[34m%s\027[m", + "\027[34m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", +} + +local function colorize_text(s) + return s +end + +local function colorize_ansi(s, t, extra) + local out = format(colortype_ansi[t], s) + if extra then out = "\027[3m"..out end + return out +end + +local irtype_ansi = setmetatable({}, + { __index = function(tab, t) + local s = colorize_ansi(irtype_text[t], t); tab[t] = s; return s; end }) + +local html_escape = { ["<"] = "<", [">"] = ">", ["&"] = "&", } + +local function colorize_html(s, t, extra) + s = gsub(s, "[<>&]", html_escape) + return format('%s', + irtype_text[t], extra and " irt_extra" or "", s) +end + +local irtype_html = setmetatable({}, + { __index = function(tab, t) + local s = colorize_html(irtype_text[t], t); tab[t] = s; return s; end }) + +local header_html = [[ + +]] + +local colorize, irtype + +-- Lookup tables to convert some literals into names. +local litname = { + ["SLOAD "] = setmetatable({}, { __index = function(t, mode) + local s = "" + if band(mode, 1) ~= 0 then s = s.."P" end + if band(mode, 2) ~= 0 then s = s.."F" end + if band(mode, 4) ~= 0 then s = s.."T" end + if band(mode, 8) ~= 0 then s = s.."C" end + if band(mode, 16) ~= 0 then s = s.."R" end + if band(mode, 32) ~= 0 then s = s.."I" end + if band(mode, 64) ~= 0 then s = s.."K" end + t[mode] = s + return s + end}), + ["XLOAD "] = { [0] = "", "R", "V", "RV", "U", "RU", "VU", "RVU", }, + ["CONV "] = setmetatable({}, { __index = function(t, mode) + local s = irtype[band(mode, 31)] + s = irtype[band(shr(mode, 5), 31)].."."..s + if band(mode, 0x800) ~= 0 then s = s.." sext" end + local c = shr(mode, 12) + if c == 1 then s = s.." none" + elseif c == 2 then s = s.." index" + elseif c == 3 then s = s.." check" end + t[mode] = s + return s + end}), + ["FLOAD "] = vmdef.irfield, + ["FREF "] = vmdef.irfield, + ["FPMATH"] = vmdef.irfpm, + ["TMPREF"] = { [0] = "", "IN", "OUT", "INOUT", "", "", "OUT2", "INOUT2" }, + ["BUFHDR"] = { [0] = "RESET", "APPEND", "WRITE" }, + ["TOSTR "] = { [0] = "INT", "NUM", "CHAR" }, +} + +local function ctlsub(c) + if c == "\n" then return "\\n" + elseif c == "\r" then return "\\r" + elseif c == "\t" then return "\\t" + else return format("\\%03d", byte(c)) + end +end + +local function fmtfunc(func, pc) + local fi = funcinfo(func, pc) + if fi.loc then + return fi.loc + elseif fi.ffid then + return vmdef.ffnames[fi.ffid] + elseif fi.addr then + return format("C:%x", fi.addr) + else + return "(?)" + end +end + +local function formatk(tr, idx, sn) + local k, t, slot = tracek(tr, idx) + local tn = type(k) + local s + if tn == "number" then + if t < 12 then + s = k == 0 and "NULL" or format("[0x%08x]", k) + elseif band(sn or 0, 0x30000) ~= 0 then + s = band(sn, 0x20000) ~= 0 and "contpc" or "ftsz" + elseif k == 2^52+2^51 then + s = "bias" + else + s = format(0 < k and k < 0x1p-1026 and "%+a" or "%+.14g", k) + end + elseif tn == "string" then + s = format(#k > 20 and '"%.20s"~' or '"%s"', gsub(k, "%c", ctlsub)) + elseif tn == "function" then + s = fmtfunc(k) + elseif tn == "table" then + s = format("{%p}", k) + elseif tn == "userdata" then + if t == 12 then + s = format("userdata:%p", k) + else + s = format("[%p]", k) + if s == "[NULL]" then s = "NULL" end + end + elseif t == 21 then -- int64_t + s = sub(tostring(k), 1, -3) + if sub(s, 1, 1) ~= "-" then s = "+"..s end + elseif sn == 0x1057fff then -- SNAP(1, SNAP_FRAME | SNAP_NORESTORE, REF_NIL) + return "----" -- Special case for LJ_FR2 slot 1. + else + s = tostring(k) -- For primitives. + end + s = colorize(format("%-4s", s), t, band(sn or 0, 0x100000) ~= 0) + if slot then + s = format("%s @%d", s, slot) + end + return s +end + +local function printsnap(tr, snap) + local n = 2 + for s=0,snap[1]-1 do + local sn = snap[n] + if shr(sn, 24) == s then + n = n + 1 + local ref = band(sn, 0xffff) - 0x8000 -- REF_BIAS + if ref < 0 then + out:write(formatk(tr, ref, sn)) + elseif band(sn, 0x80000) ~= 0 then -- SNAP_SOFTFPNUM + out:write(colorize(format("%04d/%04d", ref, ref+1), 14)) + else + local m, ot, op1, op2 = traceir(tr, ref) + out:write(colorize(format("%04d", ref), band(ot, 31), band(sn, 0x100000) ~= 0)) + end + out:write(band(sn, 0x10000) == 0 and " " or "|") -- SNAP_FRAME + else + out:write("---- ") + end + end + out:write("]\n") +end + +-- Dump snapshots (not interleaved with IR). +local function dump_snap(tr) + out:write("---- TRACE ", tr, " snapshots\n") + for i=0,1000000000 do + local snap = tracesnap(tr, i) + if not snap then break end + out:write(format("#%-3d %04d [ ", i, snap[0])) + printsnap(tr, snap) + end +end + +-- Return a register name or stack slot for a rid/sp location. +local function ridsp_name(ridsp, ins) + if not disass then disass = require("jit.dis_"..jit.arch) end + local rid, slot = band(ridsp, 0xff), shr(ridsp, 8) + if rid == 253 or rid == 254 then + return (slot == 0 or slot == 255) and " {sink" or format(" {%04d", ins-slot) + end + if ridsp > 255 then return format("[%x]", slot*4) end + if rid < 128 then return disass.regname(rid) end + return "" +end + +-- Dump CALL* function ref and return optional ctype. +local function dumpcallfunc(tr, ins) + local ctype + if ins > 0 then + local m, ot, op1, op2 = traceir(tr, ins) + if band(ot, 31) == 0 then -- nil type means CARG(func, ctype). + ins = op1 + ctype = formatk(tr, op2) + end + end + if ins < 0 then + out:write(format("[0x%x](", tonumber((tracek(tr, ins))))) + else + out:write(format("%04d (", ins)) + end + return ctype +end + +-- Recursively gather CALL* args and dump them. +local function dumpcallargs(tr, ins) + if ins < 0 then + out:write(formatk(tr, ins)) + else + local m, ot, op1, op2 = traceir(tr, ins) + local oidx = 6*shr(ot, 8) + local op = sub(vmdef.irnames, oidx+1, oidx+6) + if op == "CARG " then + dumpcallargs(tr, op1) + if op2 < 0 then + out:write(" ", formatk(tr, op2)) + else + out:write(" ", format("%04d", op2)) + end + else + out:write(format("%04d", ins)) + end + end +end + +-- Dump IR and interleaved snapshots. +local function dump_ir(tr, dumpsnap, dumpreg) + local info = traceinfo(tr) + if not info then return end + local nins = info.nins + out:write("---- TRACE ", tr, " IR\n") + local irnames = vmdef.irnames + local snapref = 65536 + local snap, snapno + if dumpsnap then + snap = tracesnap(tr, 0) + snapref = snap[0] + snapno = 0 + end + for ins=1,nins do + if ins >= snapref then + if dumpreg then + out:write(format(".... SNAP #%-3d [ ", snapno)) + else + out:write(format(".... SNAP #%-3d [ ", snapno)) + end + printsnap(tr, snap) + snapno = snapno + 1 + snap = tracesnap(tr, snapno) + snapref = snap and snap[0] or 65536 + end + local m, ot, op1, op2, ridsp = traceir(tr, ins) + local oidx, t = 6*shr(ot, 8), band(ot, 31) + local op = sub(irnames, oidx+1, oidx+6) + if op == "LOOP " then + if dumpreg then + out:write(format("%04d ------------ LOOP ------------\n", ins)) + else + out:write(format("%04d ------ LOOP ------------\n", ins)) + end + elseif op ~= "NOP " and op ~= "CARG " and + (dumpreg or op ~= "RENAME") then + local rid = band(ridsp, 255) + if dumpreg then + out:write(format("%04d %-6s", ins, ridsp_name(ridsp, ins))) + else + out:write(format("%04d ", ins)) + end + out:write(format("%s%s %s %s ", + (rid == 254 or rid == 253) and "}" or + (band(ot, 128) == 0 and " " or ">"), + band(ot, 64) == 0 and " " or "+", + irtype[t], op)) + local m1, m2 = band(m, 3), band(m, 3*4) + if sub(op, 1, 4) == "CALL" then + local ctype + if m2 == 1*4 then -- op2 == IRMlit + out:write(format("%-10s (", vmdef.ircall[op2])) + else + ctype = dumpcallfunc(tr, op2) + end + if op1 ~= -1 then dumpcallargs(tr, op1) end + out:write(")") + if ctype then out:write(" ctype ", ctype) end + elseif op == "CNEW " and op2 == -1 then + out:write(formatk(tr, op1)) + elseif m1 ~= 3 then -- op1 != IRMnone + if op1 < 0 then + out:write(formatk(tr, op1)) + else + out:write(format(m1 == 0 and "%04d" or "#%-3d", op1)) + end + if m2 ~= 3*4 then -- op2 != IRMnone + if m2 == 1*4 then -- op2 == IRMlit + local litn = litname[op] + if litn and litn[op2] then + out:write(" ", litn[op2]) + elseif op == "UREFO " or op == "UREFC " then + out:write(format(" #%-3d", shr(op2, 8))) + else + out:write(format(" #%-3d", op2)) + end + elseif op2 < 0 then + out:write(" ", formatk(tr, op2)) + else + out:write(format(" %04d", op2)) + end + end + end + out:write("\n") + end + end + if snap then + if dumpreg then + out:write(format(".... SNAP #%-3d [ ", snapno)) + else + out:write(format(".... SNAP #%-3d [ ", snapno)) + end + printsnap(tr, snap) + end +end + +------------------------------------------------------------------------------ + +local recprefix = "" +local recdepth = 0 + +-- Format trace error message. +local function fmterr(err, info) + if type(err) == "number" then + if type(info) == "function" then info = fmtfunc(info) end + err = format(vmdef.traceerr[err], info) + end + return err +end + +-- Dump trace states. +local function dump_trace(what, tr, func, pc, otr, oex) + if what == "stop" or (what == "abort" and dumpmode.a) then + if dumpmode.i then dump_ir(tr, dumpmode.s, dumpmode.r and what == "stop") + elseif dumpmode.s then dump_snap(tr) end + if dumpmode.m then dump_mcode(tr) end + end + if what == "start" then + if dumpmode.H then out:write('
\n') end
+    out:write("---- TRACE ", tr, " ", what)
+    if otr then out:write(" ", otr, "/", oex == -1 and "stitch" or oex) end
+    out:write(" ", fmtfunc(func, pc), "\n")
+  elseif what == "stop" or what == "abort" then
+    out:write("---- TRACE ", tr, " ", what)
+    if what == "abort" then
+      out:write(" ", fmtfunc(func, pc), " -- ", fmterr(otr, oex), "\n")
+    else
+      local info = traceinfo(tr)
+      local link, ltype = info.link, info.linktype
+      if link == tr or link == 0 then
+	out:write(" -> ", ltype, "\n")
+      elseif ltype == "root" then
+	out:write(" -> ", link, "\n")
+      else
+	out:write(" -> ", link, " ", ltype, "\n")
+      end
+    end
+    if dumpmode.H then out:write("
\n\n") else out:write("\n") end + else + if what == "flush" then symtab, nexitsym = {}, 0 end + out:write("---- TRACE ", what, "\n\n") + end + out:flush() +end + +-- Dump recorded bytecode. +local function dump_record(tr, func, pc, depth) + if depth ~= recdepth then + recdepth = depth + recprefix = rep(" .", depth) + end + local line + if pc >= 0 then + line = bcline(func, pc, recprefix) + if dumpmode.H then line = gsub(line, "[<>&]", html_escape) end + else + line = "0000 "..recprefix.." FUNCC \n" + end + if pc <= 0 then + out:write(sub(line, 1, -2), " ; ", fmtfunc(func), "\n") + else + out:write(line) + end + if pc >= 0 and band(funcbc(func, pc), 0xff) < 16 then -- ORDER BC + out:write(bcline(func, pc+1, recprefix)) -- Write JMP for cond. + end +end + +------------------------------------------------------------------------------ + +local gpr64 = jit.arch:match("64") +local fprmips32 = jit.arch == "mips" or jit.arch == "mipsel" + +-- Dump taken trace exits. +local function dump_texit(tr, ex, ngpr, nfpr, ...) + out:write("---- TRACE ", tr, " exit ", ex, "\n") + if dumpmode.X then + local regs = {...} + if gpr64 then + for i=1,ngpr do + out:write(format(" %016x", regs[i])) + if i % 4 == 0 then out:write("\n") end + end + else + for i=1,ngpr do + out:write(" ", tohex(regs[i])) + if i % 8 == 0 then out:write("\n") end + end + end + if fprmips32 then + for i=1,nfpr,2 do + out:write(format(" %+17.14g", regs[ngpr+i])) + if i % 8 == 7 then out:write("\n") end + end + else + for i=1,nfpr do + out:write(format(" %+17.14g", regs[ngpr+i])) + if i % 4 == 0 then out:write("\n") end + end + end + end +end + +------------------------------------------------------------------------------ + +-- Detach dump handlers. +local function dumpoff() + if active then + active = false + jit.attach(dump_texit) + jit.attach(dump_record) + jit.attach(dump_trace) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach dump handlers. +local function dumpon(opt, outfile) + if active then dumpoff() end + + local term = os.getenv("TERM") + local colormode = (term and term:match("color") or os.getenv("COLORTERM")) and "A" or "T" + if opt then + opt = gsub(opt, "[TAH]", function(mode) colormode = mode; return ""; end) + end + + local m = { t=true, b=true, i=true, m=true, } + if opt and opt ~= "" then + local o = sub(opt, 1, 1) + if o ~= "+" and o ~= "-" then m = {} end + for i=1,#opt do m[sub(opt, i, i)] = (o ~= "-") end + end + dumpmode = m + + if m.t or m.b or m.i or m.s or m.m then + jit.attach(dump_trace, "trace") + end + if m.b then + jit.attach(dump_record, "record") + if not bcline then bcline = require("jit.bc").line end + end + if m.x or m.X then + jit.attach(dump_texit, "texit") + end + + if not outfile then outfile = os.getenv("LUAJIT_DUMPFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stdout + end + + m[colormode] = true + if colormode == "A" then + colorize = colorize_ansi + irtype = irtype_ansi + elseif colormode == "H" then + colorize = colorize_html + irtype = irtype_html + out:write(header_html) + else + colorize = colorize_text + irtype = irtype_text + end + + active = true +end + +-- Public module functions. +return { + on = dumpon, + off = dumpoff, + start = dumpon -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/p.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/p.lua new file mode 100644 index 00000000..c9ec1d8b --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/p.lua @@ -0,0 +1,312 @@ +---------------------------------------------------------------------------- +-- LuaJIT profiler. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module is a simple command line interface to the built-in +-- low-overhead profiler of LuaJIT. +-- +-- The lower-level API of the profiler is accessible via the "jit.profile" +-- module or the luaJIT_profile_* C API. +-- +-- Example usage: +-- +-- luajit -jp myapp.lua +-- luajit -jp=s myapp.lua +-- luajit -jp=-s myapp.lua +-- luajit -jp=vl myapp.lua +-- luajit -jp=G,profile.txt myapp.lua +-- +-- The following dump features are available: +-- +-- f Stack dump: function name, otherwise module:line. Default mode. +-- F Stack dump: ditto, but always prepend module. +-- l Stack dump: module:line. +-- stack dump depth (callee < caller). Default: 1. +-- - Inverse stack dump depth (caller > callee). +-- s Split stack dump after first stack level. Implies abs(depth) >= 2. +-- p Show full path for module names. +-- v Show VM states. Can be combined with stack dumps, e.g. vf or fv. +-- z Show zones. Can be combined with stack dumps, e.g. zf or fz. +-- r Show raw sample counts. Default: show percentages. +-- a Annotate excerpts from source code files. +-- A Annotate complete source code files. +-- G Produce raw output suitable for graphical tools (e.g. flame graphs). +-- m Minimum sample percentage to be shown. Default: 3. +-- i Sampling interval in milliseconds. Default: 10. +-- +---------------------------------------------------------------------------- + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local profile = require("jit.profile") +local vmdef = require("jit.vmdef") +local math = math +local pairs, ipairs, tonumber, floor = pairs, ipairs, tonumber, math.floor +local sort, format = table.sort, string.format +local stdout = io.stdout +local zone -- Load jit.zone module on demand. + +-- Output file handle. +local out + +------------------------------------------------------------------------------ + +local prof_ud +local prof_states, prof_split, prof_min, prof_raw, prof_fmt, prof_depth +local prof_ann, prof_count1, prof_count2, prof_samples + +local map_vmmode = { + N = "Compiled", + I = "Interpreted", + C = "C code", + G = "Garbage Collector", + J = "JIT Compiler", +} + +-- Profiler callback. +local function prof_cb(th, samples, vmmode) + prof_samples = prof_samples + samples + local key_stack, key_stack2, key_state + -- Collect keys for sample. + if prof_states then + if prof_states == "v" then + key_state = map_vmmode[vmmode] or vmmode + else + key_state = zone:get() or "(none)" + end + end + if prof_fmt then + key_stack = profile.dumpstack(th, prof_fmt, prof_depth) + key_stack = key_stack:gsub("%[builtin#(%d+)%]", function(x) + return vmdef.ffnames[tonumber(x)] + end) + if prof_split == 2 then + local k1, k2 = key_stack:match("(.-) [<>] (.*)") + if k2 then key_stack, key_stack2 = k1, k2 end + elseif prof_split == 3 then + key_stack2 = profile.dumpstack(th, "l", 1) + end + end + -- Order keys. + local k1, k2 + if prof_split == 1 then + if key_state then + k1 = key_state + if key_stack then k2 = key_stack end + end + elseif key_stack then + k1 = key_stack + if key_stack2 then k2 = key_stack2 elseif key_state then k2 = key_state end + end + -- Coalesce samples in one or two levels. + if k1 then + local t1 = prof_count1 + t1[k1] = (t1[k1] or 0) + samples + if k2 then + local t2 = prof_count2 + local t3 = t2[k1] + if not t3 then t3 = {}; t2[k1] = t3 end + t3[k2] = (t3[k2] or 0) + samples + end + end +end + +------------------------------------------------------------------------------ + +-- Show top N list. +local function prof_top(count1, count2, samples, indent) + local t, n = {}, 0 + for k in pairs(count1) do + n = n + 1 + t[n] = k + end + sort(t, function(a, b) return count1[a] > count1[b] end) + for i=1,n do + local k = t[i] + local v = count1[k] + local pct = floor(v*100/samples + 0.5) + if pct < prof_min then break end + if not prof_raw then + out:write(format("%s%2d%% %s\n", indent, pct, k)) + elseif prof_raw == "r" then + out:write(format("%s%5d %s\n", indent, v, k)) + else + out:write(format("%s %d\n", k, v)) + end + if count2 then + local r = count2[k] + if r then + prof_top(r, nil, v, (prof_split == 3 or prof_split == 1) and " -- " or + (prof_depth < 0 and " -> " or " <- ")) + end + end + end +end + +-- Annotate source code +local function prof_annotate(count1, samples) + local files = {} + local ms = 0 + for k, v in pairs(count1) do + local pct = floor(v*100/samples + 0.5) + ms = math.max(ms, v) + if pct >= prof_min then + local file, line = k:match("^(.*):(%d+)$") + if not file then file = k; line = 0 end + local fl = files[file] + if not fl then fl = {}; files[file] = fl; files[#files+1] = file end + line = tonumber(line) + fl[line] = prof_raw and v or pct + end + end + sort(files) + local fmtv, fmtn = " %3d%% | %s\n", " | %s\n" + if prof_raw then + local n = math.max(5, math.ceil(math.log10(ms))) + fmtv = "%"..n.."d | %s\n" + fmtn = (" "):rep(n).." | %s\n" + end + local ann = prof_ann + for _, file in ipairs(files) do + local f0 = file:byte() + if f0 == 40 or f0 == 91 then + out:write(format("\n====== %s ======\n[Cannot annotate non-file]\n", file)) + break + end + local fp, err = io.open(file) + if not fp then + out:write(format("====== ERROR: %s: %s\n", file, err)) + break + end + out:write(format("\n====== %s ======\n", file)) + local fl = files[file] + local n, show = 1, false + if ann ~= 0 then + for i=1,ann do + if fl[i] then show = true; out:write("@@ 1 @@\n"); break end + end + end + for line in fp:lines() do + if line:byte() == 27 then + out:write("[Cannot annotate bytecode file]\n") + break + end + local v = fl[n] + if ann ~= 0 then + local v2 = fl[n+ann] + if show then + if v2 then show = n+ann elseif v then show = n + elseif show+ann < n then show = false end + elseif v2 then + show = n+ann + out:write(format("@@ %d @@\n", n)) + end + if not show then goto next end + end + if v then + out:write(format(fmtv, v, line)) + else + out:write(format(fmtn, line)) + end + ::next:: + n = n + 1 + end + fp:close() + end +end + +------------------------------------------------------------------------------ + +-- Finish profiling and dump result. +local function prof_finish() + if prof_ud then + profile.stop() + local samples = prof_samples + if samples == 0 then + if prof_raw ~= true then out:write("[No samples collected]\n") end + return + end + if prof_ann then + prof_annotate(prof_count1, samples) + else + prof_top(prof_count1, prof_count2, samples, "") + end + prof_count1 = nil + prof_count2 = nil + prof_ud = nil + if out ~= stdout then out:close() end + end +end + +-- Start profiling. +local function prof_start(mode) + local interval = "" + mode = mode:gsub("i%d*", function(s) interval = s; return "" end) + prof_min = 3 + mode = mode:gsub("m(%d+)", function(s) prof_min = tonumber(s); return "" end) + prof_depth = 1 + mode = mode:gsub("%-?%d+", function(s) prof_depth = tonumber(s); return "" end) + local m = {} + for c in mode:gmatch(".") do m[c] = c end + prof_states = m.z or m.v + if prof_states == "z" then zone = require("jit.zone") end + local scope = m.l or m.f or m.F or (prof_states and "" or "f") + local flags = (m.p or "") + prof_raw = m.r + if m.s then + prof_split = 2 + if prof_depth == -1 or m["-"] then prof_depth = -2 + elseif prof_depth == 1 then prof_depth = 2 end + elseif mode:find("[fF].*l") then + scope = "l" + prof_split = 3 + else + prof_split = (scope == "" or mode:find("[zv].*[lfF]")) and 1 or 0 + end + prof_ann = m.A and 0 or (m.a and 3) + if prof_ann then + scope = "l" + prof_fmt = "pl" + prof_split = 0 + prof_depth = 1 + elseif m.G and scope ~= "" then + prof_fmt = flags..scope.."Z;" + prof_depth = -100 + prof_raw = true + prof_min = 0 + elseif scope == "" then + prof_fmt = false + else + local sc = prof_split == 3 and m.f or m.F or scope + prof_fmt = flags..sc..(prof_depth >= 0 and "Z < " or "Z > ") + end + prof_count1 = {} + prof_count2 = {} + prof_samples = 0 + profile.start(scope:lower()..interval, prof_cb) + prof_ud = newproxy(true) + getmetatable(prof_ud).__gc = prof_finish +end + +------------------------------------------------------------------------------ + +local function start(mode, outfile) + if not outfile then outfile = os.getenv("LUAJIT_PROFILEFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stdout + end + prof_start(mode or "f") +end + +-- Public module functions. +return { + start = start, -- For -j command line option. + stop = prof_finish +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/v.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/v.lua new file mode 100644 index 00000000..83589143 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/v.lua @@ -0,0 +1,170 @@ +---------------------------------------------------------------------------- +-- Verbose mode of the LuaJIT compiler. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module shows verbose information about the progress of the +-- JIT compiler. It prints one line for each generated trace. This module +-- is useful to see which code has been compiled or where the compiler +-- punts and falls back to the interpreter. +-- +-- Example usage: +-- +-- luajit -jv -e "for i=1,1000 do for j=1,1000 do end end" +-- luajit -jv=myapp.out myapp.lua +-- +-- Default output is to stderr. To redirect the output to a file, pass a +-- filename as an argument (use '-' for stdout) or set the environment +-- variable LUAJIT_VERBOSEFILE. The file is overwritten every time the +-- module is started. +-- +-- The output from the first example should look like this: +-- +-- [TRACE 1 (command line):1 loop] +-- [TRACE 2 (1/3) (command line):1 -> 1] +-- +-- The first number in each line is the internal trace number. Next are +-- the file name ('(command line)') and the line number (':1') where the +-- trace has started. Side traces also show the parent trace number and +-- the exit number where they are attached to in parentheses ('(1/3)'). +-- An arrow at the end shows where the trace links to ('-> 1'), unless +-- it loops to itself. +-- +-- In this case the inner loop gets hot and is traced first, generating +-- a root trace. Then the last exit from the 1st trace gets hot, too, +-- and triggers generation of the 2nd trace. The side trace follows the +-- path along the outer loop and *around* the inner loop, back to its +-- start, and then links to the 1st trace. Yes, this may seem unusual, +-- if you know how traditional compilers work. Trace compilers are full +-- of surprises like this -- have fun! :-) +-- +-- Aborted traces are shown like this: +-- +-- [TRACE --- foo.lua:44 -- leaving loop in root trace at foo:lua:50] +-- +-- Don't worry -- trace aborts are quite common, even in programs which +-- can be fully compiled. The compiler may retry several times until it +-- finds a suitable trace. +-- +-- Of course this doesn't work with features that are not-yet-implemented +-- (NYI error messages). The VM simply falls back to the interpreter. This +-- may not matter at all if the particular trace is not very high up in +-- the CPU usage profile. Oh, and the interpreter is quite fast, too. +-- +-- Also check out the -jdump module, which prints all the gory details. +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local funcinfo, traceinfo = jutil.funcinfo, jutil.traceinfo +local type, format = type, string.format +local stdout, stderr = io.stdout, io.stderr + +-- Active flag and output file handle. +local active, out + +------------------------------------------------------------------------------ + +local startloc, startex + +local function fmtfunc(func, pc) + local fi = funcinfo(func, pc) + if fi.loc then + return fi.loc + elseif fi.ffid then + return vmdef.ffnames[fi.ffid] + elseif fi.addr then + return format("C:%x", fi.addr) + else + return "(?)" + end +end + +-- Format trace error message. +local function fmterr(err, info) + if type(err) == "number" then + if type(info) == "function" then info = fmtfunc(info) end + err = format(vmdef.traceerr[err], info) + end + return err +end + +-- Dump trace states. +local function dump_trace(what, tr, func, pc, otr, oex) + if what == "start" then + startloc = fmtfunc(func, pc) + startex = otr and "("..otr.."/"..(oex == -1 and "stitch" or oex)..") " or "" + else + if what == "abort" then + local loc = fmtfunc(func, pc) + if loc ~= startloc then + out:write(format("[TRACE --- %s%s -- %s at %s]\n", + startex, startloc, fmterr(otr, oex), loc)) + else + out:write(format("[TRACE --- %s%s -- %s]\n", + startex, startloc, fmterr(otr, oex))) + end + elseif what == "stop" then + local info = traceinfo(tr) + local link, ltype = info.link, info.linktype + if ltype == "interpreter" then + out:write(format("[TRACE %3s %s%s -- fallback to interpreter]\n", + tr, startex, startloc)) + elseif ltype == "stitch" then + out:write(format("[TRACE %3s %s%s %s %s]\n", + tr, startex, startloc, ltype, fmtfunc(func, pc))) + elseif link == tr or link == 0 then + out:write(format("[TRACE %3s %s%s %s]\n", + tr, startex, startloc, ltype)) + elseif ltype == "root" then + out:write(format("[TRACE %3s %s%s -> %d]\n", + tr, startex, startloc, link)) + else + out:write(format("[TRACE %3s %s%s -> %d %s]\n", + tr, startex, startloc, link, ltype)) + end + else + out:write(format("[TRACE %s]\n", what)) + end + out:flush() + end +end + +------------------------------------------------------------------------------ + +-- Detach dump handlers. +local function dumpoff() + if active then + active = false + jit.attach(dump_trace) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach dump handlers. +local function dumpon(outfile) + if active then dumpoff() end + if not outfile then outfile = os.getenv("LUAJIT_VERBOSEFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stderr + end + jit.attach(dump_trace, "trace") + active = true +end + +-- Public module functions. +return { + on = dumpon, + off = dumpoff, + start = dumpon -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/vmdef.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/vmdef.lua new file mode 100644 index 00000000..68ff7ba2 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/vmdef.lua @@ -0,0 +1,395 @@ +-- This is a generated file. DO NOT EDIT! + +return { + +bcnames = "ISLT ISGE ISLE ISGT ISEQV ISNEV ISEQS ISNES ISEQN ISNEN ISEQP ISNEP ISTC ISFC IST ISF ISTYPEISNUM MOV NOT UNM LEN ADDVN SUBVN MULVN DIVVN MODVN ADDNV SUBNV MULNV DIVNV MODNV ADDVV SUBVV MULVV DIVVV MODVV POW CAT KSTR KCDATAKSHORTKNUM KPRI KNIL UGET USETV USETS USETN USETP UCLO FNEW TNEW TDUP GGET GSET TGETV TGETS TGETB TGETR TSETV TSETS TSETB TSETM TSETR CALLM CALL CALLMTCALLT ITERC ITERN VARG ISNEXTRETM RET RET0 RET1 FORI JFORI FORL IFORL JFORL ITERL IITERLJITERLLOOP ILOOP JLOOP JMP FUNCF IFUNCFJFUNCFFUNCV IFUNCVJFUNCVFUNCC FUNCCW", + +irnames = "LT GE LE GT ULT UGE ULE UGT EQ NE ABC RETF NOP BASE PVAL GCSTEPHIOP LOOP USE PHI RENAMEPROF KPRI KINT KGC KPTR KKPTR KNULL KNUM KINT64KSLOT BNOT BSWAP BAND BOR BXOR BSHL BSHR BSAR BROL BROR ADD SUB MUL DIV MOD POW NEG ABS LDEXP MIN MAX FPMATHADDOV SUBOV MULOV AREF HREFK HREF NEWREFUREFO UREFC FREF TMPREFSTRREFLREF ALOAD HLOAD ULOAD FLOAD XLOAD SLOAD VLOAD ALEN ASTOREHSTOREUSTOREFSTOREXSTORESNEW XSNEW TNEW TDUP CNEW CNEWI BUFHDRBUFPUTBUFSTRTBAR OBAR XBAR CONV TOBIT TOSTR STRTO CALLN CALLA CALLL CALLS CALLXSCARG ", + +irfpm = { [0]="floor", "ceil", "trunc", "sqrt", "log", "log2", "other", }, + +irfield = { [0]="str.len", "func.env", "func.pc", "func.ffid", "thread.env", "tab.meta", "tab.array", "tab.node", "tab.asize", "tab.hmask", "tab.nomm", "udata.meta", "udata.udtype", "udata.file", "sbuf.w", "sbuf.e", "sbuf.b", "sbuf.l", "sbuf.ref", "sbuf.r", "cdata.ctypeid", "cdata.ptr", "cdata.int", "cdata.int64", "cdata.int64_4", }, + +ircall = { +[0]="lj_str_cmp", +"lj_str_find", +"lj_str_new", +"lj_strscan_num", +"lj_strfmt_int", +"lj_strfmt_num", +"lj_strfmt_char", +"lj_strfmt_putint", +"lj_strfmt_putnum", +"lj_strfmt_putquoted", +"lj_strfmt_putfxint", +"lj_strfmt_putfnum_int", +"lj_strfmt_putfnum_uint", +"lj_strfmt_putfnum", +"lj_strfmt_putfstr", +"lj_strfmt_putfchar", +"lj_buf_putmem", +"lj_buf_putstr", +"lj_buf_putchar", +"lj_buf_putstr_reverse", +"lj_buf_putstr_lower", +"lj_buf_putstr_upper", +"lj_buf_putstr_rep", +"lj_buf_puttab", +"lj_bufx_set", +"lj_bufx_more", +"lj_serialize_put", +"lj_serialize_get", +"lj_serialize_encode", +"lj_serialize_decode", +"lj_buf_tostr", +"lj_tab_new_ah", +"lj_tab_new1", +"lj_tab_dup", +"lj_tab_clear", +"lj_tab_newkey", +"lj_tab_keyindex", +"lj_vm_next", +"lj_tab_len", +"lj_tab_len_hint", +"lj_gc_step_jit", +"lj_gc_barrieruv", +"lj_mem_newgco", +"lj_prng_u64d", +"lj_vm_modi", +"log10", +"exp", +"sin", +"cos", +"tan", +"asin", +"acos", +"atan", +"sinh", +"cosh", +"tanh", +"fputc", +"fwrite", +"fflush", +"lj_vm_floor", +"lj_vm_ceil", +"lj_vm_trunc", +"sqrt", +"log", +"lj_vm_log2", +"lj_vm_powi", +"pow", +"atan2", +"ldexp", +"lj_vm_tobit", +"softfp_add", +"softfp_sub", +"softfp_mul", +"softfp_div", +"softfp_cmp", +"softfp_i2d", +"softfp_d2i", +"lj_vm_sfmin", +"lj_vm_sfmax", +"lj_vm_tointg", +"softfp_ui2d", +"softfp_f2d", +"softfp_d2ui", +"softfp_d2f", +"softfp_i2f", +"softfp_ui2f", +"softfp_f2i", +"softfp_f2ui", +"fp64_l2d", +"fp64_ul2d", +"fp64_l2f", +"fp64_ul2f", +"fp64_d2l", +"fp64_d2ul", +"fp64_f2l", +"fp64_f2ul", +"lj_carith_divi64", +"lj_carith_divu64", +"lj_carith_modi64", +"lj_carith_modu64", +"lj_carith_powi64", +"lj_carith_powu64", +"lj_cdata_newv", +"lj_cdata_setfin", +"strlen", +"memcpy", +"memset", +"lj_vm_errno", +"lj_carith_mul64", +"lj_carith_shl64", +"lj_carith_shr64", +"lj_carith_sar64", +"lj_carith_rol64", +"lj_carith_ror64", +}, + +traceerr = { +[0]="error thrown or hook called during recording", +"trace too short", +"trace too long", +"trace too deep", +"too many snapshots", +"blacklisted", +"retry recording", +"NYI: bytecode %d", +"leaving loop in root trace", +"inner loop in root trace", +"loop unroll limit reached", +"bad argument type", +"JIT compilation disabled for function", +"call unroll limit reached", +"down-recursion, restarting", +"NYI: unsupported variant of FastFunc %s", +"NYI: return to lower frame", +"store with nil or NaN key", +"missing metamethod", +"looping index lookup", +"NYI: mixed sparse/dense table", +"symbol not in cache", +"NYI: unsupported C type conversion", +"NYI: unsupported C function type", +"guard would always fail", +"too many PHIs", +"persistent type instability", +"failed to allocate mcode memory", +"machine code too long", +"hit mcode limit (retrying)", +"too many spill slots", +"inconsistent register allocation", +"NYI: cannot assemble IR instruction %d", +"NYI: PHI shuffling too complex", +"NYI: register coalescing too complex", +}, + +ffnames = { +[0]="Lua", +"C", +"assert", +"type", +"next", +"pairs", +"ipairs_aux", +"ipairs", +"getmetatable", +"setmetatable", +"getfenv", +"setfenv", +"rawget", +"rawset", +"rawequal", +"unpack", +"select", +"tonumber", +"tostring", +"error", +"pcall", +"xpcall", +"loadfile", +"load", +"loadstring", +"dofile", +"gcinfo", +"collectgarbage", +"newproxy", +"print", +"coroutine.status", +"coroutine.running", +"coroutine.isyieldable", +"coroutine.create", +"coroutine.yield", +"coroutine.resume", +"coroutine.wrap_aux", +"coroutine.wrap", +"math.abs", +"math.floor", +"math.ceil", +"math.sqrt", +"math.log10", +"math.exp", +"math.sin", +"math.cos", +"math.tan", +"math.asin", +"math.acos", +"math.atan", +"math.sinh", +"math.cosh", +"math.tanh", +"math.frexp", +"math.modf", +"math.log", +"math.atan2", +"math.pow", +"math.fmod", +"math.ldexp", +"math.min", +"math.max", +"math.random", +"math.randomseed", +"bit.tobit", +"bit.bnot", +"bit.bswap", +"bit.lshift", +"bit.rshift", +"bit.arshift", +"bit.rol", +"bit.ror", +"bit.band", +"bit.bor", +"bit.bxor", +"bit.tohex", +"string.byte", +"string.char", +"string.sub", +"string.rep", +"string.reverse", +"string.lower", +"string.upper", +"string.dump", +"string.find", +"string.match", +"string.gmatch_aux", +"string.gmatch", +"string.gsub", +"string.format", +"table.maxn", +"table.insert", +"table.concat", +"table.sort", +"table.new", +"table.clear", +"io.method.close", +"io.method.read", +"io.method.write", +"io.method.flush", +"io.method.seek", +"io.method.setvbuf", +"io.method.lines", +"io.method.__gc", +"io.method.__tostring", +"io.open", +"io.popen", +"io.tmpfile", +"io.close", +"io.read", +"io.write", +"io.flush", +"io.input", +"io.output", +"io.lines", +"io.type", +"os.execute", +"os.remove", +"os.rename", +"os.tmpname", +"os.getenv", +"os.exit", +"os.clock", +"os.date", +"os.time", +"os.difftime", +"os.setlocale", +"debug.getregistry", +"debug.getmetatable", +"debug.setmetatable", +"debug.getfenv", +"debug.setfenv", +"debug.getinfo", +"debug.getlocal", +"debug.setlocal", +"debug.getupvalue", +"debug.setupvalue", +"debug.upvalueid", +"debug.upvaluejoin", +"debug.sethook", +"debug.gethook", +"debug.debug", +"debug.traceback", +"jit.on", +"jit.off", +"jit.flush", +"jit.status", +"jit.security", +"jit.attach", +"jit.util.funcinfo", +"jit.util.funcbc", +"jit.util.funck", +"jit.util.funcuvname", +"jit.util.traceinfo", +"jit.util.traceir", +"jit.util.tracek", +"jit.util.tracesnap", +"jit.util.tracemc", +"jit.util.traceexitstub", +"jit.util.ircalladdr", +"jit.opt.start", +"jit.profile.start", +"jit.profile.stop", +"jit.profile.dumpstack", +"ffi.meta.__index", +"ffi.meta.__newindex", +"ffi.meta.__eq", +"ffi.meta.__len", +"ffi.meta.__lt", +"ffi.meta.__le", +"ffi.meta.__concat", +"ffi.meta.__call", +"ffi.meta.__add", +"ffi.meta.__sub", +"ffi.meta.__mul", +"ffi.meta.__div", +"ffi.meta.__mod", +"ffi.meta.__pow", +"ffi.meta.__unm", +"ffi.meta.__tostring", +"ffi.meta.__pairs", +"ffi.meta.__ipairs", +"ffi.clib.__index", +"ffi.clib.__newindex", +"ffi.clib.__gc", +"ffi.callback.free", +"ffi.callback.set", +"ffi.cdef", +"ffi.new", +"ffi.cast", +"ffi.typeof", +"ffi.typeinfo", +"ffi.istype", +"ffi.sizeof", +"ffi.alignof", +"ffi.offsetof", +"ffi.errno", +"ffi.string", +"ffi.copy", +"ffi.fill", +"ffi.abi", +"ffi.metatype", +"ffi.gc", +"ffi.load", +"buffer.method.free", +"buffer.method.reset", +"buffer.method.skip", +"buffer.method.set", +"buffer.method.put", +"buffer.method.putf", +"buffer.method.get", +"buffer.method.putcdata", +"buffer.method.reserve", +"buffer.method.commit", +"buffer.method.ref", +"buffer.method.encode", +"buffer.method.decode", +"buffer.method.__gc", +"buffer.method.__tostring", +"buffer.method.__len", +"buffer.new", +"buffer.encode", +"buffer.decode", +}, + +} + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/zone.lua b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/zone.lua new file mode 100644 index 00000000..94357854 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/jit/zone.lua @@ -0,0 +1,45 @@ +---------------------------------------------------------------------------- +-- LuaJIT profiler zones. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module implements a simple hierarchical zone model. +-- +-- Example usage: +-- +-- local zone = require("jit.zone") +-- zone("AI") +-- ... +-- zone("A*") +-- ... +-- print(zone:get()) --> "A*" +-- ... +-- zone() +-- ... +-- print(zone:get()) --> "AI" +-- ... +-- zone() +-- +---------------------------------------------------------------------------- + +local remove = table.remove + +return setmetatable({ + flush = function(t) + for i=#t,1,-1 do t[i] = nil end + end, + get = function(t) + return t[#t] + end +}, { + __call = function(t, zone) + if zone then + t[#t+1] = zone + else + return (assert(remove(t), "empty zone stack")) + end + end +}) + diff --git a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/libluajit.a b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/libluajit.a index 0d4e76a5..54b75396 100644 Binary files a/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/libluajit.a and b/love/src/jni/LuaJIT-2.1/android/armeabi-v7a/libluajit.a differ diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/bc.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/bc.lua new file mode 100644 index 00000000..e58a3fef --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/bc.lua @@ -0,0 +1,190 @@ +---------------------------------------------------------------------------- +-- LuaJIT bytecode listing module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module lists the bytecode of a Lua function. If it's loaded by -jbc +-- it hooks into the parser and lists all functions of a chunk as they +-- are parsed. +-- +-- Example usage: +-- +-- luajit -jbc -e 'local x=0; for i=1,1e6 do x=x+i end; print(x)' +-- luajit -jbc=- foo.lua +-- luajit -jbc=foo.list foo.lua +-- +-- Default output is to stderr. To redirect the output to a file, pass a +-- filename as an argument (use '-' for stdout) or set the environment +-- variable LUAJIT_LISTFILE. The file is overwritten every time the module +-- is started. +-- +-- This module can also be used programmatically: +-- +-- local bc = require("jit.bc") +-- +-- local function foo() print("hello") end +-- +-- bc.dump(foo) --> -- BYTECODE -- [...] +-- print(bc.line(foo, 2)) --> 0002 KSTR 1 1 ; "hello" +-- +-- local out = { +-- -- Do something with each line: +-- write = function(t, ...) io.write(...) end, +-- close = function(t) end, +-- flush = function(t) end, +-- } +-- bc.dump(foo, out) +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local bit = require("bit") +local sub, gsub, format = string.sub, string.gsub, string.format +local byte, band, shr = string.byte, bit.band, bit.rshift +local funcinfo, funcbc, funck = jutil.funcinfo, jutil.funcbc, jutil.funck +local funcuvname = jutil.funcuvname +local bcnames = vmdef.bcnames +local stdout, stderr = io.stdout, io.stderr + +------------------------------------------------------------------------------ + +local function ctlsub(c) + if c == "\n" then return "\\n" + elseif c == "\r" then return "\\r" + elseif c == "\t" then return "\\t" + else return format("\\%03d", byte(c)) + end +end + +-- Return one bytecode line. +local function bcline(func, pc, prefix) + local ins, m = funcbc(func, pc) + if not ins then return end + local ma, mb, mc = band(m, 7), band(m, 15*8), band(m, 15*128) + local a = band(shr(ins, 8), 0xff) + local oidx = 6*band(ins, 0xff) + local op = sub(bcnames, oidx+1, oidx+6) + local s = format("%04d %s %-6s %3s ", + pc, prefix or " ", op, ma == 0 and "" or a) + local d = shr(ins, 16) + if mc == 13*128 then -- BCMjump + return format("%s=> %04d\n", s, pc+d-0x7fff) + end + if mb ~= 0 then + d = band(d, 0xff) + elseif mc == 0 then + return s.."\n" + end + local kc + if mc == 10*128 then -- BCMstr + kc = funck(func, -d-1) + kc = format(#kc > 40 and '"%.40s"~' or '"%s"', gsub(kc, "%c", ctlsub)) + elseif mc == 9*128 then -- BCMnum + kc = funck(func, d) + if op == "TSETM " then kc = kc - 2^52 end + elseif mc == 12*128 then -- BCMfunc + local fi = funcinfo(funck(func, -d-1)) + if fi.ffid then + kc = vmdef.ffnames[fi.ffid] + else + kc = fi.loc + end + elseif mc == 5*128 then -- BCMuv + kc = funcuvname(func, d) + end + if ma == 5 then -- BCMuv + local ka = funcuvname(func, a) + if kc then kc = ka.." ; "..kc else kc = ka end + end + if mb ~= 0 then + local b = shr(ins, 24) + if kc then return format("%s%3d %3d ; %s\n", s, b, d, kc) end + return format("%s%3d %3d\n", s, b, d) + end + if kc then return format("%s%3d ; %s\n", s, d, kc) end + if mc == 7*128 and d > 32767 then d = d - 65536 end -- BCMlits + return format("%s%3d\n", s, d) +end + +-- Collect branch targets of a function. +local function bctargets(func) + local target = {} + for pc=1,1000000000 do + local ins, m = funcbc(func, pc) + if not ins then break end + if band(m, 15*128) == 13*128 then target[pc+shr(ins, 16)-0x7fff] = true end + end + return target +end + +-- Dump bytecode instructions of a function. +local function bcdump(func, out, all) + if not out then out = stdout end + local fi = funcinfo(func) + if all and fi.children then + for n=-1,-1000000000,-1 do + local k = funck(func, n) + if not k then break end + if type(k) == "proto" then bcdump(k, out, true) end + end + end + out:write(format("-- BYTECODE -- %s-%d\n", fi.loc, fi.lastlinedefined)) + local target = bctargets(func) + for pc=1,1000000000 do + local s = bcline(func, pc, target[pc] and "=>") + if not s then break end + out:write(s) + end + out:write("\n") + out:flush() +end + +------------------------------------------------------------------------------ + +-- Active flag and output file handle. +local active, out + +-- List handler. +local function h_list(func) + return bcdump(func, out) +end + +-- Detach list handler. +local function bclistoff() + if active then + active = false + jit.attach(h_list) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach list handler. +local function bcliston(outfile) + if active then bclistoff() end + if not outfile then outfile = os.getenv("LUAJIT_LISTFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stderr + end + jit.attach(h_list, "bc") + active = true +end + +-- Public module functions. +return { + line = bcline, + dump = bcdump, + targets = bctargets, + on = bcliston, + off = bclistoff, + start = bcliston -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/bcsave.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/bcsave.lua new file mode 100644 index 00000000..ab13667a --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/bcsave.lua @@ -0,0 +1,665 @@ +---------------------------------------------------------------------------- +-- LuaJIT module to save/list bytecode. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module saves or lists the bytecode for an input file. +-- It's run by the -b command line option. +-- +------------------------------------------------------------------------------ + +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local bit = require("bit") + +-- Symbol name prefix for LuaJIT bytecode. +local LJBC_PREFIX = "luaJIT_BC_" + +local type, assert = type, assert +local format = string.format +local tremove, tconcat = table.remove, table.concat + +------------------------------------------------------------------------------ + +local function usage() + io.stderr:write[[ +Save LuaJIT bytecode: luajit -b[options] input output + -l Only list bytecode. + -s Strip debug info (default). + -g Keep debug info. + -n name Set module name (default: auto-detect from input name). + -t type Set output file type (default: auto-detect from output name). + -a arch Override architecture for object files (default: native). + -o os Override OS for object files (default: native). + -e chunk Use chunk string as input. + -- Stop handling options. + - Use stdin as input and/or stdout as output. + +File types: c h obj o raw (default) +]] + os.exit(1) +end + +local function check(ok, ...) + if ok then return ok, ... end + io.stderr:write("luajit: ", ...) + io.stderr:write("\n") + os.exit(1) +end + +local function readfile(input) + if type(input) == "function" then return input end + if input == "-" then input = nil end + return check(loadfile(input)) +end + +local function savefile(name, mode) + if name == "-" then return io.stdout end + return check(io.open(name, mode)) +end + +------------------------------------------------------------------------------ + +local map_type = { + raw = "raw", c = "c", h = "h", o = "obj", obj = "obj", +} + +local map_arch = { + x86 = { e = "le", b = 32, m = 3, p = 0x14c, }, + x64 = { e = "le", b = 64, m = 62, p = 0x8664, }, + arm = { e = "le", b = 32, m = 40, p = 0x1c0, }, + arm64 = { e = "le", b = 64, m = 183, p = 0xaa64, }, + arm64be = { e = "be", b = 64, m = 183, }, + ppc = { e = "be", b = 32, m = 20, }, + mips = { e = "be", b = 32, m = 8, f = 0x50001006, }, + mipsel = { e = "le", b = 32, m = 8, f = 0x50001006, }, + mips64 = { e = "be", b = 64, m = 8, f = 0x80000007, }, + mips64el = { e = "le", b = 64, m = 8, f = 0x80000007, }, + mips64r6 = { e = "be", b = 64, m = 8, f = 0xa0000407, }, + mips64r6el = { e = "le", b = 64, m = 8, f = 0xa0000407, }, +} + +local map_os = { + linux = true, windows = true, osx = true, freebsd = true, netbsd = true, + openbsd = true, dragonfly = true, solaris = true, +} + +local function checkarg(str, map, err) + str = str:lower() + local s = check(map[str], "unknown ", err) + return type(s) == "string" and s or str +end + +local function detecttype(str) + local ext = str:lower():match("%.(%a+)$") + return map_type[ext] or "raw" +end + +local function checkmodname(str) + check(str:match("^[%w_.%-]+$"), "bad module name") + return str:gsub("[%.%-]", "_") +end + +local function detectmodname(str) + if type(str) == "string" then + local tail = str:match("[^/\\]+$") + if tail then str = tail end + local head = str:match("^(.*)%.[^.]*$") + if head then str = head end + str = str:match("^[%w_.%-]+") + else + str = nil + end + check(str, "cannot derive module name, use -n name") + return str:gsub("[%.%-]", "_") +end + +------------------------------------------------------------------------------ + +local function bcsave_tail(fp, output, s) + local ok, err = fp:write(s) + if ok and output ~= "-" then ok, err = fp:close() end + check(ok, "cannot write ", output, ": ", err) +end + +local function bcsave_raw(output, s) + local fp = savefile(output, "wb") + bcsave_tail(fp, output, s) +end + +local function bcsave_c(ctx, output, s) + local fp = savefile(output, "w") + if ctx.type == "c" then + fp:write(format([[ +#ifdef __cplusplus +extern "C" +#endif +#ifdef _WIN32 +__declspec(dllexport) +#endif +const unsigned char %s%s[] = { +]], LJBC_PREFIX, ctx.modname)) + else + fp:write(format([[ +#define %s%s_SIZE %d +static const unsigned char %s%s[] = { +]], LJBC_PREFIX, ctx.modname, #s, LJBC_PREFIX, ctx.modname)) + end + local t, n, m = {}, 0, 0 + for i=1,#s do + local b = tostring(string.byte(s, i)) + m = m + #b + 1 + if m > 78 then + fp:write(tconcat(t, ",", 1, n), ",\n") + n, m = 0, #b + 1 + end + n = n + 1 + t[n] = b + end + bcsave_tail(fp, output, tconcat(t, ",", 1, n).."\n};\n") +end + +local function bcsave_elfobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct { + uint8_t emagic[4], eclass, eendian, eversion, eosabi, eabiversion, epad[7]; + uint16_t type, machine; + uint32_t version; + uint32_t entry, phofs, shofs; + uint32_t flags; + uint16_t ehsize, phentsize, phnum, shentsize, shnum, shstridx; +} ELF32header; +typedef struct { + uint8_t emagic[4], eclass, eendian, eversion, eosabi, eabiversion, epad[7]; + uint16_t type, machine; + uint32_t version; + uint64_t entry, phofs, shofs; + uint32_t flags; + uint16_t ehsize, phentsize, phnum, shentsize, shnum, shstridx; +} ELF64header; +typedef struct { + uint32_t name, type, flags, addr, ofs, size, link, info, align, entsize; +} ELF32sectheader; +typedef struct { + uint32_t name, type; + uint64_t flags, addr, ofs, size; + uint32_t link, info; + uint64_t align, entsize; +} ELF64sectheader; +typedef struct { + uint32_t name, value, size; + uint8_t info, other; + uint16_t sectidx; +} ELF32symbol; +typedef struct { + uint32_t name; + uint8_t info, other; + uint16_t sectidx; + uint64_t value, size; +} ELF64symbol; +typedef struct { + ELF32header hdr; + ELF32sectheader sect[6]; + ELF32symbol sym[2]; + uint8_t space[4096]; +} ELF32obj; +typedef struct { + ELF64header hdr; + ELF64sectheader sect[6]; + ELF64symbol sym[2]; + uint8_t space[4096]; +} ELF64obj; +]] + local symname = LJBC_PREFIX..ctx.modname + local ai = assert(map_arch[ctx.arch]) + local is64, isbe = ai.b == 64, ai.e == "be" + + -- Handle different host/target endianess. + local function f32(x) return x end + local f16, fofs = f32, f32 + if ffi.abi("be") ~= isbe then + f32 = bit.bswap + function f16(x) return bit.rshift(bit.bswap(x), 16) end + if is64 then + local two32 = ffi.cast("int64_t", 2^32) + function fofs(x) return bit.bswap(x)*two32 end + else + fofs = f32 + end + end + + -- Create ELF object and fill in header. + local o = ffi.new(is64 and "ELF64obj" or "ELF32obj") + local hdr = o.hdr + if ctx.os == "bsd" or ctx.os == "other" then -- Determine native hdr.eosabi. + local bf = assert(io.open("/bin/ls", "rb")) + local bs = bf:read(9) + bf:close() + ffi.copy(o, bs, 9) + check(hdr.emagic[0] == 127, "no support for writing native object files") + else + hdr.emagic = "\127ELF" + hdr.eosabi = ({ freebsd=9, netbsd=2, openbsd=12, solaris=6 })[ctx.os] or 0 + end + hdr.eclass = is64 and 2 or 1 + hdr.eendian = isbe and 2 or 1 + hdr.eversion = 1 + hdr.type = f16(1) + hdr.machine = f16(ai.m) + hdr.flags = f32(ai.f or 0) + hdr.version = f32(1) + hdr.shofs = fofs(ffi.offsetof(o, "sect")) + hdr.ehsize = f16(ffi.sizeof(hdr)) + hdr.shentsize = f16(ffi.sizeof(o.sect[0])) + hdr.shnum = f16(6) + hdr.shstridx = f16(2) + + -- Fill in sections and symbols. + local sofs, ofs = ffi.offsetof(o, "space"), 1 + for i,name in ipairs{ + ".symtab", ".shstrtab", ".strtab", ".rodata", ".note.GNU-stack", + } do + local sect = o.sect[i] + sect.align = fofs(1) + sect.name = f32(ofs) + ffi.copy(o.space+ofs, name) + ofs = ofs + #name+1 + end + o.sect[1].type = f32(2) -- .symtab + o.sect[1].link = f32(3) + o.sect[1].info = f32(1) + o.sect[1].align = fofs(8) + o.sect[1].ofs = fofs(ffi.offsetof(o, "sym")) + o.sect[1].entsize = fofs(ffi.sizeof(o.sym[0])) + o.sect[1].size = fofs(ffi.sizeof(o.sym)) + o.sym[1].name = f32(1) + o.sym[1].sectidx = f16(4) + o.sym[1].size = fofs(#s) + o.sym[1].info = 17 + o.sect[2].type = f32(3) -- .shstrtab + o.sect[2].ofs = fofs(sofs) + o.sect[2].size = fofs(ofs) + o.sect[3].type = f32(3) -- .strtab + o.sect[3].ofs = fofs(sofs + ofs) + o.sect[3].size = fofs(#symname+2) + ffi.copy(o.space+ofs+1, symname) + ofs = ofs + #symname + 2 + o.sect[4].type = f32(1) -- .rodata + o.sect[4].flags = fofs(2) + o.sect[4].ofs = fofs(sofs + ofs) + o.sect[4].size = fofs(#s) + o.sect[5].type = f32(1) -- .note.GNU-stack + o.sect[5].ofs = fofs(sofs + ofs + #s) + + -- Write ELF object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, ffi.sizeof(o)-4096+ofs)) + bcsave_tail(fp, output, s) +end + +local function bcsave_peobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct { + uint16_t arch, nsects; + uint32_t time, symtabofs, nsyms; + uint16_t opthdrsz, flags; +} PEheader; +typedef struct { + char name[8]; + uint32_t vsize, vaddr, size, ofs, relocofs, lineofs; + uint16_t nreloc, nline; + uint32_t flags; +} PEsection; +typedef struct __attribute((packed)) { + union { + char name[8]; + uint32_t nameref[2]; + }; + uint32_t value; + int16_t sect; + uint16_t type; + uint8_t scl, naux; +} PEsym; +typedef struct __attribute((packed)) { + uint32_t size; + uint16_t nreloc, nline; + uint32_t cksum; + uint16_t assoc; + uint8_t comdatsel, unused[3]; +} PEsymaux; +typedef struct { + PEheader hdr; + PEsection sect[2]; + // Must be an even number of symbol structs. + PEsym sym0; + PEsymaux sym0aux; + PEsym sym1; + PEsymaux sym1aux; + PEsym sym2; + PEsym sym3; + uint32_t strtabsize; + uint8_t space[4096]; +} PEobj; +]] + local symname = LJBC_PREFIX..ctx.modname + local ai = assert(map_arch[ctx.arch]) + local is64 = ai.b == 64 + local symexport = " /EXPORT:"..symname..",DATA " + + -- The file format is always little-endian. Swap if the host is big-endian. + local function f32(x) return x end + local f16 = f32 + if ffi.abi("be") then + f32 = bit.bswap + function f16(x) return bit.rshift(bit.bswap(x), 16) end + end + + -- Create PE object and fill in header. + local o = ffi.new("PEobj") + local hdr = o.hdr + hdr.arch = f16(assert(ai.p)) + hdr.nsects = f16(2) + hdr.symtabofs = f32(ffi.offsetof(o, "sym0")) + hdr.nsyms = f32(6) + + -- Fill in sections and symbols. + o.sect[0].name = ".drectve" + o.sect[0].size = f32(#symexport) + o.sect[0].flags = f32(0x00100a00) + o.sym0.sect = f16(1) + o.sym0.scl = 3 + o.sym0.name = ".drectve" + o.sym0.naux = 1 + o.sym0aux.size = f32(#symexport) + o.sect[1].name = ".rdata" + o.sect[1].size = f32(#s) + o.sect[1].flags = f32(0x40300040) + o.sym1.sect = f16(2) + o.sym1.scl = 3 + o.sym1.name = ".rdata" + o.sym1.naux = 1 + o.sym1aux.size = f32(#s) + o.sym2.sect = f16(2) + o.sym2.scl = 2 + o.sym2.nameref[1] = f32(4) + o.sym3.sect = f16(-1) + o.sym3.scl = 2 + o.sym3.value = f32(1) + o.sym3.name = "@feat.00" -- Mark as SafeSEH compliant. + ffi.copy(o.space, symname) + local ofs = #symname + 1 + o.strtabsize = f32(ofs + 4) + o.sect[0].ofs = f32(ffi.offsetof(o, "space") + ofs) + ffi.copy(o.space + ofs, symexport) + ofs = ofs + #symexport + o.sect[1].ofs = f32(ffi.offsetof(o, "space") + ofs) + + -- Write PE object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, ffi.sizeof(o)-4096+ofs)) + bcsave_tail(fp, output, s) +end + +local function bcsave_machobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct +{ + uint32_t magic, cputype, cpusubtype, filetype, ncmds, sizeofcmds, flags; +} mach_header; +typedef struct +{ + mach_header; uint32_t reserved; +} mach_header_64; +typedef struct { + uint32_t cmd, cmdsize; + char segname[16]; + uint32_t vmaddr, vmsize, fileoff, filesize; + uint32_t maxprot, initprot, nsects, flags; +} mach_segment_command; +typedef struct { + uint32_t cmd, cmdsize; + char segname[16]; + uint64_t vmaddr, vmsize, fileoff, filesize; + uint32_t maxprot, initprot, nsects, flags; +} mach_segment_command_64; +typedef struct { + char sectname[16], segname[16]; + uint32_t addr, size; + uint32_t offset, align, reloff, nreloc, flags; + uint32_t reserved1, reserved2; +} mach_section; +typedef struct { + char sectname[16], segname[16]; + uint64_t addr, size; + uint32_t offset, align, reloff, nreloc, flags; + uint32_t reserved1, reserved2, reserved3; +} mach_section_64; +typedef struct { + uint32_t cmd, cmdsize, symoff, nsyms, stroff, strsize; +} mach_symtab_command; +typedef struct { + int32_t strx; + uint8_t type, sect; + int16_t desc; + uint32_t value; +} mach_nlist; +typedef struct { + uint32_t strx; + uint8_t type, sect; + uint16_t desc; + uint64_t value; +} mach_nlist_64; +typedef struct +{ + uint32_t magic, nfat_arch; +} mach_fat_header; +typedef struct +{ + uint32_t cputype, cpusubtype, offset, size, align; +} mach_fat_arch; +typedef struct { + struct { + mach_header hdr; + mach_segment_command seg; + mach_section sec; + mach_symtab_command sym; + } arch[1]; + mach_nlist sym_entry; + uint8_t space[4096]; +} mach_obj; +typedef struct { + struct { + mach_header_64 hdr; + mach_segment_command_64 seg; + mach_section_64 sec; + mach_symtab_command sym; + } arch[1]; + mach_nlist_64 sym_entry; + uint8_t space[4096]; +} mach_obj_64; +typedef struct { + mach_fat_header fat; + mach_fat_arch fat_arch[2]; + struct { + mach_header hdr; + mach_segment_command seg; + mach_section sec; + mach_symtab_command sym; + } arch[2]; + mach_nlist sym_entry; + uint8_t space[4096]; +} mach_fat_obj; +]] + local symname = '_'..LJBC_PREFIX..ctx.modname + local isfat, is64, align, mobj = false, false, 4, "mach_obj" + if ctx.arch == "x64" then + is64, align, mobj = true, 8, "mach_obj_64" + elseif ctx.arch == "arm" then + isfat, mobj = true, "mach_fat_obj" + elseif ctx.arch == "arm64" then + is64, align, isfat, mobj = true, 8, true, "mach_fat_obj" + else + check(ctx.arch == "x86", "unsupported architecture for OSX") + end + local function aligned(v, a) return bit.band(v+a-1, -a) end + local be32 = bit.bswap -- Mach-O FAT is BE, supported archs are LE. + + -- Create Mach-O object and fill in header. + local o = ffi.new(mobj) + local mach_size = aligned(ffi.offsetof(o, "space")+#symname+2, align) + local cputype = ({ x86={7}, x64={0x01000007}, arm={7,12}, arm64={0x01000007,0x0100000c} })[ctx.arch] + local cpusubtype = ({ x86={3}, x64={3}, arm={3,9}, arm64={3,0} })[ctx.arch] + if isfat then + o.fat.magic = be32(0xcafebabe) + o.fat.nfat_arch = be32(#cpusubtype) + end + + -- Fill in sections and symbols. + for i=0,#cpusubtype-1 do + local ofs = 0 + if isfat then + local a = o.fat_arch[i] + a.cputype = be32(cputype[i+1]) + a.cpusubtype = be32(cpusubtype[i+1]) + -- Subsequent slices overlap each other to share data. + ofs = ffi.offsetof(o, "arch") + i*ffi.sizeof(o.arch[0]) + a.offset = be32(ofs) + a.size = be32(mach_size-ofs+#s) + end + local a = o.arch[i] + a.hdr.magic = is64 and 0xfeedfacf or 0xfeedface + a.hdr.cputype = cputype[i+1] + a.hdr.cpusubtype = cpusubtype[i+1] + a.hdr.filetype = 1 + a.hdr.ncmds = 2 + a.hdr.sizeofcmds = ffi.sizeof(a.seg)+ffi.sizeof(a.sec)+ffi.sizeof(a.sym) + a.seg.cmd = is64 and 0x19 or 0x1 + a.seg.cmdsize = ffi.sizeof(a.seg)+ffi.sizeof(a.sec) + a.seg.vmsize = #s + a.seg.fileoff = mach_size-ofs + a.seg.filesize = #s + a.seg.maxprot = 1 + a.seg.initprot = 1 + a.seg.nsects = 1 + ffi.copy(a.sec.sectname, "__data") + ffi.copy(a.sec.segname, "__DATA") + a.sec.size = #s + a.sec.offset = mach_size-ofs + a.sym.cmd = 2 + a.sym.cmdsize = ffi.sizeof(a.sym) + a.sym.symoff = ffi.offsetof(o, "sym_entry")-ofs + a.sym.nsyms = 1 + a.sym.stroff = ffi.offsetof(o, "sym_entry")+ffi.sizeof(o.sym_entry)-ofs + a.sym.strsize = aligned(#symname+2, align) + end + o.sym_entry.type = 0xf + o.sym_entry.sect = 1 + o.sym_entry.strx = 1 + ffi.copy(o.space+1, symname) + + -- Write Macho-O object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, mach_size)) + bcsave_tail(fp, output, s) +end + +local function bcsave_obj(ctx, output, s) + local ok, ffi = pcall(require, "ffi") + check(ok, "FFI library required to write this file type") + if ctx.os == "windows" then + return bcsave_peobj(ctx, output, s, ffi) + elseif ctx.os == "osx" then + return bcsave_machobj(ctx, output, s, ffi) + else + return bcsave_elfobj(ctx, output, s, ffi) + end +end + +------------------------------------------------------------------------------ + +local function bclist(input, output) + local f = readfile(input) + require("jit.bc").dump(f, savefile(output, "w"), true) +end + +local function bcsave(ctx, input, output) + local f = readfile(input) + local s = string.dump(f, ctx.strip) + local t = ctx.type + if not t then + t = detecttype(output) + ctx.type = t + end + if t == "raw" then + bcsave_raw(output, s) + else + if not ctx.modname then ctx.modname = detectmodname(input) end + if t == "obj" then + bcsave_obj(ctx, output, s) + else + bcsave_c(ctx, output, s) + end + end +end + +local function docmd(...) + local arg = {...} + local n = 1 + local list = false + local ctx = { + strip = true, arch = jit.arch, os = jit.os:lower(), + type = false, modname = false, + } + while n <= #arg do + local a = arg[n] + if type(a) == "string" and a:sub(1, 1) == "-" and a ~= "-" then + tremove(arg, n) + if a == "--" then break end + for m=2,#a do + local opt = a:sub(m, m) + if opt == "l" then + list = true + elseif opt == "s" then + ctx.strip = true + elseif opt == "g" then + ctx.strip = false + else + if arg[n] == nil or m ~= #a then usage() end + if opt == "e" then + if n ~= 1 then usage() end + arg[1] = check(loadstring(arg[1])) + elseif opt == "n" then + ctx.modname = checkmodname(tremove(arg, n)) + elseif opt == "t" then + ctx.type = checkarg(tremove(arg, n), map_type, "file type") + elseif opt == "a" then + ctx.arch = checkarg(tremove(arg, n), map_arch, "architecture") + elseif opt == "o" then + ctx.os = checkarg(tremove(arg, n), map_os, "OS name") + else + usage() + end + end + end + else + n = n + 1 + end + end + if list then + if #arg == 0 or #arg > 2 then usage() end + bclist(arg[1], arg[2] or "-") + else + if #arg ~= 2 then usage() end + bcsave(ctx, arg[1], arg[2]) + end +end + +------------------------------------------------------------------------------ + +-- Public module functions. +return { + start = docmd -- Process -b command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm.lua new file mode 100644 index 00000000..ba79c47e --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm.lua @@ -0,0 +1,689 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles most user-mode ARMv7 instructions +-- NYI: Advanced SIMD and VFP instructions. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch = string.match, string.gmatch +local concat = table.concat +local bit = require("bit") +local band, bor, ror, tohex = bit.band, bit.bor, bit.ror, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Opcode maps +------------------------------------------------------------------------------ + +local map_loadc = { + shift = 8, mask = 15, + [10] = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "vmovFmDN", "vstmFNdr", + _ = { + shift = 21, mask = 1, + [0] = "vstrFdl", + { shift = 16, mask = 15, [13] = "vpushFdr", _ = "vstmdbFNdr", } + }, + }, + { + shift = 23, mask = 3, + [0] = "vmovFDNm", + { shift = 16, mask = 15, [13] = "vpopFdr", _ = "vldmFNdr", }, + _ = { + shift = 21, mask = 1, + [0] = "vldrFdl", "vldmdbFNdr", + }, + }, + }, + [11] = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "vmovGmDN", "vstmGNdr", + _ = { + shift = 21, mask = 1, + [0] = "vstrGdl", + { shift = 16, mask = 15, [13] = "vpushGdr", _ = "vstmdbGNdr", } + }, + }, + { + shift = 23, mask = 3, + [0] = "vmovGDNm", + { shift = 16, mask = 15, [13] = "vpopGdr", _ = "vldmGNdr", }, + _ = { + shift = 21, mask = 1, + [0] = "vldrGdl", "vldmdbGNdr", + }, + }, + }, + _ = { + shift = 0, mask = 0 -- NYI ldc, mcrr, mrrc. + }, +} + +local map_vfps = { + shift = 6, mask = 0x2c001, + [0] = "vmlaF.dnm", "vmlsF.dnm", + [0x04000] = "vnmlsF.dnm", [0x04001] = "vnmlaF.dnm", + [0x08000] = "vmulF.dnm", [0x08001] = "vnmulF.dnm", + [0x0c000] = "vaddF.dnm", [0x0c001] = "vsubF.dnm", + [0x20000] = "vdivF.dnm", + [0x24000] = "vfnmsF.dnm", [0x24001] = "vfnmaF.dnm", + [0x28000] = "vfmaF.dnm", [0x28001] = "vfmsF.dnm", + [0x2c000] = "vmovF.dY", + [0x2c001] = { + shift = 7, mask = 0x1e01, + [0] = "vmovF.dm", "vabsF.dm", + [0x0200] = "vnegF.dm", [0x0201] = "vsqrtF.dm", + [0x0800] = "vcmpF.dm", [0x0801] = "vcmpeF.dm", + [0x0a00] = "vcmpzF.d", [0x0a01] = "vcmpzeF.d", + [0x0e01] = "vcvtG.dF.m", + [0x1000] = "vcvt.f32.u32Fdm", [0x1001] = "vcvt.f32.s32Fdm", + [0x1800] = "vcvtr.u32F.dm", [0x1801] = "vcvt.u32F.dm", + [0x1a00] = "vcvtr.s32F.dm", [0x1a01] = "vcvt.s32F.dm", + }, +} + +local map_vfpd = { + shift = 6, mask = 0x2c001, + [0] = "vmlaG.dnm", "vmlsG.dnm", + [0x04000] = "vnmlsG.dnm", [0x04001] = "vnmlaG.dnm", + [0x08000] = "vmulG.dnm", [0x08001] = "vnmulG.dnm", + [0x0c000] = "vaddG.dnm", [0x0c001] = "vsubG.dnm", + [0x20000] = "vdivG.dnm", + [0x24000] = "vfnmsG.dnm", [0x24001] = "vfnmaG.dnm", + [0x28000] = "vfmaG.dnm", [0x28001] = "vfmsG.dnm", + [0x2c000] = "vmovG.dY", + [0x2c001] = { + shift = 7, mask = 0x1e01, + [0] = "vmovG.dm", "vabsG.dm", + [0x0200] = "vnegG.dm", [0x0201] = "vsqrtG.dm", + [0x0800] = "vcmpG.dm", [0x0801] = "vcmpeG.dm", + [0x0a00] = "vcmpzG.d", [0x0a01] = "vcmpzeG.d", + [0x0e01] = "vcvtF.dG.m", + [0x1000] = "vcvt.f64.u32GdFm", [0x1001] = "vcvt.f64.s32GdFm", + [0x1800] = "vcvtr.u32FdG.m", [0x1801] = "vcvt.u32FdG.m", + [0x1a00] = "vcvtr.s32FdG.m", [0x1a01] = "vcvt.s32FdG.m", + }, +} + +local map_datac = { + shift = 24, mask = 1, + [0] = { + shift = 4, mask = 1, + [0] = { + shift = 8, mask = 15, + [10] = map_vfps, + [11] = map_vfpd, + -- NYI cdp, mcr, mrc. + }, + { + shift = 8, mask = 15, + [10] = { + shift = 20, mask = 15, + [0] = "vmovFnD", "vmovFDn", + [14] = "vmsrD", + [15] = { shift = 12, mask = 15, [15] = "vmrs", _ = "vmrsD", }, + }, + }, + }, + "svcT", +} + +local map_loadcu = { + shift = 0, mask = 0, -- NYI unconditional CP load/store. +} + +local map_datacu = { + shift = 0, mask = 0, -- NYI unconditional CP data. +} + +local map_simddata = { + shift = 0, mask = 0, -- NYI SIMD data. +} + +local map_simdload = { + shift = 0, mask = 0, -- NYI SIMD load/store, preload. +} + +local map_preload = { + shift = 0, mask = 0, -- NYI preload. +} + +local map_media = { + shift = 20, mask = 31, + [0] = false, + { --01 + shift = 5, mask = 7, + [0] = "sadd16DNM", "sasxDNM", "ssaxDNM", "ssub16DNM", + "sadd8DNM", false, false, "ssub8DNM", + }, + { --02 + shift = 5, mask = 7, + [0] = "qadd16DNM", "qasxDNM", "qsaxDNM", "qsub16DNM", + "qadd8DNM", false, false, "qsub8DNM", + }, + { --03 + shift = 5, mask = 7, + [0] = "shadd16DNM", "shasxDNM", "shsaxDNM", "shsub16DNM", + "shadd8DNM", false, false, "shsub8DNM", + }, + false, + { --05 + shift = 5, mask = 7, + [0] = "uadd16DNM", "uasxDNM", "usaxDNM", "usub16DNM", + "uadd8DNM", false, false, "usub8DNM", + }, + { --06 + shift = 5, mask = 7, + [0] = "uqadd16DNM", "uqasxDNM", "uqsaxDNM", "uqsub16DNM", + "uqadd8DNM", false, false, "uqsub8DNM", + }, + { --07 + shift = 5, mask = 7, + [0] = "uhadd16DNM", "uhasxDNM", "uhsaxDNM", "uhsub16DNM", + "uhadd8DNM", false, false, "uhsub8DNM", + }, + { --08 + shift = 5, mask = 7, + [0] = "pkhbtDNMU", false, "pkhtbDNMU", + { shift = 16, mask = 15, [15] = "sxtb16DMU", _ = "sxtab16DNMU", }, + "pkhbtDNMU", "selDNM", "pkhtbDNMU", + }, + false, + { --0a + shift = 5, mask = 7, + [0] = "ssatDxMu", "ssat16DxM", "ssatDxMu", + { shift = 16, mask = 15, [15] = "sxtbDMU", _ = "sxtabDNMU", }, + "ssatDxMu", false, "ssatDxMu", + }, + { --0b + shift = 5, mask = 7, + [0] = "ssatDxMu", "revDM", "ssatDxMu", + { shift = 16, mask = 15, [15] = "sxthDMU", _ = "sxtahDNMU", }, + "ssatDxMu", "rev16DM", "ssatDxMu", + }, + { --0c + shift = 5, mask = 7, + [3] = { shift = 16, mask = 15, [15] = "uxtb16DMU", _ = "uxtab16DNMU", }, + }, + false, + { --0e + shift = 5, mask = 7, + [0] = "usatDwMu", "usat16DwM", "usatDwMu", + { shift = 16, mask = 15, [15] = "uxtbDMU", _ = "uxtabDNMU", }, + "usatDwMu", false, "usatDwMu", + }, + { --0f + shift = 5, mask = 7, + [0] = "usatDwMu", "rbitDM", "usatDwMu", + { shift = 16, mask = 15, [15] = "uxthDMU", _ = "uxtahDNMU", }, + "usatDwMu", "revshDM", "usatDwMu", + }, + { --10 + shift = 12, mask = 15, + [15] = { + shift = 5, mask = 7, + "smuadNMS", "smuadxNMS", "smusdNMS", "smusdxNMS", + }, + _ = { + shift = 5, mask = 7, + [0] = "smladNMSD", "smladxNMSD", "smlsdNMSD", "smlsdxNMSD", + }, + }, + false, false, false, + { --14 + shift = 5, mask = 7, + [0] = "smlaldDNMS", "smlaldxDNMS", "smlsldDNMS", "smlsldxDNMS", + }, + { --15 + shift = 5, mask = 7, + [0] = { shift = 12, mask = 15, [15] = "smmulNMS", _ = "smmlaNMSD", }, + { shift = 12, mask = 15, [15] = "smmulrNMS", _ = "smmlarNMSD", }, + false, false, false, false, + "smmlsNMSD", "smmlsrNMSD", + }, + false, false, + { --18 + shift = 5, mask = 7, + [0] = { shift = 12, mask = 15, [15] = "usad8NMS", _ = "usada8NMSD", }, + }, + false, + { --1a + shift = 5, mask = 3, [2] = "sbfxDMvw", + }, + { --1b + shift = 5, mask = 3, [2] = "sbfxDMvw", + }, + { --1c + shift = 5, mask = 3, + [0] = { shift = 0, mask = 15, [15] = "bfcDvX", _ = "bfiDMvX", }, + }, + { --1d + shift = 5, mask = 3, + [0] = { shift = 0, mask = 15, [15] = "bfcDvX", _ = "bfiDMvX", }, + }, + { --1e + shift = 5, mask = 3, [2] = "ubfxDMvw", + }, + { --1f + shift = 5, mask = 3, [2] = "ubfxDMvw", + }, +} + +local map_load = { + shift = 21, mask = 9, + { + shift = 20, mask = 5, + [0] = "strtDL", "ldrtDL", [4] = "strbtDL", [5] = "ldrbtDL", + }, + _ = { + shift = 20, mask = 5, + [0] = "strDL", "ldrDL", [4] = "strbDL", [5] = "ldrbDL", + } +} + +local map_load1 = { + shift = 4, mask = 1, + [0] = map_load, map_media, +} + +local map_loadm = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "stmdaNR", "stmNR", + { shift = 16, mask = 63, [45] = "pushR", _ = "stmdbNR", }, "stmibNR", + }, + { + shift = 23, mask = 3, + [0] = "ldmdaNR", { shift = 16, mask = 63, [61] = "popR", _ = "ldmNR", }, + "ldmdbNR", "ldmibNR", + }, +} + +local map_data = { + shift = 21, mask = 15, + [0] = "andDNPs", "eorDNPs", "subDNPs", "rsbDNPs", + "addDNPs", "adcDNPs", "sbcDNPs", "rscDNPs", + "tstNP", "teqNP", "cmpNP", "cmnNP", + "orrDNPs", "movDPs", "bicDNPs", "mvnDPs", +} + +local map_mul = { + shift = 21, mask = 7, + [0] = "mulNMSs", "mlaNMSDs", "umaalDNMS", "mlsDNMS", + "umullDNMSs", "umlalDNMSs", "smullDNMSs", "smlalDNMSs", +} + +local map_sync = { + shift = 20, mask = 15, -- NYI: brackets around N. R(D+1) for ldrexd/strexd. + [0] = "swpDMN", false, false, false, + "swpbDMN", false, false, false, + "strexDMN", "ldrexDN", "strexdDN", "ldrexdDN", + "strexbDMN", "ldrexbDN", "strexhDN", "ldrexhDN", +} + +local map_mulh = { + shift = 21, mask = 3, + [0] = { shift = 5, mask = 3, + [0] = "smlabbNMSD", "smlatbNMSD", "smlabtNMSD", "smlattNMSD", }, + { shift = 5, mask = 3, + [0] = "smlawbNMSD", "smulwbNMS", "smlawtNMSD", "smulwtNMS", }, + { shift = 5, mask = 3, + [0] = "smlalbbDNMS", "smlaltbDNMS", "smlalbtDNMS", "smlalttDNMS", }, + { shift = 5, mask = 3, + [0] = "smulbbNMS", "smultbNMS", "smulbtNMS", "smulttNMS", }, +} + +local map_misc = { + shift = 4, mask = 7, + -- NYI: decode PSR bits of msr. + [0] = { shift = 21, mask = 1, [0] = "mrsD", "msrM", }, + { shift = 21, mask = 3, "bxM", false, "clzDM", }, + { shift = 21, mask = 3, "bxjM", }, + { shift = 21, mask = 3, "blxM", }, + false, + { shift = 21, mask = 3, [0] = "qaddDMN", "qsubDMN", "qdaddDMN", "qdsubDMN", }, + false, + { shift = 21, mask = 3, "bkptK", }, +} + +local map_datar = { + shift = 4, mask = 9, + [9] = { + shift = 5, mask = 3, + [0] = { shift = 24, mask = 1, [0] = map_mul, map_sync, }, + { shift = 20, mask = 1, [0] = "strhDL", "ldrhDL", }, + { shift = 20, mask = 1, [0] = "ldrdDL", "ldrsbDL", }, + { shift = 20, mask = 1, [0] = "strdDL", "ldrshDL", }, + }, + _ = { + shift = 20, mask = 25, + [16] = { shift = 7, mask = 1, [0] = map_misc, map_mulh, }, + _ = { + shift = 0, mask = 0xffffffff, + [bor(0xe1a00000)] = "nop", + _ = map_data, + } + }, +} + +local map_datai = { + shift = 20, mask = 31, -- NYI: decode PSR bits of msr. Decode imm12. + [16] = "movwDW", [20] = "movtDW", + [18] = { shift = 0, mask = 0xf00ff, [0] = "nopv6", _ = "msrNW", }, + [22] = "msrNW", + _ = map_data, +} + +local map_branch = { + shift = 24, mask = 1, + [0] = "bB", "blB" +} + +local map_condins = { + [0] = map_datar, map_datai, map_load, map_load1, + map_loadm, map_branch, map_loadc, map_datac +} + +-- NYI: setend. +local map_uncondins = { + [0] = false, map_simddata, map_simdload, map_preload, + false, "blxB", map_loadcu, map_datacu, +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "sp", "lr", "pc", +} + +local map_cond = { + [0] = "eq", "ne", "hs", "lo", "mi", "pl", "vs", "vc", + "hi", "ls", "ge", "lt", "gt", "le", "al", +} + +local map_shift = { [0] = "lsl", "lsr", "asr", "ror", } + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then + extra = "\t->"..sym + elseif band(ctx.op, 0x0e000000) ~= 0x0a000000 then + extra = "\t; 0x"..tohex(ctx.rel) + end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-5s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-5s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +-- Format operand 2 of load/store opcodes. +local function fmtload(ctx, op, pos) + local base = map_gpr[band(rshift(op, 16), 15)] + local x, ofs + local ext = (band(op, 0x04000000) == 0) + if not ext and band(op, 0x02000000) == 0 then + ofs = band(op, 4095) + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + ofs = "#"..ofs + elseif ext and band(op, 0x00400000) ~= 0 then + ofs = band(op, 15) + band(rshift(op, 4), 0xf0) + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + ofs = "#"..ofs + else + ofs = map_gpr[band(op, 15)] + if ext or band(op, 0xfe0) == 0 then + elseif band(op, 0xfe0) == 0x60 then + ofs = format("%s, rrx", ofs) + else + local sh = band(rshift(op, 7), 31) + if sh == 0 then sh = 32 end + ofs = format("%s, %s #%d", ofs, map_shift[band(rshift(op, 5), 3)], sh) + end + if band(op, 0x00800000) == 0 then ofs = "-"..ofs end + end + if ofs == "#0" then + x = format("[%s]", base) + elseif band(op, 0x01000000) == 0 then + x = format("[%s], %s", base, ofs) + else + x = format("[%s, %s]", base, ofs) + end + if band(op, 0x01200000) == 0x01200000 then x = x.."!" end + return x +end + +-- Format operand 2 of vector load/store opcodes. +local function fmtvload(ctx, op, pos) + local base = map_gpr[band(rshift(op, 16), 15)] + local ofs = band(op, 255)*4 + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + if ofs == 0 then + return format("[%s]", base) + else + return format("[%s, #%d]", base, ofs) + end +end + +local function fmtvr(op, vr, sh0, sh1) + if vr == "s" then + return format("s%d", 2*band(rshift(op, sh0), 15)+band(rshift(op, sh1), 1)) + else + return format("d%d", band(rshift(op, sh0), 15)+band(rshift(op, sh1-4), 16)) + end +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) + local operands = {} + local suffix = "" + local last, name, pat + local vr + ctx.op = op + ctx.rel = nil + + local cond = rshift(op, 28) + local opat + if cond == 15 then + opat = map_uncondins[band(rshift(op, 25), 7)] + else + if cond ~= 14 then suffix = map_cond[cond] end + opat = map_condins[band(rshift(op, 25), 7)] + end + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + name, pat = match(opat, "^([a-z0-9]*)(.*)") + if sub(pat, 1, 1) == "." then + local s2, p2 = match(pat, "^([a-z0-9.]*)(.*)") + suffix = suffix..s2 + pat = p2 + end + + for p in gmatch(pat, ".") do + local x = nil + if p == "D" then + x = map_gpr[band(rshift(op, 12), 15)] + elseif p == "N" then + x = map_gpr[band(rshift(op, 16), 15)] + elseif p == "S" then + x = map_gpr[band(rshift(op, 8), 15)] + elseif p == "M" then + x = map_gpr[band(op, 15)] + elseif p == "d" then + x = fmtvr(op, vr, 12, 22) + elseif p == "n" then + x = fmtvr(op, vr, 16, 7) + elseif p == "m" then + x = fmtvr(op, vr, 0, 5) + elseif p == "P" then + if band(op, 0x02000000) ~= 0 then + x = ror(band(op, 255), 2*band(rshift(op, 8), 15)) + else + x = map_gpr[band(op, 15)] + if band(op, 0xff0) ~= 0 then + operands[#operands+1] = x + local s = map_shift[band(rshift(op, 5), 3)] + local r = nil + if band(op, 0xf90) == 0 then + if s == "ror" then s = "rrx" else r = "#32" end + elseif band(op, 0x10) == 0 then + r = "#"..band(rshift(op, 7), 31) + else + r = map_gpr[band(rshift(op, 8), 15)] + end + if name == "mov" then name = s; x = r + elseif r then x = format("%s %s", s, r) + else x = s end + end + end + elseif p == "L" then + x = fmtload(ctx, op, pos) + elseif p == "l" then + x = fmtvload(ctx, op, pos) + elseif p == "B" then + local addr = ctx.addr + pos + 8 + arshift(lshift(op, 8), 6) + if cond == 15 then addr = addr + band(rshift(op, 23), 2) end + ctx.rel = addr + x = "0x"..tohex(addr) + elseif p == "F" then + vr = "s" + elseif p == "G" then + vr = "d" + elseif p == "." then + suffix = suffix..(vr == "s" and ".f32" or ".f64") + elseif p == "R" then + if band(op, 0x00200000) ~= 0 and #operands == 1 then + operands[1] = operands[1].."!" + end + local t = {} + for i=0,15 do + if band(rshift(op, i), 1) == 1 then t[#t+1] = map_gpr[i] end + end + x = "{"..concat(t, ", ").."}" + elseif p == "r" then + if band(op, 0x00200000) ~= 0 and #operands == 2 then + operands[1] = operands[1].."!" + end + local s = tonumber(sub(last, 2)) + local n = band(op, 255) + if vr == "d" then n = rshift(n, 1) end + operands[#operands] = format("{%s-%s%d}", last, vr, s+n-1) + elseif p == "W" then + x = band(op, 0x0fff) + band(rshift(op, 4), 0xf000) + elseif p == "T" then + x = "#0x"..tohex(band(op, 0x00ffffff), 6) + elseif p == "U" then + x = band(rshift(op, 7), 31) + if x == 0 then x = nil end + elseif p == "u" then + x = band(rshift(op, 7), 31) + if band(op, 0x40) == 0 then + if x == 0 then x = nil else x = "lsl #"..x end + else + if x == 0 then x = "asr #32" else x = "asr #"..x end + end + elseif p == "v" then + x = band(rshift(op, 7), 31) + elseif p == "w" then + x = band(rshift(op, 16), 31) + elseif p == "x" then + x = band(rshift(op, 16), 31) + 1 + elseif p == "X" then + x = band(rshift(op, 16), 31) - last + 1 + elseif p == "Y" then + x = band(rshift(op, 12), 0xf0) + band(op, 0x0f) + elseif p == "K" then + x = "#0x"..tohex(band(rshift(op, 4), 0x0000fff0) + band(op, 15), 4) + elseif p == "s" then + if band(op, 0x00100000) ~= 0 then suffix = "s"..suffix end + else + assert(false) + end + if x then + last = x + if type(x) == "number" then x = "#"..x end + operands[#operands+1] = x + end + end + + return putop(ctx, name..suffix, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ctx.pos = ofs + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 16 then return map_gpr[r] end + return "d"..(r-16) +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm64.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm64.lua new file mode 100644 index 00000000..ad909fbd --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm64.lua @@ -0,0 +1,1216 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM64 disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +-- +-- Contributed by Djordje Kovacevic and Stefan Pejic from RT-RK.com. +-- Sponsored by Cisco Systems, Inc. +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles most user-mode AArch64 instructions. +-- NYI: Advanced SIMD and VFP instructions. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local concat = table.concat +local bit = require("bit") +local band, bor, bxor, tohex = bit.band, bit.bor, bit.bxor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift +local ror = bit.ror + +------------------------------------------------------------------------------ +-- Opcode maps +------------------------------------------------------------------------------ + +local map_adr = { -- PC-relative addressing. + shift = 31, mask = 1, + [0] = "adrDBx", "adrpDBx" +} + +local map_addsubi = { -- Add/subtract immediate. + shift = 29, mask = 3, + [0] = "add|movDNIg", "adds|cmnD0NIg", "subDNIg", "subs|cmpD0NIg", +} + +local map_logi = { -- Logical immediate. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "andDNig", "orr|movDN0ig", "eorDNig", "ands|tstD0Nig" + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = "andDNig", "orr|movDN0ig", "eorDNig", "ands|tstD0Nig" + } +} + +local map_movwi = { -- Move wide immediate. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "movnDWRg", false, "movz|movDYRg", "movkDWRg" + }, false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = "movnDWRg", false, "movz|movDYRg", "movkDWRg" + }, +} + +local map_bitf = { -- Bitfield. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "sbfm|sbfiz|sbfx|asr|sxtw|sxth|sxtbDN12w", + "bfm|bfi|bfxilDN13w", + "ubfm|ubfiz|ubfx|lsr|lsl|uxth|uxtbDN12w" + } + }, + { + shift = 22, mask = 1, + { + shift = 29, mask = 3, + [0] = "sbfm|sbfiz|sbfx|asr|sxtw|sxth|sxtbDN12x", + "bfm|bfi|bfxilDN13x", + "ubfm|ubfiz|ubfx|lsr|lsl|uxth|uxtbDN12x" + } + } +} + +local map_datai = { -- Data processing - immediate. + shift = 23, mask = 7, + [0] = map_adr, map_adr, map_addsubi, false, + map_logi, map_movwi, map_bitf, + { + shift = 15, mask = 0x1c0c1, + [0] = "extr|rorDNM4w", [0x10080] = "extr|rorDNM4x", + [0x10081] = "extr|rorDNM4x" + } +} + +local map_logsr = { -- Logical, shifted register. + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = { + shift = 21, mask = 7, + [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg", + "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg" + }, + { + shift = 21, mask = 7, + [0] ="orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg", + "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg" + }, + { + shift = 21, mask = 7, + [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg", + "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg" + }, + { + shift = 21, mask = 7, + [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg", + "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg" + } + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = { + shift = 21, mask = 7, + [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg", + "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg" + }, + { + shift = 21, mask = 7, + [0] = "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg", + "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg" + }, + { + shift = 21, mask = 7, + [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg", + "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg" + }, + { + shift = 21, mask = 7, + [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg", + "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg" + } + } +} + +local map_assh = { + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "addDNMSg", "addDNMSg", "addDNMSg", "addDNMg" + }, + { + shift = 22, mask = 3, + [0] = "adds|cmnD0NMSg", "adds|cmnD0NMSg", + "adds|cmnD0NMSg", "adds|cmnD0NMg" + }, + { + shift = 22, mask = 3, + [0] = "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0Mg" + }, + { + shift = 22, mask = 3, + [0] = "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0MzSg", + "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0Mzg" + }, + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "addDNMSg", "addDNMSg", "addDNMSg", "addDNMg" + }, + { + shift = 22, mask = 3, + [0] = "adds|cmnD0NMSg", "adds|cmnD0NMSg", "adds|cmnD0NMSg", + "adds|cmnD0NMg" + }, + { + shift = 22, mask = 3, + [0] = "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0Mg" + }, + { + shift = 22, mask = 3, + [0] = "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0MzSg", + "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0Mzg" + } + } +} + +local map_addsubsh = { -- Add/subtract, shifted register. + shift = 22, mask = 3, + [0] = map_assh, map_assh, map_assh +} + +local map_addsubex = { -- Add/subtract, extended register. + shift = 22, mask = 3, + [0] = { + shift = 29, mask = 3, + [0] = "addDNMXg", "adds|cmnD0NMXg", "subDNMXg", "subs|cmpD0NMzXg", + } +} + +local map_addsubc = { -- Add/subtract, with carry. + shift = 10, mask = 63, + [0] = { + shift = 29, mask = 3, + [0] = "adcDNMg", "adcsDNMg", "sbc|ngcDN0Mg", "sbcs|ngcsDN0Mg", + } +} + +local map_ccomp = { + shift = 4, mask = 1, + [0] = { + shift = 10, mask = 3, + [0] = { -- Conditional compare register. + shift = 29, mask = 3, + "ccmnNMVCg", false, "ccmpNMVCg", + }, + [2] = { -- Conditional compare immediate. + shift = 29, mask = 3, + "ccmnN5VCg", false, "ccmpN5VCg", + } + } +} + +local map_csel = { -- Conditional select. + shift = 11, mask = 1, + [0] = { + shift = 10, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "cselDNMzCg", false, "csinv|cinv|csetmDNMcg", false, + }, + { + shift = 29, mask = 3, + [0] = "csinc|cinc|csetDNMcg", false, "csneg|cnegDNMcg", false, + } + } +} + +local map_data1s = { -- Data processing, 1 source. + shift = 29, mask = 1, + [0] = { + shift = 31, mask = 1, + [0] = { + shift = 10, mask = 0x7ff, + [0] = "rbitDNg", "rev16DNg", "revDNw", false, "clzDNg", "clsDNg" + }, + { + shift = 10, mask = 0x7ff, + [0] = "rbitDNg", "rev16DNg", "rev32DNx", "revDNx", "clzDNg", "clsDNg" + } + } +} + +local map_data2s = { -- Data processing, 2 sources. + shift = 29, mask = 1, + [0] = { + shift = 10, mask = 63, + false, "udivDNMg", "sdivDNMg", false, false, false, false, "lslDNMg", + "lsrDNMg", "asrDNMg", "rorDNMg" + } +} + +local map_data3s = { -- Data processing, 3 sources. + shift = 29, mask = 7, + [0] = { + shift = 21, mask = 7, + [0] = { + shift = 15, mask = 1, + [0] = "madd|mulDNMA0g", "msub|mnegDNMA0g" + } + }, false, false, false, + { + shift = 15, mask = 1, + [0] = { + shift = 21, mask = 7, + [0] = "madd|mulDNMA0g", "smaddl|smullDxNMwA0x", "smulhDNMx", false, + false, "umaddl|umullDxNMwA0x", "umulhDNMx" + }, + { + shift = 21, mask = 7, + [0] = "msub|mnegDNMA0g", "smsubl|smneglDxNMwA0x", false, false, + false, "umsubl|umneglDxNMwA0x" + } + } +} + +local map_datar = { -- Data processing, register. + shift = 28, mask = 1, + [0] = { + shift = 24, mask = 1, + [0] = map_logsr, + { + shift = 21, mask = 1, + [0] = map_addsubsh, map_addsubex + } + }, + { + shift = 21, mask = 15, + [0] = map_addsubc, false, map_ccomp, false, map_csel, false, + { + shift = 30, mask = 1, + [0] = map_data2s, map_data1s + }, + false, map_data3s, map_data3s, map_data3s, map_data3s, map_data3s, + map_data3s, map_data3s, map_data3s + } +} + +local map_lrl = { -- Load register, literal. + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = "ldrDwB", "ldrDxB", "ldrswDxB" + }, + { + shift = 30, mask = 3, + [0] = "ldrDsB", "ldrDdB" + } +} + +local map_lsriind = { -- Load/store register, immediate pre/post-indexed. + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwzL", "ldrbDwzL", "ldrsbDxzL", "ldrsbDwzL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strhDwzL", "ldrhDwzL", "ldrshDxzL", "ldrshDwzL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strDwzL", "ldrDwzL", "ldrswDxzL" + }, + { + shift = 22, mask = 3, + [0] = "strDszL", "ldrDszL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strDxzL", "ldrDxzL" + }, + { + shift = 22, mask = 3, + [0] = "strDdzL", "ldrDdzL" + } + } +} + +local map_lsriro = { + shift = 21, mask = 1, + [0] = { -- Load/store register immediate. + shift = 10, mask = 3, + [0] = { -- Unscaled immediate. + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "sturbDwK", "ldurbDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturhDwK", "ldurhDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturDwK", "ldurDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturDxK", "ldurDxK" + } + } + }, map_lsriind, false, map_lsriind + }, + { -- Load/store register, register offset. + shift = 10, mask = 3, + [2] = { + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwO", "ldrbDwO", "ldrsbDxO", "ldrsbDwO" + }, + { + shift = 22, mask = 3, + [0] = "strhDwO", "ldrhDwO", "ldrshDxO", "ldrshDwO" + }, + { + shift = 22, mask = 3, + [0] = "strDwO", "ldrDwO", "ldrswDxO" + }, + { + shift = 22, mask = 3, + [0] = "strDxO", "ldrDxO" + } + }, + { + shift = 30, mask = 3, + [2] = { + shift = 22, mask = 3, + [0] = "strDsO", "ldrDsO" + }, + [3] = { + shift = 22, mask = 3, + [0] = "strDdO", "ldrDdO" + } + } + } + } +} + +local map_lsp = { -- Load/store register pair, offset. + shift = 22, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = "stpDzAzwP", "stpDzAzsP", + }, + { + shift = 26, mask = 1, + "stpDzAzdP" + }, + { + shift = 26, mask = 1, + [0] = "stpDzAzxP" + } + }, + { + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = "ldpDzAzwP", "ldpDzAzsP", + }, + { + shift = 26, mask = 1, + [0] = "ldpswDAxP", "ldpDzAzdP" + }, + { + shift = 26, mask = 1, + [0] = "ldpDzAzxP" + } + } +} + +local map_ls = { -- Loads and stores. + shift = 24, mask = 0x31, + [0x10] = map_lrl, [0x30] = map_lsriro, + [0x20] = { + shift = 23, mask = 3, + map_lsp, map_lsp, map_lsp + }, + [0x21] = { + shift = 23, mask = 3, + map_lsp, map_lsp, map_lsp + }, + [0x31] = { + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwzU", "ldrbDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strhDwzU", "ldrhDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strDwzU", "ldrDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strDxzU", "ldrDxzU" + } + }, + { + shift = 30, mask = 3, + [2] = { + shift = 22, mask = 3, + [0] = "strDszU", "ldrDszU" + }, + [3] = { + shift = 22, mask = 3, + [0] = "strDdzU", "ldrDdzU" + } + } + }, +} + +local map_datafp = { -- Data processing, SIMD and FP. + shift = 28, mask = 7, + { -- 001 + shift = 24, mask = 1, + [0] = { + shift = 21, mask = 1, + { + shift = 10, mask = 3, + [0] = { + shift = 12, mask = 1, + [0] = { + shift = 13, mask = 1, + [0] = { + shift = 14, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { -- FP/int conversion. + shift = 31, mask = 1, + [0] = { + shift = 16, mask = 0xff, + [0x20] = "fcvtnsDwNs", [0x21] = "fcvtnuDwNs", + [0x22] = "scvtfDsNw", [0x23] = "ucvtfDsNw", + [0x24] = "fcvtasDwNs", [0x25] = "fcvtauDwNs", + [0x26] = "fmovDwNs", [0x27] = "fmovDsNw", + [0x28] = "fcvtpsDwNs", [0x29] = "fcvtpuDwNs", + [0x30] = "fcvtmsDwNs", [0x31] = "fcvtmuDwNs", + [0x38] = "fcvtzsDwNs", [0x39] = "fcvtzuDwNs", + [0x60] = "fcvtnsDwNd", [0x61] = "fcvtnuDwNd", + [0x62] = "scvtfDdNw", [0x63] = "ucvtfDdNw", + [0x64] = "fcvtasDwNd", [0x65] = "fcvtauDwNd", + [0x68] = "fcvtpsDwNd", [0x69] = "fcvtpuDwNd", + [0x70] = "fcvtmsDwNd", [0x71] = "fcvtmuDwNd", + [0x78] = "fcvtzsDwNd", [0x79] = "fcvtzuDwNd" + }, + { + shift = 16, mask = 0xff, + [0x20] = "fcvtnsDxNs", [0x21] = "fcvtnuDxNs", + [0x22] = "scvtfDsNx", [0x23] = "ucvtfDsNx", + [0x24] = "fcvtasDxNs", [0x25] = "fcvtauDxNs", + [0x28] = "fcvtpsDxNs", [0x29] = "fcvtpuDxNs", + [0x30] = "fcvtmsDxNs", [0x31] = "fcvtmuDxNs", + [0x38] = "fcvtzsDxNs", [0x39] = "fcvtzuDxNs", + [0x60] = "fcvtnsDxNd", [0x61] = "fcvtnuDxNd", + [0x62] = "scvtfDdNx", [0x63] = "ucvtfDdNx", + [0x64] = "fcvtasDxNd", [0x65] = "fcvtauDxNd", + [0x66] = "fmovDxNd", [0x67] = "fmovDdNx", + [0x68] = "fcvtpsDxNd", [0x69] = "fcvtpuDxNd", + [0x70] = "fcvtmsDxNd", [0x71] = "fcvtmuDxNd", + [0x78] = "fcvtzsDxNd", [0x79] = "fcvtzuDxNd" + } + } + }, + { -- FP data-processing, 1 source. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = { + shift = 15, mask = 63, + [0] = "fmovDNf", "fabsDNf", "fnegDNf", + "fsqrtDNf", false, "fcvtDdNs", false, false, + "frintnDNf", "frintpDNf", "frintmDNf", "frintzDNf", + "frintaDNf", false, "frintxDNf", "frintiDNf", + }, + { + shift = 15, mask = 63, + [0] = "fmovDNf", "fabsDNf", "fnegDNf", + "fsqrtDNf", "fcvtDsNd", false, false, false, + "frintnDNf", "frintpDNf", "frintmDNf", "frintzDNf", + "frintaDNf", false, "frintxDNf", "frintiDNf", + } + } + } + }, + { -- FP compare. + shift = 31, mask = 1, + [0] = { + shift = 14, mask = 3, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 0, mask = 31, + [0] = "fcmpNMf", [8] = "fcmpNZf", + [16] = "fcmpeNMf", [24] = "fcmpeNZf", + } + } + } + } + }, + { -- FP immediate. + shift = 31, mask = 1, + [0] = { + shift = 5, mask = 31, + [0] = { + shift = 23, mask = 1, + [0] = "fmovDFf" + } + } + } + }, + { -- FP conditional compare. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 4, mask = 1, + [0] = "fccmpNMVCf", "fccmpeNMVCf" + } + } + }, + { -- FP data-processing, 2 sources. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 12, mask = 15, + [0] = "fmulDNMf", "fdivDNMf", "faddDNMf", "fsubDNMf", + "fmaxDNMf", "fminDNMf", "fmaxnmDNMf", "fminnmDNMf", + "fnmulDNMf" + } + } + }, + { -- FP conditional select. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = "fcselDNMCf" + } + } + } + }, + { -- FP data-processing, 3 sources. + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 21, mask = 5, + [0] = "fmaddDNMAf", "fnmaddDNMAf" + }, + { + shift = 21, mask = 5, + [0] = "fmsubDNMAf", "fnmsubDNMAf" + } + } + } + } +} + +local map_br = { -- Branches, exception generating and system instructions. + shift = 29, mask = 7, + [0] = "bB", + { -- Compare & branch, immediate. + shift = 24, mask = 3, + [0] = "cbzDBg", "cbnzDBg", "tbzDTBw", "tbnzDTBw" + }, + { -- Conditional branch, immediate. + shift = 24, mask = 3, + [0] = { + shift = 4, mask = 1, + [0] = { + shift = 0, mask = 15, + [0] = "beqB", "bneB", "bhsB", "bloB", "bmiB", "bplB", "bvsB", "bvcB", + "bhiB", "blsB", "bgeB", "bltB", "bgtB", "bleB", "balB" + } + } + }, false, "blB", + { -- Compare & branch, immediate. + shift = 24, mask = 3, + [0] = "cbzDBg", "cbnzDBg", "tbzDTBx", "tbnzDTBx" + }, + { + shift = 24, mask = 3, + [0] = { -- Exception generation. + shift = 0, mask = 0xe0001f, + [0x200000] = "brkW" + }, + { -- System instructions. + shift = 0, mask = 0x3fffff, + [0x03201f] = "nop" + }, + { -- Unconditional branch, register. + shift = 0, mask = 0xfffc1f, + [0x1f0000] = "brNx", [0x3f0000] = "blrNx", + [0x5f0000] = "retNx" + }, + } +} + +local map_init = { + shift = 25, mask = 15, + [0] = false, false, false, false, map_ls, map_datar, map_ls, map_datafp, + map_datai, map_datai, map_br, map_br, map_ls, map_datar, map_ls, map_datafp +} + +------------------------------------------------------------------------------ + +local map_regs = { x = {}, w = {}, d = {}, s = {} } + +for i=0,30 do + map_regs.x[i] = "x"..i + map_regs.w[i] = "w"..i + map_regs.d[i] = "d"..i + map_regs.s[i] = "s"..i +end +map_regs.x[31] = "sp" +map_regs.w[31] = "wsp" +map_regs.d[31] = "d31" +map_regs.s[31] = "s31" + +local map_cond = { + [0] = "eq", "ne", "cs", "cc", "mi", "pl", "vs", "vc", + "hi", "ls", "ge", "lt", "gt", "le", "al", +} + +local map_shift = { [0] = "lsl", "lsr", "asr", } + +local map_extend = { + [0] = "uxtb", "uxth", "uxtw", "uxtx", "sxtb", "sxth", "sxtw", "sxtx", +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then + extra = "\t->"..sym + end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-5s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-5s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +local function match_reg(p, pat, regnum) + return map_regs[match(pat, p.."%w-([xwds])")][regnum] +end + +local function fmt_hex32(x) + if x < 0 then + return tohex(x) + else + return format("%x", x) + end +end + +local imm13_rep = { 0x55555555, 0x11111111, 0x01010101, 0x00010001, 0x00000001 } + +local function decode_imm13(op) + local imms = band(rshift(op, 10), 63) + local immr = band(rshift(op, 16), 63) + if band(op, 0x00400000) == 0 then + local len = 5 + if imms >= 56 then + if imms >= 60 then len = 1 else len = 2 end + elseif imms >= 48 then len = 3 elseif imms >= 32 then len = 4 end + local l = lshift(1, len)-1 + local s = band(imms, l) + local r = band(immr, l) + local imm = ror(rshift(-1, 31-s), r) + if len ~= 5 then imm = band(imm, lshift(1, l)-1) + rshift(imm, 31-l) end + imm = imm * imm13_rep[len] + local ix = fmt_hex32(imm) + if rshift(op, 31) ~= 0 then + return ix..tohex(imm) + else + return ix + end + else + local lo, hi = -1, 0 + if imms < 32 then lo = rshift(-1, 31-imms) else hi = rshift(-1, 63-imms) end + if immr ~= 0 then + lo, hi = ror(lo, immr), ror(hi, immr) + local x = immr == 32 and 0 or band(bxor(lo, hi), lshift(-1, 32-immr)) + lo, hi = bxor(lo, x), bxor(hi, x) + if immr >= 32 then lo, hi = hi, lo end + end + if hi ~= 0 then + return fmt_hex32(hi)..tohex(lo) + else + return fmt_hex32(lo) + end + end +end + +local function parse_immpc(op, name) + if name == "b" or name == "bl" then + return arshift(lshift(op, 6), 4) + elseif name == "adr" or name == "adrp" then + local immlo = band(rshift(op, 29), 3) + local immhi = lshift(arshift(lshift(op, 8), 13), 2) + return bor(immhi, immlo) + elseif name == "tbz" or name == "tbnz" then + return lshift(arshift(lshift(op, 13), 18), 2) + else + return lshift(arshift(lshift(op, 8), 13), 2) + end +end + +local function parse_fpimm8(op) + local sign = band(op, 0x100000) == 0 and 1 or -1 + local exp = bxor(rshift(arshift(lshift(op, 12), 5), 24), 0x80) - 131 + local frac = 16+band(rshift(op, 13), 15) + return sign * frac * 2^exp +end + +local function prefer_bfx(sf, uns, imms, immr) + if imms < immr or imms == 31 or imms == 63 then + return false + end + if immr == 0 then + if sf == 0 and (imms == 7 or imms == 15) then + return false + end + if sf ~= 0 and uns == 0 and (imms == 7 or imms == 15 or imms == 31) then + return false + end + end + return true +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) + local operands = {} + local suffix = "" + local last, name, pat + local map_reg + ctx.op = op + ctx.rel = nil + last = nil + local opat + opat = map_init[band(rshift(op, 25), 15)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + name, pat = match(opat, "^([a-z0-9]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.|]*)(.*)") + if altname then pat = pat2 end + if sub(pat, 1, 1) == "." then + local s2, p2 = match(pat, "^([a-z0-9.]*)(.*)") + suffix = suffix..s2 + pat = p2 + end + + local rt = match(pat, "[gf]") + if rt then + if rt == "g" then + map_reg = band(op, 0x80000000) ~= 0 and map_regs.x or map_regs.w + else + map_reg = band(op, 0x400000) ~= 0 and map_regs.d or map_regs.s + end + end + + local second0, immr + + for p in gmatch(pat, ".") do + local x = nil + if p == "D" then + local regnum = band(op, 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "N" then + local regnum = band(rshift(op, 5), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "M" then + local regnum = band(rshift(op, 16), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "A" then + local regnum = band(rshift(op, 10), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "B" then + local addr = ctx.addr + pos + parse_immpc(op, name) + ctx.rel = addr + x = "0x"..tohex(addr) + elseif p == "T" then + x = bor(band(rshift(op, 26), 32), band(rshift(op, 19), 31)) + elseif p == "V" then + x = band(op, 15) + elseif p == "C" then + x = map_cond[band(rshift(op, 12), 15)] + elseif p == "c" then + local rn = band(rshift(op, 5), 31) + local rm = band(rshift(op, 16), 31) + local cond = band(rshift(op, 12), 15) + local invc = bxor(cond, 1) + x = map_cond[cond] + if altname and cond ~= 14 and cond ~= 15 then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if rn == rm then + local n = #operands + operands[n] = nil + x = map_cond[invc] + if rn ~= 31 then + if a1 then name = a1 else name = altname end + else + operands[n-1] = nil + name = a2 + end + end + end + elseif p == "W" then + x = band(rshift(op, 5), 0xffff) + elseif p == "Y" then + x = band(rshift(op, 5), 0xffff) + local hw = band(rshift(op, 21), 3) + if altname and (hw == 0 or x ~= 0) then + name = altname + end + elseif p == "L" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local imm9 = arshift(lshift(op, 11), 23) + if band(op, 0x800) ~= 0 then + x = "["..rn..", #"..imm9.."]!" + else + x = "["..rn.."], #"..imm9 + end + elseif p == "U" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local sz = band(rshift(op, 30), 3) + local imm12 = lshift(arshift(lshift(op, 10), 20), sz) + if imm12 ~= 0 then + x = "["..rn..", #"..imm12.."]" + else + x = "["..rn.."]" + end + elseif p == "K" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local imm9 = arshift(lshift(op, 11), 23) + if imm9 ~= 0 then + x = "["..rn..", #"..imm9.."]" + else + x = "["..rn.."]" + end + elseif p == "O" then + local rn, rm = map_regs.x[band(rshift(op, 5), 31)] + local m = band(rshift(op, 13), 1) + if m == 0 then + rm = map_regs.w[band(rshift(op, 16), 31)] + else + rm = map_regs.x[band(rshift(op, 16), 31)] + end + x = "["..rn..", "..rm + local opt = band(rshift(op, 13), 7) + local s = band(rshift(op, 12), 1) + local sz = band(rshift(op, 30), 3) + -- extension to be applied + if opt == 3 then + if s == 0 then x = x.."]" + else x = x..", lsl #"..sz.."]" end + elseif opt == 2 or opt == 6 or opt == 7 then + if s == 0 then x = x..", "..map_extend[opt].."]" + else x = x..", "..map_extend[opt].." #"..sz.."]" end + else + x = x.."]" + end + elseif p == "P" then + local opcv, sh = rshift(op, 26), 2 + if opcv >= 0x2a then sh = 4 elseif opcv >= 0x1b then sh = 3 end + local imm7 = lshift(arshift(lshift(op, 10), 25), sh) + local rn = map_regs.x[band(rshift(op, 5), 31)] + local ind = band(rshift(op, 23), 3) + if ind == 1 then + x = "["..rn.."], #"..imm7 + elseif ind == 2 then + if imm7 == 0 then + x = "["..rn.."]" + else + x = "["..rn..", #"..imm7.."]" + end + elseif ind == 3 then + x = "["..rn..", #"..imm7.."]!" + end + elseif p == "I" then + local shf = band(rshift(op, 22), 3) + local imm12 = band(rshift(op, 10), 0x0fff) + local rn, rd = band(rshift(op, 5), 31), band(op, 31) + if altname == "mov" and shf == 0 and imm12 == 0 and (rn == 31 or rd == 31) then + name = altname + x = nil + elseif shf == 0 then + x = imm12 + elseif shf == 1 then + x = imm12..", lsl #12" + end + elseif p == "i" then + x = "#0x"..decode_imm13(op) + elseif p == "1" then + immr = band(rshift(op, 16), 63) + x = immr + elseif p == "2" then + x = band(rshift(op, 10), 63) + if altname then + local a1, a2, a3, a4, a5, a6 = + match(altname, "([^|]*)|([^|]*)|([^|]*)|([^|]*)|([^|]*)|(.*)") + local sf = band(rshift(op, 26), 32) + local uns = band(rshift(op, 30), 1) + if prefer_bfx(sf, uns, x, immr) then + name = a2 + x = x - immr + 1 + elseif immr == 0 and x == 7 then + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + name = a6 + x = nil + elseif immr == 0 and x == 15 then + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + name = a5 + x = nil + elseif x == 31 or x == 63 then + if x == 31 and immr == 0 and name == "sbfm" then + name = a4 + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + else + name = a3 + end + x = nil + elseif band(x, 31) ~= 31 and immr == x+1 and name == "ubfm" then + name = a4 + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = nil + elseif x < immr then + name = a1 + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = x + 1 + end + end + elseif p == "3" then + x = band(rshift(op, 10), 63) + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if x < immr then + name = a1 + local sf = band(rshift(op, 26), 32) + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = x + 1 + else + name = a2 + x = x - immr + 1 + end + end + elseif p == "4" then + x = band(rshift(op, 10), 63) + local rn = band(rshift(op, 5), 31) + local rm = band(rshift(op, 16), 31) + if altname and rn == rm then + local n = #operands + operands[n] = nil + last = operands[n-1] + name = altname + end + elseif p == "5" then + x = band(rshift(op, 16), 31) + elseif p == "S" then + x = band(rshift(op, 10), 63) + if x == 0 then x = nil + else x = map_shift[band(rshift(op, 22), 3)].." #"..x end + elseif p == "X" then + local opt = band(rshift(op, 13), 7) + -- Width specifier . + if opt ~= 3 and opt ~= 7 then + last = map_regs.w[band(rshift(op, 16), 31)] + operands[#operands] = last + end + x = band(rshift(op, 10), 7) + -- Extension. + if opt == 2 + band(rshift(op, 31), 1) and + band(rshift(op, second0 and 5 or 0), 31) == 31 then + if x == 0 then x = nil + else x = "lsl #"..x end + else + if x == 0 then x = map_extend[band(rshift(op, 13), 7)] + else x = map_extend[band(rshift(op, 13), 7)].." #"..x end + end + elseif p == "R" then + x = band(rshift(op,21), 3) + if x == 0 then x = nil + else x = "lsl #"..x*16 end + elseif p == "z" then + local n = #operands + if operands[n] == "sp" then operands[n] = "xzr" + elseif operands[n] == "wsp" then operands[n] = "wzr" + end + elseif p == "Z" then + x = 0 + elseif p == "F" then + x = parse_fpimm8(op) + elseif p == "g" or p == "f" or p == "x" or p == "w" or + p == "d" or p == "s" then + -- These are handled in D/N/M/A. + elseif p == "0" then + if last == "sp" or last == "wsp" then + local n = #operands + operands[n] = nil + last = operands[n-1] + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if not a1 then + name = altname + elseif second0 then + name, altname = a2, a1 + else + name, altname = a1, a2 + end + end + end + second0 = true + else + assert(false) + end + if x then + last = x + if type(x) == "number" then x = "#"..x end + operands[#operands+1] = x + end + end + + return putop(ctx, name..suffix, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ctx.pos = ofs + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_regs.x[r] end + return map_regs.d[r-32] +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm64be.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm64be.lua new file mode 100644 index 00000000..edcbffa8 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_arm64be.lua @@ -0,0 +1,12 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM64BE disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- ARM64 instructions are always little-endian. So just forward to the +-- common ARM64 disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +return require((string.match(..., ".*%.") or "").."dis_arm64") + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips.lua new file mode 100644 index 00000000..6ad17f54 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips.lua @@ -0,0 +1,694 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT/X license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles all standard MIPS32R1/R2 instructions. +-- Default mode is big-endian, but see: dis_mipsel.lua +------------------------------------------------------------------------------ + +local type = type +local byte, format = string.byte, string.format +local match, gmatch = string.match, string.gmatch +local concat = table.concat +local bit = require("bit") +local band, bor, tohex = bit.band, bit.bor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Extended opcode maps common to all MIPS releases +------------------------------------------------------------------------------ + +local map_srl = { shift = 21, mask = 1, [0] = "srlDTA", "rotrDTA", } +local map_srlv = { shift = 6, mask = 1, [0] = "srlvDTS", "rotrvDTS", } + +local map_cop0 = { + shift = 25, mask = 1, + [0] = { + shift = 21, mask = 15, + [0] = "mfc0TDW", [4] = "mtc0TDW", + [10] = "rdpgprDT", + [11] = { shift = 5, mask = 1, [0] = "diT0", "eiT0", }, + [14] = "wrpgprDT", + }, { + shift = 0, mask = 63, + [1] = "tlbr", [2] = "tlbwi", [6] = "tlbwr", [8] = "tlbp", + [24] = "eret", [31] = "deret", + [32] = "wait", + }, +} + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps for MIPS R1-R5 +------------------------------------------------------------------------------ + +local map_movci = { shift = 16, mask = 1, [0] = "movfDSC", "movtDSC", } + +local map_special = { + shift = 0, mask = 63, + [0] = { shift = 0, mask = -1, [0] = "nop", _ = "sllDTA" }, + map_movci, map_srl, "sraDTA", + "sllvDTS", false, map_srlv, "sravDTS", + "jrS", "jalrD1S", "movzDST", "movnDST", + "syscallY", "breakY", false, "sync", + "mfhiD", "mthiS", "mfloD", "mtloS", + "dsllvDST", false, "dsrlvDST", "dsravDST", + "multST", "multuST", "divST", "divuST", + "dmultST", "dmultuST", "ddivST", "ddivuST", + "addDST", "addu|moveDST0", "subDST", "subu|neguDS0T", + "andDST", "or|moveDST0", "xorDST", "nor|notDST0", + false, false, "sltDST", "sltuDST", + "daddDST", "dadduDST", "dsubDST", "dsubuDST", + "tgeSTZ", "tgeuSTZ", "tltSTZ", "tltuSTZ", + "teqSTZ", false, "tneSTZ", false, + "dsllDTA", false, "dsrlDTA", "dsraDTA", + "dsll32DTA", false, "dsrl32DTA", "dsra32DTA", +} + +local map_special2 = { + shift = 0, mask = 63, + [0] = "maddST", "madduST", "mulDST", false, + "msubST", "msubuST", + [32] = "clzDS", [33] = "cloDS", + [63] = "sdbbpY", +} + +local map_bshfl = { + shift = 6, mask = 31, + [2] = "wsbhDT", + [16] = "sebDT", + [24] = "sehDT", +} + +local map_dbshfl = { + shift = 6, mask = 31, + [2] = "dsbhDT", + [5] = "dshdDT", +} + +local map_special3 = { + shift = 0, mask = 63, + [0] = "extTSAK", [1] = "dextmTSAP", [3] = "dextTSAK", + [4] = "insTSAL", [6] = "dinsuTSEQ", [7] = "dinsTSAL", + [32] = map_bshfl, [36] = map_dbshfl, [59] = "rdhwrTD", +} + +local map_regimm = { + shift = 16, mask = 31, + [0] = "bltzSB", "bgezSB", "bltzlSB", "bgezlSB", + false, false, false, false, + "tgeiSI", "tgeiuSI", "tltiSI", "tltiuSI", + "teqiSI", false, "tneiSI", false, + "bltzalSB", "bgezalSB", "bltzallSB", "bgezallSB", + false, false, false, false, + false, false, false, false, + false, false, false, "synciSO", +} + +local map_cop1s = { + shift = 0, mask = 63, + [0] = "add.sFGH", "sub.sFGH", "mul.sFGH", "div.sFGH", + "sqrt.sFG", "abs.sFG", "mov.sFG", "neg.sFG", + "round.l.sFG", "trunc.l.sFG", "ceil.l.sFG", "floor.l.sFG", + "round.w.sFG", "trunc.w.sFG", "ceil.w.sFG", "floor.w.sFG", + false, + { shift = 16, mask = 1, [0] = "movf.sFGC", "movt.sFGC" }, + "movz.sFGT", "movn.sFGT", + false, "recip.sFG", "rsqrt.sFG", false, + false, false, false, false, + false, false, false, false, + false, "cvt.d.sFG", false, false, + "cvt.w.sFG", "cvt.l.sFG", "cvt.ps.sFGH", false, + false, false, false, false, + false, false, false, false, + "c.f.sVGH", "c.un.sVGH", "c.eq.sVGH", "c.ueq.sVGH", + "c.olt.sVGH", "c.ult.sVGH", "c.ole.sVGH", "c.ule.sVGH", + "c.sf.sVGH", "c.ngle.sVGH", "c.seq.sVGH", "c.ngl.sVGH", + "c.lt.sVGH", "c.nge.sVGH", "c.le.sVGH", "c.ngt.sVGH", +} + +local map_cop1d = { + shift = 0, mask = 63, + [0] = "add.dFGH", "sub.dFGH", "mul.dFGH", "div.dFGH", + "sqrt.dFG", "abs.dFG", "mov.dFG", "neg.dFG", + "round.l.dFG", "trunc.l.dFG", "ceil.l.dFG", "floor.l.dFG", + "round.w.dFG", "trunc.w.dFG", "ceil.w.dFG", "floor.w.dFG", + false, + { shift = 16, mask = 1, [0] = "movf.dFGC", "movt.dFGC" }, + "movz.dFGT", "movn.dFGT", + false, "recip.dFG", "rsqrt.dFG", false, + false, false, false, false, + false, false, false, false, + "cvt.s.dFG", false, false, false, + "cvt.w.dFG", "cvt.l.dFG", false, false, + false, false, false, false, + false, false, false, false, + "c.f.dVGH", "c.un.dVGH", "c.eq.dVGH", "c.ueq.dVGH", + "c.olt.dVGH", "c.ult.dVGH", "c.ole.dVGH", "c.ule.dVGH", + "c.df.dVGH", "c.ngle.dVGH", "c.deq.dVGH", "c.ngl.dVGH", + "c.lt.dVGH", "c.nge.dVGH", "c.le.dVGH", "c.ngt.dVGH", +} + +local map_cop1ps = { + shift = 0, mask = 63, + [0] = "add.psFGH", "sub.psFGH", "mul.psFGH", false, + false, "abs.psFG", "mov.psFG", "neg.psFG", + false, false, false, false, + false, false, false, false, + false, + { shift = 16, mask = 1, [0] = "movf.psFGC", "movt.psFGC" }, + "movz.psFGT", "movn.psFGT", + false, false, false, false, + false, false, false, false, + false, false, false, false, + "cvt.s.puFG", false, false, false, + false, false, false, false, + "cvt.s.plFG", false, false, false, + "pll.psFGH", "plu.psFGH", "pul.psFGH", "puu.psFGH", + "c.f.psVGH", "c.un.psVGH", "c.eq.psVGH", "c.ueq.psVGH", + "c.olt.psVGH", "c.ult.psVGH", "c.ole.psVGH", "c.ule.psVGH", + "c.psf.psVGH", "c.ngle.psVGH", "c.pseq.psVGH", "c.ngl.psVGH", + "c.lt.psVGH", "c.nge.psVGH", "c.le.psVGH", "c.ngt.psVGH", +} + +local map_cop1w = { + shift = 0, mask = 63, + [32] = "cvt.s.wFG", [33] = "cvt.d.wFG", +} + +local map_cop1l = { + shift = 0, mask = 63, + [32] = "cvt.s.lFG", [33] = "cvt.d.lFG", +} + +local map_cop1bc = { + shift = 16, mask = 3, + [0] = "bc1fCB", "bc1tCB", "bc1flCB", "bc1tlCB", +} + +local map_cop1 = { + shift = 21, mask = 31, + [0] = "mfc1TG", "dmfc1TG", "cfc1TG", "mfhc1TG", + "mtc1TG", "dmtc1TG", "ctc1TG", "mthc1TG", + map_cop1bc, false, false, false, + false, false, false, false, + map_cop1s, map_cop1d, false, false, + map_cop1w, map_cop1l, map_cop1ps, +} + +local map_cop1x = { + shift = 0, mask = 63, + [0] = "lwxc1FSX", "ldxc1FSX", false, false, + false, "luxc1FSX", false, false, + "swxc1FSX", "sdxc1FSX", false, false, + false, "suxc1FSX", false, "prefxMSX", + false, false, false, false, + false, false, false, false, + false, false, false, false, + false, false, "alnv.psFGHS", false, + "madd.sFRGH", "madd.dFRGH", false, false, + false, false, "madd.psFRGH", false, + "msub.sFRGH", "msub.dFRGH", false, false, + false, false, "msub.psFRGH", false, + "nmadd.sFRGH", "nmadd.dFRGH", false, false, + false, false, "nmadd.psFRGH", false, + "nmsub.sFRGH", "nmsub.dFRGH", false, false, + false, false, "nmsub.psFRGH", false, +} + +local map_pri = { + [0] = map_special, map_regimm, "jJ", "jalJ", + "beq|beqz|bST00B", "bne|bnezST0B", "blezSB", "bgtzSB", + "addiTSI", "addiu|liTS0I", "sltiTSI", "sltiuTSI", + "andiTSU", "ori|liTS0U", "xoriTSU", "luiTU", + map_cop0, map_cop1, false, map_cop1x, + "beql|beqzlST0B", "bnel|bnezlST0B", "blezlSB", "bgtzlSB", + "daddiTSI", "daddiuTSI", false, false, + map_special2, "jalxJ", false, map_special3, + "lbTSO", "lhTSO", "lwlTSO", "lwTSO", + "lbuTSO", "lhuTSO", "lwrTSO", false, + "sbTSO", "shTSO", "swlTSO", "swTSO", + false, false, "swrTSO", "cacheNSO", + "llTSO", "lwc1HSO", "lwc2TSO", "prefNSO", + false, "ldc1HSO", "ldc2TSO", "ldTSO", + "scTSO", "swc1HSO", "swc2TSO", false, + false, "sdc1HSO", "sdc2TSO", "sdTSO", +} + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps for MIPS R6 +------------------------------------------------------------------------------ + +local map_mul_r6 = { shift = 6, mask = 3, [2] = "mulDST", [3] = "muhDST" } +local map_mulu_r6 = { shift = 6, mask = 3, [2] = "muluDST", [3] = "muhuDST" } +local map_div_r6 = { shift = 6, mask = 3, [2] = "divDST", [3] = "modDST" } +local map_divu_r6 = { shift = 6, mask = 3, [2] = "divuDST", [3] = "moduDST" } +local map_dmul_r6 = { shift = 6, mask = 3, [2] = "dmulDST", [3] = "dmuhDST" } +local map_dmulu_r6 = { shift = 6, mask = 3, [2] = "dmuluDST", [3] = "dmuhuDST" } +local map_ddiv_r6 = { shift = 6, mask = 3, [2] = "ddivDST", [3] = "dmodDST" } +local map_ddivu_r6 = { shift = 6, mask = 3, [2] = "ddivuDST", [3] = "dmoduDST" } + +local map_special_r6 = { + shift = 0, mask = 63, + [0] = { shift = 0, mask = -1, [0] = "nop", _ = "sllDTA" }, + false, map_srl, "sraDTA", + "sllvDTS", false, map_srlv, "sravDTS", + "jrS", "jalrD1S", false, false, + "syscallY", "breakY", false, "sync", + "clzDS", "cloDS", "dclzDS", "dcloDS", + "dsllvDST", "dlsaDSTA", "dsrlvDST", "dsravDST", + map_mul_r6, map_mulu_r6, map_div_r6, map_divu_r6, + map_dmul_r6, map_dmulu_r6, map_ddiv_r6, map_ddivu_r6, + "addDST", "addu|moveDST0", "subDST", "subu|neguDS0T", + "andDST", "or|moveDST0", "xorDST", "nor|notDST0", + false, false, "sltDST", "sltuDST", + "daddDST", "dadduDST", "dsubDST", "dsubuDST", + "tgeSTZ", "tgeuSTZ", "tltSTZ", "tltuSTZ", + "teqSTZ", "seleqzDST", "tneSTZ", "selnezDST", + "dsllDTA", false, "dsrlDTA", "dsraDTA", + "dsll32DTA", false, "dsrl32DTA", "dsra32DTA", +} + +local map_bshfl_r6 = { + shift = 9, mask = 3, + [1] = "alignDSTa", + _ = { + shift = 6, mask = 31, + [0] = "bitswapDT", + [2] = "wsbhDT", + [16] = "sebDT", + [24] = "sehDT", + } +} + +local map_dbshfl_r6 = { + shift = 9, mask = 3, + [1] = "dalignDSTa", + _ = { + shift = 6, mask = 31, + [0] = "dbitswapDT", + [2] = "dsbhDT", + [5] = "dshdDT", + } +} + +local map_special3_r6 = { + shift = 0, mask = 63, + [0] = "extTSAK", [1] = "dextmTSAP", [3] = "dextTSAK", + [4] = "insTSAL", [6] = "dinsuTSEQ", [7] = "dinsTSAL", + [32] = map_bshfl_r6, [36] = map_dbshfl_r6, [59] = "rdhwrTD", +} + +local map_regimm_r6 = { + shift = 16, mask = 31, + [0] = "bltzSB", [1] = "bgezSB", + [6] = "dahiSI", [30] = "datiSI", + [23] = "sigrieI", [31] = "synciSO", +} + +local map_pcrel_r6 = { + shift = 19, mask = 3, + [0] = "addiupcS2", "lwpcS2", "lwupcS2", { + shift = 18, mask = 1, + [0] = "ldpcS3", { shift = 16, mask = 3, [2] = "auipcSI", [3] = "aluipcSI" } + } +} + +local map_cop1s_r6 = { + shift = 0, mask = 63, + [0] = "add.sFGH", "sub.sFGH", "mul.sFGH", "div.sFGH", + "sqrt.sFG", "abs.sFG", "mov.sFG", "neg.sFG", + "round.l.sFG", "trunc.l.sFG", "ceil.l.sFG", "floor.l.sFG", + "round.w.sFG", "trunc.w.sFG", "ceil.w.sFG", "floor.w.sFG", + "sel.sFGH", false, false, false, + "seleqz.sFGH", "recip.sFG", "rsqrt.sFG", "selnez.sFGH", + "maddf.sFGH", "msubf.sFGH", "rint.sFG", "class.sFG", + "min.sFGH", "mina.sFGH", "max.sFGH", "maxa.sFGH", + false, "cvt.d.sFG", false, false, + "cvt.w.sFG", "cvt.l.sFG", +} + +local map_cop1d_r6 = { + shift = 0, mask = 63, + [0] = "add.dFGH", "sub.dFGH", "mul.dFGH", "div.dFGH", + "sqrt.dFG", "abs.dFG", "mov.dFG", "neg.dFG", + "round.l.dFG", "trunc.l.dFG", "ceil.l.dFG", "floor.l.dFG", + "round.w.dFG", "trunc.w.dFG", "ceil.w.dFG", "floor.w.dFG", + "sel.dFGH", false, false, false, + "seleqz.dFGH", "recip.dFG", "rsqrt.dFG", "selnez.dFGH", + "maddf.dFGH", "msubf.dFGH", "rint.dFG", "class.dFG", + "min.dFGH", "mina.dFGH", "max.dFGH", "maxa.dFGH", + "cvt.s.dFG", false, false, false, + "cvt.w.dFG", "cvt.l.dFG", +} + +local map_cop1w_r6 = { + shift = 0, mask = 63, + [0] = "cmp.af.sFGH", "cmp.un.sFGH", "cmp.eq.sFGH", "cmp.ueq.sFGH", + "cmp.lt.sFGH", "cmp.ult.sFGH", "cmp.le.sFGH", "cmp.ule.sFGH", + "cmp.saf.sFGH", "cmp.sun.sFGH", "cmp.seq.sFGH", "cmp.sueq.sFGH", + "cmp.slt.sFGH", "cmp.sult.sFGH", "cmp.sle.sFGH", "cmp.sule.sFGH", + false, "cmp.or.sFGH", "cmp.une.sFGH", "cmp.ne.sFGH", + false, false, false, false, + false, "cmp.sor.sFGH", "cmp.sune.sFGH", "cmp.sne.sFGH", + false, false, false, false, + "cvt.s.wFG", "cvt.d.wFG", +} + +local map_cop1l_r6 = { + shift = 0, mask = 63, + [0] = "cmp.af.dFGH", "cmp.un.dFGH", "cmp.eq.dFGH", "cmp.ueq.dFGH", + "cmp.lt.dFGH", "cmp.ult.dFGH", "cmp.le.dFGH", "cmp.ule.dFGH", + "cmp.saf.dFGH", "cmp.sun.dFGH", "cmp.seq.dFGH", "cmp.sueq.dFGH", + "cmp.slt.dFGH", "cmp.sult.dFGH", "cmp.sle.dFGH", "cmp.sule.dFGH", + false, "cmp.or.dFGH", "cmp.une.dFGH", "cmp.ne.dFGH", + false, false, false, false, + false, "cmp.sor.dFGH", "cmp.sune.dFGH", "cmp.sne.dFGH", + false, false, false, false, + "cvt.s.lFG", "cvt.d.lFG", +} + +local map_cop1_r6 = { + shift = 21, mask = 31, + [0] = "mfc1TG", "dmfc1TG", "cfc1TG", "mfhc1TG", + "mtc1TG", "dmtc1TG", "ctc1TG", "mthc1TG", + false, "bc1eqzHB", false, false, + false, "bc1nezHB", false, false, + map_cop1s_r6, map_cop1d_r6, false, false, + map_cop1w_r6, map_cop1l_r6, +} + +local function maprs_popTS(rs, rt) + if rt == 0 then return 0 elseif rs == 0 then return 1 + elseif rs == rt then return 2 else return 3 end +end + +local map_pop06_r6 = { + maprs = maprs_popTS, [0] = "blezSB", "blezalcTB", "bgezalcTB", "bgeucSTB" +} +local map_pop07_r6 = { + maprs = maprs_popTS, [0] = "bgtzSB", "bgtzalcTB", "bltzalcTB", "bltucSTB" +} +local map_pop26_r6 = { + maprs = maprs_popTS, "blezcTB", "bgezcTB", "bgecSTB" +} +local map_pop27_r6 = { + maprs = maprs_popTS, "bgtzcTB", "bltzcTB", "bltcSTB" +} + +local function maprs_popS(rs, rt) + if rs == 0 then return 0 else return 1 end +end + +local map_pop66_r6 = { + maprs = maprs_popS, [0] = "jicTI", "beqzcSb" +} +local map_pop76_r6 = { + maprs = maprs_popS, [0] = "jialcTI", "bnezcSb" +} + +local function maprs_popST(rs, rt) + if rs >= rt then return 0 elseif rs == 0 then return 1 else return 2 end +end + +local map_pop10_r6 = { + maprs = maprs_popST, [0] = "bovcSTB", "beqzalcTB", "beqcSTB" +} +local map_pop30_r6 = { + maprs = maprs_popST, [0] = "bnvcSTB", "bnezalcTB", "bnecSTB" +} + +local map_pri_r6 = { + [0] = map_special_r6, map_regimm_r6, "jJ", "jalJ", + "beq|beqz|bST00B", "bne|bnezST0B", map_pop06_r6, map_pop07_r6, + map_pop10_r6, "addiu|liTS0I", "sltiTSI", "sltiuTSI", + "andiTSU", "ori|liTS0U", "xoriTSU", "aui|luiTS0U", + map_cop0, map_cop1_r6, false, false, + false, false, map_pop26_r6, map_pop27_r6, + map_pop30_r6, "daddiuTSI", false, false, + false, "dauiTSI", false, map_special3_r6, + "lbTSO", "lhTSO", false, "lwTSO", + "lbuTSO", "lhuTSO", false, false, + "sbTSO", "shTSO", false, "swTSO", + false, false, false, false, + false, "lwc1HSO", "bc#", false, + false, "ldc1HSO", map_pop66_r6, "ldTSO", + false, "swc1HSO", "balc#", map_pcrel_r6, + false, "sdc1HSO", map_pop76_r6, "sdTSO", +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15", + "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23", + "r24", "r25", "r26", "r27", "r28", "sp", "r30", "ra", +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then extra = "\t->"..sym end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-7s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-7s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +local function get_be(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + return bor(lshift(b0, 24), lshift(b1, 16), lshift(b2, 8), b3) +end + +local function get_le(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + return bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local op = ctx:get() + local operands = {} + local last = nil + ctx.op = op + ctx.rel = nil + + local opat = ctx.map_pri[rshift(op, 26)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + if opat.maprs then + opat = opat[opat.maprs(band(rshift(op,21),31), band(rshift(op,16),31))] + else + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + end + local name, pat = match(opat, "^([a-z0-9_.]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.|]*)(.*)") + if altname then pat = pat2 end + + for p in gmatch(pat, ".") do + local x = nil + if p == "S" then + x = map_gpr[band(rshift(op, 21), 31)] + elseif p == "T" then + x = map_gpr[band(rshift(op, 16), 31)] + elseif p == "D" then + x = map_gpr[band(rshift(op, 11), 31)] + elseif p == "F" then + x = "f"..band(rshift(op, 6), 31) + elseif p == "G" then + x = "f"..band(rshift(op, 11), 31) + elseif p == "H" then + x = "f"..band(rshift(op, 16), 31) + elseif p == "R" then + x = "f"..band(rshift(op, 21), 31) + elseif p == "A" then + x = band(rshift(op, 6), 31) + elseif p == "a" then + x = band(rshift(op, 6), 7) + elseif p == "E" then + x = band(rshift(op, 6), 31) + 32 + elseif p == "M" then + x = band(rshift(op, 11), 31) + elseif p == "N" then + x = band(rshift(op, 16), 31) + elseif p == "C" then + x = band(rshift(op, 18), 7) + if x == 0 then x = nil end + elseif p == "K" then + x = band(rshift(op, 11), 31) + 1 + elseif p == "P" then + x = band(rshift(op, 11), 31) + 33 + elseif p == "L" then + x = band(rshift(op, 11), 31) - last + 1 + elseif p == "Q" then + x = band(rshift(op, 11), 31) - last + 33 + elseif p == "I" then + x = arshift(lshift(op, 16), 16) + elseif p == "2" then + x = arshift(lshift(op, 13), 11) + elseif p == "3" then + x = arshift(lshift(op, 14), 11) + elseif p == "U" then + x = band(op, 0xffff) + elseif p == "O" then + local disp = arshift(lshift(op, 16), 16) + operands[#operands] = format("%d(%s)", disp, last) + elseif p == "X" then + local index = map_gpr[band(rshift(op, 16), 31)] + operands[#operands] = format("%s(%s)", index, last) + elseif p == "B" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 16), 14) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "b" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 11), 9) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "#" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 6), 4) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "J" then + local a = ctx.addr + ctx.pos + x = a - band(a, 0x0fffffff) + band(op, 0x03ffffff)*4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "V" then + x = band(rshift(op, 8), 7) + if x == 0 then x = nil end + elseif p == "W" then + x = band(op, 7) + if x == 0 then x = nil end + elseif p == "Y" then + x = band(rshift(op, 6), 0x000fffff) + if x == 0 then x = nil end + elseif p == "Z" then + x = band(rshift(op, 6), 1023) + if x == 0 then x = nil end + elseif p == "0" then + if last == "r0" or last == 0 then + local n = #operands + operands[n] = nil + last = operands[n-1] + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if a1 then name, altname = a1, a2 + else name = altname end + end + end + elseif p == "1" then + if last == "ra" then + operands[#operands] = nil + end + else + assert(false) + end + if x then operands[#operands+1] = x; last = x end + end + + return putop(ctx, name, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + stop = stop - stop % 4 + ctx.pos = ofs - ofs % 4 + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + ctx.get = get_be + ctx.map_pri = map_pri + return ctx +end + +local function create_el(code, addr, out) + local ctx = create(code, addr, out) + ctx.get = get_le + return ctx +end + +local function create_r6(code, addr, out) + local ctx = create(code, addr, out) + ctx.map_pri = map_pri_r6 + return ctx +end + +local function create_r6_el(code, addr, out) + local ctx = create(code, addr, out) + ctx.get = get_le + ctx.map_pri = map_pri_r6 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +local function disass_el(code, addr, out) + create_el(code, addr, out):disass() +end + +local function disass_r6(code, addr, out) + create_r6(code, addr, out):disass() +end + +local function disass_r6_el(code, addr, out) + create_r6_el(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_gpr[r] end + return "f"..(r-32) +end + +-- Public module functions. +return { + create = create, + create_el = create_el, + create_r6 = create_r6, + create_r6_el = create_r6_el, + disass = disass, + disass_el = disass_el, + disass_r6 = disass_r6, + disass_r6_el = disass_r6_el, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64.lua new file mode 100644 index 00000000..5ad48f8f --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the big-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create, + disass = dis_mips.disass, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64el.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64el.lua new file mode 100644 index 00000000..d50e3a18 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64el.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64EL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_el, + disass = dis_mips.disass_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64r6.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64r6.lua new file mode 100644 index 00000000..921b3cbe --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64r6.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64R6 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the r6 big-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_r6, + disass = dis_mips.disass_r6, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64r6el.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64r6el.lua new file mode 100644 index 00000000..aadef9f3 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mips64r6el.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64R6EL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the r6 little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_r6_el, + disass = dis_mips.disass_r6_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mipsel.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mipsel.lua new file mode 100644 index 00000000..52cebefb --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_mipsel.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPSEL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_el, + disass = dis_mips.disass_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_ppc.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_ppc.lua new file mode 100644 index 00000000..08d742f1 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_ppc.lua @@ -0,0 +1,591 @@ +---------------------------------------------------------------------------- +-- LuaJIT PPC disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT/X license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles all common, non-privileged 32/64 bit PowerPC instructions +-- plus the e500 SPE instructions and some Cell/Xenon extensions. +-- +-- NYI: VMX, VMX128 +------------------------------------------------------------------------------ + +local type = type +local byte, format = string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local concat = table.concat +local bit = require("bit") +local band, bor, tohex = bit.band, bit.bor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps +------------------------------------------------------------------------------ + +local map_crops = { + shift = 1, mask = 1023, + [0] = "mcrfXX", + [33] = "crnor|crnotCCC=", [129] = "crandcCCC", + [193] = "crxor|crclrCCC%", [225] = "crnandCCC", + [257] = "crandCCC", [289] = "creqv|crsetCCC%", + [417] = "crorcCCC", [449] = "cror|crmoveCCC=", + [16] = "b_lrKB", [528] = "b_ctrKB", + [150] = "isync", +} + +local map_rlwinm = setmetatable({ + shift = 0, mask = -1, +}, +{ __index = function(t, x) + local rot = band(rshift(x, 11), 31) + local mb = band(rshift(x, 6), 31) + local me = band(rshift(x, 1), 31) + if mb == 0 and me == 31-rot then + return "slwiRR~A." + elseif me == 31 and mb == 32-rot then + return "srwiRR~-A." + else + return "rlwinmRR~AAA." + end + end +}) + +local map_rld = { + shift = 2, mask = 7, + [0] = "rldiclRR~HM.", "rldicrRR~HM.", "rldicRR~HM.", "rldimiRR~HM.", + { + shift = 1, mask = 1, + [0] = "rldclRR~RM.", "rldcrRR~RM.", + }, +} + +local map_ext = setmetatable({ + shift = 1, mask = 1023, + + [0] = "cmp_YLRR", [32] = "cmpl_YLRR", + [4] = "twARR", [68] = "tdARR", + + [8] = "subfcRRR.", [40] = "subfRRR.", + [104] = "negRR.", [136] = "subfeRRR.", + [200] = "subfzeRR.", [232] = "subfmeRR.", + [520] = "subfcoRRR.", [552] = "subfoRRR.", + [616] = "negoRR.", [648] = "subfeoRRR.", + [712] = "subfzeoRR.", [744] = "subfmeoRR.", + + [9] = "mulhduRRR.", [73] = "mulhdRRR.", [233] = "mulldRRR.", + [457] = "divduRRR.", [489] = "divdRRR.", + [745] = "mulldoRRR.", + [969] = "divduoRRR.", [1001] = "divdoRRR.", + + [10] = "addcRRR.", [138] = "addeRRR.", + [202] = "addzeRR.", [234] = "addmeRR.", [266] = "addRRR.", + [522] = "addcoRRR.", [650] = "addeoRRR.", + [714] = "addzeoRR.", [746] = "addmeoRR.", [778] = "addoRRR.", + + [11] = "mulhwuRRR.", [75] = "mulhwRRR.", [235] = "mullwRRR.", + [459] = "divwuRRR.", [491] = "divwRRR.", + [747] = "mullwoRRR.", + [971] = "divwouRRR.", [1003] = "divwoRRR.", + + [15] = "iselltRRR", [47] = "iselgtRRR", [79] = "iseleqRRR", + + [144] = { shift = 20, mask = 1, [0] = "mtcrfRZ~", "mtocrfRZ~", }, + [19] = { shift = 20, mask = 1, [0] = "mfcrR", "mfocrfRZ", }, + [371] = { shift = 11, mask = 1023, [392] = "mftbR", [424] = "mftbuR", }, + [339] = { + shift = 11, mask = 1023, + [32] = "mferR", [256] = "mflrR", [288] = "mfctrR", [16] = "mfspefscrR", + }, + [467] = { + shift = 11, mask = 1023, + [32] = "mtxerR", [256] = "mtlrR", [288] = "mtctrR", [16] = "mtspefscrR", + }, + + [20] = "lwarxRR0R", [84] = "ldarxRR0R", + + [21] = "ldxRR0R", [53] = "lduxRRR", + [149] = "stdxRR0R", [181] = "stduxRRR", + [341] = "lwaxRR0R", [373] = "lwauxRRR", + + [23] = "lwzxRR0R", [55] = "lwzuxRRR", + [87] = "lbzxRR0R", [119] = "lbzuxRRR", + [151] = "stwxRR0R", [183] = "stwuxRRR", + [215] = "stbxRR0R", [247] = "stbuxRRR", + [279] = "lhzxRR0R", [311] = "lhzuxRRR", + [343] = "lhaxRR0R", [375] = "lhauxRRR", + [407] = "sthxRR0R", [439] = "sthuxRRR", + + [54] = "dcbst-R0R", [86] = "dcbf-R0R", + [150] = "stwcxRR0R.", [214] = "stdcxRR0R.", + [246] = "dcbtst-R0R", [278] = "dcbt-R0R", + [310] = "eciwxRR0R", [438] = "ecowxRR0R", + [470] = "dcbi-RR", + + [598] = { + shift = 21, mask = 3, + [0] = "sync", "lwsync", "ptesync", + }, + [758] = "dcba-RR", + [854] = "eieio", [982] = "icbi-R0R", [1014] = "dcbz-R0R", + + [26] = "cntlzwRR~", [58] = "cntlzdRR~", + [122] = "popcntbRR~", + [154] = "prtywRR~", [186] = "prtydRR~", + + [28] = "andRR~R.", [60] = "andcRR~R.", [124] = "nor|notRR~R=.", + [284] = "eqvRR~R.", [316] = "xorRR~R.", + [412] = "orcRR~R.", [444] = "or|mrRR~R=.", [476] = "nandRR~R.", + [508] = "cmpbRR~R", + + [512] = "mcrxrX", + + [532] = "ldbrxRR0R", [660] = "stdbrxRR0R", + + [533] = "lswxRR0R", [597] = "lswiRR0A", + [661] = "stswxRR0R", [725] = "stswiRR0A", + + [534] = "lwbrxRR0R", [662] = "stwbrxRR0R", + [790] = "lhbrxRR0R", [918] = "sthbrxRR0R", + + [535] = "lfsxFR0R", [567] = "lfsuxFRR", + [599] = "lfdxFR0R", [631] = "lfduxFRR", + [663] = "stfsxFR0R", [695] = "stfsuxFRR", + [727] = "stfdxFR0R", [759] = "stfduxFR0R", + [855] = "lfiwaxFR0R", + [983] = "stfiwxFR0R", + + [24] = "slwRR~R.", + + [27] = "sldRR~R.", [536] = "srwRR~R.", + [792] = "srawRR~R.", [824] = "srawiRR~A.", + + [794] = "sradRR~R.", [826] = "sradiRR~H.", [827] = "sradiRR~H.", + [922] = "extshRR~.", [954] = "extsbRR~.", [986] = "extswRR~.", + + [539] = "srdRR~R.", +}, +{ __index = function(t, x) + if band(x, 31) == 15 then return "iselRRRC" end + end +}) + +local map_ld = { + shift = 0, mask = 3, + [0] = "ldRRE", "lduRRE", "lwaRRE", +} + +local map_std = { + shift = 0, mask = 3, + [0] = "stdRRE", "stduRRE", +} + +local map_fps = { + shift = 5, mask = 1, + { + shift = 1, mask = 15, + [0] = false, false, "fdivsFFF.", false, + "fsubsFFF.", "faddsFFF.", "fsqrtsF-F.", false, + "fresF-F.", "fmulsFF-F.", "frsqrtesF-F.", false, + "fmsubsFFFF~.", "fmaddsFFFF~.", "fnmsubsFFFF~.", "fnmaddsFFFF~.", + } +} + +local map_fpd = { + shift = 5, mask = 1, + [0] = { + shift = 1, mask = 1023, + [0] = "fcmpuXFF", [32] = "fcmpoXFF", [64] = "mcrfsXX", + [38] = "mtfsb1A.", [70] = "mtfsb0A.", [134] = "mtfsfiA>>-A>", + [8] = "fcpsgnFFF.", [40] = "fnegF-F.", [72] = "fmrF-F.", + [136] = "fnabsF-F.", [264] = "fabsF-F.", + [12] = "frspF-F.", + [14] = "fctiwF-F.", [15] = "fctiwzF-F.", + [583] = "mffsF.", [711] = "mtfsfZF.", + [392] = "frinF-F.", [424] = "frizF-F.", + [456] = "fripF-F.", [488] = "frimF-F.", + [814] = "fctidF-F.", [815] = "fctidzF-F.", [846] = "fcfidF-F.", + }, + { + shift = 1, mask = 15, + [0] = false, false, "fdivFFF.", false, + "fsubFFF.", "faddFFF.", "fsqrtF-F.", "fselFFFF~.", + "freF-F.", "fmulFF-F.", "frsqrteF-F.", false, + "fmsubFFFF~.", "fmaddFFFF~.", "fnmsubFFFF~.", "fnmaddFFFF~.", + } +} + +local map_spe = { + shift = 0, mask = 2047, + + [512] = "evaddwRRR", [514] = "evaddiwRAR~", + [516] = "evsubwRRR~", [518] = "evsubiwRAR~", + [520] = "evabsRR", [521] = "evnegRR", + [522] = "evextsbRR", [523] = "evextshRR", [524] = "evrndwRR", + [525] = "evcntlzwRR", [526] = "evcntlswRR", + + [527] = "brincRRR", + + [529] = "evandRRR", [530] = "evandcRRR", [534] = "evxorRRR", + [535] = "evor|evmrRRR=", [536] = "evnor|evnotRRR=", + [537] = "eveqvRRR", [539] = "evorcRRR", [542] = "evnandRRR", + + [544] = "evsrwuRRR", [545] = "evsrwsRRR", + [546] = "evsrwiuRRA", [547] = "evsrwisRRA", + [548] = "evslwRRR", [550] = "evslwiRRA", + [552] = "evrlwRRR", [553] = "evsplatiRS", + [554] = "evrlwiRRA", [555] = "evsplatfiRS", + [556] = "evmergehiRRR", [557] = "evmergeloRRR", + [558] = "evmergehiloRRR", [559] = "evmergelohiRRR", + + [560] = "evcmpgtuYRR", [561] = "evcmpgtsYRR", + [562] = "evcmpltuYRR", [563] = "evcmpltsYRR", + [564] = "evcmpeqYRR", + + [632] = "evselRRR", [633] = "evselRRRW", + [634] = "evselRRRW", [635] = "evselRRRW", + [636] = "evselRRRW", [637] = "evselRRRW", + [638] = "evselRRRW", [639] = "evselRRRW", + + [640] = "evfsaddRRR", [641] = "evfssubRRR", + [644] = "evfsabsRR", [645] = "evfsnabsRR", [646] = "evfsnegRR", + [648] = "evfsmulRRR", [649] = "evfsdivRRR", + [652] = "evfscmpgtYRR", [653] = "evfscmpltYRR", [654] = "evfscmpeqYRR", + [656] = "evfscfuiR-R", [657] = "evfscfsiR-R", + [658] = "evfscfufR-R", [659] = "evfscfsfR-R", + [660] = "evfsctuiR-R", [661] = "evfsctsiR-R", + [662] = "evfsctufR-R", [663] = "evfsctsfR-R", + [664] = "evfsctuizR-R", [666] = "evfsctsizR-R", + [668] = "evfststgtYRR", [669] = "evfststltYRR", [670] = "evfststeqYRR", + + [704] = "efsaddRRR", [705] = "efssubRRR", + [708] = "efsabsRR", [709] = "efsnabsRR", [710] = "efsnegRR", + [712] = "efsmulRRR", [713] = "efsdivRRR", + [716] = "efscmpgtYRR", [717] = "efscmpltYRR", [718] = "efscmpeqYRR", + [719] = "efscfdR-R", + [720] = "efscfuiR-R", [721] = "efscfsiR-R", + [722] = "efscfufR-R", [723] = "efscfsfR-R", + [724] = "efsctuiR-R", [725] = "efsctsiR-R", + [726] = "efsctufR-R", [727] = "efsctsfR-R", + [728] = "efsctuizR-R", [730] = "efsctsizR-R", + [732] = "efststgtYRR", [733] = "efststltYRR", [734] = "efststeqYRR", + + [736] = "efdaddRRR", [737] = "efdsubRRR", + [738] = "efdcfuidR-R", [739] = "efdcfsidR-R", + [740] = "efdabsRR", [741] = "efdnabsRR", [742] = "efdnegRR", + [744] = "efdmulRRR", [745] = "efddivRRR", + [746] = "efdctuidzR-R", [747] = "efdctsidzR-R", + [748] = "efdcmpgtYRR", [749] = "efdcmpltYRR", [750] = "efdcmpeqYRR", + [751] = "efdcfsR-R", + [752] = "efdcfuiR-R", [753] = "efdcfsiR-R", + [754] = "efdcfufR-R", [755] = "efdcfsfR-R", + [756] = "efdctuiR-R", [757] = "efdctsiR-R", + [758] = "efdctufR-R", [759] = "efdctsfR-R", + [760] = "efdctuizR-R", [762] = "efdctsizR-R", + [764] = "efdtstgtYRR", [765] = "efdtstltYRR", [766] = "efdtsteqYRR", + + [768] = "evlddxRR0R", [769] = "evlddRR8", + [770] = "evldwxRR0R", [771] = "evldwRR8", + [772] = "evldhxRR0R", [773] = "evldhRR8", + [776] = "evlhhesplatxRR0R", [777] = "evlhhesplatRR2", + [780] = "evlhhousplatxRR0R", [781] = "evlhhousplatRR2", + [782] = "evlhhossplatxRR0R", [783] = "evlhhossplatRR2", + [784] = "evlwhexRR0R", [785] = "evlwheRR4", + [788] = "evlwhouxRR0R", [789] = "evlwhouRR4", + [790] = "evlwhosxRR0R", [791] = "evlwhosRR4", + [792] = "evlwwsplatxRR0R", [793] = "evlwwsplatRR4", + [796] = "evlwhsplatxRR0R", [797] = "evlwhsplatRR4", + + [800] = "evstddxRR0R", [801] = "evstddRR8", + [802] = "evstdwxRR0R", [803] = "evstdwRR8", + [804] = "evstdhxRR0R", [805] = "evstdhRR8", + [816] = "evstwhexRR0R", [817] = "evstwheRR4", + [820] = "evstwhoxRR0R", [821] = "evstwhoRR4", + [824] = "evstwwexRR0R", [825] = "evstwweRR4", + [828] = "evstwwoxRR0R", [829] = "evstwwoRR4", + + [1027] = "evmhessfRRR", [1031] = "evmhossfRRR", [1032] = "evmheumiRRR", + [1033] = "evmhesmiRRR", [1035] = "evmhesmfRRR", [1036] = "evmhoumiRRR", + [1037] = "evmhosmiRRR", [1039] = "evmhosmfRRR", [1059] = "evmhessfaRRR", + [1063] = "evmhossfaRRR", [1064] = "evmheumiaRRR", [1065] = "evmhesmiaRRR", + [1067] = "evmhesmfaRRR", [1068] = "evmhoumiaRRR", [1069] = "evmhosmiaRRR", + [1071] = "evmhosmfaRRR", [1095] = "evmwhssfRRR", [1096] = "evmwlumiRRR", + [1100] = "evmwhumiRRR", [1101] = "evmwhsmiRRR", [1103] = "evmwhsmfRRR", + [1107] = "evmwssfRRR", [1112] = "evmwumiRRR", [1113] = "evmwsmiRRR", + [1115] = "evmwsmfRRR", [1127] = "evmwhssfaRRR", [1128] = "evmwlumiaRRR", + [1132] = "evmwhumiaRRR", [1133] = "evmwhsmiaRRR", [1135] = "evmwhsmfaRRR", + [1139] = "evmwssfaRRR", [1144] = "evmwumiaRRR", [1145] = "evmwsmiaRRR", + [1147] = "evmwsmfaRRR", + + [1216] = "evaddusiaawRR", [1217] = "evaddssiaawRR", + [1218] = "evsubfusiaawRR", [1219] = "evsubfssiaawRR", + [1220] = "evmraRR", + [1222] = "evdivwsRRR", [1223] = "evdivwuRRR", + [1224] = "evaddumiaawRR", [1225] = "evaddsmiaawRR", + [1226] = "evsubfumiaawRR", [1227] = "evsubfsmiaawRR", + + [1280] = "evmheusiaawRRR", [1281] = "evmhessiaawRRR", + [1283] = "evmhessfaawRRR", [1284] = "evmhousiaawRRR", + [1285] = "evmhossiaawRRR", [1287] = "evmhossfaawRRR", + [1288] = "evmheumiaawRRR", [1289] = "evmhesmiaawRRR", + [1291] = "evmhesmfaawRRR", [1292] = "evmhoumiaawRRR", + [1293] = "evmhosmiaawRRR", [1295] = "evmhosmfaawRRR", + [1320] = "evmhegumiaaRRR", [1321] = "evmhegsmiaaRRR", + [1323] = "evmhegsmfaaRRR", [1324] = "evmhogumiaaRRR", + [1325] = "evmhogsmiaaRRR", [1327] = "evmhogsmfaaRRR", + [1344] = "evmwlusiaawRRR", [1345] = "evmwlssiaawRRR", + [1352] = "evmwlumiaawRRR", [1353] = "evmwlsmiaawRRR", + [1363] = "evmwssfaaRRR", [1368] = "evmwumiaaRRR", + [1369] = "evmwsmiaaRRR", [1371] = "evmwsmfaaRRR", + [1408] = "evmheusianwRRR", [1409] = "evmhessianwRRR", + [1411] = "evmhessfanwRRR", [1412] = "evmhousianwRRR", + [1413] = "evmhossianwRRR", [1415] = "evmhossfanwRRR", + [1416] = "evmheumianwRRR", [1417] = "evmhesmianwRRR", + [1419] = "evmhesmfanwRRR", [1420] = "evmhoumianwRRR", + [1421] = "evmhosmianwRRR", [1423] = "evmhosmfanwRRR", + [1448] = "evmhegumianRRR", [1449] = "evmhegsmianRRR", + [1451] = "evmhegsmfanRRR", [1452] = "evmhogumianRRR", + [1453] = "evmhogsmianRRR", [1455] = "evmhogsmfanRRR", + [1472] = "evmwlusianwRRR", [1473] = "evmwlssianwRRR", + [1480] = "evmwlumianwRRR", [1481] = "evmwlsmianwRRR", + [1491] = "evmwssfanRRR", [1496] = "evmwumianRRR", + [1497] = "evmwsmianRRR", [1499] = "evmwsmfanRRR", +} + +local map_pri = { + [0] = false, false, "tdiARI", "twiARI", + map_spe, false, false, "mulliRRI", + "subficRRI", false, "cmpl_iYLRU", "cmp_iYLRI", + "addicRRI", "addic.RRI", "addi|liRR0I", "addis|lisRR0I", + "b_KBJ", "sc", "bKJ", map_crops, + "rlwimiRR~AAA.", map_rlwinm, false, "rlwnmRR~RAA.", + "oriNRR~U", "orisRR~U", "xoriRR~U", "xorisRR~U", + "andi.RR~U", "andis.RR~U", map_rld, map_ext, + "lwzRRD", "lwzuRRD", "lbzRRD", "lbzuRRD", + "stwRRD", "stwuRRD", "stbRRD", "stbuRRD", + "lhzRRD", "lhzuRRD", "lhaRRD", "lhauRRD", + "sthRRD", "sthuRRD", "lmwRRD", "stmwRRD", + "lfsFRD", "lfsuFRD", "lfdFRD", "lfduFRD", + "stfsFRD", "stfsuFRD", "stfdFRD", "stfduFRD", + false, false, map_ld, map_fps, + false, false, map_std, map_fpd, +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "sp", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15", + "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23", + "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31", +} + +local map_cond = { [0] = "lt", "gt", "eq", "so", "ge", "le", "ne", "ns", } + +-- Format a condition bit. +local function condfmt(cond) + if cond <= 3 then + return map_cond[band(cond, 3)] + else + return format("4*cr%d+%s", rshift(cond, 2), map_cond[band(cond, 3)]) + end +end + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then extra = "\t->"..sym end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-7s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-7s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b0, 24), lshift(b1, 16), lshift(b2, 8), b3) + local operands = {} + local last = nil + local rs = 21 + ctx.op = op + ctx.rel = nil + + local opat = map_pri[rshift(b0, 2)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] + end + local name, pat = match(opat, "^([a-z0-9_.]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.]*)(.*)") + if altname then pat = pat2 end + + for p in gmatch(pat, ".") do + local x = nil + if p == "R" then + x = map_gpr[band(rshift(op, rs), 31)] + rs = rs - 5 + elseif p == "F" then + x = "f"..band(rshift(op, rs), 31) + rs = rs - 5 + elseif p == "A" then + x = band(rshift(op, rs), 31) + rs = rs - 5 + elseif p == "S" then + x = arshift(lshift(op, 27-rs), 27) + rs = rs - 5 + elseif p == "I" then + x = arshift(lshift(op, 16), 16) + elseif p == "U" then + x = band(op, 0xffff) + elseif p == "D" or p == "E" then + local disp = arshift(lshift(op, 16), 16) + if p == "E" then disp = band(disp, -4) end + if last == "r0" then last = "0" end + operands[#operands] = format("%d(%s)", disp, last) + elseif p >= "2" and p <= "8" then + local disp = band(rshift(op, rs), 31) * p + if last == "r0" then last = "0" end + operands[#operands] = format("%d(%s)", disp, last) + elseif p == "H" then + x = band(rshift(op, rs), 31) + lshift(band(op, 2), 4) + rs = rs - 5 + elseif p == "M" then + x = band(rshift(op, rs), 31) + band(op, 0x20) + elseif p == "C" then + x = condfmt(band(rshift(op, rs), 31)) + rs = rs - 5 + elseif p == "B" then + local bo = rshift(op, 21) + local cond = band(rshift(op, 16), 31) + local cn = "" + rs = rs - 10 + if band(bo, 4) == 0 then + cn = band(bo, 2) == 0 and "dnz" or "dz" + if band(bo, 0x10) == 0 then + cn = cn..(band(bo, 8) == 0 and "f" or "t") + end + if band(bo, 0x10) == 0 then x = condfmt(cond) end + name = name..(band(bo, 1) == band(rshift(op, 15), 1) and "-" or "+") + elseif band(bo, 0x10) == 0 then + cn = map_cond[band(cond, 3) + (band(bo, 8) == 0 and 4 or 0)] + if cond > 3 then x = "cr"..rshift(cond, 2) end + name = name..(band(bo, 1) == band(rshift(op, 15), 1) and "-" or "+") + end + name = gsub(name, "_", cn) + elseif p == "J" then + x = arshift(lshift(op, 27-rs), 29-rs)*4 + if band(op, 2) == 0 then x = ctx.addr + pos + x end + ctx.rel = x + x = "0x"..tohex(x) + elseif p == "K" then + if band(op, 1) ~= 0 then name = name.."l" end + if band(op, 2) ~= 0 then name = name.."a" end + elseif p == "X" or p == "Y" then + x = band(rshift(op, rs+2), 7) + if x == 0 and p == "Y" then x = nil else x = "cr"..x end + rs = rs - 5 + elseif p == "W" then + x = "cr"..band(op, 7) + elseif p == "Z" then + x = band(rshift(op, rs-4), 255) + rs = rs - 10 + elseif p == ">" then + operands[#operands] = rshift(operands[#operands], 1) + elseif p == "0" then + if last == "r0" then + operands[#operands] = nil + if altname then name = altname end + end + elseif p == "L" then + name = gsub(name, "_", band(op, 0x00200000) ~= 0 and "d" or "w") + elseif p == "." then + if band(op, 1) == 1 then name = name.."." end + elseif p == "N" then + if op == 0x60000000 then name = "nop"; break end + elseif p == "~" then + local n = #operands + operands[n-1], operands[n] = operands[n], operands[n-1] + elseif p == "=" then + local n = #operands + if last == operands[n-1] then + operands[n] = nil + name = altname + end + elseif p == "%" then + local n = #operands + if last == operands[n-1] and last == operands[n-2] then + operands[n] = nil + operands[n-1] = nil + name = altname + end + elseif p == "-" then + rs = rs - 5 + else + assert(false) + end + if x then operands[#operands+1] = x; last = x end + end + + return putop(ctx, name, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + stop = stop - stop % 4 + ctx.pos = ofs - ofs % 4 + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_gpr[r] end + return "f"..(r-32) +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_x64.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_x64.lua new file mode 100644 index 00000000..2d37423e --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_x64.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT x64 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the 64 bit functions from the combined +-- x86/x64 disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_x86 = require((string.match(..., ".*%.") or "").."dis_x86") +return { + create = dis_x86.create64, + disass = dis_x86.disass64, + regname = dis_x86.regname64 +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_x86.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_x86.lua new file mode 100644 index 00000000..5480854c --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dis_x86.lua @@ -0,0 +1,953 @@ +---------------------------------------------------------------------------- +-- LuaJIT x86/x64 disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- Sending small code snippets to an external disassembler and mixing the +-- output with our own stuff was too fragile. So I had to bite the bullet +-- and write yet another x86 disassembler. Oh well ... +-- +-- The output format is very similar to what ndisasm generates. But it has +-- been developed independently by looking at the opcode tables from the +-- Intel and AMD manuals. The supported instruction set is quite extensive +-- and reflects what a current generation Intel or AMD CPU implements in +-- 32 bit and 64 bit mode. Yes, this includes MMX, SSE, SSE2, SSE3, SSSE3, +-- SSE4.1, SSE4.2, SSE4a, AVX, AVX2 and even privileged and hypervisor +-- (VMX/SVM) instructions. +-- +-- Notes: +-- * The (useless) a16 prefix, 3DNow and pre-586 opcodes are unsupported. +-- * No attempt at optimization has been made -- it's fast enough for my needs. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local lower, rep = string.lower, string.rep +local bit = require("bit") +local tohex = bit.tohex + +-- Map for 1st opcode byte in 32 bit mode. Ugly? Well ... read on. +local map_opc1_32 = { +--0x +[0]="addBmr","addVmr","addBrm","addVrm","addBai","addVai","push es","pop es", +"orBmr","orVmr","orBrm","orVrm","orBai","orVai","push cs","opc2*", +--1x +"adcBmr","adcVmr","adcBrm","adcVrm","adcBai","adcVai","push ss","pop ss", +"sbbBmr","sbbVmr","sbbBrm","sbbVrm","sbbBai","sbbVai","push ds","pop ds", +--2x +"andBmr","andVmr","andBrm","andVrm","andBai","andVai","es:seg","daa", +"subBmr","subVmr","subBrm","subVrm","subBai","subVai","cs:seg","das", +--3x +"xorBmr","xorVmr","xorBrm","xorVrm","xorBai","xorVai","ss:seg","aaa", +"cmpBmr","cmpVmr","cmpBrm","cmpVrm","cmpBai","cmpVai","ds:seg","aas", +--4x +"incVR","incVR","incVR","incVR","incVR","incVR","incVR","incVR", +"decVR","decVR","decVR","decVR","decVR","decVR","decVR","decVR", +--5x +"pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR", +"popUR","popUR","popUR","popUR","popUR","popUR","popUR","popUR", +--6x +"sz*pushaw,pusha","sz*popaw,popa","boundVrm","arplWmr", +"fs:seg","gs:seg","o16:","a16", +"pushUi","imulVrmi","pushBs","imulVrms", +"insb","insVS","outsb","outsVS", +--7x +"joBj","jnoBj","jbBj","jnbBj","jzBj","jnzBj","jbeBj","jaBj", +"jsBj","jnsBj","jpeBj","jpoBj","jlBj","jgeBj","jleBj","jgBj", +--8x +"arith!Bmi","arith!Vmi","arith!Bmi","arith!Vms", +"testBmr","testVmr","xchgBrm","xchgVrm", +"movBmr","movVmr","movBrm","movVrm", +"movVmg","leaVrm","movWgm","popUm", +--9x +"nop*xchgVaR|pause|xchgWaR|repne nop","xchgVaR","xchgVaR","xchgVaR", +"xchgVaR","xchgVaR","xchgVaR","xchgVaR", +"sz*cbw,cwde,cdqe","sz*cwd,cdq,cqo","call farViw","wait", +"sz*pushfw,pushf","sz*popfw,popf","sahf","lahf", +--Ax +"movBao","movVao","movBoa","movVoa", +"movsb","movsVS","cmpsb","cmpsVS", +"testBai","testVai","stosb","stosVS", +"lodsb","lodsVS","scasb","scasVS", +--Bx +"movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi", +"movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI", +--Cx +"shift!Bmu","shift!Vmu","retBw","ret","vex*3$lesVrm","vex*2$ldsVrm","movBmi","movVmi", +"enterBwu","leave","retfBw","retf","int3","intBu","into","iretVS", +--Dx +"shift!Bm1","shift!Vm1","shift!Bmc","shift!Vmc","aamBu","aadBu","salc","xlatb", +"fp*0","fp*1","fp*2","fp*3","fp*4","fp*5","fp*6","fp*7", +--Ex +"loopneBj","loopeBj","loopBj","sz*jcxzBj,jecxzBj,jrcxzBj", +"inBau","inVau","outBua","outVua", +"callVj","jmpVj","jmp farViw","jmpBj","inBad","inVad","outBda","outVda", +--Fx +"lock:","int1","repne:rep","rep:","hlt","cmc","testb!Bm","testv!Vm", +"clc","stc","cli","sti","cld","std","incb!Bm","incd!Vm", +} +assert(#map_opc1_32 == 255) + +-- Map for 1st opcode byte in 64 bit mode (overrides only). +local map_opc1_64 = setmetatable({ + [0x06]=false, [0x07]=false, [0x0e]=false, + [0x16]=false, [0x17]=false, [0x1e]=false, [0x1f]=false, + [0x27]=false, [0x2f]=false, [0x37]=false, [0x3f]=false, + [0x60]=false, [0x61]=false, [0x62]=false, [0x63]="movsxdVrDmt", [0x67]="a32:", + [0x40]="rex*", [0x41]="rex*b", [0x42]="rex*x", [0x43]="rex*xb", + [0x44]="rex*r", [0x45]="rex*rb", [0x46]="rex*rx", [0x47]="rex*rxb", + [0x48]="rex*w", [0x49]="rex*wb", [0x4a]="rex*wx", [0x4b]="rex*wxb", + [0x4c]="rex*wr", [0x4d]="rex*wrb", [0x4e]="rex*wrx", [0x4f]="rex*wrxb", + [0x82]=false, [0x9a]=false, [0xc4]="vex*3", [0xc5]="vex*2", [0xce]=false, + [0xd4]=false, [0xd5]=false, [0xd6]=false, [0xea]=false, +}, { __index = map_opc1_32 }) + +-- Map for 2nd opcode byte (0F xx). True CISC hell. Hey, I told you. +-- Prefix dependent MMX/SSE opcodes: (none)|rep|o16|repne, -|F3|66|F2 +local map_opc2 = { +--0x +[0]="sldt!Dmp","sgdt!Ump","larVrm","lslVrm",nil,"syscall","clts","sysret", +"invd","wbinvd",nil,"ud1",nil,"$prefetch!Bm","femms","3dnowMrmu", +--1x +"movupsXrm|movssXrvm|movupdXrm|movsdXrvm", +"movupsXmr|movssXmvr|movupdXmr|movsdXmvr", +"movhlpsXrm$movlpsXrm|movsldupXrm|movlpdXrm|movddupXrm", +"movlpsXmr||movlpdXmr", +"unpcklpsXrvm||unpcklpdXrvm", +"unpckhpsXrvm||unpckhpdXrvm", +"movlhpsXrm$movhpsXrm|movshdupXrm|movhpdXrm", +"movhpsXmr||movhpdXmr", +"$prefetcht!Bm","hintnopVm","hintnopVm","hintnopVm", +"hintnopVm","hintnopVm","hintnopVm","hintnopVm", +--2x +"movUmx$","movUmy$","movUxm$","movUym$","movUmz$",nil,"movUzm$",nil, +"movapsXrm||movapdXrm", +"movapsXmr||movapdXmr", +"cvtpi2psXrMm|cvtsi2ssXrvVmt|cvtpi2pdXrMm|cvtsi2sdXrvVmt", +"movntpsXmr|movntssXmr|movntpdXmr|movntsdXmr", +"cvttps2piMrXm|cvttss2siVrXm|cvttpd2piMrXm|cvttsd2siVrXm", +"cvtps2piMrXm|cvtss2siVrXm|cvtpd2piMrXm|cvtsd2siVrXm", +"ucomissXrm||ucomisdXrm", +"comissXrm||comisdXrm", +--3x +"wrmsr","rdtsc","rdmsr","rdpmc","sysenter","sysexit",nil,"getsec", +"opc3*38",nil,"opc3*3a",nil,nil,nil,nil,nil, +--4x +"cmovoVrm","cmovnoVrm","cmovbVrm","cmovnbVrm", +"cmovzVrm","cmovnzVrm","cmovbeVrm","cmovaVrm", +"cmovsVrm","cmovnsVrm","cmovpeVrm","cmovpoVrm", +"cmovlVrm","cmovgeVrm","cmovleVrm","cmovgVrm", +--5x +"movmskpsVrXm$||movmskpdVrXm$","sqrtpsXrm|sqrtssXrm|sqrtpdXrm|sqrtsdXrm", +"rsqrtpsXrm|rsqrtssXrvm","rcppsXrm|rcpssXrvm", +"andpsXrvm||andpdXrvm","andnpsXrvm||andnpdXrvm", +"orpsXrvm||orpdXrvm","xorpsXrvm||xorpdXrvm", +"addpsXrvm|addssXrvm|addpdXrvm|addsdXrvm","mulpsXrvm|mulssXrvm|mulpdXrvm|mulsdXrvm", +"cvtps2pdXrm|cvtss2sdXrvm|cvtpd2psXrm|cvtsd2ssXrvm", +"cvtdq2psXrm|cvttps2dqXrm|cvtps2dqXrm", +"subpsXrvm|subssXrvm|subpdXrvm|subsdXrvm","minpsXrvm|minssXrvm|minpdXrvm|minsdXrvm", +"divpsXrvm|divssXrvm|divpdXrvm|divsdXrvm","maxpsXrvm|maxssXrvm|maxpdXrvm|maxsdXrvm", +--6x +"punpcklbwPrvm","punpcklwdPrvm","punpckldqPrvm","packsswbPrvm", +"pcmpgtbPrvm","pcmpgtwPrvm","pcmpgtdPrvm","packuswbPrvm", +"punpckhbwPrvm","punpckhwdPrvm","punpckhdqPrvm","packssdwPrvm", +"||punpcklqdqXrvm","||punpckhqdqXrvm", +"movPrVSm","movqMrm|movdquXrm|movdqaXrm", +--7x +"pshufwMrmu|pshufhwXrmu|pshufdXrmu|pshuflwXrmu","pshiftw!Pvmu", +"pshiftd!Pvmu","pshiftq!Mvmu||pshiftdq!Xvmu", +"pcmpeqbPrvm","pcmpeqwPrvm","pcmpeqdPrvm","emms*|", +"vmreadUmr||extrqXmuu$|insertqXrmuu$","vmwriteUrm||extrqXrm$|insertqXrm$", +nil,nil, +"||haddpdXrvm|haddpsXrvm","||hsubpdXrvm|hsubpsXrvm", +"movVSmMr|movqXrm|movVSmXr","movqMmr|movdquXmr|movdqaXmr", +--8x +"joVj","jnoVj","jbVj","jnbVj","jzVj","jnzVj","jbeVj","jaVj", +"jsVj","jnsVj","jpeVj","jpoVj","jlVj","jgeVj","jleVj","jgVj", +--9x +"setoBm","setnoBm","setbBm","setnbBm","setzBm","setnzBm","setbeBm","setaBm", +"setsBm","setnsBm","setpeBm","setpoBm","setlBm","setgeBm","setleBm","setgBm", +--Ax +"push fs","pop fs","cpuid","btVmr","shldVmru","shldVmrc",nil,nil, +"push gs","pop gs","rsm","btsVmr","shrdVmru","shrdVmrc","fxsave!Dmp","imulVrm", +--Bx +"cmpxchgBmr","cmpxchgVmr","$lssVrm","btrVmr", +"$lfsVrm","$lgsVrm","movzxVrBmt","movzxVrWmt", +"|popcntVrm","ud2Dp","bt!Vmu","btcVmr", +"bsfVrm","bsrVrm|lzcntVrm|bsrWrm","movsxVrBmt","movsxVrWmt", +--Cx +"xaddBmr","xaddVmr", +"cmppsXrvmu|cmpssXrvmu|cmppdXrvmu|cmpsdXrvmu","$movntiVmr|", +"pinsrwPrvWmu","pextrwDrPmu", +"shufpsXrvmu||shufpdXrvmu","$cmpxchg!Qmp", +"bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR", +--Dx +"||addsubpdXrvm|addsubpsXrvm","psrlwPrvm","psrldPrvm","psrlqPrvm", +"paddqPrvm","pmullwPrvm", +"|movq2dqXrMm|movqXmr|movdq2qMrXm$","pmovmskbVrMm||pmovmskbVrXm", +"psubusbPrvm","psubuswPrvm","pminubPrvm","pandPrvm", +"paddusbPrvm","padduswPrvm","pmaxubPrvm","pandnPrvm", +--Ex +"pavgbPrvm","psrawPrvm","psradPrvm","pavgwPrvm", +"pmulhuwPrvm","pmulhwPrvm", +"|cvtdq2pdXrm|cvttpd2dqXrm|cvtpd2dqXrm","$movntqMmr||$movntdqXmr", +"psubsbPrvm","psubswPrvm","pminswPrvm","porPrvm", +"paddsbPrvm","paddswPrvm","pmaxswPrvm","pxorPrvm", +--Fx +"|||lddquXrm","psllwPrvm","pslldPrvm","psllqPrvm", +"pmuludqPrvm","pmaddwdPrvm","psadbwPrvm","maskmovqMrm||maskmovdquXrm$", +"psubbPrvm","psubwPrvm","psubdPrvm","psubqPrvm", +"paddbPrvm","paddwPrvm","padddPrvm","ud", +} +assert(map_opc2[255] == "ud") + +-- Map for three-byte opcodes. Can't wait for their next invention. +local map_opc3 = { +["38"] = { -- [66] 0f 38 xx +--0x +[0]="pshufbPrvm","phaddwPrvm","phadddPrvm","phaddswPrvm", +"pmaddubswPrvm","phsubwPrvm","phsubdPrvm","phsubswPrvm", +"psignbPrvm","psignwPrvm","psigndPrvm","pmulhrswPrvm", +"||permilpsXrvm","||permilpdXrvm",nil,nil, +--1x +"||pblendvbXrma",nil,nil,nil, +"||blendvpsXrma","||blendvpdXrma","||permpsXrvm","||ptestXrm", +"||broadcastssXrm","||broadcastsdXrm","||broadcastf128XrlXm",nil, +"pabsbPrm","pabswPrm","pabsdPrm",nil, +--2x +"||pmovsxbwXrm","||pmovsxbdXrm","||pmovsxbqXrm","||pmovsxwdXrm", +"||pmovsxwqXrm","||pmovsxdqXrm",nil,nil, +"||pmuldqXrvm","||pcmpeqqXrvm","||$movntdqaXrm","||packusdwXrvm", +"||maskmovpsXrvm","||maskmovpdXrvm","||maskmovpsXmvr","||maskmovpdXmvr", +--3x +"||pmovzxbwXrm","||pmovzxbdXrm","||pmovzxbqXrm","||pmovzxwdXrm", +"||pmovzxwqXrm","||pmovzxdqXrm","||permdXrvm","||pcmpgtqXrvm", +"||pminsbXrvm","||pminsdXrvm","||pminuwXrvm","||pminudXrvm", +"||pmaxsbXrvm","||pmaxsdXrvm","||pmaxuwXrvm","||pmaxudXrvm", +--4x +"||pmulddXrvm","||phminposuwXrm",nil,nil, +nil,"||psrlvVSXrvm","||psravdXrvm","||psllvVSXrvm", +--5x +[0x58] = "||pbroadcastdXrlXm",[0x59] = "||pbroadcastqXrlXm", +[0x5a] = "||broadcasti128XrlXm", +--7x +[0x78] = "||pbroadcastbXrlXm",[0x79] = "||pbroadcastwXrlXm", +--8x +[0x8c] = "||pmaskmovXrvVSm", +[0x8e] = "||pmaskmovVSmXvr", +--9x +[0x96] = "||fmaddsub132pHXrvm",[0x97] = "||fmsubadd132pHXrvm", +[0x98] = "||fmadd132pHXrvm",[0x99] = "||fmadd132sHXrvm", +[0x9a] = "||fmsub132pHXrvm",[0x9b] = "||fmsub132sHXrvm", +[0x9c] = "||fnmadd132pHXrvm",[0x9d] = "||fnmadd132sHXrvm", +[0x9e] = "||fnmsub132pHXrvm",[0x9f] = "||fnmsub132sHXrvm", +--Ax +[0xa6] = "||fmaddsub213pHXrvm",[0xa7] = "||fmsubadd213pHXrvm", +[0xa8] = "||fmadd213pHXrvm",[0xa9] = "||fmadd213sHXrvm", +[0xaa] = "||fmsub213pHXrvm",[0xab] = "||fmsub213sHXrvm", +[0xac] = "||fnmadd213pHXrvm",[0xad] = "||fnmadd213sHXrvm", +[0xae] = "||fnmsub213pHXrvm",[0xaf] = "||fnmsub213sHXrvm", +--Bx +[0xb6] = "||fmaddsub231pHXrvm",[0xb7] = "||fmsubadd231pHXrvm", +[0xb8] = "||fmadd231pHXrvm",[0xb9] = "||fmadd231sHXrvm", +[0xba] = "||fmsub231pHXrvm",[0xbb] = "||fmsub231sHXrvm", +[0xbc] = "||fnmadd231pHXrvm",[0xbd] = "||fnmadd231sHXrvm", +[0xbe] = "||fnmsub231pHXrvm",[0xbf] = "||fnmsub231sHXrvm", +--Dx +[0xdc] = "||aesencXrvm", [0xdd] = "||aesenclastXrvm", +[0xde] = "||aesdecXrvm", [0xdf] = "||aesdeclastXrvm", +--Fx +[0xf0] = "|||crc32TrBmt",[0xf1] = "|||crc32TrVmt", +[0xf7] = "| sarxVrmv| shlxVrmv| shrxVrmv", +}, + +["3a"] = { -- [66] 0f 3a xx +--0x +[0x00]="||permqXrmu","||permpdXrmu","||pblenddXrvmu",nil, +"||permilpsXrmu","||permilpdXrmu","||perm2f128Xrvmu",nil, +"||roundpsXrmu","||roundpdXrmu","||roundssXrvmu","||roundsdXrvmu", +"||blendpsXrvmu","||blendpdXrvmu","||pblendwXrvmu","palignrPrvmu", +--1x +nil,nil,nil,nil, +"||pextrbVmXru","||pextrwVmXru","||pextrVmSXru","||extractpsVmXru", +"||insertf128XrvlXmu","||extractf128XlXmYru",nil,nil, +nil,nil,nil,nil, +--2x +"||pinsrbXrvVmu","||insertpsXrvmu","||pinsrXrvVmuS",nil, +--3x +[0x38] = "||inserti128Xrvmu",[0x39] = "||extracti128XlXmYru", +--4x +[0x40] = "||dppsXrvmu", +[0x41] = "||dppdXrvmu", +[0x42] = "||mpsadbwXrvmu", +[0x44] = "||pclmulqdqXrvmu", +[0x46] = "||perm2i128Xrvmu", +[0x4a] = "||blendvpsXrvmb",[0x4b] = "||blendvpdXrvmb", +[0x4c] = "||pblendvbXrvmb", +--6x +[0x60] = "||pcmpestrmXrmu",[0x61] = "||pcmpestriXrmu", +[0x62] = "||pcmpistrmXrmu",[0x63] = "||pcmpistriXrmu", +[0xdf] = "||aeskeygenassistXrmu", +--Fx +[0xf0] = "||| rorxVrmu", +}, +} + +-- Map for VMX/SVM opcodes 0F 01 C0-FF (sgdt group with register operands). +local map_opcvm = { +[0xc1]="vmcall",[0xc2]="vmlaunch",[0xc3]="vmresume",[0xc4]="vmxoff", +[0xc8]="monitor",[0xc9]="mwait", +[0xd8]="vmrun",[0xd9]="vmmcall",[0xda]="vmload",[0xdb]="vmsave", +[0xdc]="stgi",[0xdd]="clgi",[0xde]="skinit",[0xdf]="invlpga", +[0xf8]="swapgs",[0xf9]="rdtscp", +} + +-- Map for FP opcodes. And you thought stack machines are simple? +local map_opcfp = { +-- D8-DF 00-BF: opcodes with a memory operand. +-- D8 +[0]="faddFm","fmulFm","fcomFm","fcompFm","fsubFm","fsubrFm","fdivFm","fdivrFm", +"fldFm",nil,"fstFm","fstpFm","fldenvVm","fldcwWm","fnstenvVm","fnstcwWm", +-- DA +"fiaddDm","fimulDm","ficomDm","ficompDm", +"fisubDm","fisubrDm","fidivDm","fidivrDm", +-- DB +"fildDm","fisttpDm","fistDm","fistpDm",nil,"fld twordFmp",nil,"fstp twordFmp", +-- DC +"faddGm","fmulGm","fcomGm","fcompGm","fsubGm","fsubrGm","fdivGm","fdivrGm", +-- DD +"fldGm","fisttpQm","fstGm","fstpGm","frstorDmp",nil,"fnsaveDmp","fnstswWm", +-- DE +"fiaddWm","fimulWm","ficomWm","ficompWm", +"fisubWm","fisubrWm","fidivWm","fidivrWm", +-- DF +"fildWm","fisttpWm","fistWm","fistpWm", +"fbld twordFmp","fildQm","fbstp twordFmp","fistpQm", +-- xx C0-FF: opcodes with a pseudo-register operand. +-- D8 +"faddFf","fmulFf","fcomFf","fcompFf","fsubFf","fsubrFf","fdivFf","fdivrFf", +-- D9 +"fldFf","fxchFf",{"fnop"},nil, +{"fchs","fabs",nil,nil,"ftst","fxam"}, +{"fld1","fldl2t","fldl2e","fldpi","fldlg2","fldln2","fldz"}, +{"f2xm1","fyl2x","fptan","fpatan","fxtract","fprem1","fdecstp","fincstp"}, +{"fprem","fyl2xp1","fsqrt","fsincos","frndint","fscale","fsin","fcos"}, +-- DA +"fcmovbFf","fcmoveFf","fcmovbeFf","fcmovuFf",nil,{nil,"fucompp"},nil,nil, +-- DB +"fcmovnbFf","fcmovneFf","fcmovnbeFf","fcmovnuFf", +{nil,nil,"fnclex","fninit"},"fucomiFf","fcomiFf",nil, +-- DC +"fadd toFf","fmul toFf",nil,nil, +"fsub toFf","fsubr toFf","fdivr toFf","fdiv toFf", +-- DD +"ffreeFf",nil,"fstFf","fstpFf","fucomFf","fucompFf",nil,nil, +-- DE +"faddpFf","fmulpFf",nil,{nil,"fcompp"}, +"fsubrpFf","fsubpFf","fdivrpFf","fdivpFf", +-- DF +nil,nil,nil,nil,{"fnstsw ax"},"fucomipFf","fcomipFf",nil, +} +assert(map_opcfp[126] == "fcomipFf") + +-- Map for opcode groups. The subkey is sp from the ModRM byte. +local map_opcgroup = { + arith = { "add", "or", "adc", "sbb", "and", "sub", "xor", "cmp" }, + shift = { "rol", "ror", "rcl", "rcr", "shl", "shr", "sal", "sar" }, + testb = { "testBmi", "testBmi", "not", "neg", "mul", "imul", "div", "idiv" }, + testv = { "testVmi", "testVmi", "not", "neg", "mul", "imul", "div", "idiv" }, + incb = { "inc", "dec" }, + incd = { "inc", "dec", "callUmp", "$call farDmp", + "jmpUmp", "$jmp farDmp", "pushUm" }, + sldt = { "sldt", "str", "lldt", "ltr", "verr", "verw" }, + sgdt = { "vm*$sgdt", "vm*$sidt", "$lgdt", "vm*$lidt", + "smsw", nil, "lmsw", "vm*$invlpg" }, + bt = { nil, nil, nil, nil, "bt", "bts", "btr", "btc" }, + cmpxchg = { nil, "sz*,cmpxchg8bQmp,cmpxchg16bXmp", nil, nil, + nil, nil, "vmptrld|vmxon|vmclear", "vmptrst" }, + pshiftw = { nil, nil, "psrlw", nil, "psraw", nil, "psllw" }, + pshiftd = { nil, nil, "psrld", nil, "psrad", nil, "pslld" }, + pshiftq = { nil, nil, "psrlq", nil, nil, nil, "psllq" }, + pshiftdq = { nil, nil, "psrlq", "psrldq", nil, nil, "psllq", "pslldq" }, + fxsave = { "$fxsave", "$fxrstor", "$ldmxcsr", "$stmxcsr", + nil, "lfenceDp$", "mfenceDp$", "sfenceDp$clflush" }, + prefetch = { "prefetch", "prefetchw" }, + prefetcht = { "prefetchnta", "prefetcht0", "prefetcht1", "prefetcht2" }, +} + +------------------------------------------------------------------------------ + +-- Maps for register names. +local map_regs = { + B = { "al", "cl", "dl", "bl", "ah", "ch", "dh", "bh", + "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, + B64 = { "al", "cl", "dl", "bl", "spl", "bpl", "sil", "dil", + "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, + W = { "ax", "cx", "dx", "bx", "sp", "bp", "si", "di", + "r8w", "r9w", "r10w", "r11w", "r12w", "r13w", "r14w", "r15w" }, + D = { "eax", "ecx", "edx", "ebx", "esp", "ebp", "esi", "edi", + "r8d", "r9d", "r10d", "r11d", "r12d", "r13d", "r14d", "r15d" }, + Q = { "rax", "rcx", "rdx", "rbx", "rsp", "rbp", "rsi", "rdi", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15" }, + M = { "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7", + "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7" }, -- No x64 ext! + X = { "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7", + "xmm8", "xmm9", "xmm10", "xmm11", "xmm12", "xmm13", "xmm14", "xmm15" }, + Y = { "ymm0", "ymm1", "ymm2", "ymm3", "ymm4", "ymm5", "ymm6", "ymm7", + "ymm8", "ymm9", "ymm10", "ymm11", "ymm12", "ymm13", "ymm14", "ymm15" }, +} +local map_segregs = { "es", "cs", "ss", "ds", "fs", "gs", "segr6", "segr7" } + +-- Maps for size names. +local map_sz2n = { + B = 1, W = 2, D = 4, Q = 8, M = 8, X = 16, Y = 32, +} +local map_sz2prefix = { + B = "byte", W = "word", D = "dword", + Q = "qword", + M = "qword", X = "xword", Y = "yword", + F = "dword", G = "qword", -- No need for sizes/register names for these two. +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local code, pos, hex = ctx.code, ctx.pos, "" + local hmax = ctx.hexdump + if hmax > 0 then + for i=ctx.start,pos-1 do + hex = hex..format("%02X", byte(code, i, i)) + end + if #hex > hmax then hex = sub(hex, 1, hmax)..". " + else hex = hex..rep(" ", hmax-#hex+2) end + end + if operands then text = text.." "..operands end + if ctx.o16 then text = "o16 "..text; ctx.o16 = false end + if ctx.a32 then text = "a32 "..text; ctx.a32 = false end + if ctx.rep then text = ctx.rep.." "..text; ctx.rep = false end + if ctx.rex then + local t = (ctx.rexw and "w" or "")..(ctx.rexr and "r" or "").. + (ctx.rexx and "x" or "")..(ctx.rexb and "b" or "").. + (ctx.vexl and "l" or "") + if ctx.vexv and ctx.vexv ~= 0 then t = t.."v"..ctx.vexv end + if t ~= "" then text = ctx.rex.."."..t.." "..gsub(text, "^ ", "") + elseif ctx.rex == "vex" then text = gsub("v"..text, "^v ", "") end + ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false + ctx.rex = false; ctx.vexl = false; ctx.vexv = false + end + if ctx.seg then + local text2, n = gsub(text, "%[", "["..ctx.seg..":") + if n == 0 then text = ctx.seg.." "..text else text = text2 end + ctx.seg = false + end + if ctx.lock then text = "lock "..text; ctx.lock = false end + local imm = ctx.imm + if imm then + local sym = ctx.symtab[imm] + if sym then text = text.."\t->"..sym end + end + ctx.out(format("%08x %s%s\n", ctx.addr+ctx.start, hex, text)) + ctx.mrm = false + ctx.vexv = false + ctx.start = pos + ctx.imm = nil +end + +-- Clear all prefix flags. +local function clearprefixes(ctx) + ctx.o16 = false; ctx.seg = false; ctx.lock = false; ctx.rep = false + ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false + ctx.rex = false; ctx.a32 = false; ctx.vexl = false +end + +-- Fallback for incomplete opcodes at the end. +local function incomplete(ctx) + ctx.pos = ctx.stop+1 + clearprefixes(ctx) + return putop(ctx, "(incomplete)") +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + clearprefixes(ctx) + return putop(ctx, "(unknown)") +end + +-- Return an immediate of the specified size. +local function getimm(ctx, pos, n) + if pos+n-1 > ctx.stop then return incomplete(ctx) end + local code = ctx.code + if n == 1 then + local b1 = byte(code, pos, pos) + return b1 + elseif n == 2 then + local b1, b2 = byte(code, pos, pos+1) + return b1+b2*256 + else + local b1, b2, b3, b4 = byte(code, pos, pos+3) + local imm = b1+b2*256+b3*65536+b4*16777216 + ctx.imm = imm + return imm + end +end + +-- Process pattern string and generate the operands. +local function putpat(ctx, name, pat) + local operands, regs, sz, mode, sp, rm, sc, rx, sdisp + local code, pos, stop, vexl = ctx.code, ctx.pos, ctx.stop, ctx.vexl + + -- Chars used: 1DFGHIMPQRSTUVWXYabcdfgijlmoprstuvwxyz + for p in gmatch(pat, ".") do + local x = nil + if p == "V" or p == "U" then + if ctx.rexw then sz = "Q"; ctx.rexw = false + elseif ctx.o16 then sz = "W"; ctx.o16 = false + elseif p == "U" and ctx.x64 then sz = "Q" + else sz = "D" end + regs = map_regs[sz] + elseif p == "T" then + if ctx.rexw then sz = "Q"; ctx.rexw = false else sz = "D" end + regs = map_regs[sz] + elseif p == "B" then + sz = "B" + regs = ctx.rex and map_regs.B64 or map_regs.B + elseif match(p, "[WDQMXYFG]") then + sz = p + if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end + regs = map_regs[sz] + elseif p == "P" then + sz = ctx.o16 and "X" or "M"; ctx.o16 = false + if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end + regs = map_regs[sz] + elseif p == "H" then + name = name..(ctx.rexw and "d" or "s") + ctx.rexw = false + elseif p == "S" then + name = name..lower(sz) + elseif p == "s" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = imm <= 127 and format("+0x%02x", imm) + or format("-0x%02x", 256-imm) + pos = pos+1 + elseif p == "u" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = format("0x%02x", imm) + pos = pos+1 + elseif p == "b" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = regs[imm/16+1] + pos = pos+1 + elseif p == "w" then + local imm = getimm(ctx, pos, 2); if not imm then return end + x = format("0x%x", imm) + pos = pos+2 + elseif p == "o" then -- [offset] + if ctx.x64 then + local imm1 = getimm(ctx, pos, 4); if not imm1 then return end + local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end + x = format("[0x%08x%08x]", imm2, imm1) + pos = pos+8 + else + local imm = getimm(ctx, pos, 4); if not imm then return end + x = format("[0x%08x]", imm) + pos = pos+4 + end + elseif p == "i" or p == "I" then + local n = map_sz2n[sz] + if n == 8 and ctx.x64 and p == "I" then + local imm1 = getimm(ctx, pos, 4); if not imm1 then return end + local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end + x = format("0x%08x%08x", imm2, imm1) + else + if n == 8 then n = 4 end + local imm = getimm(ctx, pos, n); if not imm then return end + if sz == "Q" and (imm < 0 or imm > 0x7fffffff) then + imm = (0xffffffff+1)-imm + x = format(imm > 65535 and "-0x%08x" or "-0x%x", imm) + else + x = format(imm > 65535 and "0x%08x" or "0x%x", imm) + end + end + pos = pos+n + elseif p == "j" then + local n = map_sz2n[sz] + if n == 8 then n = 4 end + local imm = getimm(ctx, pos, n); if not imm then return end + if sz == "B" and imm > 127 then imm = imm-256 + elseif imm > 2147483647 then imm = imm-4294967296 end + pos = pos+n + imm = imm + pos + ctx.addr + if imm > 4294967295 and not ctx.x64 then imm = imm-4294967296 end + ctx.imm = imm + if sz == "W" then + x = format("word 0x%04x", imm%65536) + elseif ctx.x64 then + local lo = imm % 0x1000000 + x = format("0x%02x%06x", (imm-lo) / 0x1000000, lo) + else + x = "0x"..tohex(imm) + end + elseif p == "R" then + local r = byte(code, pos-1, pos-1)%8 + if ctx.rexb then r = r + 8; ctx.rexb = false end + x = regs[r+1] + elseif p == "a" then x = regs[1] + elseif p == "c" then x = "cl" + elseif p == "d" then x = "dx" + elseif p == "1" then x = "1" + else + if not mode then + mode = ctx.mrm + if not mode then + if pos > stop then return incomplete(ctx) end + mode = byte(code, pos, pos) + pos = pos+1 + end + rm = mode%8; mode = (mode-rm)/8 + sp = mode%8; mode = (mode-sp)/8 + sdisp = "" + if mode < 3 then + if rm == 4 then + if pos > stop then return incomplete(ctx) end + sc = byte(code, pos, pos) + pos = pos+1 + rm = sc%8; sc = (sc-rm)/8 + rx = sc%8; sc = (sc-rx)/8 + if ctx.rexx then rx = rx + 8; ctx.rexx = false end + if rx == 4 then rx = nil end + end + if mode > 0 or rm == 5 then + local dsz = mode + if dsz ~= 1 then dsz = 4 end + local disp = getimm(ctx, pos, dsz); if not disp then return end + if mode == 0 then rm = nil end + if rm or rx or (not sc and ctx.x64 and not ctx.a32) then + if dsz == 1 and disp > 127 then + sdisp = format("-0x%x", 256-disp) + elseif disp >= 0 and disp <= 0x7fffffff then + sdisp = format("+0x%x", disp) + else + sdisp = format("-0x%x", (0xffffffff+1)-disp) + end + else + sdisp = format(ctx.x64 and not ctx.a32 and + not (disp >= 0 and disp <= 0x7fffffff) + and "0xffffffff%08x" or "0x%08x", disp) + end + pos = pos+dsz + end + end + if rm and ctx.rexb then rm = rm + 8; ctx.rexb = false end + if ctx.rexr then sp = sp + 8; ctx.rexr = false end + end + if p == "m" then + if mode == 3 then x = regs[rm+1] + else + local aregs = ctx.a32 and map_regs.D or ctx.aregs + local srm, srx = "", "" + if rm then srm = aregs[rm+1] + elseif not sc and ctx.x64 and not ctx.a32 then srm = "rip" end + ctx.a32 = false + if rx then + if rm then srm = srm.."+" end + srx = aregs[rx+1] + if sc > 0 then srx = srx.."*"..(2^sc) end + end + x = format("[%s%s%s]", srm, srx, sdisp) + end + if mode < 3 and + (not match(pat, "[aRrgp]") or match(pat, "t")) then -- Yuck. + x = map_sz2prefix[sz].." "..x + end + elseif p == "r" then x = regs[sp+1] + elseif p == "g" then x = map_segregs[sp+1] + elseif p == "p" then -- Suppress prefix. + elseif p == "f" then x = "st"..rm + elseif p == "x" then + if sp == 0 and ctx.lock and not ctx.x64 then + x = "CR8"; ctx.lock = false + else + x = "CR"..sp + end + elseif p == "v" then + if ctx.vexv then + x = regs[ctx.vexv+1]; ctx.vexv = false + end + elseif p == "y" then x = "DR"..sp + elseif p == "z" then x = "TR"..sp + elseif p == "l" then vexl = false + elseif p == "t" then + else + error("bad pattern `"..pat.."'") + end + end + if x then operands = operands and operands..", "..x or x end + end + ctx.pos = pos + return putop(ctx, name, operands) +end + +-- Forward declaration. +local map_act + +-- Fetch and cache MRM byte. +local function getmrm(ctx) + local mrm = ctx.mrm + if not mrm then + local pos = ctx.pos + if pos > ctx.stop then return nil end + mrm = byte(ctx.code, pos, pos) + ctx.pos = pos+1 + ctx.mrm = mrm + end + return mrm +end + +-- Dispatch to handler depending on pattern. +local function dispatch(ctx, opat, patgrp) + if not opat then return unknown(ctx) end + if match(opat, "%|") then -- MMX/SSE variants depending on prefix. + local p + if ctx.rep then + p = ctx.rep=="rep" and "%|([^%|]*)" or "%|[^%|]*%|[^%|]*%|([^%|]*)" + ctx.rep = false + elseif ctx.o16 then p = "%|[^%|]*%|([^%|]*)"; ctx.o16 = false + else p = "^[^%|]*" end + opat = match(opat, p) + if not opat then return unknown(ctx) end +-- ctx.rep = false; ctx.o16 = false + --XXX fails for 66 f2 0f 38 f1 06 crc32 eax,WORD PTR [esi] + --XXX remove in branches? + end + if match(opat, "%$") then -- reg$mem variants. + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + opat = match(opat, mrm >= 192 and "^[^%$]*" or "%$(.*)") + if opat == "" then return unknown(ctx) end + end + if opat == "" then return unknown(ctx) end + local name, pat = match(opat, "^([a-z0-9 ]*)(.*)") + if pat == "" and patgrp then pat = patgrp end + return map_act[sub(pat, 1, 1)](ctx, name, pat) +end + +-- Get a pattern from an opcode map and dispatch to handler. +local function dispatchmap(ctx, opcmap) + local pos = ctx.pos + local opat = opcmap[byte(ctx.code, pos, pos)] + pos = pos + 1 + ctx.pos = pos + return dispatch(ctx, opat) +end + +-- Map for action codes. The key is the first char after the name. +map_act = { + -- Simple opcodes without operands. + [""] = function(ctx, name, pat) + return putop(ctx, name) + end, + + -- Operand size chars fall right through. + B = putpat, W = putpat, D = putpat, Q = putpat, + V = putpat, U = putpat, T = putpat, + M = putpat, X = putpat, P = putpat, + F = putpat, G = putpat, Y = putpat, + H = putpat, + + -- Collect prefixes. + [":"] = function(ctx, name, pat) + ctx[pat == ":" and name or sub(pat, 2)] = name + if ctx.pos - ctx.start > 5 then return unknown(ctx) end -- Limit #prefixes. + end, + + -- Chain to special handler specified by name. + ["*"] = function(ctx, name, pat) + return map_act[name](ctx, name, sub(pat, 2)) + end, + + -- Use named subtable for opcode group. + ["!"] = function(ctx, name, pat) + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + return dispatch(ctx, map_opcgroup[name][((mrm-(mrm%8))/8)%8+1], sub(pat, 2)) + end, + + -- o16,o32[,o64] variants. + sz = function(ctx, name, pat) + if ctx.o16 then ctx.o16 = false + else + pat = match(pat, ",(.*)") + if ctx.rexw then + local p = match(pat, ",(.*)") + if p then pat = p; ctx.rexw = false end + end + end + pat = match(pat, "^[^,]*") + return dispatch(ctx, pat) + end, + + -- Two-byte opcode dispatch. + opc2 = function(ctx, name, pat) + return dispatchmap(ctx, map_opc2) + end, + + -- Three-byte opcode dispatch. + opc3 = function(ctx, name, pat) + return dispatchmap(ctx, map_opc3[pat]) + end, + + -- VMX/SVM dispatch. + vm = function(ctx, name, pat) + return dispatch(ctx, map_opcvm[ctx.mrm]) + end, + + -- Floating point opcode dispatch. + fp = function(ctx, name, pat) + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + local rm = mrm%8 + local idx = pat*8 + ((mrm-rm)/8)%8 + if mrm >= 192 then idx = idx + 64 end + local opat = map_opcfp[idx] + if type(opat) == "table" then opat = opat[rm+1] end + return dispatch(ctx, opat) + end, + + -- REX prefix. + rex = function(ctx, name, pat) + if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. + for p in gmatch(pat, ".") do ctx["rex"..p] = true end + ctx.rex = "rex" + end, + + -- VEX prefix. + vex = function(ctx, name, pat) + if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. + ctx.rex = "vex" + local pos = ctx.pos + if ctx.mrm then + ctx.mrm = nil + pos = pos-1 + end + local b = byte(ctx.code, pos, pos) + if not b then return incomplete(ctx) end + pos = pos+1 + if b < 128 then ctx.rexr = true end + local m = 1 + if pat == "3" then + m = b%32; b = (b-m)/32 + local nb = b%2; b = (b-nb)/2 + if nb == 0 then ctx.rexb = true end + local nx = b%2 + if nx == 0 then ctx.rexx = true end + b = byte(ctx.code, pos, pos) + if not b then return incomplete(ctx) end + pos = pos+1 + if b >= 128 then ctx.rexw = true end + end + ctx.pos = pos + local map + if m == 1 then map = map_opc2 + elseif m == 2 then map = map_opc3["38"] + elseif m == 3 then map = map_opc3["3a"] + else return unknown(ctx) end + local p = b%4; b = (b-p)/4 + if p == 1 then ctx.o16 = "o16" + elseif p == 2 then ctx.rep = "rep" + elseif p == 3 then ctx.rep = "repne" end + local l = b%2; b = (b-l)/2 + if l ~= 0 then ctx.vexl = true end + ctx.vexv = (-1-b)%16 + return dispatchmap(ctx, map) + end, + + -- Special case for nop with REX prefix. + nop = function(ctx, name, pat) + return dispatch(ctx, ctx.rex and pat or "nop") + end, + + -- Special case for 0F 77. + emms = function(ctx, name, pat) + if ctx.rex ~= "vex" then + return putop(ctx, "emms") + elseif ctx.vexl then + ctx.vexl = false + return putop(ctx, "zeroall") + else + return putop(ctx, "zeroupper") + end + end, +} + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ofs = ofs + 1 + ctx.start = ofs + ctx.pos = ofs + ctx.stop = stop + ctx.imm = nil + ctx.mrm = false + clearprefixes(ctx) + while ctx.pos <= stop do dispatchmap(ctx, ctx.map1) end + if ctx.pos ~= ctx.start then incomplete(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = (addr or 0) - 1 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 16 + ctx.x64 = false + ctx.map1 = map_opc1_32 + ctx.aregs = map_regs.D + return ctx +end + +local function create64(code, addr, out) + local ctx = create(code, addr, out) + ctx.x64 = true + ctx.map1 = map_opc1_64 + ctx.aregs = map_regs.Q + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +local function disass64(code, addr, out) + create64(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 8 then return map_regs.D[r+1] end + return map_regs.X[r-7] +end + +local function regname64(r) + if r < 16 then return map_regs.Q[r+1] end + return map_regs.X[r-15] +end + +-- Public module functions. +return { + create = create, + create64 = create64, + disass = disass, + disass64 = disass64, + regname = regname, + regname64 = regname64 +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/dump.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/dump.lua new file mode 100644 index 00000000..9eda08c4 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/dump.lua @@ -0,0 +1,726 @@ +---------------------------------------------------------------------------- +-- LuaJIT compiler dump module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module can be used to debug the JIT compiler itself. It dumps the +-- code representations and structures used in various compiler stages. +-- +-- Example usage: +-- +-- luajit -jdump -e "local x=0; for i=1,1e6 do x=x+i end; print(x)" +-- luajit -jdump=im -e "for i=1,1000 do for j=1,1000 do end end" | less -R +-- luajit -jdump=is myapp.lua | less -R +-- luajit -jdump=-b myapp.lua +-- luajit -jdump=+aH,myapp.html myapp.lua +-- luajit -jdump=ixT,myapp.dump myapp.lua +-- +-- The first argument specifies the dump mode. The second argument gives +-- the output file name. Default output is to stdout, unless the environment +-- variable LUAJIT_DUMPFILE is set. The file is overwritten every time the +-- module is started. +-- +-- Different features can be turned on or off with the dump mode. If the +-- mode starts with a '+', the following features are added to the default +-- set of features; a '-' removes them. Otherwise the features are replaced. +-- +-- The following dump features are available (* marks the default): +-- +-- * t Print a line for each started, ended or aborted trace (see also -jv). +-- * b Dump the traced bytecode. +-- * i Dump the IR (intermediate representation). +-- r Augment the IR with register/stack slots. +-- s Dump the snapshot map. +-- * m Dump the generated machine code. +-- x Print each taken trace exit. +-- X Print each taken trace exit and the contents of all registers. +-- a Print the IR of aborted traces, too. +-- +-- The output format can be set with the following characters: +-- +-- T Plain text output. +-- A ANSI-colored text output +-- H Colorized HTML + CSS output. +-- +-- The default output format is plain text. It's set to ANSI-colored text +-- if the COLORTERM variable is set. Note: this is independent of any output +-- redirection, which is actually considered a feature. +-- +-- You probably want to use less -R to enjoy viewing ANSI-colored text from +-- a pipe or a file. Add this to your ~/.bashrc: export LESS="-R" +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local funcinfo, funcbc = jutil.funcinfo, jutil.funcbc +local traceinfo, traceir, tracek = jutil.traceinfo, jutil.traceir, jutil.tracek +local tracemc, tracesnap = jutil.tracemc, jutil.tracesnap +local traceexitstub, ircalladdr = jutil.traceexitstub, jutil.ircalladdr +local bit = require("bit") +local band, shr, tohex = bit.band, bit.rshift, bit.tohex +local sub, gsub, format = string.sub, string.gsub, string.format +local byte, rep = string.byte, string.rep +local type, tostring = type, tostring +local stdout, stderr = io.stdout, io.stderr + +-- Load other modules on-demand. +local bcline, disass + +-- Active flag, output file handle and dump mode. +local active, out, dumpmode + +------------------------------------------------------------------------------ + +local symtabmt = { __index = false } +local symtab = {} +local nexitsym = 0 + +-- Fill nested symbol table with per-trace exit stub addresses. +local function fillsymtab_tr(tr, nexit) + local t = {} + symtabmt.__index = t + if jit.arch:sub(1, 4) == "mips" then + t[traceexitstub(tr, 0)] = "exit" + return + end + for i=0,nexit-1 do + local addr = traceexitstub(tr, i) + if addr < 0 then addr = addr + 2^32 end + t[addr] = tostring(i) + end + local addr = traceexitstub(tr, nexit) + if addr then t[addr] = "stack_check" end +end + +-- Fill symbol table with trace exit stub addresses. +local function fillsymtab(tr, nexit) + local t = symtab + if nexitsym == 0 then + local maskaddr = jit.arch == "arm" and -2 + local ircall = vmdef.ircall + for i=0,#ircall do + local addr = ircalladdr(i) + if addr ~= 0 then + if maskaddr then addr = band(addr, maskaddr) end + if addr < 0 then addr = addr + 2^32 end + t[addr] = ircall[i] + end + end + end + if nexitsym == 1000000 then -- Per-trace exit stubs. + fillsymtab_tr(tr, nexit) + elseif nexit > nexitsym then -- Shared exit stubs. + for i=nexitsym,nexit-1 do + local addr = traceexitstub(i) + if addr == nil then -- Fall back to per-trace exit stubs. + fillsymtab_tr(tr, nexit) + setmetatable(symtab, symtabmt) + nexit = 1000000 + break + end + if addr < 0 then addr = addr + 2^32 end + t[addr] = tostring(i) + end + nexitsym = nexit + end + return t +end + +local function dumpwrite(s) + out:write(s) +end + +-- Disassemble machine code. +local function dump_mcode(tr) + local info = traceinfo(tr) + if not info then return end + local mcode, addr, loop = tracemc(tr) + if not mcode then return end + if not disass then disass = require("jit.dis_"..jit.arch) end + if addr < 0 then addr = addr + 2^32 end + out:write("---- TRACE ", tr, " mcode ", #mcode, "\n") + local ctx = disass.create(mcode, addr, dumpwrite) + ctx.hexdump = 0 + ctx.symtab = fillsymtab(tr, info.nexit) + if loop ~= 0 then + symtab[addr+loop] = "LOOP" + ctx:disass(0, loop) + out:write("->LOOP:\n") + ctx:disass(loop, #mcode-loop) + symtab[addr+loop] = nil + else + ctx:disass(0, #mcode) + end +end + +------------------------------------------------------------------------------ + +local irtype_text = { + [0] = "nil", + "fal", + "tru", + "lud", + "str", + "p32", + "thr", + "pro", + "fun", + "p64", + "cdt", + "tab", + "udt", + "flt", + "num", + "i8 ", + "u8 ", + "i16", + "u16", + "int", + "u32", + "i64", + "u64", + "sfp", +} + +local colortype_ansi = { + [0] = "%s", + "%s", + "%s", + "\027[36m%s\027[m", + "\027[32m%s\027[m", + "%s", + "\027[1m%s\027[m", + "%s", + "\027[1m%s\027[m", + "%s", + "\027[33m%s\027[m", + "\027[31m%s\027[m", + "\027[36m%s\027[m", + "\027[34m%s\027[m", + "\027[34m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", +} + +local function colorize_text(s) + return s +end + +local function colorize_ansi(s, t, extra) + local out = format(colortype_ansi[t], s) + if extra then out = "\027[3m"..out end + return out +end + +local irtype_ansi = setmetatable({}, + { __index = function(tab, t) + local s = colorize_ansi(irtype_text[t], t); tab[t] = s; return s; end }) + +local html_escape = { ["<"] = "<", [">"] = ">", ["&"] = "&", } + +local function colorize_html(s, t, extra) + s = gsub(s, "[<>&]", html_escape) + return format('%s', + irtype_text[t], extra and " irt_extra" or "", s) +end + +local irtype_html = setmetatable({}, + { __index = function(tab, t) + local s = colorize_html(irtype_text[t], t); tab[t] = s; return s; end }) + +local header_html = [[ + +]] + +local colorize, irtype + +-- Lookup tables to convert some literals into names. +local litname = { + ["SLOAD "] = setmetatable({}, { __index = function(t, mode) + local s = "" + if band(mode, 1) ~= 0 then s = s.."P" end + if band(mode, 2) ~= 0 then s = s.."F" end + if band(mode, 4) ~= 0 then s = s.."T" end + if band(mode, 8) ~= 0 then s = s.."C" end + if band(mode, 16) ~= 0 then s = s.."R" end + if band(mode, 32) ~= 0 then s = s.."I" end + if band(mode, 64) ~= 0 then s = s.."K" end + t[mode] = s + return s + end}), + ["XLOAD "] = { [0] = "", "R", "V", "RV", "U", "RU", "VU", "RVU", }, + ["CONV "] = setmetatable({}, { __index = function(t, mode) + local s = irtype[band(mode, 31)] + s = irtype[band(shr(mode, 5), 31)].."."..s + if band(mode, 0x800) ~= 0 then s = s.." sext" end + local c = shr(mode, 12) + if c == 1 then s = s.." none" + elseif c == 2 then s = s.." index" + elseif c == 3 then s = s.." check" end + t[mode] = s + return s + end}), + ["FLOAD "] = vmdef.irfield, + ["FREF "] = vmdef.irfield, + ["FPMATH"] = vmdef.irfpm, + ["TMPREF"] = { [0] = "", "IN", "OUT", "INOUT", "", "", "OUT2", "INOUT2" }, + ["BUFHDR"] = { [0] = "RESET", "APPEND", "WRITE" }, + ["TOSTR "] = { [0] = "INT", "NUM", "CHAR" }, +} + +local function ctlsub(c) + if c == "\n" then return "\\n" + elseif c == "\r" then return "\\r" + elseif c == "\t" then return "\\t" + else return format("\\%03d", byte(c)) + end +end + +local function fmtfunc(func, pc) + local fi = funcinfo(func, pc) + if fi.loc then + return fi.loc + elseif fi.ffid then + return vmdef.ffnames[fi.ffid] + elseif fi.addr then + return format("C:%x", fi.addr) + else + return "(?)" + end +end + +local function formatk(tr, idx, sn) + local k, t, slot = tracek(tr, idx) + local tn = type(k) + local s + if tn == "number" then + if t < 12 then + s = k == 0 and "NULL" or format("[0x%08x]", k) + elseif band(sn or 0, 0x30000) ~= 0 then + s = band(sn, 0x20000) ~= 0 and "contpc" or "ftsz" + elseif k == 2^52+2^51 then + s = "bias" + else + s = format(0 < k and k < 0x1p-1026 and "%+a" or "%+.14g", k) + end + elseif tn == "string" then + s = format(#k > 20 and '"%.20s"~' or '"%s"', gsub(k, "%c", ctlsub)) + elseif tn == "function" then + s = fmtfunc(k) + elseif tn == "table" then + s = format("{%p}", k) + elseif tn == "userdata" then + if t == 12 then + s = format("userdata:%p", k) + else + s = format("[%p]", k) + if s == "[NULL]" then s = "NULL" end + end + elseif t == 21 then -- int64_t + s = sub(tostring(k), 1, -3) + if sub(s, 1, 1) ~= "-" then s = "+"..s end + elseif sn == 0x1057fff then -- SNAP(1, SNAP_FRAME | SNAP_NORESTORE, REF_NIL) + return "----" -- Special case for LJ_FR2 slot 1. + else + s = tostring(k) -- For primitives. + end + s = colorize(format("%-4s", s), t, band(sn or 0, 0x100000) ~= 0) + if slot then + s = format("%s @%d", s, slot) + end + return s +end + +local function printsnap(tr, snap) + local n = 2 + for s=0,snap[1]-1 do + local sn = snap[n] + if shr(sn, 24) == s then + n = n + 1 + local ref = band(sn, 0xffff) - 0x8000 -- REF_BIAS + if ref < 0 then + out:write(formatk(tr, ref, sn)) + elseif band(sn, 0x80000) ~= 0 then -- SNAP_SOFTFPNUM + out:write(colorize(format("%04d/%04d", ref, ref+1), 14)) + else + local m, ot, op1, op2 = traceir(tr, ref) + out:write(colorize(format("%04d", ref), band(ot, 31), band(sn, 0x100000) ~= 0)) + end + out:write(band(sn, 0x10000) == 0 and " " or "|") -- SNAP_FRAME + else + out:write("---- ") + end + end + out:write("]\n") +end + +-- Dump snapshots (not interleaved with IR). +local function dump_snap(tr) + out:write("---- TRACE ", tr, " snapshots\n") + for i=0,1000000000 do + local snap = tracesnap(tr, i) + if not snap then break end + out:write(format("#%-3d %04d [ ", i, snap[0])) + printsnap(tr, snap) + end +end + +-- Return a register name or stack slot for a rid/sp location. +local function ridsp_name(ridsp, ins) + if not disass then disass = require("jit.dis_"..jit.arch) end + local rid, slot = band(ridsp, 0xff), shr(ridsp, 8) + if rid == 253 or rid == 254 then + return (slot == 0 or slot == 255) and " {sink" or format(" {%04d", ins-slot) + end + if ridsp > 255 then return format("[%x]", slot*4) end + if rid < 128 then return disass.regname(rid) end + return "" +end + +-- Dump CALL* function ref and return optional ctype. +local function dumpcallfunc(tr, ins) + local ctype + if ins > 0 then + local m, ot, op1, op2 = traceir(tr, ins) + if band(ot, 31) == 0 then -- nil type means CARG(func, ctype). + ins = op1 + ctype = formatk(tr, op2) + end + end + if ins < 0 then + out:write(format("[0x%x](", tonumber((tracek(tr, ins))))) + else + out:write(format("%04d (", ins)) + end + return ctype +end + +-- Recursively gather CALL* args and dump them. +local function dumpcallargs(tr, ins) + if ins < 0 then + out:write(formatk(tr, ins)) + else + local m, ot, op1, op2 = traceir(tr, ins) + local oidx = 6*shr(ot, 8) + local op = sub(vmdef.irnames, oidx+1, oidx+6) + if op == "CARG " then + dumpcallargs(tr, op1) + if op2 < 0 then + out:write(" ", formatk(tr, op2)) + else + out:write(" ", format("%04d", op2)) + end + else + out:write(format("%04d", ins)) + end + end +end + +-- Dump IR and interleaved snapshots. +local function dump_ir(tr, dumpsnap, dumpreg) + local info = traceinfo(tr) + if not info then return end + local nins = info.nins + out:write("---- TRACE ", tr, " IR\n") + local irnames = vmdef.irnames + local snapref = 65536 + local snap, snapno + if dumpsnap then + snap = tracesnap(tr, 0) + snapref = snap[0] + snapno = 0 + end + for ins=1,nins do + if ins >= snapref then + if dumpreg then + out:write(format(".... SNAP #%-3d [ ", snapno)) + else + out:write(format(".... SNAP #%-3d [ ", snapno)) + end + printsnap(tr, snap) + snapno = snapno + 1 + snap = tracesnap(tr, snapno) + snapref = snap and snap[0] or 65536 + end + local m, ot, op1, op2, ridsp = traceir(tr, ins) + local oidx, t = 6*shr(ot, 8), band(ot, 31) + local op = sub(irnames, oidx+1, oidx+6) + if op == "LOOP " then + if dumpreg then + out:write(format("%04d ------------ LOOP ------------\n", ins)) + else + out:write(format("%04d ------ LOOP ------------\n", ins)) + end + elseif op ~= "NOP " and op ~= "CARG " and + (dumpreg or op ~= "RENAME") then + local rid = band(ridsp, 255) + if dumpreg then + out:write(format("%04d %-6s", ins, ridsp_name(ridsp, ins))) + else + out:write(format("%04d ", ins)) + end + out:write(format("%s%s %s %s ", + (rid == 254 or rid == 253) and "}" or + (band(ot, 128) == 0 and " " or ">"), + band(ot, 64) == 0 and " " or "+", + irtype[t], op)) + local m1, m2 = band(m, 3), band(m, 3*4) + if sub(op, 1, 4) == "CALL" then + local ctype + if m2 == 1*4 then -- op2 == IRMlit + out:write(format("%-10s (", vmdef.ircall[op2])) + else + ctype = dumpcallfunc(tr, op2) + end + if op1 ~= -1 then dumpcallargs(tr, op1) end + out:write(")") + if ctype then out:write(" ctype ", ctype) end + elseif op == "CNEW " and op2 == -1 then + out:write(formatk(tr, op1)) + elseif m1 ~= 3 then -- op1 != IRMnone + if op1 < 0 then + out:write(formatk(tr, op1)) + else + out:write(format(m1 == 0 and "%04d" or "#%-3d", op1)) + end + if m2 ~= 3*4 then -- op2 != IRMnone + if m2 == 1*4 then -- op2 == IRMlit + local litn = litname[op] + if litn and litn[op2] then + out:write(" ", litn[op2]) + elseif op == "UREFO " or op == "UREFC " then + out:write(format(" #%-3d", shr(op2, 8))) + else + out:write(format(" #%-3d", op2)) + end + elseif op2 < 0 then + out:write(" ", formatk(tr, op2)) + else + out:write(format(" %04d", op2)) + end + end + end + out:write("\n") + end + end + if snap then + if dumpreg then + out:write(format(".... SNAP #%-3d [ ", snapno)) + else + out:write(format(".... SNAP #%-3d [ ", snapno)) + end + printsnap(tr, snap) + end +end + +------------------------------------------------------------------------------ + +local recprefix = "" +local recdepth = 0 + +-- Format trace error message. +local function fmterr(err, info) + if type(err) == "number" then + if type(info) == "function" then info = fmtfunc(info) end + err = format(vmdef.traceerr[err], info) + end + return err +end + +-- Dump trace states. +local function dump_trace(what, tr, func, pc, otr, oex) + if what == "stop" or (what == "abort" and dumpmode.a) then + if dumpmode.i then dump_ir(tr, dumpmode.s, dumpmode.r and what == "stop") + elseif dumpmode.s then dump_snap(tr) end + if dumpmode.m then dump_mcode(tr) end + end + if what == "start" then + if dumpmode.H then out:write('
\n') end
+    out:write("---- TRACE ", tr, " ", what)
+    if otr then out:write(" ", otr, "/", oex == -1 and "stitch" or oex) end
+    out:write(" ", fmtfunc(func, pc), "\n")
+  elseif what == "stop" or what == "abort" then
+    out:write("---- TRACE ", tr, " ", what)
+    if what == "abort" then
+      out:write(" ", fmtfunc(func, pc), " -- ", fmterr(otr, oex), "\n")
+    else
+      local info = traceinfo(tr)
+      local link, ltype = info.link, info.linktype
+      if link == tr or link == 0 then
+	out:write(" -> ", ltype, "\n")
+      elseif ltype == "root" then
+	out:write(" -> ", link, "\n")
+      else
+	out:write(" -> ", link, " ", ltype, "\n")
+      end
+    end
+    if dumpmode.H then out:write("
\n\n") else out:write("\n") end + else + if what == "flush" then symtab, nexitsym = {}, 0 end + out:write("---- TRACE ", what, "\n\n") + end + out:flush() +end + +-- Dump recorded bytecode. +local function dump_record(tr, func, pc, depth) + if depth ~= recdepth then + recdepth = depth + recprefix = rep(" .", depth) + end + local line + if pc >= 0 then + line = bcline(func, pc, recprefix) + if dumpmode.H then line = gsub(line, "[<>&]", html_escape) end + else + line = "0000 "..recprefix.." FUNCC \n" + end + if pc <= 0 then + out:write(sub(line, 1, -2), " ; ", fmtfunc(func), "\n") + else + out:write(line) + end + if pc >= 0 and band(funcbc(func, pc), 0xff) < 16 then -- ORDER BC + out:write(bcline(func, pc+1, recprefix)) -- Write JMP for cond. + end +end + +------------------------------------------------------------------------------ + +local gpr64 = jit.arch:match("64") +local fprmips32 = jit.arch == "mips" or jit.arch == "mipsel" + +-- Dump taken trace exits. +local function dump_texit(tr, ex, ngpr, nfpr, ...) + out:write("---- TRACE ", tr, " exit ", ex, "\n") + if dumpmode.X then + local regs = {...} + if gpr64 then + for i=1,ngpr do + out:write(format(" %016x", regs[i])) + if i % 4 == 0 then out:write("\n") end + end + else + for i=1,ngpr do + out:write(" ", tohex(regs[i])) + if i % 8 == 0 then out:write("\n") end + end + end + if fprmips32 then + for i=1,nfpr,2 do + out:write(format(" %+17.14g", regs[ngpr+i])) + if i % 8 == 7 then out:write("\n") end + end + else + for i=1,nfpr do + out:write(format(" %+17.14g", regs[ngpr+i])) + if i % 4 == 0 then out:write("\n") end + end + end + end +end + +------------------------------------------------------------------------------ + +-- Detach dump handlers. +local function dumpoff() + if active then + active = false + jit.attach(dump_texit) + jit.attach(dump_record) + jit.attach(dump_trace) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach dump handlers. +local function dumpon(opt, outfile) + if active then dumpoff() end + + local term = os.getenv("TERM") + local colormode = (term and term:match("color") or os.getenv("COLORTERM")) and "A" or "T" + if opt then + opt = gsub(opt, "[TAH]", function(mode) colormode = mode; return ""; end) + end + + local m = { t=true, b=true, i=true, m=true, } + if opt and opt ~= "" then + local o = sub(opt, 1, 1) + if o ~= "+" and o ~= "-" then m = {} end + for i=1,#opt do m[sub(opt, i, i)] = (o ~= "-") end + end + dumpmode = m + + if m.t or m.b or m.i or m.s or m.m then + jit.attach(dump_trace, "trace") + end + if m.b then + jit.attach(dump_record, "record") + if not bcline then bcline = require("jit.bc").line end + end + if m.x or m.X then + jit.attach(dump_texit, "texit") + end + + if not outfile then outfile = os.getenv("LUAJIT_DUMPFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stdout + end + + m[colormode] = true + if colormode == "A" then + colorize = colorize_ansi + irtype = irtype_ansi + elseif colormode == "H" then + colorize = colorize_html + irtype = irtype_html + out:write(header_html) + else + colorize = colorize_text + irtype = irtype_text + end + + active = true +end + +-- Public module functions. +return { + on = dumpon, + off = dumpoff, + start = dumpon -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/p.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/p.lua new file mode 100644 index 00000000..c9ec1d8b --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/p.lua @@ -0,0 +1,312 @@ +---------------------------------------------------------------------------- +-- LuaJIT profiler. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module is a simple command line interface to the built-in +-- low-overhead profiler of LuaJIT. +-- +-- The lower-level API of the profiler is accessible via the "jit.profile" +-- module or the luaJIT_profile_* C API. +-- +-- Example usage: +-- +-- luajit -jp myapp.lua +-- luajit -jp=s myapp.lua +-- luajit -jp=-s myapp.lua +-- luajit -jp=vl myapp.lua +-- luajit -jp=G,profile.txt myapp.lua +-- +-- The following dump features are available: +-- +-- f Stack dump: function name, otherwise module:line. Default mode. +-- F Stack dump: ditto, but always prepend module. +-- l Stack dump: module:line. +-- stack dump depth (callee < caller). Default: 1. +-- - Inverse stack dump depth (caller > callee). +-- s Split stack dump after first stack level. Implies abs(depth) >= 2. +-- p Show full path for module names. +-- v Show VM states. Can be combined with stack dumps, e.g. vf or fv. +-- z Show zones. Can be combined with stack dumps, e.g. zf or fz. +-- r Show raw sample counts. Default: show percentages. +-- a Annotate excerpts from source code files. +-- A Annotate complete source code files. +-- G Produce raw output suitable for graphical tools (e.g. flame graphs). +-- m Minimum sample percentage to be shown. Default: 3. +-- i Sampling interval in milliseconds. Default: 10. +-- +---------------------------------------------------------------------------- + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local profile = require("jit.profile") +local vmdef = require("jit.vmdef") +local math = math +local pairs, ipairs, tonumber, floor = pairs, ipairs, tonumber, math.floor +local sort, format = table.sort, string.format +local stdout = io.stdout +local zone -- Load jit.zone module on demand. + +-- Output file handle. +local out + +------------------------------------------------------------------------------ + +local prof_ud +local prof_states, prof_split, prof_min, prof_raw, prof_fmt, prof_depth +local prof_ann, prof_count1, prof_count2, prof_samples + +local map_vmmode = { + N = "Compiled", + I = "Interpreted", + C = "C code", + G = "Garbage Collector", + J = "JIT Compiler", +} + +-- Profiler callback. +local function prof_cb(th, samples, vmmode) + prof_samples = prof_samples + samples + local key_stack, key_stack2, key_state + -- Collect keys for sample. + if prof_states then + if prof_states == "v" then + key_state = map_vmmode[vmmode] or vmmode + else + key_state = zone:get() or "(none)" + end + end + if prof_fmt then + key_stack = profile.dumpstack(th, prof_fmt, prof_depth) + key_stack = key_stack:gsub("%[builtin#(%d+)%]", function(x) + return vmdef.ffnames[tonumber(x)] + end) + if prof_split == 2 then + local k1, k2 = key_stack:match("(.-) [<>] (.*)") + if k2 then key_stack, key_stack2 = k1, k2 end + elseif prof_split == 3 then + key_stack2 = profile.dumpstack(th, "l", 1) + end + end + -- Order keys. + local k1, k2 + if prof_split == 1 then + if key_state then + k1 = key_state + if key_stack then k2 = key_stack end + end + elseif key_stack then + k1 = key_stack + if key_stack2 then k2 = key_stack2 elseif key_state then k2 = key_state end + end + -- Coalesce samples in one or two levels. + if k1 then + local t1 = prof_count1 + t1[k1] = (t1[k1] or 0) + samples + if k2 then + local t2 = prof_count2 + local t3 = t2[k1] + if not t3 then t3 = {}; t2[k1] = t3 end + t3[k2] = (t3[k2] or 0) + samples + end + end +end + +------------------------------------------------------------------------------ + +-- Show top N list. +local function prof_top(count1, count2, samples, indent) + local t, n = {}, 0 + for k in pairs(count1) do + n = n + 1 + t[n] = k + end + sort(t, function(a, b) return count1[a] > count1[b] end) + for i=1,n do + local k = t[i] + local v = count1[k] + local pct = floor(v*100/samples + 0.5) + if pct < prof_min then break end + if not prof_raw then + out:write(format("%s%2d%% %s\n", indent, pct, k)) + elseif prof_raw == "r" then + out:write(format("%s%5d %s\n", indent, v, k)) + else + out:write(format("%s %d\n", k, v)) + end + if count2 then + local r = count2[k] + if r then + prof_top(r, nil, v, (prof_split == 3 or prof_split == 1) and " -- " or + (prof_depth < 0 and " -> " or " <- ")) + end + end + end +end + +-- Annotate source code +local function prof_annotate(count1, samples) + local files = {} + local ms = 0 + for k, v in pairs(count1) do + local pct = floor(v*100/samples + 0.5) + ms = math.max(ms, v) + if pct >= prof_min then + local file, line = k:match("^(.*):(%d+)$") + if not file then file = k; line = 0 end + local fl = files[file] + if not fl then fl = {}; files[file] = fl; files[#files+1] = file end + line = tonumber(line) + fl[line] = prof_raw and v or pct + end + end + sort(files) + local fmtv, fmtn = " %3d%% | %s\n", " | %s\n" + if prof_raw then + local n = math.max(5, math.ceil(math.log10(ms))) + fmtv = "%"..n.."d | %s\n" + fmtn = (" "):rep(n).." | %s\n" + end + local ann = prof_ann + for _, file in ipairs(files) do + local f0 = file:byte() + if f0 == 40 or f0 == 91 then + out:write(format("\n====== %s ======\n[Cannot annotate non-file]\n", file)) + break + end + local fp, err = io.open(file) + if not fp then + out:write(format("====== ERROR: %s: %s\n", file, err)) + break + end + out:write(format("\n====== %s ======\n", file)) + local fl = files[file] + local n, show = 1, false + if ann ~= 0 then + for i=1,ann do + if fl[i] then show = true; out:write("@@ 1 @@\n"); break end + end + end + for line in fp:lines() do + if line:byte() == 27 then + out:write("[Cannot annotate bytecode file]\n") + break + end + local v = fl[n] + if ann ~= 0 then + local v2 = fl[n+ann] + if show then + if v2 then show = n+ann elseif v then show = n + elseif show+ann < n then show = false end + elseif v2 then + show = n+ann + out:write(format("@@ %d @@\n", n)) + end + if not show then goto next end + end + if v then + out:write(format(fmtv, v, line)) + else + out:write(format(fmtn, line)) + end + ::next:: + n = n + 1 + end + fp:close() + end +end + +------------------------------------------------------------------------------ + +-- Finish profiling and dump result. +local function prof_finish() + if prof_ud then + profile.stop() + local samples = prof_samples + if samples == 0 then + if prof_raw ~= true then out:write("[No samples collected]\n") end + return + end + if prof_ann then + prof_annotate(prof_count1, samples) + else + prof_top(prof_count1, prof_count2, samples, "") + end + prof_count1 = nil + prof_count2 = nil + prof_ud = nil + if out ~= stdout then out:close() end + end +end + +-- Start profiling. +local function prof_start(mode) + local interval = "" + mode = mode:gsub("i%d*", function(s) interval = s; return "" end) + prof_min = 3 + mode = mode:gsub("m(%d+)", function(s) prof_min = tonumber(s); return "" end) + prof_depth = 1 + mode = mode:gsub("%-?%d+", function(s) prof_depth = tonumber(s); return "" end) + local m = {} + for c in mode:gmatch(".") do m[c] = c end + prof_states = m.z or m.v + if prof_states == "z" then zone = require("jit.zone") end + local scope = m.l or m.f or m.F or (prof_states and "" or "f") + local flags = (m.p or "") + prof_raw = m.r + if m.s then + prof_split = 2 + if prof_depth == -1 or m["-"] then prof_depth = -2 + elseif prof_depth == 1 then prof_depth = 2 end + elseif mode:find("[fF].*l") then + scope = "l" + prof_split = 3 + else + prof_split = (scope == "" or mode:find("[zv].*[lfF]")) and 1 or 0 + end + prof_ann = m.A and 0 or (m.a and 3) + if prof_ann then + scope = "l" + prof_fmt = "pl" + prof_split = 0 + prof_depth = 1 + elseif m.G and scope ~= "" then + prof_fmt = flags..scope.."Z;" + prof_depth = -100 + prof_raw = true + prof_min = 0 + elseif scope == "" then + prof_fmt = false + else + local sc = prof_split == 3 and m.f or m.F or scope + prof_fmt = flags..sc..(prof_depth >= 0 and "Z < " or "Z > ") + end + prof_count1 = {} + prof_count2 = {} + prof_samples = 0 + profile.start(scope:lower()..interval, prof_cb) + prof_ud = newproxy(true) + getmetatable(prof_ud).__gc = prof_finish +end + +------------------------------------------------------------------------------ + +local function start(mode, outfile) + if not outfile then outfile = os.getenv("LUAJIT_PROFILEFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stdout + end + prof_start(mode or "f") +end + +-- Public module functions. +return { + start = start, -- For -j command line option. + stop = prof_finish +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/v.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/v.lua new file mode 100644 index 00000000..83589143 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/v.lua @@ -0,0 +1,170 @@ +---------------------------------------------------------------------------- +-- Verbose mode of the LuaJIT compiler. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module shows verbose information about the progress of the +-- JIT compiler. It prints one line for each generated trace. This module +-- is useful to see which code has been compiled or where the compiler +-- punts and falls back to the interpreter. +-- +-- Example usage: +-- +-- luajit -jv -e "for i=1,1000 do for j=1,1000 do end end" +-- luajit -jv=myapp.out myapp.lua +-- +-- Default output is to stderr. To redirect the output to a file, pass a +-- filename as an argument (use '-' for stdout) or set the environment +-- variable LUAJIT_VERBOSEFILE. The file is overwritten every time the +-- module is started. +-- +-- The output from the first example should look like this: +-- +-- [TRACE 1 (command line):1 loop] +-- [TRACE 2 (1/3) (command line):1 -> 1] +-- +-- The first number in each line is the internal trace number. Next are +-- the file name ('(command line)') and the line number (':1') where the +-- trace has started. Side traces also show the parent trace number and +-- the exit number where they are attached to in parentheses ('(1/3)'). +-- An arrow at the end shows where the trace links to ('-> 1'), unless +-- it loops to itself. +-- +-- In this case the inner loop gets hot and is traced first, generating +-- a root trace. Then the last exit from the 1st trace gets hot, too, +-- and triggers generation of the 2nd trace. The side trace follows the +-- path along the outer loop and *around* the inner loop, back to its +-- start, and then links to the 1st trace. Yes, this may seem unusual, +-- if you know how traditional compilers work. Trace compilers are full +-- of surprises like this -- have fun! :-) +-- +-- Aborted traces are shown like this: +-- +-- [TRACE --- foo.lua:44 -- leaving loop in root trace at foo:lua:50] +-- +-- Don't worry -- trace aborts are quite common, even in programs which +-- can be fully compiled. The compiler may retry several times until it +-- finds a suitable trace. +-- +-- Of course this doesn't work with features that are not-yet-implemented +-- (NYI error messages). The VM simply falls back to the interpreter. This +-- may not matter at all if the particular trace is not very high up in +-- the CPU usage profile. Oh, and the interpreter is quite fast, too. +-- +-- Also check out the -jdump module, which prints all the gory details. +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local funcinfo, traceinfo = jutil.funcinfo, jutil.traceinfo +local type, format = type, string.format +local stdout, stderr = io.stdout, io.stderr + +-- Active flag and output file handle. +local active, out + +------------------------------------------------------------------------------ + +local startloc, startex + +local function fmtfunc(func, pc) + local fi = funcinfo(func, pc) + if fi.loc then + return fi.loc + elseif fi.ffid then + return vmdef.ffnames[fi.ffid] + elseif fi.addr then + return format("C:%x", fi.addr) + else + return "(?)" + end +end + +-- Format trace error message. +local function fmterr(err, info) + if type(err) == "number" then + if type(info) == "function" then info = fmtfunc(info) end + err = format(vmdef.traceerr[err], info) + end + return err +end + +-- Dump trace states. +local function dump_trace(what, tr, func, pc, otr, oex) + if what == "start" then + startloc = fmtfunc(func, pc) + startex = otr and "("..otr.."/"..(oex == -1 and "stitch" or oex)..") " or "" + else + if what == "abort" then + local loc = fmtfunc(func, pc) + if loc ~= startloc then + out:write(format("[TRACE --- %s%s -- %s at %s]\n", + startex, startloc, fmterr(otr, oex), loc)) + else + out:write(format("[TRACE --- %s%s -- %s]\n", + startex, startloc, fmterr(otr, oex))) + end + elseif what == "stop" then + local info = traceinfo(tr) + local link, ltype = info.link, info.linktype + if ltype == "interpreter" then + out:write(format("[TRACE %3s %s%s -- fallback to interpreter]\n", + tr, startex, startloc)) + elseif ltype == "stitch" then + out:write(format("[TRACE %3s %s%s %s %s]\n", + tr, startex, startloc, ltype, fmtfunc(func, pc))) + elseif link == tr or link == 0 then + out:write(format("[TRACE %3s %s%s %s]\n", + tr, startex, startloc, ltype)) + elseif ltype == "root" then + out:write(format("[TRACE %3s %s%s -> %d]\n", + tr, startex, startloc, link)) + else + out:write(format("[TRACE %3s %s%s -> %d %s]\n", + tr, startex, startloc, link, ltype)) + end + else + out:write(format("[TRACE %s]\n", what)) + end + out:flush() + end +end + +------------------------------------------------------------------------------ + +-- Detach dump handlers. +local function dumpoff() + if active then + active = false + jit.attach(dump_trace) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach dump handlers. +local function dumpon(outfile) + if active then dumpoff() end + if not outfile then outfile = os.getenv("LUAJIT_VERBOSEFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stderr + end + jit.attach(dump_trace, "trace") + active = true +end + +-- Public module functions. +return { + on = dumpon, + off = dumpoff, + start = dumpon -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/vmdef.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/vmdef.lua new file mode 100644 index 00000000..68ff7ba2 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/vmdef.lua @@ -0,0 +1,395 @@ +-- This is a generated file. DO NOT EDIT! + +return { + +bcnames = "ISLT ISGE ISLE ISGT ISEQV ISNEV ISEQS ISNES ISEQN ISNEN ISEQP ISNEP ISTC ISFC IST ISF ISTYPEISNUM MOV NOT UNM LEN ADDVN SUBVN MULVN DIVVN MODVN ADDNV SUBNV MULNV DIVNV MODNV ADDVV SUBVV MULVV DIVVV MODVV POW CAT KSTR KCDATAKSHORTKNUM KPRI KNIL UGET USETV USETS USETN USETP UCLO FNEW TNEW TDUP GGET GSET TGETV TGETS TGETB TGETR TSETV TSETS TSETB TSETM TSETR CALLM CALL CALLMTCALLT ITERC ITERN VARG ISNEXTRETM RET RET0 RET1 FORI JFORI FORL IFORL JFORL ITERL IITERLJITERLLOOP ILOOP JLOOP JMP FUNCF IFUNCFJFUNCFFUNCV IFUNCVJFUNCVFUNCC FUNCCW", + +irnames = "LT GE LE GT ULT UGE ULE UGT EQ NE ABC RETF NOP BASE PVAL GCSTEPHIOP LOOP USE PHI RENAMEPROF KPRI KINT KGC KPTR KKPTR KNULL KNUM KINT64KSLOT BNOT BSWAP BAND BOR BXOR BSHL BSHR BSAR BROL BROR ADD SUB MUL DIV MOD POW NEG ABS LDEXP MIN MAX FPMATHADDOV SUBOV MULOV AREF HREFK HREF NEWREFUREFO UREFC FREF TMPREFSTRREFLREF ALOAD HLOAD ULOAD FLOAD XLOAD SLOAD VLOAD ALEN ASTOREHSTOREUSTOREFSTOREXSTORESNEW XSNEW TNEW TDUP CNEW CNEWI BUFHDRBUFPUTBUFSTRTBAR OBAR XBAR CONV TOBIT TOSTR STRTO CALLN CALLA CALLL CALLS CALLXSCARG ", + +irfpm = { [0]="floor", "ceil", "trunc", "sqrt", "log", "log2", "other", }, + +irfield = { [0]="str.len", "func.env", "func.pc", "func.ffid", "thread.env", "tab.meta", "tab.array", "tab.node", "tab.asize", "tab.hmask", "tab.nomm", "udata.meta", "udata.udtype", "udata.file", "sbuf.w", "sbuf.e", "sbuf.b", "sbuf.l", "sbuf.ref", "sbuf.r", "cdata.ctypeid", "cdata.ptr", "cdata.int", "cdata.int64", "cdata.int64_4", }, + +ircall = { +[0]="lj_str_cmp", +"lj_str_find", +"lj_str_new", +"lj_strscan_num", +"lj_strfmt_int", +"lj_strfmt_num", +"lj_strfmt_char", +"lj_strfmt_putint", +"lj_strfmt_putnum", +"lj_strfmt_putquoted", +"lj_strfmt_putfxint", +"lj_strfmt_putfnum_int", +"lj_strfmt_putfnum_uint", +"lj_strfmt_putfnum", +"lj_strfmt_putfstr", +"lj_strfmt_putfchar", +"lj_buf_putmem", +"lj_buf_putstr", +"lj_buf_putchar", +"lj_buf_putstr_reverse", +"lj_buf_putstr_lower", +"lj_buf_putstr_upper", +"lj_buf_putstr_rep", +"lj_buf_puttab", +"lj_bufx_set", +"lj_bufx_more", +"lj_serialize_put", +"lj_serialize_get", +"lj_serialize_encode", +"lj_serialize_decode", +"lj_buf_tostr", +"lj_tab_new_ah", +"lj_tab_new1", +"lj_tab_dup", +"lj_tab_clear", +"lj_tab_newkey", +"lj_tab_keyindex", +"lj_vm_next", +"lj_tab_len", +"lj_tab_len_hint", +"lj_gc_step_jit", +"lj_gc_barrieruv", +"lj_mem_newgco", +"lj_prng_u64d", +"lj_vm_modi", +"log10", +"exp", +"sin", +"cos", +"tan", +"asin", +"acos", +"atan", +"sinh", +"cosh", +"tanh", +"fputc", +"fwrite", +"fflush", +"lj_vm_floor", +"lj_vm_ceil", +"lj_vm_trunc", +"sqrt", +"log", +"lj_vm_log2", +"lj_vm_powi", +"pow", +"atan2", +"ldexp", +"lj_vm_tobit", +"softfp_add", +"softfp_sub", +"softfp_mul", +"softfp_div", +"softfp_cmp", +"softfp_i2d", +"softfp_d2i", +"lj_vm_sfmin", +"lj_vm_sfmax", +"lj_vm_tointg", +"softfp_ui2d", +"softfp_f2d", +"softfp_d2ui", +"softfp_d2f", +"softfp_i2f", +"softfp_ui2f", +"softfp_f2i", +"softfp_f2ui", +"fp64_l2d", +"fp64_ul2d", +"fp64_l2f", +"fp64_ul2f", +"fp64_d2l", +"fp64_d2ul", +"fp64_f2l", +"fp64_f2ul", +"lj_carith_divi64", +"lj_carith_divu64", +"lj_carith_modi64", +"lj_carith_modu64", +"lj_carith_powi64", +"lj_carith_powu64", +"lj_cdata_newv", +"lj_cdata_setfin", +"strlen", +"memcpy", +"memset", +"lj_vm_errno", +"lj_carith_mul64", +"lj_carith_shl64", +"lj_carith_shr64", +"lj_carith_sar64", +"lj_carith_rol64", +"lj_carith_ror64", +}, + +traceerr = { +[0]="error thrown or hook called during recording", +"trace too short", +"trace too long", +"trace too deep", +"too many snapshots", +"blacklisted", +"retry recording", +"NYI: bytecode %d", +"leaving loop in root trace", +"inner loop in root trace", +"loop unroll limit reached", +"bad argument type", +"JIT compilation disabled for function", +"call unroll limit reached", +"down-recursion, restarting", +"NYI: unsupported variant of FastFunc %s", +"NYI: return to lower frame", +"store with nil or NaN key", +"missing metamethod", +"looping index lookup", +"NYI: mixed sparse/dense table", +"symbol not in cache", +"NYI: unsupported C type conversion", +"NYI: unsupported C function type", +"guard would always fail", +"too many PHIs", +"persistent type instability", +"failed to allocate mcode memory", +"machine code too long", +"hit mcode limit (retrying)", +"too many spill slots", +"inconsistent register allocation", +"NYI: cannot assemble IR instruction %d", +"NYI: PHI shuffling too complex", +"NYI: register coalescing too complex", +}, + +ffnames = { +[0]="Lua", +"C", +"assert", +"type", +"next", +"pairs", +"ipairs_aux", +"ipairs", +"getmetatable", +"setmetatable", +"getfenv", +"setfenv", +"rawget", +"rawset", +"rawequal", +"unpack", +"select", +"tonumber", +"tostring", +"error", +"pcall", +"xpcall", +"loadfile", +"load", +"loadstring", +"dofile", +"gcinfo", +"collectgarbage", +"newproxy", +"print", +"coroutine.status", +"coroutine.running", +"coroutine.isyieldable", +"coroutine.create", +"coroutine.yield", +"coroutine.resume", +"coroutine.wrap_aux", +"coroutine.wrap", +"math.abs", +"math.floor", +"math.ceil", +"math.sqrt", +"math.log10", +"math.exp", +"math.sin", +"math.cos", +"math.tan", +"math.asin", +"math.acos", +"math.atan", +"math.sinh", +"math.cosh", +"math.tanh", +"math.frexp", +"math.modf", +"math.log", +"math.atan2", +"math.pow", +"math.fmod", +"math.ldexp", +"math.min", +"math.max", +"math.random", +"math.randomseed", +"bit.tobit", +"bit.bnot", +"bit.bswap", +"bit.lshift", +"bit.rshift", +"bit.arshift", +"bit.rol", +"bit.ror", +"bit.band", +"bit.bor", +"bit.bxor", +"bit.tohex", +"string.byte", +"string.char", +"string.sub", +"string.rep", +"string.reverse", +"string.lower", +"string.upper", +"string.dump", +"string.find", +"string.match", +"string.gmatch_aux", +"string.gmatch", +"string.gsub", +"string.format", +"table.maxn", +"table.insert", +"table.concat", +"table.sort", +"table.new", +"table.clear", +"io.method.close", +"io.method.read", +"io.method.write", +"io.method.flush", +"io.method.seek", +"io.method.setvbuf", +"io.method.lines", +"io.method.__gc", +"io.method.__tostring", +"io.open", +"io.popen", +"io.tmpfile", +"io.close", +"io.read", +"io.write", +"io.flush", +"io.input", +"io.output", +"io.lines", +"io.type", +"os.execute", +"os.remove", +"os.rename", +"os.tmpname", +"os.getenv", +"os.exit", +"os.clock", +"os.date", +"os.time", +"os.difftime", +"os.setlocale", +"debug.getregistry", +"debug.getmetatable", +"debug.setmetatable", +"debug.getfenv", +"debug.setfenv", +"debug.getinfo", +"debug.getlocal", +"debug.setlocal", +"debug.getupvalue", +"debug.setupvalue", +"debug.upvalueid", +"debug.upvaluejoin", +"debug.sethook", +"debug.gethook", +"debug.debug", +"debug.traceback", +"jit.on", +"jit.off", +"jit.flush", +"jit.status", +"jit.security", +"jit.attach", +"jit.util.funcinfo", +"jit.util.funcbc", +"jit.util.funck", +"jit.util.funcuvname", +"jit.util.traceinfo", +"jit.util.traceir", +"jit.util.tracek", +"jit.util.tracesnap", +"jit.util.tracemc", +"jit.util.traceexitstub", +"jit.util.ircalladdr", +"jit.opt.start", +"jit.profile.start", +"jit.profile.stop", +"jit.profile.dumpstack", +"ffi.meta.__index", +"ffi.meta.__newindex", +"ffi.meta.__eq", +"ffi.meta.__len", +"ffi.meta.__lt", +"ffi.meta.__le", +"ffi.meta.__concat", +"ffi.meta.__call", +"ffi.meta.__add", +"ffi.meta.__sub", +"ffi.meta.__mul", +"ffi.meta.__div", +"ffi.meta.__mod", +"ffi.meta.__pow", +"ffi.meta.__unm", +"ffi.meta.__tostring", +"ffi.meta.__pairs", +"ffi.meta.__ipairs", +"ffi.clib.__index", +"ffi.clib.__newindex", +"ffi.clib.__gc", +"ffi.callback.free", +"ffi.callback.set", +"ffi.cdef", +"ffi.new", +"ffi.cast", +"ffi.typeof", +"ffi.typeinfo", +"ffi.istype", +"ffi.sizeof", +"ffi.alignof", +"ffi.offsetof", +"ffi.errno", +"ffi.string", +"ffi.copy", +"ffi.fill", +"ffi.abi", +"ffi.metatype", +"ffi.gc", +"ffi.load", +"buffer.method.free", +"buffer.method.reset", +"buffer.method.skip", +"buffer.method.set", +"buffer.method.put", +"buffer.method.putf", +"buffer.method.get", +"buffer.method.putcdata", +"buffer.method.reserve", +"buffer.method.commit", +"buffer.method.ref", +"buffer.method.encode", +"buffer.method.decode", +"buffer.method.__gc", +"buffer.method.__tostring", +"buffer.method.__len", +"buffer.new", +"buffer.encode", +"buffer.decode", +}, + +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/jit/zone.lua b/love/src/jni/LuaJIT-2.1/android/x86/jit/zone.lua new file mode 100644 index 00000000..94357854 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86/jit/zone.lua @@ -0,0 +1,45 @@ +---------------------------------------------------------------------------- +-- LuaJIT profiler zones. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module implements a simple hierarchical zone model. +-- +-- Example usage: +-- +-- local zone = require("jit.zone") +-- zone("AI") +-- ... +-- zone("A*") +-- ... +-- print(zone:get()) --> "A*" +-- ... +-- zone() +-- ... +-- print(zone:get()) --> "AI" +-- ... +-- zone() +-- +---------------------------------------------------------------------------- + +local remove = table.remove + +return setmetatable({ + flush = function(t) + for i=#t,1,-1 do t[i] = nil end + end, + get = function(t) + return t[#t] + end +}, { + __call = function(t, zone) + if zone then + t[#t+1] = zone + else + return (assert(remove(t), "empty zone stack")) + end + end +}) + diff --git a/love/src/jni/LuaJIT-2.1/android/x86/libluajit.a b/love/src/jni/LuaJIT-2.1/android/x86/libluajit.a index ec49d56b..aea69707 100644 Binary files a/love/src/jni/LuaJIT-2.1/android/x86/libluajit.a and b/love/src/jni/LuaJIT-2.1/android/x86/libluajit.a differ diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/bc.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/bc.lua new file mode 100644 index 00000000..e58a3fef --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/bc.lua @@ -0,0 +1,190 @@ +---------------------------------------------------------------------------- +-- LuaJIT bytecode listing module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module lists the bytecode of a Lua function. If it's loaded by -jbc +-- it hooks into the parser and lists all functions of a chunk as they +-- are parsed. +-- +-- Example usage: +-- +-- luajit -jbc -e 'local x=0; for i=1,1e6 do x=x+i end; print(x)' +-- luajit -jbc=- foo.lua +-- luajit -jbc=foo.list foo.lua +-- +-- Default output is to stderr. To redirect the output to a file, pass a +-- filename as an argument (use '-' for stdout) or set the environment +-- variable LUAJIT_LISTFILE. The file is overwritten every time the module +-- is started. +-- +-- This module can also be used programmatically: +-- +-- local bc = require("jit.bc") +-- +-- local function foo() print("hello") end +-- +-- bc.dump(foo) --> -- BYTECODE -- [...] +-- print(bc.line(foo, 2)) --> 0002 KSTR 1 1 ; "hello" +-- +-- local out = { +-- -- Do something with each line: +-- write = function(t, ...) io.write(...) end, +-- close = function(t) end, +-- flush = function(t) end, +-- } +-- bc.dump(foo, out) +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local bit = require("bit") +local sub, gsub, format = string.sub, string.gsub, string.format +local byte, band, shr = string.byte, bit.band, bit.rshift +local funcinfo, funcbc, funck = jutil.funcinfo, jutil.funcbc, jutil.funck +local funcuvname = jutil.funcuvname +local bcnames = vmdef.bcnames +local stdout, stderr = io.stdout, io.stderr + +------------------------------------------------------------------------------ + +local function ctlsub(c) + if c == "\n" then return "\\n" + elseif c == "\r" then return "\\r" + elseif c == "\t" then return "\\t" + else return format("\\%03d", byte(c)) + end +end + +-- Return one bytecode line. +local function bcline(func, pc, prefix) + local ins, m = funcbc(func, pc) + if not ins then return end + local ma, mb, mc = band(m, 7), band(m, 15*8), band(m, 15*128) + local a = band(shr(ins, 8), 0xff) + local oidx = 6*band(ins, 0xff) + local op = sub(bcnames, oidx+1, oidx+6) + local s = format("%04d %s %-6s %3s ", + pc, prefix or " ", op, ma == 0 and "" or a) + local d = shr(ins, 16) + if mc == 13*128 then -- BCMjump + return format("%s=> %04d\n", s, pc+d-0x7fff) + end + if mb ~= 0 then + d = band(d, 0xff) + elseif mc == 0 then + return s.."\n" + end + local kc + if mc == 10*128 then -- BCMstr + kc = funck(func, -d-1) + kc = format(#kc > 40 and '"%.40s"~' or '"%s"', gsub(kc, "%c", ctlsub)) + elseif mc == 9*128 then -- BCMnum + kc = funck(func, d) + if op == "TSETM " then kc = kc - 2^52 end + elseif mc == 12*128 then -- BCMfunc + local fi = funcinfo(funck(func, -d-1)) + if fi.ffid then + kc = vmdef.ffnames[fi.ffid] + else + kc = fi.loc + end + elseif mc == 5*128 then -- BCMuv + kc = funcuvname(func, d) + end + if ma == 5 then -- BCMuv + local ka = funcuvname(func, a) + if kc then kc = ka.." ; "..kc else kc = ka end + end + if mb ~= 0 then + local b = shr(ins, 24) + if kc then return format("%s%3d %3d ; %s\n", s, b, d, kc) end + return format("%s%3d %3d\n", s, b, d) + end + if kc then return format("%s%3d ; %s\n", s, d, kc) end + if mc == 7*128 and d > 32767 then d = d - 65536 end -- BCMlits + return format("%s%3d\n", s, d) +end + +-- Collect branch targets of a function. +local function bctargets(func) + local target = {} + for pc=1,1000000000 do + local ins, m = funcbc(func, pc) + if not ins then break end + if band(m, 15*128) == 13*128 then target[pc+shr(ins, 16)-0x7fff] = true end + end + return target +end + +-- Dump bytecode instructions of a function. +local function bcdump(func, out, all) + if not out then out = stdout end + local fi = funcinfo(func) + if all and fi.children then + for n=-1,-1000000000,-1 do + local k = funck(func, n) + if not k then break end + if type(k) == "proto" then bcdump(k, out, true) end + end + end + out:write(format("-- BYTECODE -- %s-%d\n", fi.loc, fi.lastlinedefined)) + local target = bctargets(func) + for pc=1,1000000000 do + local s = bcline(func, pc, target[pc] and "=>") + if not s then break end + out:write(s) + end + out:write("\n") + out:flush() +end + +------------------------------------------------------------------------------ + +-- Active flag and output file handle. +local active, out + +-- List handler. +local function h_list(func) + return bcdump(func, out) +end + +-- Detach list handler. +local function bclistoff() + if active then + active = false + jit.attach(h_list) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach list handler. +local function bcliston(outfile) + if active then bclistoff() end + if not outfile then outfile = os.getenv("LUAJIT_LISTFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stderr + end + jit.attach(h_list, "bc") + active = true +end + +-- Public module functions. +return { + line = bcline, + dump = bcdump, + targets = bctargets, + on = bcliston, + off = bclistoff, + start = bcliston -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/bcsave.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/bcsave.lua new file mode 100644 index 00000000..ab13667a --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/bcsave.lua @@ -0,0 +1,665 @@ +---------------------------------------------------------------------------- +-- LuaJIT module to save/list bytecode. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module saves or lists the bytecode for an input file. +-- It's run by the -b command line option. +-- +------------------------------------------------------------------------------ + +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local bit = require("bit") + +-- Symbol name prefix for LuaJIT bytecode. +local LJBC_PREFIX = "luaJIT_BC_" + +local type, assert = type, assert +local format = string.format +local tremove, tconcat = table.remove, table.concat + +------------------------------------------------------------------------------ + +local function usage() + io.stderr:write[[ +Save LuaJIT bytecode: luajit -b[options] input output + -l Only list bytecode. + -s Strip debug info (default). + -g Keep debug info. + -n name Set module name (default: auto-detect from input name). + -t type Set output file type (default: auto-detect from output name). + -a arch Override architecture for object files (default: native). + -o os Override OS for object files (default: native). + -e chunk Use chunk string as input. + -- Stop handling options. + - Use stdin as input and/or stdout as output. + +File types: c h obj o raw (default) +]] + os.exit(1) +end + +local function check(ok, ...) + if ok then return ok, ... end + io.stderr:write("luajit: ", ...) + io.stderr:write("\n") + os.exit(1) +end + +local function readfile(input) + if type(input) == "function" then return input end + if input == "-" then input = nil end + return check(loadfile(input)) +end + +local function savefile(name, mode) + if name == "-" then return io.stdout end + return check(io.open(name, mode)) +end + +------------------------------------------------------------------------------ + +local map_type = { + raw = "raw", c = "c", h = "h", o = "obj", obj = "obj", +} + +local map_arch = { + x86 = { e = "le", b = 32, m = 3, p = 0x14c, }, + x64 = { e = "le", b = 64, m = 62, p = 0x8664, }, + arm = { e = "le", b = 32, m = 40, p = 0x1c0, }, + arm64 = { e = "le", b = 64, m = 183, p = 0xaa64, }, + arm64be = { e = "be", b = 64, m = 183, }, + ppc = { e = "be", b = 32, m = 20, }, + mips = { e = "be", b = 32, m = 8, f = 0x50001006, }, + mipsel = { e = "le", b = 32, m = 8, f = 0x50001006, }, + mips64 = { e = "be", b = 64, m = 8, f = 0x80000007, }, + mips64el = { e = "le", b = 64, m = 8, f = 0x80000007, }, + mips64r6 = { e = "be", b = 64, m = 8, f = 0xa0000407, }, + mips64r6el = { e = "le", b = 64, m = 8, f = 0xa0000407, }, +} + +local map_os = { + linux = true, windows = true, osx = true, freebsd = true, netbsd = true, + openbsd = true, dragonfly = true, solaris = true, +} + +local function checkarg(str, map, err) + str = str:lower() + local s = check(map[str], "unknown ", err) + return type(s) == "string" and s or str +end + +local function detecttype(str) + local ext = str:lower():match("%.(%a+)$") + return map_type[ext] or "raw" +end + +local function checkmodname(str) + check(str:match("^[%w_.%-]+$"), "bad module name") + return str:gsub("[%.%-]", "_") +end + +local function detectmodname(str) + if type(str) == "string" then + local tail = str:match("[^/\\]+$") + if tail then str = tail end + local head = str:match("^(.*)%.[^.]*$") + if head then str = head end + str = str:match("^[%w_.%-]+") + else + str = nil + end + check(str, "cannot derive module name, use -n name") + return str:gsub("[%.%-]", "_") +end + +------------------------------------------------------------------------------ + +local function bcsave_tail(fp, output, s) + local ok, err = fp:write(s) + if ok and output ~= "-" then ok, err = fp:close() end + check(ok, "cannot write ", output, ": ", err) +end + +local function bcsave_raw(output, s) + local fp = savefile(output, "wb") + bcsave_tail(fp, output, s) +end + +local function bcsave_c(ctx, output, s) + local fp = savefile(output, "w") + if ctx.type == "c" then + fp:write(format([[ +#ifdef __cplusplus +extern "C" +#endif +#ifdef _WIN32 +__declspec(dllexport) +#endif +const unsigned char %s%s[] = { +]], LJBC_PREFIX, ctx.modname)) + else + fp:write(format([[ +#define %s%s_SIZE %d +static const unsigned char %s%s[] = { +]], LJBC_PREFIX, ctx.modname, #s, LJBC_PREFIX, ctx.modname)) + end + local t, n, m = {}, 0, 0 + for i=1,#s do + local b = tostring(string.byte(s, i)) + m = m + #b + 1 + if m > 78 then + fp:write(tconcat(t, ",", 1, n), ",\n") + n, m = 0, #b + 1 + end + n = n + 1 + t[n] = b + end + bcsave_tail(fp, output, tconcat(t, ",", 1, n).."\n};\n") +end + +local function bcsave_elfobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct { + uint8_t emagic[4], eclass, eendian, eversion, eosabi, eabiversion, epad[7]; + uint16_t type, machine; + uint32_t version; + uint32_t entry, phofs, shofs; + uint32_t flags; + uint16_t ehsize, phentsize, phnum, shentsize, shnum, shstridx; +} ELF32header; +typedef struct { + uint8_t emagic[4], eclass, eendian, eversion, eosabi, eabiversion, epad[7]; + uint16_t type, machine; + uint32_t version; + uint64_t entry, phofs, shofs; + uint32_t flags; + uint16_t ehsize, phentsize, phnum, shentsize, shnum, shstridx; +} ELF64header; +typedef struct { + uint32_t name, type, flags, addr, ofs, size, link, info, align, entsize; +} ELF32sectheader; +typedef struct { + uint32_t name, type; + uint64_t flags, addr, ofs, size; + uint32_t link, info; + uint64_t align, entsize; +} ELF64sectheader; +typedef struct { + uint32_t name, value, size; + uint8_t info, other; + uint16_t sectidx; +} ELF32symbol; +typedef struct { + uint32_t name; + uint8_t info, other; + uint16_t sectidx; + uint64_t value, size; +} ELF64symbol; +typedef struct { + ELF32header hdr; + ELF32sectheader sect[6]; + ELF32symbol sym[2]; + uint8_t space[4096]; +} ELF32obj; +typedef struct { + ELF64header hdr; + ELF64sectheader sect[6]; + ELF64symbol sym[2]; + uint8_t space[4096]; +} ELF64obj; +]] + local symname = LJBC_PREFIX..ctx.modname + local ai = assert(map_arch[ctx.arch]) + local is64, isbe = ai.b == 64, ai.e == "be" + + -- Handle different host/target endianess. + local function f32(x) return x end + local f16, fofs = f32, f32 + if ffi.abi("be") ~= isbe then + f32 = bit.bswap + function f16(x) return bit.rshift(bit.bswap(x), 16) end + if is64 then + local two32 = ffi.cast("int64_t", 2^32) + function fofs(x) return bit.bswap(x)*two32 end + else + fofs = f32 + end + end + + -- Create ELF object and fill in header. + local o = ffi.new(is64 and "ELF64obj" or "ELF32obj") + local hdr = o.hdr + if ctx.os == "bsd" or ctx.os == "other" then -- Determine native hdr.eosabi. + local bf = assert(io.open("/bin/ls", "rb")) + local bs = bf:read(9) + bf:close() + ffi.copy(o, bs, 9) + check(hdr.emagic[0] == 127, "no support for writing native object files") + else + hdr.emagic = "\127ELF" + hdr.eosabi = ({ freebsd=9, netbsd=2, openbsd=12, solaris=6 })[ctx.os] or 0 + end + hdr.eclass = is64 and 2 or 1 + hdr.eendian = isbe and 2 or 1 + hdr.eversion = 1 + hdr.type = f16(1) + hdr.machine = f16(ai.m) + hdr.flags = f32(ai.f or 0) + hdr.version = f32(1) + hdr.shofs = fofs(ffi.offsetof(o, "sect")) + hdr.ehsize = f16(ffi.sizeof(hdr)) + hdr.shentsize = f16(ffi.sizeof(o.sect[0])) + hdr.shnum = f16(6) + hdr.shstridx = f16(2) + + -- Fill in sections and symbols. + local sofs, ofs = ffi.offsetof(o, "space"), 1 + for i,name in ipairs{ + ".symtab", ".shstrtab", ".strtab", ".rodata", ".note.GNU-stack", + } do + local sect = o.sect[i] + sect.align = fofs(1) + sect.name = f32(ofs) + ffi.copy(o.space+ofs, name) + ofs = ofs + #name+1 + end + o.sect[1].type = f32(2) -- .symtab + o.sect[1].link = f32(3) + o.sect[1].info = f32(1) + o.sect[1].align = fofs(8) + o.sect[1].ofs = fofs(ffi.offsetof(o, "sym")) + o.sect[1].entsize = fofs(ffi.sizeof(o.sym[0])) + o.sect[1].size = fofs(ffi.sizeof(o.sym)) + o.sym[1].name = f32(1) + o.sym[1].sectidx = f16(4) + o.sym[1].size = fofs(#s) + o.sym[1].info = 17 + o.sect[2].type = f32(3) -- .shstrtab + o.sect[2].ofs = fofs(sofs) + o.sect[2].size = fofs(ofs) + o.sect[3].type = f32(3) -- .strtab + o.sect[3].ofs = fofs(sofs + ofs) + o.sect[3].size = fofs(#symname+2) + ffi.copy(o.space+ofs+1, symname) + ofs = ofs + #symname + 2 + o.sect[4].type = f32(1) -- .rodata + o.sect[4].flags = fofs(2) + o.sect[4].ofs = fofs(sofs + ofs) + o.sect[4].size = fofs(#s) + o.sect[5].type = f32(1) -- .note.GNU-stack + o.sect[5].ofs = fofs(sofs + ofs + #s) + + -- Write ELF object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, ffi.sizeof(o)-4096+ofs)) + bcsave_tail(fp, output, s) +end + +local function bcsave_peobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct { + uint16_t arch, nsects; + uint32_t time, symtabofs, nsyms; + uint16_t opthdrsz, flags; +} PEheader; +typedef struct { + char name[8]; + uint32_t vsize, vaddr, size, ofs, relocofs, lineofs; + uint16_t nreloc, nline; + uint32_t flags; +} PEsection; +typedef struct __attribute((packed)) { + union { + char name[8]; + uint32_t nameref[2]; + }; + uint32_t value; + int16_t sect; + uint16_t type; + uint8_t scl, naux; +} PEsym; +typedef struct __attribute((packed)) { + uint32_t size; + uint16_t nreloc, nline; + uint32_t cksum; + uint16_t assoc; + uint8_t comdatsel, unused[3]; +} PEsymaux; +typedef struct { + PEheader hdr; + PEsection sect[2]; + // Must be an even number of symbol structs. + PEsym sym0; + PEsymaux sym0aux; + PEsym sym1; + PEsymaux sym1aux; + PEsym sym2; + PEsym sym3; + uint32_t strtabsize; + uint8_t space[4096]; +} PEobj; +]] + local symname = LJBC_PREFIX..ctx.modname + local ai = assert(map_arch[ctx.arch]) + local is64 = ai.b == 64 + local symexport = " /EXPORT:"..symname..",DATA " + + -- The file format is always little-endian. Swap if the host is big-endian. + local function f32(x) return x end + local f16 = f32 + if ffi.abi("be") then + f32 = bit.bswap + function f16(x) return bit.rshift(bit.bswap(x), 16) end + end + + -- Create PE object and fill in header. + local o = ffi.new("PEobj") + local hdr = o.hdr + hdr.arch = f16(assert(ai.p)) + hdr.nsects = f16(2) + hdr.symtabofs = f32(ffi.offsetof(o, "sym0")) + hdr.nsyms = f32(6) + + -- Fill in sections and symbols. + o.sect[0].name = ".drectve" + o.sect[0].size = f32(#symexport) + o.sect[0].flags = f32(0x00100a00) + o.sym0.sect = f16(1) + o.sym0.scl = 3 + o.sym0.name = ".drectve" + o.sym0.naux = 1 + o.sym0aux.size = f32(#symexport) + o.sect[1].name = ".rdata" + o.sect[1].size = f32(#s) + o.sect[1].flags = f32(0x40300040) + o.sym1.sect = f16(2) + o.sym1.scl = 3 + o.sym1.name = ".rdata" + o.sym1.naux = 1 + o.sym1aux.size = f32(#s) + o.sym2.sect = f16(2) + o.sym2.scl = 2 + o.sym2.nameref[1] = f32(4) + o.sym3.sect = f16(-1) + o.sym3.scl = 2 + o.sym3.value = f32(1) + o.sym3.name = "@feat.00" -- Mark as SafeSEH compliant. + ffi.copy(o.space, symname) + local ofs = #symname + 1 + o.strtabsize = f32(ofs + 4) + o.sect[0].ofs = f32(ffi.offsetof(o, "space") + ofs) + ffi.copy(o.space + ofs, symexport) + ofs = ofs + #symexport + o.sect[1].ofs = f32(ffi.offsetof(o, "space") + ofs) + + -- Write PE object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, ffi.sizeof(o)-4096+ofs)) + bcsave_tail(fp, output, s) +end + +local function bcsave_machobj(ctx, output, s, ffi) + ffi.cdef[[ +typedef struct +{ + uint32_t magic, cputype, cpusubtype, filetype, ncmds, sizeofcmds, flags; +} mach_header; +typedef struct +{ + mach_header; uint32_t reserved; +} mach_header_64; +typedef struct { + uint32_t cmd, cmdsize; + char segname[16]; + uint32_t vmaddr, vmsize, fileoff, filesize; + uint32_t maxprot, initprot, nsects, flags; +} mach_segment_command; +typedef struct { + uint32_t cmd, cmdsize; + char segname[16]; + uint64_t vmaddr, vmsize, fileoff, filesize; + uint32_t maxprot, initprot, nsects, flags; +} mach_segment_command_64; +typedef struct { + char sectname[16], segname[16]; + uint32_t addr, size; + uint32_t offset, align, reloff, nreloc, flags; + uint32_t reserved1, reserved2; +} mach_section; +typedef struct { + char sectname[16], segname[16]; + uint64_t addr, size; + uint32_t offset, align, reloff, nreloc, flags; + uint32_t reserved1, reserved2, reserved3; +} mach_section_64; +typedef struct { + uint32_t cmd, cmdsize, symoff, nsyms, stroff, strsize; +} mach_symtab_command; +typedef struct { + int32_t strx; + uint8_t type, sect; + int16_t desc; + uint32_t value; +} mach_nlist; +typedef struct { + uint32_t strx; + uint8_t type, sect; + uint16_t desc; + uint64_t value; +} mach_nlist_64; +typedef struct +{ + uint32_t magic, nfat_arch; +} mach_fat_header; +typedef struct +{ + uint32_t cputype, cpusubtype, offset, size, align; +} mach_fat_arch; +typedef struct { + struct { + mach_header hdr; + mach_segment_command seg; + mach_section sec; + mach_symtab_command sym; + } arch[1]; + mach_nlist sym_entry; + uint8_t space[4096]; +} mach_obj; +typedef struct { + struct { + mach_header_64 hdr; + mach_segment_command_64 seg; + mach_section_64 sec; + mach_symtab_command sym; + } arch[1]; + mach_nlist_64 sym_entry; + uint8_t space[4096]; +} mach_obj_64; +typedef struct { + mach_fat_header fat; + mach_fat_arch fat_arch[2]; + struct { + mach_header hdr; + mach_segment_command seg; + mach_section sec; + mach_symtab_command sym; + } arch[2]; + mach_nlist sym_entry; + uint8_t space[4096]; +} mach_fat_obj; +]] + local symname = '_'..LJBC_PREFIX..ctx.modname + local isfat, is64, align, mobj = false, false, 4, "mach_obj" + if ctx.arch == "x64" then + is64, align, mobj = true, 8, "mach_obj_64" + elseif ctx.arch == "arm" then + isfat, mobj = true, "mach_fat_obj" + elseif ctx.arch == "arm64" then + is64, align, isfat, mobj = true, 8, true, "mach_fat_obj" + else + check(ctx.arch == "x86", "unsupported architecture for OSX") + end + local function aligned(v, a) return bit.band(v+a-1, -a) end + local be32 = bit.bswap -- Mach-O FAT is BE, supported archs are LE. + + -- Create Mach-O object and fill in header. + local o = ffi.new(mobj) + local mach_size = aligned(ffi.offsetof(o, "space")+#symname+2, align) + local cputype = ({ x86={7}, x64={0x01000007}, arm={7,12}, arm64={0x01000007,0x0100000c} })[ctx.arch] + local cpusubtype = ({ x86={3}, x64={3}, arm={3,9}, arm64={3,0} })[ctx.arch] + if isfat then + o.fat.magic = be32(0xcafebabe) + o.fat.nfat_arch = be32(#cpusubtype) + end + + -- Fill in sections and symbols. + for i=0,#cpusubtype-1 do + local ofs = 0 + if isfat then + local a = o.fat_arch[i] + a.cputype = be32(cputype[i+1]) + a.cpusubtype = be32(cpusubtype[i+1]) + -- Subsequent slices overlap each other to share data. + ofs = ffi.offsetof(o, "arch") + i*ffi.sizeof(o.arch[0]) + a.offset = be32(ofs) + a.size = be32(mach_size-ofs+#s) + end + local a = o.arch[i] + a.hdr.magic = is64 and 0xfeedfacf or 0xfeedface + a.hdr.cputype = cputype[i+1] + a.hdr.cpusubtype = cpusubtype[i+1] + a.hdr.filetype = 1 + a.hdr.ncmds = 2 + a.hdr.sizeofcmds = ffi.sizeof(a.seg)+ffi.sizeof(a.sec)+ffi.sizeof(a.sym) + a.seg.cmd = is64 and 0x19 or 0x1 + a.seg.cmdsize = ffi.sizeof(a.seg)+ffi.sizeof(a.sec) + a.seg.vmsize = #s + a.seg.fileoff = mach_size-ofs + a.seg.filesize = #s + a.seg.maxprot = 1 + a.seg.initprot = 1 + a.seg.nsects = 1 + ffi.copy(a.sec.sectname, "__data") + ffi.copy(a.sec.segname, "__DATA") + a.sec.size = #s + a.sec.offset = mach_size-ofs + a.sym.cmd = 2 + a.sym.cmdsize = ffi.sizeof(a.sym) + a.sym.symoff = ffi.offsetof(o, "sym_entry")-ofs + a.sym.nsyms = 1 + a.sym.stroff = ffi.offsetof(o, "sym_entry")+ffi.sizeof(o.sym_entry)-ofs + a.sym.strsize = aligned(#symname+2, align) + end + o.sym_entry.type = 0xf + o.sym_entry.sect = 1 + o.sym_entry.strx = 1 + ffi.copy(o.space+1, symname) + + -- Write Macho-O object file. + local fp = savefile(output, "wb") + fp:write(ffi.string(o, mach_size)) + bcsave_tail(fp, output, s) +end + +local function bcsave_obj(ctx, output, s) + local ok, ffi = pcall(require, "ffi") + check(ok, "FFI library required to write this file type") + if ctx.os == "windows" then + return bcsave_peobj(ctx, output, s, ffi) + elseif ctx.os == "osx" then + return bcsave_machobj(ctx, output, s, ffi) + else + return bcsave_elfobj(ctx, output, s, ffi) + end +end + +------------------------------------------------------------------------------ + +local function bclist(input, output) + local f = readfile(input) + require("jit.bc").dump(f, savefile(output, "w"), true) +end + +local function bcsave(ctx, input, output) + local f = readfile(input) + local s = string.dump(f, ctx.strip) + local t = ctx.type + if not t then + t = detecttype(output) + ctx.type = t + end + if t == "raw" then + bcsave_raw(output, s) + else + if not ctx.modname then ctx.modname = detectmodname(input) end + if t == "obj" then + bcsave_obj(ctx, output, s) + else + bcsave_c(ctx, output, s) + end + end +end + +local function docmd(...) + local arg = {...} + local n = 1 + local list = false + local ctx = { + strip = true, arch = jit.arch, os = jit.os:lower(), + type = false, modname = false, + } + while n <= #arg do + local a = arg[n] + if type(a) == "string" and a:sub(1, 1) == "-" and a ~= "-" then + tremove(arg, n) + if a == "--" then break end + for m=2,#a do + local opt = a:sub(m, m) + if opt == "l" then + list = true + elseif opt == "s" then + ctx.strip = true + elseif opt == "g" then + ctx.strip = false + else + if arg[n] == nil or m ~= #a then usage() end + if opt == "e" then + if n ~= 1 then usage() end + arg[1] = check(loadstring(arg[1])) + elseif opt == "n" then + ctx.modname = checkmodname(tremove(arg, n)) + elseif opt == "t" then + ctx.type = checkarg(tremove(arg, n), map_type, "file type") + elseif opt == "a" then + ctx.arch = checkarg(tremove(arg, n), map_arch, "architecture") + elseif opt == "o" then + ctx.os = checkarg(tremove(arg, n), map_os, "OS name") + else + usage() + end + end + end + else + n = n + 1 + end + end + if list then + if #arg == 0 or #arg > 2 then usage() end + bclist(arg[1], arg[2] or "-") + else + if #arg ~= 2 then usage() end + bcsave(ctx, arg[1], arg[2]) + end +end + +------------------------------------------------------------------------------ + +-- Public module functions. +return { + start = docmd -- Process -b command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm.lua new file mode 100644 index 00000000..ba79c47e --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm.lua @@ -0,0 +1,689 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles most user-mode ARMv7 instructions +-- NYI: Advanced SIMD and VFP instructions. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch = string.match, string.gmatch +local concat = table.concat +local bit = require("bit") +local band, bor, ror, tohex = bit.band, bit.bor, bit.ror, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Opcode maps +------------------------------------------------------------------------------ + +local map_loadc = { + shift = 8, mask = 15, + [10] = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "vmovFmDN", "vstmFNdr", + _ = { + shift = 21, mask = 1, + [0] = "vstrFdl", + { shift = 16, mask = 15, [13] = "vpushFdr", _ = "vstmdbFNdr", } + }, + }, + { + shift = 23, mask = 3, + [0] = "vmovFDNm", + { shift = 16, mask = 15, [13] = "vpopFdr", _ = "vldmFNdr", }, + _ = { + shift = 21, mask = 1, + [0] = "vldrFdl", "vldmdbFNdr", + }, + }, + }, + [11] = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "vmovGmDN", "vstmGNdr", + _ = { + shift = 21, mask = 1, + [0] = "vstrGdl", + { shift = 16, mask = 15, [13] = "vpushGdr", _ = "vstmdbGNdr", } + }, + }, + { + shift = 23, mask = 3, + [0] = "vmovGDNm", + { shift = 16, mask = 15, [13] = "vpopGdr", _ = "vldmGNdr", }, + _ = { + shift = 21, mask = 1, + [0] = "vldrGdl", "vldmdbGNdr", + }, + }, + }, + _ = { + shift = 0, mask = 0 -- NYI ldc, mcrr, mrrc. + }, +} + +local map_vfps = { + shift = 6, mask = 0x2c001, + [0] = "vmlaF.dnm", "vmlsF.dnm", + [0x04000] = "vnmlsF.dnm", [0x04001] = "vnmlaF.dnm", + [0x08000] = "vmulF.dnm", [0x08001] = "vnmulF.dnm", + [0x0c000] = "vaddF.dnm", [0x0c001] = "vsubF.dnm", + [0x20000] = "vdivF.dnm", + [0x24000] = "vfnmsF.dnm", [0x24001] = "vfnmaF.dnm", + [0x28000] = "vfmaF.dnm", [0x28001] = "vfmsF.dnm", + [0x2c000] = "vmovF.dY", + [0x2c001] = { + shift = 7, mask = 0x1e01, + [0] = "vmovF.dm", "vabsF.dm", + [0x0200] = "vnegF.dm", [0x0201] = "vsqrtF.dm", + [0x0800] = "vcmpF.dm", [0x0801] = "vcmpeF.dm", + [0x0a00] = "vcmpzF.d", [0x0a01] = "vcmpzeF.d", + [0x0e01] = "vcvtG.dF.m", + [0x1000] = "vcvt.f32.u32Fdm", [0x1001] = "vcvt.f32.s32Fdm", + [0x1800] = "vcvtr.u32F.dm", [0x1801] = "vcvt.u32F.dm", + [0x1a00] = "vcvtr.s32F.dm", [0x1a01] = "vcvt.s32F.dm", + }, +} + +local map_vfpd = { + shift = 6, mask = 0x2c001, + [0] = "vmlaG.dnm", "vmlsG.dnm", + [0x04000] = "vnmlsG.dnm", [0x04001] = "vnmlaG.dnm", + [0x08000] = "vmulG.dnm", [0x08001] = "vnmulG.dnm", + [0x0c000] = "vaddG.dnm", [0x0c001] = "vsubG.dnm", + [0x20000] = "vdivG.dnm", + [0x24000] = "vfnmsG.dnm", [0x24001] = "vfnmaG.dnm", + [0x28000] = "vfmaG.dnm", [0x28001] = "vfmsG.dnm", + [0x2c000] = "vmovG.dY", + [0x2c001] = { + shift = 7, mask = 0x1e01, + [0] = "vmovG.dm", "vabsG.dm", + [0x0200] = "vnegG.dm", [0x0201] = "vsqrtG.dm", + [0x0800] = "vcmpG.dm", [0x0801] = "vcmpeG.dm", + [0x0a00] = "vcmpzG.d", [0x0a01] = "vcmpzeG.d", + [0x0e01] = "vcvtF.dG.m", + [0x1000] = "vcvt.f64.u32GdFm", [0x1001] = "vcvt.f64.s32GdFm", + [0x1800] = "vcvtr.u32FdG.m", [0x1801] = "vcvt.u32FdG.m", + [0x1a00] = "vcvtr.s32FdG.m", [0x1a01] = "vcvt.s32FdG.m", + }, +} + +local map_datac = { + shift = 24, mask = 1, + [0] = { + shift = 4, mask = 1, + [0] = { + shift = 8, mask = 15, + [10] = map_vfps, + [11] = map_vfpd, + -- NYI cdp, mcr, mrc. + }, + { + shift = 8, mask = 15, + [10] = { + shift = 20, mask = 15, + [0] = "vmovFnD", "vmovFDn", + [14] = "vmsrD", + [15] = { shift = 12, mask = 15, [15] = "vmrs", _ = "vmrsD", }, + }, + }, + }, + "svcT", +} + +local map_loadcu = { + shift = 0, mask = 0, -- NYI unconditional CP load/store. +} + +local map_datacu = { + shift = 0, mask = 0, -- NYI unconditional CP data. +} + +local map_simddata = { + shift = 0, mask = 0, -- NYI SIMD data. +} + +local map_simdload = { + shift = 0, mask = 0, -- NYI SIMD load/store, preload. +} + +local map_preload = { + shift = 0, mask = 0, -- NYI preload. +} + +local map_media = { + shift = 20, mask = 31, + [0] = false, + { --01 + shift = 5, mask = 7, + [0] = "sadd16DNM", "sasxDNM", "ssaxDNM", "ssub16DNM", + "sadd8DNM", false, false, "ssub8DNM", + }, + { --02 + shift = 5, mask = 7, + [0] = "qadd16DNM", "qasxDNM", "qsaxDNM", "qsub16DNM", + "qadd8DNM", false, false, "qsub8DNM", + }, + { --03 + shift = 5, mask = 7, + [0] = "shadd16DNM", "shasxDNM", "shsaxDNM", "shsub16DNM", + "shadd8DNM", false, false, "shsub8DNM", + }, + false, + { --05 + shift = 5, mask = 7, + [0] = "uadd16DNM", "uasxDNM", "usaxDNM", "usub16DNM", + "uadd8DNM", false, false, "usub8DNM", + }, + { --06 + shift = 5, mask = 7, + [0] = "uqadd16DNM", "uqasxDNM", "uqsaxDNM", "uqsub16DNM", + "uqadd8DNM", false, false, "uqsub8DNM", + }, + { --07 + shift = 5, mask = 7, + [0] = "uhadd16DNM", "uhasxDNM", "uhsaxDNM", "uhsub16DNM", + "uhadd8DNM", false, false, "uhsub8DNM", + }, + { --08 + shift = 5, mask = 7, + [0] = "pkhbtDNMU", false, "pkhtbDNMU", + { shift = 16, mask = 15, [15] = "sxtb16DMU", _ = "sxtab16DNMU", }, + "pkhbtDNMU", "selDNM", "pkhtbDNMU", + }, + false, + { --0a + shift = 5, mask = 7, + [0] = "ssatDxMu", "ssat16DxM", "ssatDxMu", + { shift = 16, mask = 15, [15] = "sxtbDMU", _ = "sxtabDNMU", }, + "ssatDxMu", false, "ssatDxMu", + }, + { --0b + shift = 5, mask = 7, + [0] = "ssatDxMu", "revDM", "ssatDxMu", + { shift = 16, mask = 15, [15] = "sxthDMU", _ = "sxtahDNMU", }, + "ssatDxMu", "rev16DM", "ssatDxMu", + }, + { --0c + shift = 5, mask = 7, + [3] = { shift = 16, mask = 15, [15] = "uxtb16DMU", _ = "uxtab16DNMU", }, + }, + false, + { --0e + shift = 5, mask = 7, + [0] = "usatDwMu", "usat16DwM", "usatDwMu", + { shift = 16, mask = 15, [15] = "uxtbDMU", _ = "uxtabDNMU", }, + "usatDwMu", false, "usatDwMu", + }, + { --0f + shift = 5, mask = 7, + [0] = "usatDwMu", "rbitDM", "usatDwMu", + { shift = 16, mask = 15, [15] = "uxthDMU", _ = "uxtahDNMU", }, + "usatDwMu", "revshDM", "usatDwMu", + }, + { --10 + shift = 12, mask = 15, + [15] = { + shift = 5, mask = 7, + "smuadNMS", "smuadxNMS", "smusdNMS", "smusdxNMS", + }, + _ = { + shift = 5, mask = 7, + [0] = "smladNMSD", "smladxNMSD", "smlsdNMSD", "smlsdxNMSD", + }, + }, + false, false, false, + { --14 + shift = 5, mask = 7, + [0] = "smlaldDNMS", "smlaldxDNMS", "smlsldDNMS", "smlsldxDNMS", + }, + { --15 + shift = 5, mask = 7, + [0] = { shift = 12, mask = 15, [15] = "smmulNMS", _ = "smmlaNMSD", }, + { shift = 12, mask = 15, [15] = "smmulrNMS", _ = "smmlarNMSD", }, + false, false, false, false, + "smmlsNMSD", "smmlsrNMSD", + }, + false, false, + { --18 + shift = 5, mask = 7, + [0] = { shift = 12, mask = 15, [15] = "usad8NMS", _ = "usada8NMSD", }, + }, + false, + { --1a + shift = 5, mask = 3, [2] = "sbfxDMvw", + }, + { --1b + shift = 5, mask = 3, [2] = "sbfxDMvw", + }, + { --1c + shift = 5, mask = 3, + [0] = { shift = 0, mask = 15, [15] = "bfcDvX", _ = "bfiDMvX", }, + }, + { --1d + shift = 5, mask = 3, + [0] = { shift = 0, mask = 15, [15] = "bfcDvX", _ = "bfiDMvX", }, + }, + { --1e + shift = 5, mask = 3, [2] = "ubfxDMvw", + }, + { --1f + shift = 5, mask = 3, [2] = "ubfxDMvw", + }, +} + +local map_load = { + shift = 21, mask = 9, + { + shift = 20, mask = 5, + [0] = "strtDL", "ldrtDL", [4] = "strbtDL", [5] = "ldrbtDL", + }, + _ = { + shift = 20, mask = 5, + [0] = "strDL", "ldrDL", [4] = "strbDL", [5] = "ldrbDL", + } +} + +local map_load1 = { + shift = 4, mask = 1, + [0] = map_load, map_media, +} + +local map_loadm = { + shift = 20, mask = 1, + [0] = { + shift = 23, mask = 3, + [0] = "stmdaNR", "stmNR", + { shift = 16, mask = 63, [45] = "pushR", _ = "stmdbNR", }, "stmibNR", + }, + { + shift = 23, mask = 3, + [0] = "ldmdaNR", { shift = 16, mask = 63, [61] = "popR", _ = "ldmNR", }, + "ldmdbNR", "ldmibNR", + }, +} + +local map_data = { + shift = 21, mask = 15, + [0] = "andDNPs", "eorDNPs", "subDNPs", "rsbDNPs", + "addDNPs", "adcDNPs", "sbcDNPs", "rscDNPs", + "tstNP", "teqNP", "cmpNP", "cmnNP", + "orrDNPs", "movDPs", "bicDNPs", "mvnDPs", +} + +local map_mul = { + shift = 21, mask = 7, + [0] = "mulNMSs", "mlaNMSDs", "umaalDNMS", "mlsDNMS", + "umullDNMSs", "umlalDNMSs", "smullDNMSs", "smlalDNMSs", +} + +local map_sync = { + shift = 20, mask = 15, -- NYI: brackets around N. R(D+1) for ldrexd/strexd. + [0] = "swpDMN", false, false, false, + "swpbDMN", false, false, false, + "strexDMN", "ldrexDN", "strexdDN", "ldrexdDN", + "strexbDMN", "ldrexbDN", "strexhDN", "ldrexhDN", +} + +local map_mulh = { + shift = 21, mask = 3, + [0] = { shift = 5, mask = 3, + [0] = "smlabbNMSD", "smlatbNMSD", "smlabtNMSD", "smlattNMSD", }, + { shift = 5, mask = 3, + [0] = "smlawbNMSD", "smulwbNMS", "smlawtNMSD", "smulwtNMS", }, + { shift = 5, mask = 3, + [0] = "smlalbbDNMS", "smlaltbDNMS", "smlalbtDNMS", "smlalttDNMS", }, + { shift = 5, mask = 3, + [0] = "smulbbNMS", "smultbNMS", "smulbtNMS", "smulttNMS", }, +} + +local map_misc = { + shift = 4, mask = 7, + -- NYI: decode PSR bits of msr. + [0] = { shift = 21, mask = 1, [0] = "mrsD", "msrM", }, + { shift = 21, mask = 3, "bxM", false, "clzDM", }, + { shift = 21, mask = 3, "bxjM", }, + { shift = 21, mask = 3, "blxM", }, + false, + { shift = 21, mask = 3, [0] = "qaddDMN", "qsubDMN", "qdaddDMN", "qdsubDMN", }, + false, + { shift = 21, mask = 3, "bkptK", }, +} + +local map_datar = { + shift = 4, mask = 9, + [9] = { + shift = 5, mask = 3, + [0] = { shift = 24, mask = 1, [0] = map_mul, map_sync, }, + { shift = 20, mask = 1, [0] = "strhDL", "ldrhDL", }, + { shift = 20, mask = 1, [0] = "ldrdDL", "ldrsbDL", }, + { shift = 20, mask = 1, [0] = "strdDL", "ldrshDL", }, + }, + _ = { + shift = 20, mask = 25, + [16] = { shift = 7, mask = 1, [0] = map_misc, map_mulh, }, + _ = { + shift = 0, mask = 0xffffffff, + [bor(0xe1a00000)] = "nop", + _ = map_data, + } + }, +} + +local map_datai = { + shift = 20, mask = 31, -- NYI: decode PSR bits of msr. Decode imm12. + [16] = "movwDW", [20] = "movtDW", + [18] = { shift = 0, mask = 0xf00ff, [0] = "nopv6", _ = "msrNW", }, + [22] = "msrNW", + _ = map_data, +} + +local map_branch = { + shift = 24, mask = 1, + [0] = "bB", "blB" +} + +local map_condins = { + [0] = map_datar, map_datai, map_load, map_load1, + map_loadm, map_branch, map_loadc, map_datac +} + +-- NYI: setend. +local map_uncondins = { + [0] = false, map_simddata, map_simdload, map_preload, + false, "blxB", map_loadcu, map_datacu, +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "sp", "lr", "pc", +} + +local map_cond = { + [0] = "eq", "ne", "hs", "lo", "mi", "pl", "vs", "vc", + "hi", "ls", "ge", "lt", "gt", "le", "al", +} + +local map_shift = { [0] = "lsl", "lsr", "asr", "ror", } + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then + extra = "\t->"..sym + elseif band(ctx.op, 0x0e000000) ~= 0x0a000000 then + extra = "\t; 0x"..tohex(ctx.rel) + end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-5s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-5s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +-- Format operand 2 of load/store opcodes. +local function fmtload(ctx, op, pos) + local base = map_gpr[band(rshift(op, 16), 15)] + local x, ofs + local ext = (band(op, 0x04000000) == 0) + if not ext and band(op, 0x02000000) == 0 then + ofs = band(op, 4095) + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + ofs = "#"..ofs + elseif ext and band(op, 0x00400000) ~= 0 then + ofs = band(op, 15) + band(rshift(op, 4), 0xf0) + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + ofs = "#"..ofs + else + ofs = map_gpr[band(op, 15)] + if ext or band(op, 0xfe0) == 0 then + elseif band(op, 0xfe0) == 0x60 then + ofs = format("%s, rrx", ofs) + else + local sh = band(rshift(op, 7), 31) + if sh == 0 then sh = 32 end + ofs = format("%s, %s #%d", ofs, map_shift[band(rshift(op, 5), 3)], sh) + end + if band(op, 0x00800000) == 0 then ofs = "-"..ofs end + end + if ofs == "#0" then + x = format("[%s]", base) + elseif band(op, 0x01000000) == 0 then + x = format("[%s], %s", base, ofs) + else + x = format("[%s, %s]", base, ofs) + end + if band(op, 0x01200000) == 0x01200000 then x = x.."!" end + return x +end + +-- Format operand 2 of vector load/store opcodes. +local function fmtvload(ctx, op, pos) + local base = map_gpr[band(rshift(op, 16), 15)] + local ofs = band(op, 255)*4 + if band(op, 0x00800000) == 0 then ofs = -ofs end + if base == "pc" then ctx.rel = ctx.addr + pos + 8 + ofs end + if ofs == 0 then + return format("[%s]", base) + else + return format("[%s, #%d]", base, ofs) + end +end + +local function fmtvr(op, vr, sh0, sh1) + if vr == "s" then + return format("s%d", 2*band(rshift(op, sh0), 15)+band(rshift(op, sh1), 1)) + else + return format("d%d", band(rshift(op, sh0), 15)+band(rshift(op, sh1-4), 16)) + end +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) + local operands = {} + local suffix = "" + local last, name, pat + local vr + ctx.op = op + ctx.rel = nil + + local cond = rshift(op, 28) + local opat + if cond == 15 then + opat = map_uncondins[band(rshift(op, 25), 7)] + else + if cond ~= 14 then suffix = map_cond[cond] end + opat = map_condins[band(rshift(op, 25), 7)] + end + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + name, pat = match(opat, "^([a-z0-9]*)(.*)") + if sub(pat, 1, 1) == "." then + local s2, p2 = match(pat, "^([a-z0-9.]*)(.*)") + suffix = suffix..s2 + pat = p2 + end + + for p in gmatch(pat, ".") do + local x = nil + if p == "D" then + x = map_gpr[band(rshift(op, 12), 15)] + elseif p == "N" then + x = map_gpr[band(rshift(op, 16), 15)] + elseif p == "S" then + x = map_gpr[band(rshift(op, 8), 15)] + elseif p == "M" then + x = map_gpr[band(op, 15)] + elseif p == "d" then + x = fmtvr(op, vr, 12, 22) + elseif p == "n" then + x = fmtvr(op, vr, 16, 7) + elseif p == "m" then + x = fmtvr(op, vr, 0, 5) + elseif p == "P" then + if band(op, 0x02000000) ~= 0 then + x = ror(band(op, 255), 2*band(rshift(op, 8), 15)) + else + x = map_gpr[band(op, 15)] + if band(op, 0xff0) ~= 0 then + operands[#operands+1] = x + local s = map_shift[band(rshift(op, 5), 3)] + local r = nil + if band(op, 0xf90) == 0 then + if s == "ror" then s = "rrx" else r = "#32" end + elseif band(op, 0x10) == 0 then + r = "#"..band(rshift(op, 7), 31) + else + r = map_gpr[band(rshift(op, 8), 15)] + end + if name == "mov" then name = s; x = r + elseif r then x = format("%s %s", s, r) + else x = s end + end + end + elseif p == "L" then + x = fmtload(ctx, op, pos) + elseif p == "l" then + x = fmtvload(ctx, op, pos) + elseif p == "B" then + local addr = ctx.addr + pos + 8 + arshift(lshift(op, 8), 6) + if cond == 15 then addr = addr + band(rshift(op, 23), 2) end + ctx.rel = addr + x = "0x"..tohex(addr) + elseif p == "F" then + vr = "s" + elseif p == "G" then + vr = "d" + elseif p == "." then + suffix = suffix..(vr == "s" and ".f32" or ".f64") + elseif p == "R" then + if band(op, 0x00200000) ~= 0 and #operands == 1 then + operands[1] = operands[1].."!" + end + local t = {} + for i=0,15 do + if band(rshift(op, i), 1) == 1 then t[#t+1] = map_gpr[i] end + end + x = "{"..concat(t, ", ").."}" + elseif p == "r" then + if band(op, 0x00200000) ~= 0 and #operands == 2 then + operands[1] = operands[1].."!" + end + local s = tonumber(sub(last, 2)) + local n = band(op, 255) + if vr == "d" then n = rshift(n, 1) end + operands[#operands] = format("{%s-%s%d}", last, vr, s+n-1) + elseif p == "W" then + x = band(op, 0x0fff) + band(rshift(op, 4), 0xf000) + elseif p == "T" then + x = "#0x"..tohex(band(op, 0x00ffffff), 6) + elseif p == "U" then + x = band(rshift(op, 7), 31) + if x == 0 then x = nil end + elseif p == "u" then + x = band(rshift(op, 7), 31) + if band(op, 0x40) == 0 then + if x == 0 then x = nil else x = "lsl #"..x end + else + if x == 0 then x = "asr #32" else x = "asr #"..x end + end + elseif p == "v" then + x = band(rshift(op, 7), 31) + elseif p == "w" then + x = band(rshift(op, 16), 31) + elseif p == "x" then + x = band(rshift(op, 16), 31) + 1 + elseif p == "X" then + x = band(rshift(op, 16), 31) - last + 1 + elseif p == "Y" then + x = band(rshift(op, 12), 0xf0) + band(op, 0x0f) + elseif p == "K" then + x = "#0x"..tohex(band(rshift(op, 4), 0x0000fff0) + band(op, 15), 4) + elseif p == "s" then + if band(op, 0x00100000) ~= 0 then suffix = "s"..suffix end + else + assert(false) + end + if x then + last = x + if type(x) == "number" then x = "#"..x end + operands[#operands+1] = x + end + end + + return putop(ctx, name..suffix, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ctx.pos = ofs + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 16 then return map_gpr[r] end + return "d"..(r-16) +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm64.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm64.lua new file mode 100644 index 00000000..ad909fbd --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm64.lua @@ -0,0 +1,1216 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM64 disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +-- +-- Contributed by Djordje Kovacevic and Stefan Pejic from RT-RK.com. +-- Sponsored by Cisco Systems, Inc. +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles most user-mode AArch64 instructions. +-- NYI: Advanced SIMD and VFP instructions. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local concat = table.concat +local bit = require("bit") +local band, bor, bxor, tohex = bit.band, bit.bor, bit.bxor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift +local ror = bit.ror + +------------------------------------------------------------------------------ +-- Opcode maps +------------------------------------------------------------------------------ + +local map_adr = { -- PC-relative addressing. + shift = 31, mask = 1, + [0] = "adrDBx", "adrpDBx" +} + +local map_addsubi = { -- Add/subtract immediate. + shift = 29, mask = 3, + [0] = "add|movDNIg", "adds|cmnD0NIg", "subDNIg", "subs|cmpD0NIg", +} + +local map_logi = { -- Logical immediate. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "andDNig", "orr|movDN0ig", "eorDNig", "ands|tstD0Nig" + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = "andDNig", "orr|movDN0ig", "eorDNig", "ands|tstD0Nig" + } +} + +local map_movwi = { -- Move wide immediate. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "movnDWRg", false, "movz|movDYRg", "movkDWRg" + }, false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = "movnDWRg", false, "movz|movDYRg", "movkDWRg" + }, +} + +local map_bitf = { -- Bitfield. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "sbfm|sbfiz|sbfx|asr|sxtw|sxth|sxtbDN12w", + "bfm|bfi|bfxilDN13w", + "ubfm|ubfiz|ubfx|lsr|lsl|uxth|uxtbDN12w" + } + }, + { + shift = 22, mask = 1, + { + shift = 29, mask = 3, + [0] = "sbfm|sbfiz|sbfx|asr|sxtw|sxth|sxtbDN12x", + "bfm|bfi|bfxilDN13x", + "ubfm|ubfiz|ubfx|lsr|lsl|uxth|uxtbDN12x" + } + } +} + +local map_datai = { -- Data processing - immediate. + shift = 23, mask = 7, + [0] = map_adr, map_adr, map_addsubi, false, + map_logi, map_movwi, map_bitf, + { + shift = 15, mask = 0x1c0c1, + [0] = "extr|rorDNM4w", [0x10080] = "extr|rorDNM4x", + [0x10081] = "extr|rorDNM4x" + } +} + +local map_logsr = { -- Logical, shifted register. + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = { + shift = 21, mask = 7, + [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg", + "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg" + }, + { + shift = 21, mask = 7, + [0] ="orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg", + "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg" + }, + { + shift = 21, mask = 7, + [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg", + "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg" + }, + { + shift = 21, mask = 7, + [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg", + "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg" + } + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = { + shift = 21, mask = 7, + [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg", + "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg" + }, + { + shift = 21, mask = 7, + [0] = "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg", + "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg" + }, + { + shift = 21, mask = 7, + [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg", + "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg" + }, + { + shift = 21, mask = 7, + [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg", + "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg" + } + } +} + +local map_assh = { + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "addDNMSg", "addDNMSg", "addDNMSg", "addDNMg" + }, + { + shift = 22, mask = 3, + [0] = "adds|cmnD0NMSg", "adds|cmnD0NMSg", + "adds|cmnD0NMSg", "adds|cmnD0NMg" + }, + { + shift = 22, mask = 3, + [0] = "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0Mg" + }, + { + shift = 22, mask = 3, + [0] = "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0MzSg", + "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0Mzg" + }, + }, + false -- unallocated + }, + { + shift = 29, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "addDNMSg", "addDNMSg", "addDNMSg", "addDNMg" + }, + { + shift = 22, mask = 3, + [0] = "adds|cmnD0NMSg", "adds|cmnD0NMSg", "adds|cmnD0NMSg", + "adds|cmnD0NMg" + }, + { + shift = 22, mask = 3, + [0] = "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0MSg", "sub|negDN0Mg" + }, + { + shift = 22, mask = 3, + [0] = "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0MzSg", + "subs|cmp|negsD0N0MzSg", "subs|cmp|negsD0N0Mzg" + } + } +} + +local map_addsubsh = { -- Add/subtract, shifted register. + shift = 22, mask = 3, + [0] = map_assh, map_assh, map_assh +} + +local map_addsubex = { -- Add/subtract, extended register. + shift = 22, mask = 3, + [0] = { + shift = 29, mask = 3, + [0] = "addDNMXg", "adds|cmnD0NMXg", "subDNMXg", "subs|cmpD0NMzXg", + } +} + +local map_addsubc = { -- Add/subtract, with carry. + shift = 10, mask = 63, + [0] = { + shift = 29, mask = 3, + [0] = "adcDNMg", "adcsDNMg", "sbc|ngcDN0Mg", "sbcs|ngcsDN0Mg", + } +} + +local map_ccomp = { + shift = 4, mask = 1, + [0] = { + shift = 10, mask = 3, + [0] = { -- Conditional compare register. + shift = 29, mask = 3, + "ccmnNMVCg", false, "ccmpNMVCg", + }, + [2] = { -- Conditional compare immediate. + shift = 29, mask = 3, + "ccmnN5VCg", false, "ccmpN5VCg", + } + } +} + +local map_csel = { -- Conditional select. + shift = 11, mask = 1, + [0] = { + shift = 10, mask = 1, + [0] = { + shift = 29, mask = 3, + [0] = "cselDNMzCg", false, "csinv|cinv|csetmDNMcg", false, + }, + { + shift = 29, mask = 3, + [0] = "csinc|cinc|csetDNMcg", false, "csneg|cnegDNMcg", false, + } + } +} + +local map_data1s = { -- Data processing, 1 source. + shift = 29, mask = 1, + [0] = { + shift = 31, mask = 1, + [0] = { + shift = 10, mask = 0x7ff, + [0] = "rbitDNg", "rev16DNg", "revDNw", false, "clzDNg", "clsDNg" + }, + { + shift = 10, mask = 0x7ff, + [0] = "rbitDNg", "rev16DNg", "rev32DNx", "revDNx", "clzDNg", "clsDNg" + } + } +} + +local map_data2s = { -- Data processing, 2 sources. + shift = 29, mask = 1, + [0] = { + shift = 10, mask = 63, + false, "udivDNMg", "sdivDNMg", false, false, false, false, "lslDNMg", + "lsrDNMg", "asrDNMg", "rorDNMg" + } +} + +local map_data3s = { -- Data processing, 3 sources. + shift = 29, mask = 7, + [0] = { + shift = 21, mask = 7, + [0] = { + shift = 15, mask = 1, + [0] = "madd|mulDNMA0g", "msub|mnegDNMA0g" + } + }, false, false, false, + { + shift = 15, mask = 1, + [0] = { + shift = 21, mask = 7, + [0] = "madd|mulDNMA0g", "smaddl|smullDxNMwA0x", "smulhDNMx", false, + false, "umaddl|umullDxNMwA0x", "umulhDNMx" + }, + { + shift = 21, mask = 7, + [0] = "msub|mnegDNMA0g", "smsubl|smneglDxNMwA0x", false, false, + false, "umsubl|umneglDxNMwA0x" + } + } +} + +local map_datar = { -- Data processing, register. + shift = 28, mask = 1, + [0] = { + shift = 24, mask = 1, + [0] = map_logsr, + { + shift = 21, mask = 1, + [0] = map_addsubsh, map_addsubex + } + }, + { + shift = 21, mask = 15, + [0] = map_addsubc, false, map_ccomp, false, map_csel, false, + { + shift = 30, mask = 1, + [0] = map_data2s, map_data1s + }, + false, map_data3s, map_data3s, map_data3s, map_data3s, map_data3s, + map_data3s, map_data3s, map_data3s + } +} + +local map_lrl = { -- Load register, literal. + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = "ldrDwB", "ldrDxB", "ldrswDxB" + }, + { + shift = 30, mask = 3, + [0] = "ldrDsB", "ldrDdB" + } +} + +local map_lsriind = { -- Load/store register, immediate pre/post-indexed. + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwzL", "ldrbDwzL", "ldrsbDxzL", "ldrsbDwzL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strhDwzL", "ldrhDwzL", "ldrshDxzL", "ldrshDwzL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strDwzL", "ldrDwzL", "ldrswDxzL" + }, + { + shift = 22, mask = 3, + [0] = "strDszL", "ldrDszL" + } + }, + { + shift = 26, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = "strDxzL", "ldrDxzL" + }, + { + shift = 22, mask = 3, + [0] = "strDdzL", "ldrDdzL" + } + } +} + +local map_lsriro = { + shift = 21, mask = 1, + [0] = { -- Load/store register immediate. + shift = 10, mask = 3, + [0] = { -- Unscaled immediate. + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "sturbDwK", "ldurbDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturhDwK", "ldurhDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturDwK", "ldurDwK" + }, + { + shift = 22, mask = 3, + [0] = "sturDxK", "ldurDxK" + } + } + }, map_lsriind, false, map_lsriind + }, + { -- Load/store register, register offset. + shift = 10, mask = 3, + [2] = { + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwO", "ldrbDwO", "ldrsbDxO", "ldrsbDwO" + }, + { + shift = 22, mask = 3, + [0] = "strhDwO", "ldrhDwO", "ldrshDxO", "ldrshDwO" + }, + { + shift = 22, mask = 3, + [0] = "strDwO", "ldrDwO", "ldrswDxO" + }, + { + shift = 22, mask = 3, + [0] = "strDxO", "ldrDxO" + } + }, + { + shift = 30, mask = 3, + [2] = { + shift = 22, mask = 3, + [0] = "strDsO", "ldrDsO" + }, + [3] = { + shift = 22, mask = 3, + [0] = "strDdO", "ldrDdO" + } + } + } + } +} + +local map_lsp = { -- Load/store register pair, offset. + shift = 22, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = "stpDzAzwP", "stpDzAzsP", + }, + { + shift = 26, mask = 1, + "stpDzAzdP" + }, + { + shift = 26, mask = 1, + [0] = "stpDzAzxP" + } + }, + { + shift = 30, mask = 3, + [0] = { + shift = 26, mask = 1, + [0] = "ldpDzAzwP", "ldpDzAzsP", + }, + { + shift = 26, mask = 1, + [0] = "ldpswDAxP", "ldpDzAzdP" + }, + { + shift = 26, mask = 1, + [0] = "ldpDzAzxP" + } + } +} + +local map_ls = { -- Loads and stores. + shift = 24, mask = 0x31, + [0x10] = map_lrl, [0x30] = map_lsriro, + [0x20] = { + shift = 23, mask = 3, + map_lsp, map_lsp, map_lsp + }, + [0x21] = { + shift = 23, mask = 3, + map_lsp, map_lsp, map_lsp + }, + [0x31] = { + shift = 26, mask = 1, + [0] = { + shift = 30, mask = 3, + [0] = { + shift = 22, mask = 3, + [0] = "strbDwzU", "ldrbDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strhDwzU", "ldrhDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strDwzU", "ldrDwzU" + }, + { + shift = 22, mask = 3, + [0] = "strDxzU", "ldrDxzU" + } + }, + { + shift = 30, mask = 3, + [2] = { + shift = 22, mask = 3, + [0] = "strDszU", "ldrDszU" + }, + [3] = { + shift = 22, mask = 3, + [0] = "strDdzU", "ldrDdzU" + } + } + }, +} + +local map_datafp = { -- Data processing, SIMD and FP. + shift = 28, mask = 7, + { -- 001 + shift = 24, mask = 1, + [0] = { + shift = 21, mask = 1, + { + shift = 10, mask = 3, + [0] = { + shift = 12, mask = 1, + [0] = { + shift = 13, mask = 1, + [0] = { + shift = 14, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { -- FP/int conversion. + shift = 31, mask = 1, + [0] = { + shift = 16, mask = 0xff, + [0x20] = "fcvtnsDwNs", [0x21] = "fcvtnuDwNs", + [0x22] = "scvtfDsNw", [0x23] = "ucvtfDsNw", + [0x24] = "fcvtasDwNs", [0x25] = "fcvtauDwNs", + [0x26] = "fmovDwNs", [0x27] = "fmovDsNw", + [0x28] = "fcvtpsDwNs", [0x29] = "fcvtpuDwNs", + [0x30] = "fcvtmsDwNs", [0x31] = "fcvtmuDwNs", + [0x38] = "fcvtzsDwNs", [0x39] = "fcvtzuDwNs", + [0x60] = "fcvtnsDwNd", [0x61] = "fcvtnuDwNd", + [0x62] = "scvtfDdNw", [0x63] = "ucvtfDdNw", + [0x64] = "fcvtasDwNd", [0x65] = "fcvtauDwNd", + [0x68] = "fcvtpsDwNd", [0x69] = "fcvtpuDwNd", + [0x70] = "fcvtmsDwNd", [0x71] = "fcvtmuDwNd", + [0x78] = "fcvtzsDwNd", [0x79] = "fcvtzuDwNd" + }, + { + shift = 16, mask = 0xff, + [0x20] = "fcvtnsDxNs", [0x21] = "fcvtnuDxNs", + [0x22] = "scvtfDsNx", [0x23] = "ucvtfDsNx", + [0x24] = "fcvtasDxNs", [0x25] = "fcvtauDxNs", + [0x28] = "fcvtpsDxNs", [0x29] = "fcvtpuDxNs", + [0x30] = "fcvtmsDxNs", [0x31] = "fcvtmuDxNs", + [0x38] = "fcvtzsDxNs", [0x39] = "fcvtzuDxNs", + [0x60] = "fcvtnsDxNd", [0x61] = "fcvtnuDxNd", + [0x62] = "scvtfDdNx", [0x63] = "ucvtfDdNx", + [0x64] = "fcvtasDxNd", [0x65] = "fcvtauDxNd", + [0x66] = "fmovDxNd", [0x67] = "fmovDdNx", + [0x68] = "fcvtpsDxNd", [0x69] = "fcvtpuDxNd", + [0x70] = "fcvtmsDxNd", [0x71] = "fcvtmuDxNd", + [0x78] = "fcvtzsDxNd", [0x79] = "fcvtzuDxNd" + } + } + }, + { -- FP data-processing, 1 source. + shift = 31, mask = 1, + [0] = { + shift = 22, mask = 3, + [0] = { + shift = 15, mask = 63, + [0] = "fmovDNf", "fabsDNf", "fnegDNf", + "fsqrtDNf", false, "fcvtDdNs", false, false, + "frintnDNf", "frintpDNf", "frintmDNf", "frintzDNf", + "frintaDNf", false, "frintxDNf", "frintiDNf", + }, + { + shift = 15, mask = 63, + [0] = "fmovDNf", "fabsDNf", "fnegDNf", + "fsqrtDNf", "fcvtDsNd", false, false, false, + "frintnDNf", "frintpDNf", "frintmDNf", "frintzDNf", + "frintaDNf", false, "frintxDNf", "frintiDNf", + } + } + } + }, + { -- FP compare. + shift = 31, mask = 1, + [0] = { + shift = 14, mask = 3, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 0, mask = 31, + [0] = "fcmpNMf", [8] = "fcmpNZf", + [16] = "fcmpeNMf", [24] = "fcmpeNZf", + } + } + } + } + }, + { -- FP immediate. + shift = 31, mask = 1, + [0] = { + shift = 5, mask = 31, + [0] = { + shift = 23, mask = 1, + [0] = "fmovDFf" + } + } + } + }, + { -- FP conditional compare. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 4, mask = 1, + [0] = "fccmpNMVCf", "fccmpeNMVCf" + } + } + }, + { -- FP data-processing, 2 sources. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = { + shift = 12, mask = 15, + [0] = "fmulDNMf", "fdivDNMf", "faddDNMf", "fsubDNMf", + "fmaxDNMf", "fminDNMf", "fmaxnmDNMf", "fminnmDNMf", + "fnmulDNMf" + } + } + }, + { -- FP conditional select. + shift = 31, mask = 1, + [0] = { + shift = 23, mask = 1, + [0] = "fcselDNMCf" + } + } + } + }, + { -- FP data-processing, 3 sources. + shift = 31, mask = 1, + [0] = { + shift = 15, mask = 1, + [0] = { + shift = 21, mask = 5, + [0] = "fmaddDNMAf", "fnmaddDNMAf" + }, + { + shift = 21, mask = 5, + [0] = "fmsubDNMAf", "fnmsubDNMAf" + } + } + } + } +} + +local map_br = { -- Branches, exception generating and system instructions. + shift = 29, mask = 7, + [0] = "bB", + { -- Compare & branch, immediate. + shift = 24, mask = 3, + [0] = "cbzDBg", "cbnzDBg", "tbzDTBw", "tbnzDTBw" + }, + { -- Conditional branch, immediate. + shift = 24, mask = 3, + [0] = { + shift = 4, mask = 1, + [0] = { + shift = 0, mask = 15, + [0] = "beqB", "bneB", "bhsB", "bloB", "bmiB", "bplB", "bvsB", "bvcB", + "bhiB", "blsB", "bgeB", "bltB", "bgtB", "bleB", "balB" + } + } + }, false, "blB", + { -- Compare & branch, immediate. + shift = 24, mask = 3, + [0] = "cbzDBg", "cbnzDBg", "tbzDTBx", "tbnzDTBx" + }, + { + shift = 24, mask = 3, + [0] = { -- Exception generation. + shift = 0, mask = 0xe0001f, + [0x200000] = "brkW" + }, + { -- System instructions. + shift = 0, mask = 0x3fffff, + [0x03201f] = "nop" + }, + { -- Unconditional branch, register. + shift = 0, mask = 0xfffc1f, + [0x1f0000] = "brNx", [0x3f0000] = "blrNx", + [0x5f0000] = "retNx" + }, + } +} + +local map_init = { + shift = 25, mask = 15, + [0] = false, false, false, false, map_ls, map_datar, map_ls, map_datafp, + map_datai, map_datai, map_br, map_br, map_ls, map_datar, map_ls, map_datafp +} + +------------------------------------------------------------------------------ + +local map_regs = { x = {}, w = {}, d = {}, s = {} } + +for i=0,30 do + map_regs.x[i] = "x"..i + map_regs.w[i] = "w"..i + map_regs.d[i] = "d"..i + map_regs.s[i] = "s"..i +end +map_regs.x[31] = "sp" +map_regs.w[31] = "wsp" +map_regs.d[31] = "d31" +map_regs.s[31] = "s31" + +local map_cond = { + [0] = "eq", "ne", "cs", "cc", "mi", "pl", "vs", "vc", + "hi", "ls", "ge", "lt", "gt", "le", "al", +} + +local map_shift = { [0] = "lsl", "lsr", "asr", } + +local map_extend = { + [0] = "uxtb", "uxth", "uxtw", "uxtx", "sxtb", "sxth", "sxtw", "sxtx", +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then + extra = "\t->"..sym + end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-5s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-5s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +local function match_reg(p, pat, regnum) + return map_regs[match(pat, p.."%w-([xwds])")][regnum] +end + +local function fmt_hex32(x) + if x < 0 then + return tohex(x) + else + return format("%x", x) + end +end + +local imm13_rep = { 0x55555555, 0x11111111, 0x01010101, 0x00010001, 0x00000001 } + +local function decode_imm13(op) + local imms = band(rshift(op, 10), 63) + local immr = band(rshift(op, 16), 63) + if band(op, 0x00400000) == 0 then + local len = 5 + if imms >= 56 then + if imms >= 60 then len = 1 else len = 2 end + elseif imms >= 48 then len = 3 elseif imms >= 32 then len = 4 end + local l = lshift(1, len)-1 + local s = band(imms, l) + local r = band(immr, l) + local imm = ror(rshift(-1, 31-s), r) + if len ~= 5 then imm = band(imm, lshift(1, l)-1) + rshift(imm, 31-l) end + imm = imm * imm13_rep[len] + local ix = fmt_hex32(imm) + if rshift(op, 31) ~= 0 then + return ix..tohex(imm) + else + return ix + end + else + local lo, hi = -1, 0 + if imms < 32 then lo = rshift(-1, 31-imms) else hi = rshift(-1, 63-imms) end + if immr ~= 0 then + lo, hi = ror(lo, immr), ror(hi, immr) + local x = immr == 32 and 0 or band(bxor(lo, hi), lshift(-1, 32-immr)) + lo, hi = bxor(lo, x), bxor(hi, x) + if immr >= 32 then lo, hi = hi, lo end + end + if hi ~= 0 then + return fmt_hex32(hi)..tohex(lo) + else + return fmt_hex32(lo) + end + end +end + +local function parse_immpc(op, name) + if name == "b" or name == "bl" then + return arshift(lshift(op, 6), 4) + elseif name == "adr" or name == "adrp" then + local immlo = band(rshift(op, 29), 3) + local immhi = lshift(arshift(lshift(op, 8), 13), 2) + return bor(immhi, immlo) + elseif name == "tbz" or name == "tbnz" then + return lshift(arshift(lshift(op, 13), 18), 2) + else + return lshift(arshift(lshift(op, 8), 13), 2) + end +end + +local function parse_fpimm8(op) + local sign = band(op, 0x100000) == 0 and 1 or -1 + local exp = bxor(rshift(arshift(lshift(op, 12), 5), 24), 0x80) - 131 + local frac = 16+band(rshift(op, 13), 15) + return sign * frac * 2^exp +end + +local function prefer_bfx(sf, uns, imms, immr) + if imms < immr or imms == 31 or imms == 63 then + return false + end + if immr == 0 then + if sf == 0 and (imms == 7 or imms == 15) then + return false + end + if sf ~= 0 and uns == 0 and (imms == 7 or imms == 15 or imms == 31) then + return false + end + end + return true +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) + local operands = {} + local suffix = "" + local last, name, pat + local map_reg + ctx.op = op + ctx.rel = nil + last = nil + local opat + opat = map_init[band(rshift(op, 25), 15)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + name, pat = match(opat, "^([a-z0-9]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.|]*)(.*)") + if altname then pat = pat2 end + if sub(pat, 1, 1) == "." then + local s2, p2 = match(pat, "^([a-z0-9.]*)(.*)") + suffix = suffix..s2 + pat = p2 + end + + local rt = match(pat, "[gf]") + if rt then + if rt == "g" then + map_reg = band(op, 0x80000000) ~= 0 and map_regs.x or map_regs.w + else + map_reg = band(op, 0x400000) ~= 0 and map_regs.d or map_regs.s + end + end + + local second0, immr + + for p in gmatch(pat, ".") do + local x = nil + if p == "D" then + local regnum = band(op, 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "N" then + local regnum = band(rshift(op, 5), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "M" then + local regnum = band(rshift(op, 16), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "A" then + local regnum = band(rshift(op, 10), 31) + x = rt and map_reg[regnum] or match_reg(p, pat, regnum) + elseif p == "B" then + local addr = ctx.addr + pos + parse_immpc(op, name) + ctx.rel = addr + x = "0x"..tohex(addr) + elseif p == "T" then + x = bor(band(rshift(op, 26), 32), band(rshift(op, 19), 31)) + elseif p == "V" then + x = band(op, 15) + elseif p == "C" then + x = map_cond[band(rshift(op, 12), 15)] + elseif p == "c" then + local rn = band(rshift(op, 5), 31) + local rm = band(rshift(op, 16), 31) + local cond = band(rshift(op, 12), 15) + local invc = bxor(cond, 1) + x = map_cond[cond] + if altname and cond ~= 14 and cond ~= 15 then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if rn == rm then + local n = #operands + operands[n] = nil + x = map_cond[invc] + if rn ~= 31 then + if a1 then name = a1 else name = altname end + else + operands[n-1] = nil + name = a2 + end + end + end + elseif p == "W" then + x = band(rshift(op, 5), 0xffff) + elseif p == "Y" then + x = band(rshift(op, 5), 0xffff) + local hw = band(rshift(op, 21), 3) + if altname and (hw == 0 or x ~= 0) then + name = altname + end + elseif p == "L" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local imm9 = arshift(lshift(op, 11), 23) + if band(op, 0x800) ~= 0 then + x = "["..rn..", #"..imm9.."]!" + else + x = "["..rn.."], #"..imm9 + end + elseif p == "U" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local sz = band(rshift(op, 30), 3) + local imm12 = lshift(arshift(lshift(op, 10), 20), sz) + if imm12 ~= 0 then + x = "["..rn..", #"..imm12.."]" + else + x = "["..rn.."]" + end + elseif p == "K" then + local rn = map_regs.x[band(rshift(op, 5), 31)] + local imm9 = arshift(lshift(op, 11), 23) + if imm9 ~= 0 then + x = "["..rn..", #"..imm9.."]" + else + x = "["..rn.."]" + end + elseif p == "O" then + local rn, rm = map_regs.x[band(rshift(op, 5), 31)] + local m = band(rshift(op, 13), 1) + if m == 0 then + rm = map_regs.w[band(rshift(op, 16), 31)] + else + rm = map_regs.x[band(rshift(op, 16), 31)] + end + x = "["..rn..", "..rm + local opt = band(rshift(op, 13), 7) + local s = band(rshift(op, 12), 1) + local sz = band(rshift(op, 30), 3) + -- extension to be applied + if opt == 3 then + if s == 0 then x = x.."]" + else x = x..", lsl #"..sz.."]" end + elseif opt == 2 or opt == 6 or opt == 7 then + if s == 0 then x = x..", "..map_extend[opt].."]" + else x = x..", "..map_extend[opt].." #"..sz.."]" end + else + x = x.."]" + end + elseif p == "P" then + local opcv, sh = rshift(op, 26), 2 + if opcv >= 0x2a then sh = 4 elseif opcv >= 0x1b then sh = 3 end + local imm7 = lshift(arshift(lshift(op, 10), 25), sh) + local rn = map_regs.x[band(rshift(op, 5), 31)] + local ind = band(rshift(op, 23), 3) + if ind == 1 then + x = "["..rn.."], #"..imm7 + elseif ind == 2 then + if imm7 == 0 then + x = "["..rn.."]" + else + x = "["..rn..", #"..imm7.."]" + end + elseif ind == 3 then + x = "["..rn..", #"..imm7.."]!" + end + elseif p == "I" then + local shf = band(rshift(op, 22), 3) + local imm12 = band(rshift(op, 10), 0x0fff) + local rn, rd = band(rshift(op, 5), 31), band(op, 31) + if altname == "mov" and shf == 0 and imm12 == 0 and (rn == 31 or rd == 31) then + name = altname + x = nil + elseif shf == 0 then + x = imm12 + elseif shf == 1 then + x = imm12..", lsl #12" + end + elseif p == "i" then + x = "#0x"..decode_imm13(op) + elseif p == "1" then + immr = band(rshift(op, 16), 63) + x = immr + elseif p == "2" then + x = band(rshift(op, 10), 63) + if altname then + local a1, a2, a3, a4, a5, a6 = + match(altname, "([^|]*)|([^|]*)|([^|]*)|([^|]*)|([^|]*)|(.*)") + local sf = band(rshift(op, 26), 32) + local uns = band(rshift(op, 30), 1) + if prefer_bfx(sf, uns, x, immr) then + name = a2 + x = x - immr + 1 + elseif immr == 0 and x == 7 then + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + name = a6 + x = nil + elseif immr == 0 and x == 15 then + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + name = a5 + x = nil + elseif x == 31 or x == 63 then + if x == 31 and immr == 0 and name == "sbfm" then + name = a4 + local n = #operands + operands[n] = nil + if sf ~= 0 then + operands[n-1] = gsub(operands[n-1], "x", "w") + end + last = operands[n-1] + else + name = a3 + end + x = nil + elseif band(x, 31) ~= 31 and immr == x+1 and name == "ubfm" then + name = a4 + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = nil + elseif x < immr then + name = a1 + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = x + 1 + end + end + elseif p == "3" then + x = band(rshift(op, 10), 63) + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if x < immr then + name = a1 + local sf = band(rshift(op, 26), 32) + last = "#"..(sf+32 - immr) + operands[#operands] = last + x = x + 1 + else + name = a2 + x = x - immr + 1 + end + end + elseif p == "4" then + x = band(rshift(op, 10), 63) + local rn = band(rshift(op, 5), 31) + local rm = band(rshift(op, 16), 31) + if altname and rn == rm then + local n = #operands + operands[n] = nil + last = operands[n-1] + name = altname + end + elseif p == "5" then + x = band(rshift(op, 16), 31) + elseif p == "S" then + x = band(rshift(op, 10), 63) + if x == 0 then x = nil + else x = map_shift[band(rshift(op, 22), 3)].." #"..x end + elseif p == "X" then + local opt = band(rshift(op, 13), 7) + -- Width specifier . + if opt ~= 3 and opt ~= 7 then + last = map_regs.w[band(rshift(op, 16), 31)] + operands[#operands] = last + end + x = band(rshift(op, 10), 7) + -- Extension. + if opt == 2 + band(rshift(op, 31), 1) and + band(rshift(op, second0 and 5 or 0), 31) == 31 then + if x == 0 then x = nil + else x = "lsl #"..x end + else + if x == 0 then x = map_extend[band(rshift(op, 13), 7)] + else x = map_extend[band(rshift(op, 13), 7)].." #"..x end + end + elseif p == "R" then + x = band(rshift(op,21), 3) + if x == 0 then x = nil + else x = "lsl #"..x*16 end + elseif p == "z" then + local n = #operands + if operands[n] == "sp" then operands[n] = "xzr" + elseif operands[n] == "wsp" then operands[n] = "wzr" + end + elseif p == "Z" then + x = 0 + elseif p == "F" then + x = parse_fpimm8(op) + elseif p == "g" or p == "f" or p == "x" or p == "w" or + p == "d" or p == "s" then + -- These are handled in D/N/M/A. + elseif p == "0" then + if last == "sp" or last == "wsp" then + local n = #operands + operands[n] = nil + last = operands[n-1] + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if not a1 then + name = altname + elseif second0 then + name, altname = a2, a1 + else + name, altname = a1, a2 + end + end + end + second0 = true + else + assert(false) + end + if x then + last = x + if type(x) == "number" then x = "#"..x end + operands[#operands+1] = x + end + end + + return putop(ctx, name..suffix, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ctx.pos = ofs + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_regs.x[r] end + return map_regs.d[r-32] +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm64be.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm64be.lua new file mode 100644 index 00000000..edcbffa8 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_arm64be.lua @@ -0,0 +1,12 @@ +---------------------------------------------------------------------------- +-- LuaJIT ARM64BE disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- ARM64 instructions are always little-endian. So just forward to the +-- common ARM64 disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +return require((string.match(..., ".*%.") or "").."dis_arm64") + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips.lua new file mode 100644 index 00000000..6ad17f54 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips.lua @@ -0,0 +1,694 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT/X license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles all standard MIPS32R1/R2 instructions. +-- Default mode is big-endian, but see: dis_mipsel.lua +------------------------------------------------------------------------------ + +local type = type +local byte, format = string.byte, string.format +local match, gmatch = string.match, string.gmatch +local concat = table.concat +local bit = require("bit") +local band, bor, tohex = bit.band, bit.bor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Extended opcode maps common to all MIPS releases +------------------------------------------------------------------------------ + +local map_srl = { shift = 21, mask = 1, [0] = "srlDTA", "rotrDTA", } +local map_srlv = { shift = 6, mask = 1, [0] = "srlvDTS", "rotrvDTS", } + +local map_cop0 = { + shift = 25, mask = 1, + [0] = { + shift = 21, mask = 15, + [0] = "mfc0TDW", [4] = "mtc0TDW", + [10] = "rdpgprDT", + [11] = { shift = 5, mask = 1, [0] = "diT0", "eiT0", }, + [14] = "wrpgprDT", + }, { + shift = 0, mask = 63, + [1] = "tlbr", [2] = "tlbwi", [6] = "tlbwr", [8] = "tlbp", + [24] = "eret", [31] = "deret", + [32] = "wait", + }, +} + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps for MIPS R1-R5 +------------------------------------------------------------------------------ + +local map_movci = { shift = 16, mask = 1, [0] = "movfDSC", "movtDSC", } + +local map_special = { + shift = 0, mask = 63, + [0] = { shift = 0, mask = -1, [0] = "nop", _ = "sllDTA" }, + map_movci, map_srl, "sraDTA", + "sllvDTS", false, map_srlv, "sravDTS", + "jrS", "jalrD1S", "movzDST", "movnDST", + "syscallY", "breakY", false, "sync", + "mfhiD", "mthiS", "mfloD", "mtloS", + "dsllvDST", false, "dsrlvDST", "dsravDST", + "multST", "multuST", "divST", "divuST", + "dmultST", "dmultuST", "ddivST", "ddivuST", + "addDST", "addu|moveDST0", "subDST", "subu|neguDS0T", + "andDST", "or|moveDST0", "xorDST", "nor|notDST0", + false, false, "sltDST", "sltuDST", + "daddDST", "dadduDST", "dsubDST", "dsubuDST", + "tgeSTZ", "tgeuSTZ", "tltSTZ", "tltuSTZ", + "teqSTZ", false, "tneSTZ", false, + "dsllDTA", false, "dsrlDTA", "dsraDTA", + "dsll32DTA", false, "dsrl32DTA", "dsra32DTA", +} + +local map_special2 = { + shift = 0, mask = 63, + [0] = "maddST", "madduST", "mulDST", false, + "msubST", "msubuST", + [32] = "clzDS", [33] = "cloDS", + [63] = "sdbbpY", +} + +local map_bshfl = { + shift = 6, mask = 31, + [2] = "wsbhDT", + [16] = "sebDT", + [24] = "sehDT", +} + +local map_dbshfl = { + shift = 6, mask = 31, + [2] = "dsbhDT", + [5] = "dshdDT", +} + +local map_special3 = { + shift = 0, mask = 63, + [0] = "extTSAK", [1] = "dextmTSAP", [3] = "dextTSAK", + [4] = "insTSAL", [6] = "dinsuTSEQ", [7] = "dinsTSAL", + [32] = map_bshfl, [36] = map_dbshfl, [59] = "rdhwrTD", +} + +local map_regimm = { + shift = 16, mask = 31, + [0] = "bltzSB", "bgezSB", "bltzlSB", "bgezlSB", + false, false, false, false, + "tgeiSI", "tgeiuSI", "tltiSI", "tltiuSI", + "teqiSI", false, "tneiSI", false, + "bltzalSB", "bgezalSB", "bltzallSB", "bgezallSB", + false, false, false, false, + false, false, false, false, + false, false, false, "synciSO", +} + +local map_cop1s = { + shift = 0, mask = 63, + [0] = "add.sFGH", "sub.sFGH", "mul.sFGH", "div.sFGH", + "sqrt.sFG", "abs.sFG", "mov.sFG", "neg.sFG", + "round.l.sFG", "trunc.l.sFG", "ceil.l.sFG", "floor.l.sFG", + "round.w.sFG", "trunc.w.sFG", "ceil.w.sFG", "floor.w.sFG", + false, + { shift = 16, mask = 1, [0] = "movf.sFGC", "movt.sFGC" }, + "movz.sFGT", "movn.sFGT", + false, "recip.sFG", "rsqrt.sFG", false, + false, false, false, false, + false, false, false, false, + false, "cvt.d.sFG", false, false, + "cvt.w.sFG", "cvt.l.sFG", "cvt.ps.sFGH", false, + false, false, false, false, + false, false, false, false, + "c.f.sVGH", "c.un.sVGH", "c.eq.sVGH", "c.ueq.sVGH", + "c.olt.sVGH", "c.ult.sVGH", "c.ole.sVGH", "c.ule.sVGH", + "c.sf.sVGH", "c.ngle.sVGH", "c.seq.sVGH", "c.ngl.sVGH", + "c.lt.sVGH", "c.nge.sVGH", "c.le.sVGH", "c.ngt.sVGH", +} + +local map_cop1d = { + shift = 0, mask = 63, + [0] = "add.dFGH", "sub.dFGH", "mul.dFGH", "div.dFGH", + "sqrt.dFG", "abs.dFG", "mov.dFG", "neg.dFG", + "round.l.dFG", "trunc.l.dFG", "ceil.l.dFG", "floor.l.dFG", + "round.w.dFG", "trunc.w.dFG", "ceil.w.dFG", "floor.w.dFG", + false, + { shift = 16, mask = 1, [0] = "movf.dFGC", "movt.dFGC" }, + "movz.dFGT", "movn.dFGT", + false, "recip.dFG", "rsqrt.dFG", false, + false, false, false, false, + false, false, false, false, + "cvt.s.dFG", false, false, false, + "cvt.w.dFG", "cvt.l.dFG", false, false, + false, false, false, false, + false, false, false, false, + "c.f.dVGH", "c.un.dVGH", "c.eq.dVGH", "c.ueq.dVGH", + "c.olt.dVGH", "c.ult.dVGH", "c.ole.dVGH", "c.ule.dVGH", + "c.df.dVGH", "c.ngle.dVGH", "c.deq.dVGH", "c.ngl.dVGH", + "c.lt.dVGH", "c.nge.dVGH", "c.le.dVGH", "c.ngt.dVGH", +} + +local map_cop1ps = { + shift = 0, mask = 63, + [0] = "add.psFGH", "sub.psFGH", "mul.psFGH", false, + false, "abs.psFG", "mov.psFG", "neg.psFG", + false, false, false, false, + false, false, false, false, + false, + { shift = 16, mask = 1, [0] = "movf.psFGC", "movt.psFGC" }, + "movz.psFGT", "movn.psFGT", + false, false, false, false, + false, false, false, false, + false, false, false, false, + "cvt.s.puFG", false, false, false, + false, false, false, false, + "cvt.s.plFG", false, false, false, + "pll.psFGH", "plu.psFGH", "pul.psFGH", "puu.psFGH", + "c.f.psVGH", "c.un.psVGH", "c.eq.psVGH", "c.ueq.psVGH", + "c.olt.psVGH", "c.ult.psVGH", "c.ole.psVGH", "c.ule.psVGH", + "c.psf.psVGH", "c.ngle.psVGH", "c.pseq.psVGH", "c.ngl.psVGH", + "c.lt.psVGH", "c.nge.psVGH", "c.le.psVGH", "c.ngt.psVGH", +} + +local map_cop1w = { + shift = 0, mask = 63, + [32] = "cvt.s.wFG", [33] = "cvt.d.wFG", +} + +local map_cop1l = { + shift = 0, mask = 63, + [32] = "cvt.s.lFG", [33] = "cvt.d.lFG", +} + +local map_cop1bc = { + shift = 16, mask = 3, + [0] = "bc1fCB", "bc1tCB", "bc1flCB", "bc1tlCB", +} + +local map_cop1 = { + shift = 21, mask = 31, + [0] = "mfc1TG", "dmfc1TG", "cfc1TG", "mfhc1TG", + "mtc1TG", "dmtc1TG", "ctc1TG", "mthc1TG", + map_cop1bc, false, false, false, + false, false, false, false, + map_cop1s, map_cop1d, false, false, + map_cop1w, map_cop1l, map_cop1ps, +} + +local map_cop1x = { + shift = 0, mask = 63, + [0] = "lwxc1FSX", "ldxc1FSX", false, false, + false, "luxc1FSX", false, false, + "swxc1FSX", "sdxc1FSX", false, false, + false, "suxc1FSX", false, "prefxMSX", + false, false, false, false, + false, false, false, false, + false, false, false, false, + false, false, "alnv.psFGHS", false, + "madd.sFRGH", "madd.dFRGH", false, false, + false, false, "madd.psFRGH", false, + "msub.sFRGH", "msub.dFRGH", false, false, + false, false, "msub.psFRGH", false, + "nmadd.sFRGH", "nmadd.dFRGH", false, false, + false, false, "nmadd.psFRGH", false, + "nmsub.sFRGH", "nmsub.dFRGH", false, false, + false, false, "nmsub.psFRGH", false, +} + +local map_pri = { + [0] = map_special, map_regimm, "jJ", "jalJ", + "beq|beqz|bST00B", "bne|bnezST0B", "blezSB", "bgtzSB", + "addiTSI", "addiu|liTS0I", "sltiTSI", "sltiuTSI", + "andiTSU", "ori|liTS0U", "xoriTSU", "luiTU", + map_cop0, map_cop1, false, map_cop1x, + "beql|beqzlST0B", "bnel|bnezlST0B", "blezlSB", "bgtzlSB", + "daddiTSI", "daddiuTSI", false, false, + map_special2, "jalxJ", false, map_special3, + "lbTSO", "lhTSO", "lwlTSO", "lwTSO", + "lbuTSO", "lhuTSO", "lwrTSO", false, + "sbTSO", "shTSO", "swlTSO", "swTSO", + false, false, "swrTSO", "cacheNSO", + "llTSO", "lwc1HSO", "lwc2TSO", "prefNSO", + false, "ldc1HSO", "ldc2TSO", "ldTSO", + "scTSO", "swc1HSO", "swc2TSO", false, + false, "sdc1HSO", "sdc2TSO", "sdTSO", +} + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps for MIPS R6 +------------------------------------------------------------------------------ + +local map_mul_r6 = { shift = 6, mask = 3, [2] = "mulDST", [3] = "muhDST" } +local map_mulu_r6 = { shift = 6, mask = 3, [2] = "muluDST", [3] = "muhuDST" } +local map_div_r6 = { shift = 6, mask = 3, [2] = "divDST", [3] = "modDST" } +local map_divu_r6 = { shift = 6, mask = 3, [2] = "divuDST", [3] = "moduDST" } +local map_dmul_r6 = { shift = 6, mask = 3, [2] = "dmulDST", [3] = "dmuhDST" } +local map_dmulu_r6 = { shift = 6, mask = 3, [2] = "dmuluDST", [3] = "dmuhuDST" } +local map_ddiv_r6 = { shift = 6, mask = 3, [2] = "ddivDST", [3] = "dmodDST" } +local map_ddivu_r6 = { shift = 6, mask = 3, [2] = "ddivuDST", [3] = "dmoduDST" } + +local map_special_r6 = { + shift = 0, mask = 63, + [0] = { shift = 0, mask = -1, [0] = "nop", _ = "sllDTA" }, + false, map_srl, "sraDTA", + "sllvDTS", false, map_srlv, "sravDTS", + "jrS", "jalrD1S", false, false, + "syscallY", "breakY", false, "sync", + "clzDS", "cloDS", "dclzDS", "dcloDS", + "dsllvDST", "dlsaDSTA", "dsrlvDST", "dsravDST", + map_mul_r6, map_mulu_r6, map_div_r6, map_divu_r6, + map_dmul_r6, map_dmulu_r6, map_ddiv_r6, map_ddivu_r6, + "addDST", "addu|moveDST0", "subDST", "subu|neguDS0T", + "andDST", "or|moveDST0", "xorDST", "nor|notDST0", + false, false, "sltDST", "sltuDST", + "daddDST", "dadduDST", "dsubDST", "dsubuDST", + "tgeSTZ", "tgeuSTZ", "tltSTZ", "tltuSTZ", + "teqSTZ", "seleqzDST", "tneSTZ", "selnezDST", + "dsllDTA", false, "dsrlDTA", "dsraDTA", + "dsll32DTA", false, "dsrl32DTA", "dsra32DTA", +} + +local map_bshfl_r6 = { + shift = 9, mask = 3, + [1] = "alignDSTa", + _ = { + shift = 6, mask = 31, + [0] = "bitswapDT", + [2] = "wsbhDT", + [16] = "sebDT", + [24] = "sehDT", + } +} + +local map_dbshfl_r6 = { + shift = 9, mask = 3, + [1] = "dalignDSTa", + _ = { + shift = 6, mask = 31, + [0] = "dbitswapDT", + [2] = "dsbhDT", + [5] = "dshdDT", + } +} + +local map_special3_r6 = { + shift = 0, mask = 63, + [0] = "extTSAK", [1] = "dextmTSAP", [3] = "dextTSAK", + [4] = "insTSAL", [6] = "dinsuTSEQ", [7] = "dinsTSAL", + [32] = map_bshfl_r6, [36] = map_dbshfl_r6, [59] = "rdhwrTD", +} + +local map_regimm_r6 = { + shift = 16, mask = 31, + [0] = "bltzSB", [1] = "bgezSB", + [6] = "dahiSI", [30] = "datiSI", + [23] = "sigrieI", [31] = "synciSO", +} + +local map_pcrel_r6 = { + shift = 19, mask = 3, + [0] = "addiupcS2", "lwpcS2", "lwupcS2", { + shift = 18, mask = 1, + [0] = "ldpcS3", { shift = 16, mask = 3, [2] = "auipcSI", [3] = "aluipcSI" } + } +} + +local map_cop1s_r6 = { + shift = 0, mask = 63, + [0] = "add.sFGH", "sub.sFGH", "mul.sFGH", "div.sFGH", + "sqrt.sFG", "abs.sFG", "mov.sFG", "neg.sFG", + "round.l.sFG", "trunc.l.sFG", "ceil.l.sFG", "floor.l.sFG", + "round.w.sFG", "trunc.w.sFG", "ceil.w.sFG", "floor.w.sFG", + "sel.sFGH", false, false, false, + "seleqz.sFGH", "recip.sFG", "rsqrt.sFG", "selnez.sFGH", + "maddf.sFGH", "msubf.sFGH", "rint.sFG", "class.sFG", + "min.sFGH", "mina.sFGH", "max.sFGH", "maxa.sFGH", + false, "cvt.d.sFG", false, false, + "cvt.w.sFG", "cvt.l.sFG", +} + +local map_cop1d_r6 = { + shift = 0, mask = 63, + [0] = "add.dFGH", "sub.dFGH", "mul.dFGH", "div.dFGH", + "sqrt.dFG", "abs.dFG", "mov.dFG", "neg.dFG", + "round.l.dFG", "trunc.l.dFG", "ceil.l.dFG", "floor.l.dFG", + "round.w.dFG", "trunc.w.dFG", "ceil.w.dFG", "floor.w.dFG", + "sel.dFGH", false, false, false, + "seleqz.dFGH", "recip.dFG", "rsqrt.dFG", "selnez.dFGH", + "maddf.dFGH", "msubf.dFGH", "rint.dFG", "class.dFG", + "min.dFGH", "mina.dFGH", "max.dFGH", "maxa.dFGH", + "cvt.s.dFG", false, false, false, + "cvt.w.dFG", "cvt.l.dFG", +} + +local map_cop1w_r6 = { + shift = 0, mask = 63, + [0] = "cmp.af.sFGH", "cmp.un.sFGH", "cmp.eq.sFGH", "cmp.ueq.sFGH", + "cmp.lt.sFGH", "cmp.ult.sFGH", "cmp.le.sFGH", "cmp.ule.sFGH", + "cmp.saf.sFGH", "cmp.sun.sFGH", "cmp.seq.sFGH", "cmp.sueq.sFGH", + "cmp.slt.sFGH", "cmp.sult.sFGH", "cmp.sle.sFGH", "cmp.sule.sFGH", + false, "cmp.or.sFGH", "cmp.une.sFGH", "cmp.ne.sFGH", + false, false, false, false, + false, "cmp.sor.sFGH", "cmp.sune.sFGH", "cmp.sne.sFGH", + false, false, false, false, + "cvt.s.wFG", "cvt.d.wFG", +} + +local map_cop1l_r6 = { + shift = 0, mask = 63, + [0] = "cmp.af.dFGH", "cmp.un.dFGH", "cmp.eq.dFGH", "cmp.ueq.dFGH", + "cmp.lt.dFGH", "cmp.ult.dFGH", "cmp.le.dFGH", "cmp.ule.dFGH", + "cmp.saf.dFGH", "cmp.sun.dFGH", "cmp.seq.dFGH", "cmp.sueq.dFGH", + "cmp.slt.dFGH", "cmp.sult.dFGH", "cmp.sle.dFGH", "cmp.sule.dFGH", + false, "cmp.or.dFGH", "cmp.une.dFGH", "cmp.ne.dFGH", + false, false, false, false, + false, "cmp.sor.dFGH", "cmp.sune.dFGH", "cmp.sne.dFGH", + false, false, false, false, + "cvt.s.lFG", "cvt.d.lFG", +} + +local map_cop1_r6 = { + shift = 21, mask = 31, + [0] = "mfc1TG", "dmfc1TG", "cfc1TG", "mfhc1TG", + "mtc1TG", "dmtc1TG", "ctc1TG", "mthc1TG", + false, "bc1eqzHB", false, false, + false, "bc1nezHB", false, false, + map_cop1s_r6, map_cop1d_r6, false, false, + map_cop1w_r6, map_cop1l_r6, +} + +local function maprs_popTS(rs, rt) + if rt == 0 then return 0 elseif rs == 0 then return 1 + elseif rs == rt then return 2 else return 3 end +end + +local map_pop06_r6 = { + maprs = maprs_popTS, [0] = "blezSB", "blezalcTB", "bgezalcTB", "bgeucSTB" +} +local map_pop07_r6 = { + maprs = maprs_popTS, [0] = "bgtzSB", "bgtzalcTB", "bltzalcTB", "bltucSTB" +} +local map_pop26_r6 = { + maprs = maprs_popTS, "blezcTB", "bgezcTB", "bgecSTB" +} +local map_pop27_r6 = { + maprs = maprs_popTS, "bgtzcTB", "bltzcTB", "bltcSTB" +} + +local function maprs_popS(rs, rt) + if rs == 0 then return 0 else return 1 end +end + +local map_pop66_r6 = { + maprs = maprs_popS, [0] = "jicTI", "beqzcSb" +} +local map_pop76_r6 = { + maprs = maprs_popS, [0] = "jialcTI", "bnezcSb" +} + +local function maprs_popST(rs, rt) + if rs >= rt then return 0 elseif rs == 0 then return 1 else return 2 end +end + +local map_pop10_r6 = { + maprs = maprs_popST, [0] = "bovcSTB", "beqzalcTB", "beqcSTB" +} +local map_pop30_r6 = { + maprs = maprs_popST, [0] = "bnvcSTB", "bnezalcTB", "bnecSTB" +} + +local map_pri_r6 = { + [0] = map_special_r6, map_regimm_r6, "jJ", "jalJ", + "beq|beqz|bST00B", "bne|bnezST0B", map_pop06_r6, map_pop07_r6, + map_pop10_r6, "addiu|liTS0I", "sltiTSI", "sltiuTSI", + "andiTSU", "ori|liTS0U", "xoriTSU", "aui|luiTS0U", + map_cop0, map_cop1_r6, false, false, + false, false, map_pop26_r6, map_pop27_r6, + map_pop30_r6, "daddiuTSI", false, false, + false, "dauiTSI", false, map_special3_r6, + "lbTSO", "lhTSO", false, "lwTSO", + "lbuTSO", "lhuTSO", false, false, + "sbTSO", "shTSO", false, "swTSO", + false, false, false, false, + false, "lwc1HSO", "bc#", false, + false, "ldc1HSO", map_pop66_r6, "ldTSO", + false, "swc1HSO", "balc#", map_pcrel_r6, + false, "sdc1HSO", map_pop76_r6, "sdTSO", +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15", + "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23", + "r24", "r25", "r26", "r27", "r28", "sp", "r30", "ra", +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then extra = "\t->"..sym end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-7s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-7s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +local function get_be(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + return bor(lshift(b0, 24), lshift(b1, 16), lshift(b2, 8), b3) +end + +local function get_le(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + return bor(lshift(b3, 24), lshift(b2, 16), lshift(b1, 8), b0) +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local op = ctx:get() + local operands = {} + local last = nil + ctx.op = op + ctx.rel = nil + + local opat = ctx.map_pri[rshift(op, 26)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + if opat.maprs then + opat = opat[opat.maprs(band(rshift(op,21),31), band(rshift(op,16),31))] + else + opat = opat[band(rshift(op, opat.shift), opat.mask)] or opat._ + end + end + local name, pat = match(opat, "^([a-z0-9_.]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.|]*)(.*)") + if altname then pat = pat2 end + + for p in gmatch(pat, ".") do + local x = nil + if p == "S" then + x = map_gpr[band(rshift(op, 21), 31)] + elseif p == "T" then + x = map_gpr[band(rshift(op, 16), 31)] + elseif p == "D" then + x = map_gpr[band(rshift(op, 11), 31)] + elseif p == "F" then + x = "f"..band(rshift(op, 6), 31) + elseif p == "G" then + x = "f"..band(rshift(op, 11), 31) + elseif p == "H" then + x = "f"..band(rshift(op, 16), 31) + elseif p == "R" then + x = "f"..band(rshift(op, 21), 31) + elseif p == "A" then + x = band(rshift(op, 6), 31) + elseif p == "a" then + x = band(rshift(op, 6), 7) + elseif p == "E" then + x = band(rshift(op, 6), 31) + 32 + elseif p == "M" then + x = band(rshift(op, 11), 31) + elseif p == "N" then + x = band(rshift(op, 16), 31) + elseif p == "C" then + x = band(rshift(op, 18), 7) + if x == 0 then x = nil end + elseif p == "K" then + x = band(rshift(op, 11), 31) + 1 + elseif p == "P" then + x = band(rshift(op, 11), 31) + 33 + elseif p == "L" then + x = band(rshift(op, 11), 31) - last + 1 + elseif p == "Q" then + x = band(rshift(op, 11), 31) - last + 33 + elseif p == "I" then + x = arshift(lshift(op, 16), 16) + elseif p == "2" then + x = arshift(lshift(op, 13), 11) + elseif p == "3" then + x = arshift(lshift(op, 14), 11) + elseif p == "U" then + x = band(op, 0xffff) + elseif p == "O" then + local disp = arshift(lshift(op, 16), 16) + operands[#operands] = format("%d(%s)", disp, last) + elseif p == "X" then + local index = map_gpr[band(rshift(op, 16), 31)] + operands[#operands] = format("%s(%s)", index, last) + elseif p == "B" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 16), 14) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "b" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 11), 9) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "#" then + x = ctx.addr + ctx.pos + arshift(lshift(op, 6), 4) + 4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "J" then + local a = ctx.addr + ctx.pos + x = a - band(a, 0x0fffffff) + band(op, 0x03ffffff)*4 + ctx.rel = x + x = format("0x%08x", x) + elseif p == "V" then + x = band(rshift(op, 8), 7) + if x == 0 then x = nil end + elseif p == "W" then + x = band(op, 7) + if x == 0 then x = nil end + elseif p == "Y" then + x = band(rshift(op, 6), 0x000fffff) + if x == 0 then x = nil end + elseif p == "Z" then + x = band(rshift(op, 6), 1023) + if x == 0 then x = nil end + elseif p == "0" then + if last == "r0" or last == 0 then + local n = #operands + operands[n] = nil + last = operands[n-1] + if altname then + local a1, a2 = match(altname, "([^|]*)|(.*)") + if a1 then name, altname = a1, a2 + else name = altname end + end + end + elseif p == "1" then + if last == "ra" then + operands[#operands] = nil + end + else + assert(false) + end + if x then operands[#operands+1] = x; last = x end + end + + return putop(ctx, name, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + stop = stop - stop % 4 + ctx.pos = ofs - ofs % 4 + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + ctx.get = get_be + ctx.map_pri = map_pri + return ctx +end + +local function create_el(code, addr, out) + local ctx = create(code, addr, out) + ctx.get = get_le + return ctx +end + +local function create_r6(code, addr, out) + local ctx = create(code, addr, out) + ctx.map_pri = map_pri_r6 + return ctx +end + +local function create_r6_el(code, addr, out) + local ctx = create(code, addr, out) + ctx.get = get_le + ctx.map_pri = map_pri_r6 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +local function disass_el(code, addr, out) + create_el(code, addr, out):disass() +end + +local function disass_r6(code, addr, out) + create_r6(code, addr, out):disass() +end + +local function disass_r6_el(code, addr, out) + create_r6_el(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_gpr[r] end + return "f"..(r-32) +end + +-- Public module functions. +return { + create = create, + create_el = create_el, + create_r6 = create_r6, + create_r6_el = create_r6_el, + disass = disass, + disass_el = disass_el, + disass_r6 = disass_r6, + disass_r6_el = disass_r6_el, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64.lua new file mode 100644 index 00000000..5ad48f8f --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the big-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create, + disass = dis_mips.disass, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64el.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64el.lua new file mode 100644 index 00000000..d50e3a18 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64el.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64EL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_el, + disass = dis_mips.disass_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64r6.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64r6.lua new file mode 100644 index 00000000..921b3cbe --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64r6.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64R6 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the r6 big-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_r6, + disass = dis_mips.disass_r6, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64r6el.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64r6el.lua new file mode 100644 index 00000000..aadef9f3 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mips64r6el.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPS64R6EL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the r6 little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_r6_el, + disass = dis_mips.disass_r6_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mipsel.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mipsel.lua new file mode 100644 index 00000000..52cebefb --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_mipsel.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT MIPSEL disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the little-endian functions from the +-- MIPS disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_mips = require((string.match(..., ".*%.") or "").."dis_mips") +return { + create = dis_mips.create_el, + disass = dis_mips.disass_el, + regname = dis_mips.regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_ppc.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_ppc.lua new file mode 100644 index 00000000..08d742f1 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_ppc.lua @@ -0,0 +1,591 @@ +---------------------------------------------------------------------------- +-- LuaJIT PPC disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT/X license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- It disassembles all common, non-privileged 32/64 bit PowerPC instructions +-- plus the e500 SPE instructions and some Cell/Xenon extensions. +-- +-- NYI: VMX, VMX128 +------------------------------------------------------------------------------ + +local type = type +local byte, format = string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local concat = table.concat +local bit = require("bit") +local band, bor, tohex = bit.band, bit.bor, bit.tohex +local lshift, rshift, arshift = bit.lshift, bit.rshift, bit.arshift + +------------------------------------------------------------------------------ +-- Primary and extended opcode maps +------------------------------------------------------------------------------ + +local map_crops = { + shift = 1, mask = 1023, + [0] = "mcrfXX", + [33] = "crnor|crnotCCC=", [129] = "crandcCCC", + [193] = "crxor|crclrCCC%", [225] = "crnandCCC", + [257] = "crandCCC", [289] = "creqv|crsetCCC%", + [417] = "crorcCCC", [449] = "cror|crmoveCCC=", + [16] = "b_lrKB", [528] = "b_ctrKB", + [150] = "isync", +} + +local map_rlwinm = setmetatable({ + shift = 0, mask = -1, +}, +{ __index = function(t, x) + local rot = band(rshift(x, 11), 31) + local mb = band(rshift(x, 6), 31) + local me = band(rshift(x, 1), 31) + if mb == 0 and me == 31-rot then + return "slwiRR~A." + elseif me == 31 and mb == 32-rot then + return "srwiRR~-A." + else + return "rlwinmRR~AAA." + end + end +}) + +local map_rld = { + shift = 2, mask = 7, + [0] = "rldiclRR~HM.", "rldicrRR~HM.", "rldicRR~HM.", "rldimiRR~HM.", + { + shift = 1, mask = 1, + [0] = "rldclRR~RM.", "rldcrRR~RM.", + }, +} + +local map_ext = setmetatable({ + shift = 1, mask = 1023, + + [0] = "cmp_YLRR", [32] = "cmpl_YLRR", + [4] = "twARR", [68] = "tdARR", + + [8] = "subfcRRR.", [40] = "subfRRR.", + [104] = "negRR.", [136] = "subfeRRR.", + [200] = "subfzeRR.", [232] = "subfmeRR.", + [520] = "subfcoRRR.", [552] = "subfoRRR.", + [616] = "negoRR.", [648] = "subfeoRRR.", + [712] = "subfzeoRR.", [744] = "subfmeoRR.", + + [9] = "mulhduRRR.", [73] = "mulhdRRR.", [233] = "mulldRRR.", + [457] = "divduRRR.", [489] = "divdRRR.", + [745] = "mulldoRRR.", + [969] = "divduoRRR.", [1001] = "divdoRRR.", + + [10] = "addcRRR.", [138] = "addeRRR.", + [202] = "addzeRR.", [234] = "addmeRR.", [266] = "addRRR.", + [522] = "addcoRRR.", [650] = "addeoRRR.", + [714] = "addzeoRR.", [746] = "addmeoRR.", [778] = "addoRRR.", + + [11] = "mulhwuRRR.", [75] = "mulhwRRR.", [235] = "mullwRRR.", + [459] = "divwuRRR.", [491] = "divwRRR.", + [747] = "mullwoRRR.", + [971] = "divwouRRR.", [1003] = "divwoRRR.", + + [15] = "iselltRRR", [47] = "iselgtRRR", [79] = "iseleqRRR", + + [144] = { shift = 20, mask = 1, [0] = "mtcrfRZ~", "mtocrfRZ~", }, + [19] = { shift = 20, mask = 1, [0] = "mfcrR", "mfocrfRZ", }, + [371] = { shift = 11, mask = 1023, [392] = "mftbR", [424] = "mftbuR", }, + [339] = { + shift = 11, mask = 1023, + [32] = "mferR", [256] = "mflrR", [288] = "mfctrR", [16] = "mfspefscrR", + }, + [467] = { + shift = 11, mask = 1023, + [32] = "mtxerR", [256] = "mtlrR", [288] = "mtctrR", [16] = "mtspefscrR", + }, + + [20] = "lwarxRR0R", [84] = "ldarxRR0R", + + [21] = "ldxRR0R", [53] = "lduxRRR", + [149] = "stdxRR0R", [181] = "stduxRRR", + [341] = "lwaxRR0R", [373] = "lwauxRRR", + + [23] = "lwzxRR0R", [55] = "lwzuxRRR", + [87] = "lbzxRR0R", [119] = "lbzuxRRR", + [151] = "stwxRR0R", [183] = "stwuxRRR", + [215] = "stbxRR0R", [247] = "stbuxRRR", + [279] = "lhzxRR0R", [311] = "lhzuxRRR", + [343] = "lhaxRR0R", [375] = "lhauxRRR", + [407] = "sthxRR0R", [439] = "sthuxRRR", + + [54] = "dcbst-R0R", [86] = "dcbf-R0R", + [150] = "stwcxRR0R.", [214] = "stdcxRR0R.", + [246] = "dcbtst-R0R", [278] = "dcbt-R0R", + [310] = "eciwxRR0R", [438] = "ecowxRR0R", + [470] = "dcbi-RR", + + [598] = { + shift = 21, mask = 3, + [0] = "sync", "lwsync", "ptesync", + }, + [758] = "dcba-RR", + [854] = "eieio", [982] = "icbi-R0R", [1014] = "dcbz-R0R", + + [26] = "cntlzwRR~", [58] = "cntlzdRR~", + [122] = "popcntbRR~", + [154] = "prtywRR~", [186] = "prtydRR~", + + [28] = "andRR~R.", [60] = "andcRR~R.", [124] = "nor|notRR~R=.", + [284] = "eqvRR~R.", [316] = "xorRR~R.", + [412] = "orcRR~R.", [444] = "or|mrRR~R=.", [476] = "nandRR~R.", + [508] = "cmpbRR~R", + + [512] = "mcrxrX", + + [532] = "ldbrxRR0R", [660] = "stdbrxRR0R", + + [533] = "lswxRR0R", [597] = "lswiRR0A", + [661] = "stswxRR0R", [725] = "stswiRR0A", + + [534] = "lwbrxRR0R", [662] = "stwbrxRR0R", + [790] = "lhbrxRR0R", [918] = "sthbrxRR0R", + + [535] = "lfsxFR0R", [567] = "lfsuxFRR", + [599] = "lfdxFR0R", [631] = "lfduxFRR", + [663] = "stfsxFR0R", [695] = "stfsuxFRR", + [727] = "stfdxFR0R", [759] = "stfduxFR0R", + [855] = "lfiwaxFR0R", + [983] = "stfiwxFR0R", + + [24] = "slwRR~R.", + + [27] = "sldRR~R.", [536] = "srwRR~R.", + [792] = "srawRR~R.", [824] = "srawiRR~A.", + + [794] = "sradRR~R.", [826] = "sradiRR~H.", [827] = "sradiRR~H.", + [922] = "extshRR~.", [954] = "extsbRR~.", [986] = "extswRR~.", + + [539] = "srdRR~R.", +}, +{ __index = function(t, x) + if band(x, 31) == 15 then return "iselRRRC" end + end +}) + +local map_ld = { + shift = 0, mask = 3, + [0] = "ldRRE", "lduRRE", "lwaRRE", +} + +local map_std = { + shift = 0, mask = 3, + [0] = "stdRRE", "stduRRE", +} + +local map_fps = { + shift = 5, mask = 1, + { + shift = 1, mask = 15, + [0] = false, false, "fdivsFFF.", false, + "fsubsFFF.", "faddsFFF.", "fsqrtsF-F.", false, + "fresF-F.", "fmulsFF-F.", "frsqrtesF-F.", false, + "fmsubsFFFF~.", "fmaddsFFFF~.", "fnmsubsFFFF~.", "fnmaddsFFFF~.", + } +} + +local map_fpd = { + shift = 5, mask = 1, + [0] = { + shift = 1, mask = 1023, + [0] = "fcmpuXFF", [32] = "fcmpoXFF", [64] = "mcrfsXX", + [38] = "mtfsb1A.", [70] = "mtfsb0A.", [134] = "mtfsfiA>>-A>", + [8] = "fcpsgnFFF.", [40] = "fnegF-F.", [72] = "fmrF-F.", + [136] = "fnabsF-F.", [264] = "fabsF-F.", + [12] = "frspF-F.", + [14] = "fctiwF-F.", [15] = "fctiwzF-F.", + [583] = "mffsF.", [711] = "mtfsfZF.", + [392] = "frinF-F.", [424] = "frizF-F.", + [456] = "fripF-F.", [488] = "frimF-F.", + [814] = "fctidF-F.", [815] = "fctidzF-F.", [846] = "fcfidF-F.", + }, + { + shift = 1, mask = 15, + [0] = false, false, "fdivFFF.", false, + "fsubFFF.", "faddFFF.", "fsqrtF-F.", "fselFFFF~.", + "freF-F.", "fmulFF-F.", "frsqrteF-F.", false, + "fmsubFFFF~.", "fmaddFFFF~.", "fnmsubFFFF~.", "fnmaddFFFF~.", + } +} + +local map_spe = { + shift = 0, mask = 2047, + + [512] = "evaddwRRR", [514] = "evaddiwRAR~", + [516] = "evsubwRRR~", [518] = "evsubiwRAR~", + [520] = "evabsRR", [521] = "evnegRR", + [522] = "evextsbRR", [523] = "evextshRR", [524] = "evrndwRR", + [525] = "evcntlzwRR", [526] = "evcntlswRR", + + [527] = "brincRRR", + + [529] = "evandRRR", [530] = "evandcRRR", [534] = "evxorRRR", + [535] = "evor|evmrRRR=", [536] = "evnor|evnotRRR=", + [537] = "eveqvRRR", [539] = "evorcRRR", [542] = "evnandRRR", + + [544] = "evsrwuRRR", [545] = "evsrwsRRR", + [546] = "evsrwiuRRA", [547] = "evsrwisRRA", + [548] = "evslwRRR", [550] = "evslwiRRA", + [552] = "evrlwRRR", [553] = "evsplatiRS", + [554] = "evrlwiRRA", [555] = "evsplatfiRS", + [556] = "evmergehiRRR", [557] = "evmergeloRRR", + [558] = "evmergehiloRRR", [559] = "evmergelohiRRR", + + [560] = "evcmpgtuYRR", [561] = "evcmpgtsYRR", + [562] = "evcmpltuYRR", [563] = "evcmpltsYRR", + [564] = "evcmpeqYRR", + + [632] = "evselRRR", [633] = "evselRRRW", + [634] = "evselRRRW", [635] = "evselRRRW", + [636] = "evselRRRW", [637] = "evselRRRW", + [638] = "evselRRRW", [639] = "evselRRRW", + + [640] = "evfsaddRRR", [641] = "evfssubRRR", + [644] = "evfsabsRR", [645] = "evfsnabsRR", [646] = "evfsnegRR", + [648] = "evfsmulRRR", [649] = "evfsdivRRR", + [652] = "evfscmpgtYRR", [653] = "evfscmpltYRR", [654] = "evfscmpeqYRR", + [656] = "evfscfuiR-R", [657] = "evfscfsiR-R", + [658] = "evfscfufR-R", [659] = "evfscfsfR-R", + [660] = "evfsctuiR-R", [661] = "evfsctsiR-R", + [662] = "evfsctufR-R", [663] = "evfsctsfR-R", + [664] = "evfsctuizR-R", [666] = "evfsctsizR-R", + [668] = "evfststgtYRR", [669] = "evfststltYRR", [670] = "evfststeqYRR", + + [704] = "efsaddRRR", [705] = "efssubRRR", + [708] = "efsabsRR", [709] = "efsnabsRR", [710] = "efsnegRR", + [712] = "efsmulRRR", [713] = "efsdivRRR", + [716] = "efscmpgtYRR", [717] = "efscmpltYRR", [718] = "efscmpeqYRR", + [719] = "efscfdR-R", + [720] = "efscfuiR-R", [721] = "efscfsiR-R", + [722] = "efscfufR-R", [723] = "efscfsfR-R", + [724] = "efsctuiR-R", [725] = "efsctsiR-R", + [726] = "efsctufR-R", [727] = "efsctsfR-R", + [728] = "efsctuizR-R", [730] = "efsctsizR-R", + [732] = "efststgtYRR", [733] = "efststltYRR", [734] = "efststeqYRR", + + [736] = "efdaddRRR", [737] = "efdsubRRR", + [738] = "efdcfuidR-R", [739] = "efdcfsidR-R", + [740] = "efdabsRR", [741] = "efdnabsRR", [742] = "efdnegRR", + [744] = "efdmulRRR", [745] = "efddivRRR", + [746] = "efdctuidzR-R", [747] = "efdctsidzR-R", + [748] = "efdcmpgtYRR", [749] = "efdcmpltYRR", [750] = "efdcmpeqYRR", + [751] = "efdcfsR-R", + [752] = "efdcfuiR-R", [753] = "efdcfsiR-R", + [754] = "efdcfufR-R", [755] = "efdcfsfR-R", + [756] = "efdctuiR-R", [757] = "efdctsiR-R", + [758] = "efdctufR-R", [759] = "efdctsfR-R", + [760] = "efdctuizR-R", [762] = "efdctsizR-R", + [764] = "efdtstgtYRR", [765] = "efdtstltYRR", [766] = "efdtsteqYRR", + + [768] = "evlddxRR0R", [769] = "evlddRR8", + [770] = "evldwxRR0R", [771] = "evldwRR8", + [772] = "evldhxRR0R", [773] = "evldhRR8", + [776] = "evlhhesplatxRR0R", [777] = "evlhhesplatRR2", + [780] = "evlhhousplatxRR0R", [781] = "evlhhousplatRR2", + [782] = "evlhhossplatxRR0R", [783] = "evlhhossplatRR2", + [784] = "evlwhexRR0R", [785] = "evlwheRR4", + [788] = "evlwhouxRR0R", [789] = "evlwhouRR4", + [790] = "evlwhosxRR0R", [791] = "evlwhosRR4", + [792] = "evlwwsplatxRR0R", [793] = "evlwwsplatRR4", + [796] = "evlwhsplatxRR0R", [797] = "evlwhsplatRR4", + + [800] = "evstddxRR0R", [801] = "evstddRR8", + [802] = "evstdwxRR0R", [803] = "evstdwRR8", + [804] = "evstdhxRR0R", [805] = "evstdhRR8", + [816] = "evstwhexRR0R", [817] = "evstwheRR4", + [820] = "evstwhoxRR0R", [821] = "evstwhoRR4", + [824] = "evstwwexRR0R", [825] = "evstwweRR4", + [828] = "evstwwoxRR0R", [829] = "evstwwoRR4", + + [1027] = "evmhessfRRR", [1031] = "evmhossfRRR", [1032] = "evmheumiRRR", + [1033] = "evmhesmiRRR", [1035] = "evmhesmfRRR", [1036] = "evmhoumiRRR", + [1037] = "evmhosmiRRR", [1039] = "evmhosmfRRR", [1059] = "evmhessfaRRR", + [1063] = "evmhossfaRRR", [1064] = "evmheumiaRRR", [1065] = "evmhesmiaRRR", + [1067] = "evmhesmfaRRR", [1068] = "evmhoumiaRRR", [1069] = "evmhosmiaRRR", + [1071] = "evmhosmfaRRR", [1095] = "evmwhssfRRR", [1096] = "evmwlumiRRR", + [1100] = "evmwhumiRRR", [1101] = "evmwhsmiRRR", [1103] = "evmwhsmfRRR", + [1107] = "evmwssfRRR", [1112] = "evmwumiRRR", [1113] = "evmwsmiRRR", + [1115] = "evmwsmfRRR", [1127] = "evmwhssfaRRR", [1128] = "evmwlumiaRRR", + [1132] = "evmwhumiaRRR", [1133] = "evmwhsmiaRRR", [1135] = "evmwhsmfaRRR", + [1139] = "evmwssfaRRR", [1144] = "evmwumiaRRR", [1145] = "evmwsmiaRRR", + [1147] = "evmwsmfaRRR", + + [1216] = "evaddusiaawRR", [1217] = "evaddssiaawRR", + [1218] = "evsubfusiaawRR", [1219] = "evsubfssiaawRR", + [1220] = "evmraRR", + [1222] = "evdivwsRRR", [1223] = "evdivwuRRR", + [1224] = "evaddumiaawRR", [1225] = "evaddsmiaawRR", + [1226] = "evsubfumiaawRR", [1227] = "evsubfsmiaawRR", + + [1280] = "evmheusiaawRRR", [1281] = "evmhessiaawRRR", + [1283] = "evmhessfaawRRR", [1284] = "evmhousiaawRRR", + [1285] = "evmhossiaawRRR", [1287] = "evmhossfaawRRR", + [1288] = "evmheumiaawRRR", [1289] = "evmhesmiaawRRR", + [1291] = "evmhesmfaawRRR", [1292] = "evmhoumiaawRRR", + [1293] = "evmhosmiaawRRR", [1295] = "evmhosmfaawRRR", + [1320] = "evmhegumiaaRRR", [1321] = "evmhegsmiaaRRR", + [1323] = "evmhegsmfaaRRR", [1324] = "evmhogumiaaRRR", + [1325] = "evmhogsmiaaRRR", [1327] = "evmhogsmfaaRRR", + [1344] = "evmwlusiaawRRR", [1345] = "evmwlssiaawRRR", + [1352] = "evmwlumiaawRRR", [1353] = "evmwlsmiaawRRR", + [1363] = "evmwssfaaRRR", [1368] = "evmwumiaaRRR", + [1369] = "evmwsmiaaRRR", [1371] = "evmwsmfaaRRR", + [1408] = "evmheusianwRRR", [1409] = "evmhessianwRRR", + [1411] = "evmhessfanwRRR", [1412] = "evmhousianwRRR", + [1413] = "evmhossianwRRR", [1415] = "evmhossfanwRRR", + [1416] = "evmheumianwRRR", [1417] = "evmhesmianwRRR", + [1419] = "evmhesmfanwRRR", [1420] = "evmhoumianwRRR", + [1421] = "evmhosmianwRRR", [1423] = "evmhosmfanwRRR", + [1448] = "evmhegumianRRR", [1449] = "evmhegsmianRRR", + [1451] = "evmhegsmfanRRR", [1452] = "evmhogumianRRR", + [1453] = "evmhogsmianRRR", [1455] = "evmhogsmfanRRR", + [1472] = "evmwlusianwRRR", [1473] = "evmwlssianwRRR", + [1480] = "evmwlumianwRRR", [1481] = "evmwlsmianwRRR", + [1491] = "evmwssfanRRR", [1496] = "evmwumianRRR", + [1497] = "evmwsmianRRR", [1499] = "evmwsmfanRRR", +} + +local map_pri = { + [0] = false, false, "tdiARI", "twiARI", + map_spe, false, false, "mulliRRI", + "subficRRI", false, "cmpl_iYLRU", "cmp_iYLRI", + "addicRRI", "addic.RRI", "addi|liRR0I", "addis|lisRR0I", + "b_KBJ", "sc", "bKJ", map_crops, + "rlwimiRR~AAA.", map_rlwinm, false, "rlwnmRR~RAA.", + "oriNRR~U", "orisRR~U", "xoriRR~U", "xorisRR~U", + "andi.RR~U", "andis.RR~U", map_rld, map_ext, + "lwzRRD", "lwzuRRD", "lbzRRD", "lbzuRRD", + "stwRRD", "stwuRRD", "stbRRD", "stbuRRD", + "lhzRRD", "lhzuRRD", "lhaRRD", "lhauRRD", + "sthRRD", "sthuRRD", "lmwRRD", "stmwRRD", + "lfsFRD", "lfsuFRD", "lfdFRD", "lfduFRD", + "stfsFRD", "stfsuFRD", "stfdFRD", "stfduFRD", + false, false, map_ld, map_fps, + false, false, map_std, map_fpd, +} + +------------------------------------------------------------------------------ + +local map_gpr = { + [0] = "r0", "sp", "r2", "r3", "r4", "r5", "r6", "r7", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15", + "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23", + "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31", +} + +local map_cond = { [0] = "lt", "gt", "eq", "so", "ge", "le", "ne", "ns", } + +-- Format a condition bit. +local function condfmt(cond) + if cond <= 3 then + return map_cond[band(cond, 3)] + else + return format("4*cr%d+%s", rshift(cond, 2), map_cond[band(cond, 3)]) + end +end + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local pos = ctx.pos + local extra = "" + if ctx.rel then + local sym = ctx.symtab[ctx.rel] + if sym then extra = "\t->"..sym end + end + if ctx.hexdump > 0 then + ctx.out(format("%08x %s %-7s %s%s\n", + ctx.addr+pos, tohex(ctx.op), text, concat(operands, ", "), extra)) + else + ctx.out(format("%08x %-7s %s%s\n", + ctx.addr+pos, text, concat(operands, ", "), extra)) + end + ctx.pos = pos + 4 +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + return putop(ctx, ".long", { "0x"..tohex(ctx.op) }) +end + +-- Disassemble a single instruction. +local function disass_ins(ctx) + local pos = ctx.pos + local b0, b1, b2, b3 = byte(ctx.code, pos+1, pos+4) + local op = bor(lshift(b0, 24), lshift(b1, 16), lshift(b2, 8), b3) + local operands = {} + local last = nil + local rs = 21 + ctx.op = op + ctx.rel = nil + + local opat = map_pri[rshift(b0, 2)] + while type(opat) ~= "string" do + if not opat then return unknown(ctx) end + opat = opat[band(rshift(op, opat.shift), opat.mask)] + end + local name, pat = match(opat, "^([a-z0-9_.]*)(.*)") + local altname, pat2 = match(pat, "|([a-z0-9_.]*)(.*)") + if altname then pat = pat2 end + + for p in gmatch(pat, ".") do + local x = nil + if p == "R" then + x = map_gpr[band(rshift(op, rs), 31)] + rs = rs - 5 + elseif p == "F" then + x = "f"..band(rshift(op, rs), 31) + rs = rs - 5 + elseif p == "A" then + x = band(rshift(op, rs), 31) + rs = rs - 5 + elseif p == "S" then + x = arshift(lshift(op, 27-rs), 27) + rs = rs - 5 + elseif p == "I" then + x = arshift(lshift(op, 16), 16) + elseif p == "U" then + x = band(op, 0xffff) + elseif p == "D" or p == "E" then + local disp = arshift(lshift(op, 16), 16) + if p == "E" then disp = band(disp, -4) end + if last == "r0" then last = "0" end + operands[#operands] = format("%d(%s)", disp, last) + elseif p >= "2" and p <= "8" then + local disp = band(rshift(op, rs), 31) * p + if last == "r0" then last = "0" end + operands[#operands] = format("%d(%s)", disp, last) + elseif p == "H" then + x = band(rshift(op, rs), 31) + lshift(band(op, 2), 4) + rs = rs - 5 + elseif p == "M" then + x = band(rshift(op, rs), 31) + band(op, 0x20) + elseif p == "C" then + x = condfmt(band(rshift(op, rs), 31)) + rs = rs - 5 + elseif p == "B" then + local bo = rshift(op, 21) + local cond = band(rshift(op, 16), 31) + local cn = "" + rs = rs - 10 + if band(bo, 4) == 0 then + cn = band(bo, 2) == 0 and "dnz" or "dz" + if band(bo, 0x10) == 0 then + cn = cn..(band(bo, 8) == 0 and "f" or "t") + end + if band(bo, 0x10) == 0 then x = condfmt(cond) end + name = name..(band(bo, 1) == band(rshift(op, 15), 1) and "-" or "+") + elseif band(bo, 0x10) == 0 then + cn = map_cond[band(cond, 3) + (band(bo, 8) == 0 and 4 or 0)] + if cond > 3 then x = "cr"..rshift(cond, 2) end + name = name..(band(bo, 1) == band(rshift(op, 15), 1) and "-" or "+") + end + name = gsub(name, "_", cn) + elseif p == "J" then + x = arshift(lshift(op, 27-rs), 29-rs)*4 + if band(op, 2) == 0 then x = ctx.addr + pos + x end + ctx.rel = x + x = "0x"..tohex(x) + elseif p == "K" then + if band(op, 1) ~= 0 then name = name.."l" end + if band(op, 2) ~= 0 then name = name.."a" end + elseif p == "X" or p == "Y" then + x = band(rshift(op, rs+2), 7) + if x == 0 and p == "Y" then x = nil else x = "cr"..x end + rs = rs - 5 + elseif p == "W" then + x = "cr"..band(op, 7) + elseif p == "Z" then + x = band(rshift(op, rs-4), 255) + rs = rs - 10 + elseif p == ">" then + operands[#operands] = rshift(operands[#operands], 1) + elseif p == "0" then + if last == "r0" then + operands[#operands] = nil + if altname then name = altname end + end + elseif p == "L" then + name = gsub(name, "_", band(op, 0x00200000) ~= 0 and "d" or "w") + elseif p == "." then + if band(op, 1) == 1 then name = name.."." end + elseif p == "N" then + if op == 0x60000000 then name = "nop"; break end + elseif p == "~" then + local n = #operands + operands[n-1], operands[n] = operands[n], operands[n-1] + elseif p == "=" then + local n = #operands + if last == operands[n-1] then + operands[n] = nil + name = altname + end + elseif p == "%" then + local n = #operands + if last == operands[n-1] and last == operands[n-2] then + operands[n] = nil + operands[n-1] = nil + name = altname + end + elseif p == "-" then + rs = rs - 5 + else + assert(false) + end + if x then operands[#operands+1] = x; last = x end + end + + return putop(ctx, name, operands) +end + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + stop = stop - stop % 4 + ctx.pos = ofs - ofs % 4 + ctx.rel = nil + while ctx.pos < stop do disass_ins(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = addr or 0 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 8 + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 32 then return map_gpr[r] end + return "f"..(r-32) +end + +-- Public module functions. +return { + create = create, + disass = disass, + regname = regname +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_x64.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_x64.lua new file mode 100644 index 00000000..2d37423e --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_x64.lua @@ -0,0 +1,17 @@ +---------------------------------------------------------------------------- +-- LuaJIT x64 disassembler wrapper module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This module just exports the 64 bit functions from the combined +-- x86/x64 disassembler module. All the interesting stuff is there. +------------------------------------------------------------------------------ + +local dis_x86 = require((string.match(..., ".*%.") or "").."dis_x86") +return { + create = dis_x86.create64, + disass = dis_x86.disass64, + regname = dis_x86.regname64 +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_x86.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_x86.lua new file mode 100644 index 00000000..5480854c --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dis_x86.lua @@ -0,0 +1,953 @@ +---------------------------------------------------------------------------- +-- LuaJIT x86/x64 disassembler module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- This is a helper module used by the LuaJIT machine code dumper module. +-- +-- Sending small code snippets to an external disassembler and mixing the +-- output with our own stuff was too fragile. So I had to bite the bullet +-- and write yet another x86 disassembler. Oh well ... +-- +-- The output format is very similar to what ndisasm generates. But it has +-- been developed independently by looking at the opcode tables from the +-- Intel and AMD manuals. The supported instruction set is quite extensive +-- and reflects what a current generation Intel or AMD CPU implements in +-- 32 bit and 64 bit mode. Yes, this includes MMX, SSE, SSE2, SSE3, SSSE3, +-- SSE4.1, SSE4.2, SSE4a, AVX, AVX2 and even privileged and hypervisor +-- (VMX/SVM) instructions. +-- +-- Notes: +-- * The (useless) a16 prefix, 3DNow and pre-586 opcodes are unsupported. +-- * No attempt at optimization has been made -- it's fast enough for my needs. +------------------------------------------------------------------------------ + +local type = type +local sub, byte, format = string.sub, string.byte, string.format +local match, gmatch, gsub = string.match, string.gmatch, string.gsub +local lower, rep = string.lower, string.rep +local bit = require("bit") +local tohex = bit.tohex + +-- Map for 1st opcode byte in 32 bit mode. Ugly? Well ... read on. +local map_opc1_32 = { +--0x +[0]="addBmr","addVmr","addBrm","addVrm","addBai","addVai","push es","pop es", +"orBmr","orVmr","orBrm","orVrm","orBai","orVai","push cs","opc2*", +--1x +"adcBmr","adcVmr","adcBrm","adcVrm","adcBai","adcVai","push ss","pop ss", +"sbbBmr","sbbVmr","sbbBrm","sbbVrm","sbbBai","sbbVai","push ds","pop ds", +--2x +"andBmr","andVmr","andBrm","andVrm","andBai","andVai","es:seg","daa", +"subBmr","subVmr","subBrm","subVrm","subBai","subVai","cs:seg","das", +--3x +"xorBmr","xorVmr","xorBrm","xorVrm","xorBai","xorVai","ss:seg","aaa", +"cmpBmr","cmpVmr","cmpBrm","cmpVrm","cmpBai","cmpVai","ds:seg","aas", +--4x +"incVR","incVR","incVR","incVR","incVR","incVR","incVR","incVR", +"decVR","decVR","decVR","decVR","decVR","decVR","decVR","decVR", +--5x +"pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR", +"popUR","popUR","popUR","popUR","popUR","popUR","popUR","popUR", +--6x +"sz*pushaw,pusha","sz*popaw,popa","boundVrm","arplWmr", +"fs:seg","gs:seg","o16:","a16", +"pushUi","imulVrmi","pushBs","imulVrms", +"insb","insVS","outsb","outsVS", +--7x +"joBj","jnoBj","jbBj","jnbBj","jzBj","jnzBj","jbeBj","jaBj", +"jsBj","jnsBj","jpeBj","jpoBj","jlBj","jgeBj","jleBj","jgBj", +--8x +"arith!Bmi","arith!Vmi","arith!Bmi","arith!Vms", +"testBmr","testVmr","xchgBrm","xchgVrm", +"movBmr","movVmr","movBrm","movVrm", +"movVmg","leaVrm","movWgm","popUm", +--9x +"nop*xchgVaR|pause|xchgWaR|repne nop","xchgVaR","xchgVaR","xchgVaR", +"xchgVaR","xchgVaR","xchgVaR","xchgVaR", +"sz*cbw,cwde,cdqe","sz*cwd,cdq,cqo","call farViw","wait", +"sz*pushfw,pushf","sz*popfw,popf","sahf","lahf", +--Ax +"movBao","movVao","movBoa","movVoa", +"movsb","movsVS","cmpsb","cmpsVS", +"testBai","testVai","stosb","stosVS", +"lodsb","lodsVS","scasb","scasVS", +--Bx +"movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi", +"movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI", +--Cx +"shift!Bmu","shift!Vmu","retBw","ret","vex*3$lesVrm","vex*2$ldsVrm","movBmi","movVmi", +"enterBwu","leave","retfBw","retf","int3","intBu","into","iretVS", +--Dx +"shift!Bm1","shift!Vm1","shift!Bmc","shift!Vmc","aamBu","aadBu","salc","xlatb", +"fp*0","fp*1","fp*2","fp*3","fp*4","fp*5","fp*6","fp*7", +--Ex +"loopneBj","loopeBj","loopBj","sz*jcxzBj,jecxzBj,jrcxzBj", +"inBau","inVau","outBua","outVua", +"callVj","jmpVj","jmp farViw","jmpBj","inBad","inVad","outBda","outVda", +--Fx +"lock:","int1","repne:rep","rep:","hlt","cmc","testb!Bm","testv!Vm", +"clc","stc","cli","sti","cld","std","incb!Bm","incd!Vm", +} +assert(#map_opc1_32 == 255) + +-- Map for 1st opcode byte in 64 bit mode (overrides only). +local map_opc1_64 = setmetatable({ + [0x06]=false, [0x07]=false, [0x0e]=false, + [0x16]=false, [0x17]=false, [0x1e]=false, [0x1f]=false, + [0x27]=false, [0x2f]=false, [0x37]=false, [0x3f]=false, + [0x60]=false, [0x61]=false, [0x62]=false, [0x63]="movsxdVrDmt", [0x67]="a32:", + [0x40]="rex*", [0x41]="rex*b", [0x42]="rex*x", [0x43]="rex*xb", + [0x44]="rex*r", [0x45]="rex*rb", [0x46]="rex*rx", [0x47]="rex*rxb", + [0x48]="rex*w", [0x49]="rex*wb", [0x4a]="rex*wx", [0x4b]="rex*wxb", + [0x4c]="rex*wr", [0x4d]="rex*wrb", [0x4e]="rex*wrx", [0x4f]="rex*wrxb", + [0x82]=false, [0x9a]=false, [0xc4]="vex*3", [0xc5]="vex*2", [0xce]=false, + [0xd4]=false, [0xd5]=false, [0xd6]=false, [0xea]=false, +}, { __index = map_opc1_32 }) + +-- Map for 2nd opcode byte (0F xx). True CISC hell. Hey, I told you. +-- Prefix dependent MMX/SSE opcodes: (none)|rep|o16|repne, -|F3|66|F2 +local map_opc2 = { +--0x +[0]="sldt!Dmp","sgdt!Ump","larVrm","lslVrm",nil,"syscall","clts","sysret", +"invd","wbinvd",nil,"ud1",nil,"$prefetch!Bm","femms","3dnowMrmu", +--1x +"movupsXrm|movssXrvm|movupdXrm|movsdXrvm", +"movupsXmr|movssXmvr|movupdXmr|movsdXmvr", +"movhlpsXrm$movlpsXrm|movsldupXrm|movlpdXrm|movddupXrm", +"movlpsXmr||movlpdXmr", +"unpcklpsXrvm||unpcklpdXrvm", +"unpckhpsXrvm||unpckhpdXrvm", +"movlhpsXrm$movhpsXrm|movshdupXrm|movhpdXrm", +"movhpsXmr||movhpdXmr", +"$prefetcht!Bm","hintnopVm","hintnopVm","hintnopVm", +"hintnopVm","hintnopVm","hintnopVm","hintnopVm", +--2x +"movUmx$","movUmy$","movUxm$","movUym$","movUmz$",nil,"movUzm$",nil, +"movapsXrm||movapdXrm", +"movapsXmr||movapdXmr", +"cvtpi2psXrMm|cvtsi2ssXrvVmt|cvtpi2pdXrMm|cvtsi2sdXrvVmt", +"movntpsXmr|movntssXmr|movntpdXmr|movntsdXmr", +"cvttps2piMrXm|cvttss2siVrXm|cvttpd2piMrXm|cvttsd2siVrXm", +"cvtps2piMrXm|cvtss2siVrXm|cvtpd2piMrXm|cvtsd2siVrXm", +"ucomissXrm||ucomisdXrm", +"comissXrm||comisdXrm", +--3x +"wrmsr","rdtsc","rdmsr","rdpmc","sysenter","sysexit",nil,"getsec", +"opc3*38",nil,"opc3*3a",nil,nil,nil,nil,nil, +--4x +"cmovoVrm","cmovnoVrm","cmovbVrm","cmovnbVrm", +"cmovzVrm","cmovnzVrm","cmovbeVrm","cmovaVrm", +"cmovsVrm","cmovnsVrm","cmovpeVrm","cmovpoVrm", +"cmovlVrm","cmovgeVrm","cmovleVrm","cmovgVrm", +--5x +"movmskpsVrXm$||movmskpdVrXm$","sqrtpsXrm|sqrtssXrm|sqrtpdXrm|sqrtsdXrm", +"rsqrtpsXrm|rsqrtssXrvm","rcppsXrm|rcpssXrvm", +"andpsXrvm||andpdXrvm","andnpsXrvm||andnpdXrvm", +"orpsXrvm||orpdXrvm","xorpsXrvm||xorpdXrvm", +"addpsXrvm|addssXrvm|addpdXrvm|addsdXrvm","mulpsXrvm|mulssXrvm|mulpdXrvm|mulsdXrvm", +"cvtps2pdXrm|cvtss2sdXrvm|cvtpd2psXrm|cvtsd2ssXrvm", +"cvtdq2psXrm|cvttps2dqXrm|cvtps2dqXrm", +"subpsXrvm|subssXrvm|subpdXrvm|subsdXrvm","minpsXrvm|minssXrvm|minpdXrvm|minsdXrvm", +"divpsXrvm|divssXrvm|divpdXrvm|divsdXrvm","maxpsXrvm|maxssXrvm|maxpdXrvm|maxsdXrvm", +--6x +"punpcklbwPrvm","punpcklwdPrvm","punpckldqPrvm","packsswbPrvm", +"pcmpgtbPrvm","pcmpgtwPrvm","pcmpgtdPrvm","packuswbPrvm", +"punpckhbwPrvm","punpckhwdPrvm","punpckhdqPrvm","packssdwPrvm", +"||punpcklqdqXrvm","||punpckhqdqXrvm", +"movPrVSm","movqMrm|movdquXrm|movdqaXrm", +--7x +"pshufwMrmu|pshufhwXrmu|pshufdXrmu|pshuflwXrmu","pshiftw!Pvmu", +"pshiftd!Pvmu","pshiftq!Mvmu||pshiftdq!Xvmu", +"pcmpeqbPrvm","pcmpeqwPrvm","pcmpeqdPrvm","emms*|", +"vmreadUmr||extrqXmuu$|insertqXrmuu$","vmwriteUrm||extrqXrm$|insertqXrm$", +nil,nil, +"||haddpdXrvm|haddpsXrvm","||hsubpdXrvm|hsubpsXrvm", +"movVSmMr|movqXrm|movVSmXr","movqMmr|movdquXmr|movdqaXmr", +--8x +"joVj","jnoVj","jbVj","jnbVj","jzVj","jnzVj","jbeVj","jaVj", +"jsVj","jnsVj","jpeVj","jpoVj","jlVj","jgeVj","jleVj","jgVj", +--9x +"setoBm","setnoBm","setbBm","setnbBm","setzBm","setnzBm","setbeBm","setaBm", +"setsBm","setnsBm","setpeBm","setpoBm","setlBm","setgeBm","setleBm","setgBm", +--Ax +"push fs","pop fs","cpuid","btVmr","shldVmru","shldVmrc",nil,nil, +"push gs","pop gs","rsm","btsVmr","shrdVmru","shrdVmrc","fxsave!Dmp","imulVrm", +--Bx +"cmpxchgBmr","cmpxchgVmr","$lssVrm","btrVmr", +"$lfsVrm","$lgsVrm","movzxVrBmt","movzxVrWmt", +"|popcntVrm","ud2Dp","bt!Vmu","btcVmr", +"bsfVrm","bsrVrm|lzcntVrm|bsrWrm","movsxVrBmt","movsxVrWmt", +--Cx +"xaddBmr","xaddVmr", +"cmppsXrvmu|cmpssXrvmu|cmppdXrvmu|cmpsdXrvmu","$movntiVmr|", +"pinsrwPrvWmu","pextrwDrPmu", +"shufpsXrvmu||shufpdXrvmu","$cmpxchg!Qmp", +"bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR", +--Dx +"||addsubpdXrvm|addsubpsXrvm","psrlwPrvm","psrldPrvm","psrlqPrvm", +"paddqPrvm","pmullwPrvm", +"|movq2dqXrMm|movqXmr|movdq2qMrXm$","pmovmskbVrMm||pmovmskbVrXm", +"psubusbPrvm","psubuswPrvm","pminubPrvm","pandPrvm", +"paddusbPrvm","padduswPrvm","pmaxubPrvm","pandnPrvm", +--Ex +"pavgbPrvm","psrawPrvm","psradPrvm","pavgwPrvm", +"pmulhuwPrvm","pmulhwPrvm", +"|cvtdq2pdXrm|cvttpd2dqXrm|cvtpd2dqXrm","$movntqMmr||$movntdqXmr", +"psubsbPrvm","psubswPrvm","pminswPrvm","porPrvm", +"paddsbPrvm","paddswPrvm","pmaxswPrvm","pxorPrvm", +--Fx +"|||lddquXrm","psllwPrvm","pslldPrvm","psllqPrvm", +"pmuludqPrvm","pmaddwdPrvm","psadbwPrvm","maskmovqMrm||maskmovdquXrm$", +"psubbPrvm","psubwPrvm","psubdPrvm","psubqPrvm", +"paddbPrvm","paddwPrvm","padddPrvm","ud", +} +assert(map_opc2[255] == "ud") + +-- Map for three-byte opcodes. Can't wait for their next invention. +local map_opc3 = { +["38"] = { -- [66] 0f 38 xx +--0x +[0]="pshufbPrvm","phaddwPrvm","phadddPrvm","phaddswPrvm", +"pmaddubswPrvm","phsubwPrvm","phsubdPrvm","phsubswPrvm", +"psignbPrvm","psignwPrvm","psigndPrvm","pmulhrswPrvm", +"||permilpsXrvm","||permilpdXrvm",nil,nil, +--1x +"||pblendvbXrma",nil,nil,nil, +"||blendvpsXrma","||blendvpdXrma","||permpsXrvm","||ptestXrm", +"||broadcastssXrm","||broadcastsdXrm","||broadcastf128XrlXm",nil, +"pabsbPrm","pabswPrm","pabsdPrm",nil, +--2x +"||pmovsxbwXrm","||pmovsxbdXrm","||pmovsxbqXrm","||pmovsxwdXrm", +"||pmovsxwqXrm","||pmovsxdqXrm",nil,nil, +"||pmuldqXrvm","||pcmpeqqXrvm","||$movntdqaXrm","||packusdwXrvm", +"||maskmovpsXrvm","||maskmovpdXrvm","||maskmovpsXmvr","||maskmovpdXmvr", +--3x +"||pmovzxbwXrm","||pmovzxbdXrm","||pmovzxbqXrm","||pmovzxwdXrm", +"||pmovzxwqXrm","||pmovzxdqXrm","||permdXrvm","||pcmpgtqXrvm", +"||pminsbXrvm","||pminsdXrvm","||pminuwXrvm","||pminudXrvm", +"||pmaxsbXrvm","||pmaxsdXrvm","||pmaxuwXrvm","||pmaxudXrvm", +--4x +"||pmulddXrvm","||phminposuwXrm",nil,nil, +nil,"||psrlvVSXrvm","||psravdXrvm","||psllvVSXrvm", +--5x +[0x58] = "||pbroadcastdXrlXm",[0x59] = "||pbroadcastqXrlXm", +[0x5a] = "||broadcasti128XrlXm", +--7x +[0x78] = "||pbroadcastbXrlXm",[0x79] = "||pbroadcastwXrlXm", +--8x +[0x8c] = "||pmaskmovXrvVSm", +[0x8e] = "||pmaskmovVSmXvr", +--9x +[0x96] = "||fmaddsub132pHXrvm",[0x97] = "||fmsubadd132pHXrvm", +[0x98] = "||fmadd132pHXrvm",[0x99] = "||fmadd132sHXrvm", +[0x9a] = "||fmsub132pHXrvm",[0x9b] = "||fmsub132sHXrvm", +[0x9c] = "||fnmadd132pHXrvm",[0x9d] = "||fnmadd132sHXrvm", +[0x9e] = "||fnmsub132pHXrvm",[0x9f] = "||fnmsub132sHXrvm", +--Ax +[0xa6] = "||fmaddsub213pHXrvm",[0xa7] = "||fmsubadd213pHXrvm", +[0xa8] = "||fmadd213pHXrvm",[0xa9] = "||fmadd213sHXrvm", +[0xaa] = "||fmsub213pHXrvm",[0xab] = "||fmsub213sHXrvm", +[0xac] = "||fnmadd213pHXrvm",[0xad] = "||fnmadd213sHXrvm", +[0xae] = "||fnmsub213pHXrvm",[0xaf] = "||fnmsub213sHXrvm", +--Bx +[0xb6] = "||fmaddsub231pHXrvm",[0xb7] = "||fmsubadd231pHXrvm", +[0xb8] = "||fmadd231pHXrvm",[0xb9] = "||fmadd231sHXrvm", +[0xba] = "||fmsub231pHXrvm",[0xbb] = "||fmsub231sHXrvm", +[0xbc] = "||fnmadd231pHXrvm",[0xbd] = "||fnmadd231sHXrvm", +[0xbe] = "||fnmsub231pHXrvm",[0xbf] = "||fnmsub231sHXrvm", +--Dx +[0xdc] = "||aesencXrvm", [0xdd] = "||aesenclastXrvm", +[0xde] = "||aesdecXrvm", [0xdf] = "||aesdeclastXrvm", +--Fx +[0xf0] = "|||crc32TrBmt",[0xf1] = "|||crc32TrVmt", +[0xf7] = "| sarxVrmv| shlxVrmv| shrxVrmv", +}, + +["3a"] = { -- [66] 0f 3a xx +--0x +[0x00]="||permqXrmu","||permpdXrmu","||pblenddXrvmu",nil, +"||permilpsXrmu","||permilpdXrmu","||perm2f128Xrvmu",nil, +"||roundpsXrmu","||roundpdXrmu","||roundssXrvmu","||roundsdXrvmu", +"||blendpsXrvmu","||blendpdXrvmu","||pblendwXrvmu","palignrPrvmu", +--1x +nil,nil,nil,nil, +"||pextrbVmXru","||pextrwVmXru","||pextrVmSXru","||extractpsVmXru", +"||insertf128XrvlXmu","||extractf128XlXmYru",nil,nil, +nil,nil,nil,nil, +--2x +"||pinsrbXrvVmu","||insertpsXrvmu","||pinsrXrvVmuS",nil, +--3x +[0x38] = "||inserti128Xrvmu",[0x39] = "||extracti128XlXmYru", +--4x +[0x40] = "||dppsXrvmu", +[0x41] = "||dppdXrvmu", +[0x42] = "||mpsadbwXrvmu", +[0x44] = "||pclmulqdqXrvmu", +[0x46] = "||perm2i128Xrvmu", +[0x4a] = "||blendvpsXrvmb",[0x4b] = "||blendvpdXrvmb", +[0x4c] = "||pblendvbXrvmb", +--6x +[0x60] = "||pcmpestrmXrmu",[0x61] = "||pcmpestriXrmu", +[0x62] = "||pcmpistrmXrmu",[0x63] = "||pcmpistriXrmu", +[0xdf] = "||aeskeygenassistXrmu", +--Fx +[0xf0] = "||| rorxVrmu", +}, +} + +-- Map for VMX/SVM opcodes 0F 01 C0-FF (sgdt group with register operands). +local map_opcvm = { +[0xc1]="vmcall",[0xc2]="vmlaunch",[0xc3]="vmresume",[0xc4]="vmxoff", +[0xc8]="monitor",[0xc9]="mwait", +[0xd8]="vmrun",[0xd9]="vmmcall",[0xda]="vmload",[0xdb]="vmsave", +[0xdc]="stgi",[0xdd]="clgi",[0xde]="skinit",[0xdf]="invlpga", +[0xf8]="swapgs",[0xf9]="rdtscp", +} + +-- Map for FP opcodes. And you thought stack machines are simple? +local map_opcfp = { +-- D8-DF 00-BF: opcodes with a memory operand. +-- D8 +[0]="faddFm","fmulFm","fcomFm","fcompFm","fsubFm","fsubrFm","fdivFm","fdivrFm", +"fldFm",nil,"fstFm","fstpFm","fldenvVm","fldcwWm","fnstenvVm","fnstcwWm", +-- DA +"fiaddDm","fimulDm","ficomDm","ficompDm", +"fisubDm","fisubrDm","fidivDm","fidivrDm", +-- DB +"fildDm","fisttpDm","fistDm","fistpDm",nil,"fld twordFmp",nil,"fstp twordFmp", +-- DC +"faddGm","fmulGm","fcomGm","fcompGm","fsubGm","fsubrGm","fdivGm","fdivrGm", +-- DD +"fldGm","fisttpQm","fstGm","fstpGm","frstorDmp",nil,"fnsaveDmp","fnstswWm", +-- DE +"fiaddWm","fimulWm","ficomWm","ficompWm", +"fisubWm","fisubrWm","fidivWm","fidivrWm", +-- DF +"fildWm","fisttpWm","fistWm","fistpWm", +"fbld twordFmp","fildQm","fbstp twordFmp","fistpQm", +-- xx C0-FF: opcodes with a pseudo-register operand. +-- D8 +"faddFf","fmulFf","fcomFf","fcompFf","fsubFf","fsubrFf","fdivFf","fdivrFf", +-- D9 +"fldFf","fxchFf",{"fnop"},nil, +{"fchs","fabs",nil,nil,"ftst","fxam"}, +{"fld1","fldl2t","fldl2e","fldpi","fldlg2","fldln2","fldz"}, +{"f2xm1","fyl2x","fptan","fpatan","fxtract","fprem1","fdecstp","fincstp"}, +{"fprem","fyl2xp1","fsqrt","fsincos","frndint","fscale","fsin","fcos"}, +-- DA +"fcmovbFf","fcmoveFf","fcmovbeFf","fcmovuFf",nil,{nil,"fucompp"},nil,nil, +-- DB +"fcmovnbFf","fcmovneFf","fcmovnbeFf","fcmovnuFf", +{nil,nil,"fnclex","fninit"},"fucomiFf","fcomiFf",nil, +-- DC +"fadd toFf","fmul toFf",nil,nil, +"fsub toFf","fsubr toFf","fdivr toFf","fdiv toFf", +-- DD +"ffreeFf",nil,"fstFf","fstpFf","fucomFf","fucompFf",nil,nil, +-- DE +"faddpFf","fmulpFf",nil,{nil,"fcompp"}, +"fsubrpFf","fsubpFf","fdivrpFf","fdivpFf", +-- DF +nil,nil,nil,nil,{"fnstsw ax"},"fucomipFf","fcomipFf",nil, +} +assert(map_opcfp[126] == "fcomipFf") + +-- Map for opcode groups. The subkey is sp from the ModRM byte. +local map_opcgroup = { + arith = { "add", "or", "adc", "sbb", "and", "sub", "xor", "cmp" }, + shift = { "rol", "ror", "rcl", "rcr", "shl", "shr", "sal", "sar" }, + testb = { "testBmi", "testBmi", "not", "neg", "mul", "imul", "div", "idiv" }, + testv = { "testVmi", "testVmi", "not", "neg", "mul", "imul", "div", "idiv" }, + incb = { "inc", "dec" }, + incd = { "inc", "dec", "callUmp", "$call farDmp", + "jmpUmp", "$jmp farDmp", "pushUm" }, + sldt = { "sldt", "str", "lldt", "ltr", "verr", "verw" }, + sgdt = { "vm*$sgdt", "vm*$sidt", "$lgdt", "vm*$lidt", + "smsw", nil, "lmsw", "vm*$invlpg" }, + bt = { nil, nil, nil, nil, "bt", "bts", "btr", "btc" }, + cmpxchg = { nil, "sz*,cmpxchg8bQmp,cmpxchg16bXmp", nil, nil, + nil, nil, "vmptrld|vmxon|vmclear", "vmptrst" }, + pshiftw = { nil, nil, "psrlw", nil, "psraw", nil, "psllw" }, + pshiftd = { nil, nil, "psrld", nil, "psrad", nil, "pslld" }, + pshiftq = { nil, nil, "psrlq", nil, nil, nil, "psllq" }, + pshiftdq = { nil, nil, "psrlq", "psrldq", nil, nil, "psllq", "pslldq" }, + fxsave = { "$fxsave", "$fxrstor", "$ldmxcsr", "$stmxcsr", + nil, "lfenceDp$", "mfenceDp$", "sfenceDp$clflush" }, + prefetch = { "prefetch", "prefetchw" }, + prefetcht = { "prefetchnta", "prefetcht0", "prefetcht1", "prefetcht2" }, +} + +------------------------------------------------------------------------------ + +-- Maps for register names. +local map_regs = { + B = { "al", "cl", "dl", "bl", "ah", "ch", "dh", "bh", + "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, + B64 = { "al", "cl", "dl", "bl", "spl", "bpl", "sil", "dil", + "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, + W = { "ax", "cx", "dx", "bx", "sp", "bp", "si", "di", + "r8w", "r9w", "r10w", "r11w", "r12w", "r13w", "r14w", "r15w" }, + D = { "eax", "ecx", "edx", "ebx", "esp", "ebp", "esi", "edi", + "r8d", "r9d", "r10d", "r11d", "r12d", "r13d", "r14d", "r15d" }, + Q = { "rax", "rcx", "rdx", "rbx", "rsp", "rbp", "rsi", "rdi", + "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15" }, + M = { "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7", + "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7" }, -- No x64 ext! + X = { "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7", + "xmm8", "xmm9", "xmm10", "xmm11", "xmm12", "xmm13", "xmm14", "xmm15" }, + Y = { "ymm0", "ymm1", "ymm2", "ymm3", "ymm4", "ymm5", "ymm6", "ymm7", + "ymm8", "ymm9", "ymm10", "ymm11", "ymm12", "ymm13", "ymm14", "ymm15" }, +} +local map_segregs = { "es", "cs", "ss", "ds", "fs", "gs", "segr6", "segr7" } + +-- Maps for size names. +local map_sz2n = { + B = 1, W = 2, D = 4, Q = 8, M = 8, X = 16, Y = 32, +} +local map_sz2prefix = { + B = "byte", W = "word", D = "dword", + Q = "qword", + M = "qword", X = "xword", Y = "yword", + F = "dword", G = "qword", -- No need for sizes/register names for these two. +} + +------------------------------------------------------------------------------ + +-- Output a nicely formatted line with an opcode and operands. +local function putop(ctx, text, operands) + local code, pos, hex = ctx.code, ctx.pos, "" + local hmax = ctx.hexdump + if hmax > 0 then + for i=ctx.start,pos-1 do + hex = hex..format("%02X", byte(code, i, i)) + end + if #hex > hmax then hex = sub(hex, 1, hmax)..". " + else hex = hex..rep(" ", hmax-#hex+2) end + end + if operands then text = text.." "..operands end + if ctx.o16 then text = "o16 "..text; ctx.o16 = false end + if ctx.a32 then text = "a32 "..text; ctx.a32 = false end + if ctx.rep then text = ctx.rep.." "..text; ctx.rep = false end + if ctx.rex then + local t = (ctx.rexw and "w" or "")..(ctx.rexr and "r" or "").. + (ctx.rexx and "x" or "")..(ctx.rexb and "b" or "").. + (ctx.vexl and "l" or "") + if ctx.vexv and ctx.vexv ~= 0 then t = t.."v"..ctx.vexv end + if t ~= "" then text = ctx.rex.."."..t.." "..gsub(text, "^ ", "") + elseif ctx.rex == "vex" then text = gsub("v"..text, "^v ", "") end + ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false + ctx.rex = false; ctx.vexl = false; ctx.vexv = false + end + if ctx.seg then + local text2, n = gsub(text, "%[", "["..ctx.seg..":") + if n == 0 then text = ctx.seg.." "..text else text = text2 end + ctx.seg = false + end + if ctx.lock then text = "lock "..text; ctx.lock = false end + local imm = ctx.imm + if imm then + local sym = ctx.symtab[imm] + if sym then text = text.."\t->"..sym end + end + ctx.out(format("%08x %s%s\n", ctx.addr+ctx.start, hex, text)) + ctx.mrm = false + ctx.vexv = false + ctx.start = pos + ctx.imm = nil +end + +-- Clear all prefix flags. +local function clearprefixes(ctx) + ctx.o16 = false; ctx.seg = false; ctx.lock = false; ctx.rep = false + ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false + ctx.rex = false; ctx.a32 = false; ctx.vexl = false +end + +-- Fallback for incomplete opcodes at the end. +local function incomplete(ctx) + ctx.pos = ctx.stop+1 + clearprefixes(ctx) + return putop(ctx, "(incomplete)") +end + +-- Fallback for unknown opcodes. +local function unknown(ctx) + clearprefixes(ctx) + return putop(ctx, "(unknown)") +end + +-- Return an immediate of the specified size. +local function getimm(ctx, pos, n) + if pos+n-1 > ctx.stop then return incomplete(ctx) end + local code = ctx.code + if n == 1 then + local b1 = byte(code, pos, pos) + return b1 + elseif n == 2 then + local b1, b2 = byte(code, pos, pos+1) + return b1+b2*256 + else + local b1, b2, b3, b4 = byte(code, pos, pos+3) + local imm = b1+b2*256+b3*65536+b4*16777216 + ctx.imm = imm + return imm + end +end + +-- Process pattern string and generate the operands. +local function putpat(ctx, name, pat) + local operands, regs, sz, mode, sp, rm, sc, rx, sdisp + local code, pos, stop, vexl = ctx.code, ctx.pos, ctx.stop, ctx.vexl + + -- Chars used: 1DFGHIMPQRSTUVWXYabcdfgijlmoprstuvwxyz + for p in gmatch(pat, ".") do + local x = nil + if p == "V" or p == "U" then + if ctx.rexw then sz = "Q"; ctx.rexw = false + elseif ctx.o16 then sz = "W"; ctx.o16 = false + elseif p == "U" and ctx.x64 then sz = "Q" + else sz = "D" end + regs = map_regs[sz] + elseif p == "T" then + if ctx.rexw then sz = "Q"; ctx.rexw = false else sz = "D" end + regs = map_regs[sz] + elseif p == "B" then + sz = "B" + regs = ctx.rex and map_regs.B64 or map_regs.B + elseif match(p, "[WDQMXYFG]") then + sz = p + if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end + regs = map_regs[sz] + elseif p == "P" then + sz = ctx.o16 and "X" or "M"; ctx.o16 = false + if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end + regs = map_regs[sz] + elseif p == "H" then + name = name..(ctx.rexw and "d" or "s") + ctx.rexw = false + elseif p == "S" then + name = name..lower(sz) + elseif p == "s" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = imm <= 127 and format("+0x%02x", imm) + or format("-0x%02x", 256-imm) + pos = pos+1 + elseif p == "u" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = format("0x%02x", imm) + pos = pos+1 + elseif p == "b" then + local imm = getimm(ctx, pos, 1); if not imm then return end + x = regs[imm/16+1] + pos = pos+1 + elseif p == "w" then + local imm = getimm(ctx, pos, 2); if not imm then return end + x = format("0x%x", imm) + pos = pos+2 + elseif p == "o" then -- [offset] + if ctx.x64 then + local imm1 = getimm(ctx, pos, 4); if not imm1 then return end + local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end + x = format("[0x%08x%08x]", imm2, imm1) + pos = pos+8 + else + local imm = getimm(ctx, pos, 4); if not imm then return end + x = format("[0x%08x]", imm) + pos = pos+4 + end + elseif p == "i" or p == "I" then + local n = map_sz2n[sz] + if n == 8 and ctx.x64 and p == "I" then + local imm1 = getimm(ctx, pos, 4); if not imm1 then return end + local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end + x = format("0x%08x%08x", imm2, imm1) + else + if n == 8 then n = 4 end + local imm = getimm(ctx, pos, n); if not imm then return end + if sz == "Q" and (imm < 0 or imm > 0x7fffffff) then + imm = (0xffffffff+1)-imm + x = format(imm > 65535 and "-0x%08x" or "-0x%x", imm) + else + x = format(imm > 65535 and "0x%08x" or "0x%x", imm) + end + end + pos = pos+n + elseif p == "j" then + local n = map_sz2n[sz] + if n == 8 then n = 4 end + local imm = getimm(ctx, pos, n); if not imm then return end + if sz == "B" and imm > 127 then imm = imm-256 + elseif imm > 2147483647 then imm = imm-4294967296 end + pos = pos+n + imm = imm + pos + ctx.addr + if imm > 4294967295 and not ctx.x64 then imm = imm-4294967296 end + ctx.imm = imm + if sz == "W" then + x = format("word 0x%04x", imm%65536) + elseif ctx.x64 then + local lo = imm % 0x1000000 + x = format("0x%02x%06x", (imm-lo) / 0x1000000, lo) + else + x = "0x"..tohex(imm) + end + elseif p == "R" then + local r = byte(code, pos-1, pos-1)%8 + if ctx.rexb then r = r + 8; ctx.rexb = false end + x = regs[r+1] + elseif p == "a" then x = regs[1] + elseif p == "c" then x = "cl" + elseif p == "d" then x = "dx" + elseif p == "1" then x = "1" + else + if not mode then + mode = ctx.mrm + if not mode then + if pos > stop then return incomplete(ctx) end + mode = byte(code, pos, pos) + pos = pos+1 + end + rm = mode%8; mode = (mode-rm)/8 + sp = mode%8; mode = (mode-sp)/8 + sdisp = "" + if mode < 3 then + if rm == 4 then + if pos > stop then return incomplete(ctx) end + sc = byte(code, pos, pos) + pos = pos+1 + rm = sc%8; sc = (sc-rm)/8 + rx = sc%8; sc = (sc-rx)/8 + if ctx.rexx then rx = rx + 8; ctx.rexx = false end + if rx == 4 then rx = nil end + end + if mode > 0 or rm == 5 then + local dsz = mode + if dsz ~= 1 then dsz = 4 end + local disp = getimm(ctx, pos, dsz); if not disp then return end + if mode == 0 then rm = nil end + if rm or rx or (not sc and ctx.x64 and not ctx.a32) then + if dsz == 1 and disp > 127 then + sdisp = format("-0x%x", 256-disp) + elseif disp >= 0 and disp <= 0x7fffffff then + sdisp = format("+0x%x", disp) + else + sdisp = format("-0x%x", (0xffffffff+1)-disp) + end + else + sdisp = format(ctx.x64 and not ctx.a32 and + not (disp >= 0 and disp <= 0x7fffffff) + and "0xffffffff%08x" or "0x%08x", disp) + end + pos = pos+dsz + end + end + if rm and ctx.rexb then rm = rm + 8; ctx.rexb = false end + if ctx.rexr then sp = sp + 8; ctx.rexr = false end + end + if p == "m" then + if mode == 3 then x = regs[rm+1] + else + local aregs = ctx.a32 and map_regs.D or ctx.aregs + local srm, srx = "", "" + if rm then srm = aregs[rm+1] + elseif not sc and ctx.x64 and not ctx.a32 then srm = "rip" end + ctx.a32 = false + if rx then + if rm then srm = srm.."+" end + srx = aregs[rx+1] + if sc > 0 then srx = srx.."*"..(2^sc) end + end + x = format("[%s%s%s]", srm, srx, sdisp) + end + if mode < 3 and + (not match(pat, "[aRrgp]") or match(pat, "t")) then -- Yuck. + x = map_sz2prefix[sz].." "..x + end + elseif p == "r" then x = regs[sp+1] + elseif p == "g" then x = map_segregs[sp+1] + elseif p == "p" then -- Suppress prefix. + elseif p == "f" then x = "st"..rm + elseif p == "x" then + if sp == 0 and ctx.lock and not ctx.x64 then + x = "CR8"; ctx.lock = false + else + x = "CR"..sp + end + elseif p == "v" then + if ctx.vexv then + x = regs[ctx.vexv+1]; ctx.vexv = false + end + elseif p == "y" then x = "DR"..sp + elseif p == "z" then x = "TR"..sp + elseif p == "l" then vexl = false + elseif p == "t" then + else + error("bad pattern `"..pat.."'") + end + end + if x then operands = operands and operands..", "..x or x end + end + ctx.pos = pos + return putop(ctx, name, operands) +end + +-- Forward declaration. +local map_act + +-- Fetch and cache MRM byte. +local function getmrm(ctx) + local mrm = ctx.mrm + if not mrm then + local pos = ctx.pos + if pos > ctx.stop then return nil end + mrm = byte(ctx.code, pos, pos) + ctx.pos = pos+1 + ctx.mrm = mrm + end + return mrm +end + +-- Dispatch to handler depending on pattern. +local function dispatch(ctx, opat, patgrp) + if not opat then return unknown(ctx) end + if match(opat, "%|") then -- MMX/SSE variants depending on prefix. + local p + if ctx.rep then + p = ctx.rep=="rep" and "%|([^%|]*)" or "%|[^%|]*%|[^%|]*%|([^%|]*)" + ctx.rep = false + elseif ctx.o16 then p = "%|[^%|]*%|([^%|]*)"; ctx.o16 = false + else p = "^[^%|]*" end + opat = match(opat, p) + if not opat then return unknown(ctx) end +-- ctx.rep = false; ctx.o16 = false + --XXX fails for 66 f2 0f 38 f1 06 crc32 eax,WORD PTR [esi] + --XXX remove in branches? + end + if match(opat, "%$") then -- reg$mem variants. + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + opat = match(opat, mrm >= 192 and "^[^%$]*" or "%$(.*)") + if opat == "" then return unknown(ctx) end + end + if opat == "" then return unknown(ctx) end + local name, pat = match(opat, "^([a-z0-9 ]*)(.*)") + if pat == "" and patgrp then pat = patgrp end + return map_act[sub(pat, 1, 1)](ctx, name, pat) +end + +-- Get a pattern from an opcode map and dispatch to handler. +local function dispatchmap(ctx, opcmap) + local pos = ctx.pos + local opat = opcmap[byte(ctx.code, pos, pos)] + pos = pos + 1 + ctx.pos = pos + return dispatch(ctx, opat) +end + +-- Map for action codes. The key is the first char after the name. +map_act = { + -- Simple opcodes without operands. + [""] = function(ctx, name, pat) + return putop(ctx, name) + end, + + -- Operand size chars fall right through. + B = putpat, W = putpat, D = putpat, Q = putpat, + V = putpat, U = putpat, T = putpat, + M = putpat, X = putpat, P = putpat, + F = putpat, G = putpat, Y = putpat, + H = putpat, + + -- Collect prefixes. + [":"] = function(ctx, name, pat) + ctx[pat == ":" and name or sub(pat, 2)] = name + if ctx.pos - ctx.start > 5 then return unknown(ctx) end -- Limit #prefixes. + end, + + -- Chain to special handler specified by name. + ["*"] = function(ctx, name, pat) + return map_act[name](ctx, name, sub(pat, 2)) + end, + + -- Use named subtable for opcode group. + ["!"] = function(ctx, name, pat) + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + return dispatch(ctx, map_opcgroup[name][((mrm-(mrm%8))/8)%8+1], sub(pat, 2)) + end, + + -- o16,o32[,o64] variants. + sz = function(ctx, name, pat) + if ctx.o16 then ctx.o16 = false + else + pat = match(pat, ",(.*)") + if ctx.rexw then + local p = match(pat, ",(.*)") + if p then pat = p; ctx.rexw = false end + end + end + pat = match(pat, "^[^,]*") + return dispatch(ctx, pat) + end, + + -- Two-byte opcode dispatch. + opc2 = function(ctx, name, pat) + return dispatchmap(ctx, map_opc2) + end, + + -- Three-byte opcode dispatch. + opc3 = function(ctx, name, pat) + return dispatchmap(ctx, map_opc3[pat]) + end, + + -- VMX/SVM dispatch. + vm = function(ctx, name, pat) + return dispatch(ctx, map_opcvm[ctx.mrm]) + end, + + -- Floating point opcode dispatch. + fp = function(ctx, name, pat) + local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end + local rm = mrm%8 + local idx = pat*8 + ((mrm-rm)/8)%8 + if mrm >= 192 then idx = idx + 64 end + local opat = map_opcfp[idx] + if type(opat) == "table" then opat = opat[rm+1] end + return dispatch(ctx, opat) + end, + + -- REX prefix. + rex = function(ctx, name, pat) + if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. + for p in gmatch(pat, ".") do ctx["rex"..p] = true end + ctx.rex = "rex" + end, + + -- VEX prefix. + vex = function(ctx, name, pat) + if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. + ctx.rex = "vex" + local pos = ctx.pos + if ctx.mrm then + ctx.mrm = nil + pos = pos-1 + end + local b = byte(ctx.code, pos, pos) + if not b then return incomplete(ctx) end + pos = pos+1 + if b < 128 then ctx.rexr = true end + local m = 1 + if pat == "3" then + m = b%32; b = (b-m)/32 + local nb = b%2; b = (b-nb)/2 + if nb == 0 then ctx.rexb = true end + local nx = b%2 + if nx == 0 then ctx.rexx = true end + b = byte(ctx.code, pos, pos) + if not b then return incomplete(ctx) end + pos = pos+1 + if b >= 128 then ctx.rexw = true end + end + ctx.pos = pos + local map + if m == 1 then map = map_opc2 + elseif m == 2 then map = map_opc3["38"] + elseif m == 3 then map = map_opc3["3a"] + else return unknown(ctx) end + local p = b%4; b = (b-p)/4 + if p == 1 then ctx.o16 = "o16" + elseif p == 2 then ctx.rep = "rep" + elseif p == 3 then ctx.rep = "repne" end + local l = b%2; b = (b-l)/2 + if l ~= 0 then ctx.vexl = true end + ctx.vexv = (-1-b)%16 + return dispatchmap(ctx, map) + end, + + -- Special case for nop with REX prefix. + nop = function(ctx, name, pat) + return dispatch(ctx, ctx.rex and pat or "nop") + end, + + -- Special case for 0F 77. + emms = function(ctx, name, pat) + if ctx.rex ~= "vex" then + return putop(ctx, "emms") + elseif ctx.vexl then + ctx.vexl = false + return putop(ctx, "zeroall") + else + return putop(ctx, "zeroupper") + end + end, +} + +------------------------------------------------------------------------------ + +-- Disassemble a block of code. +local function disass_block(ctx, ofs, len) + if not ofs then ofs = 0 end + local stop = len and ofs+len or #ctx.code + ofs = ofs + 1 + ctx.start = ofs + ctx.pos = ofs + ctx.stop = stop + ctx.imm = nil + ctx.mrm = false + clearprefixes(ctx) + while ctx.pos <= stop do dispatchmap(ctx, ctx.map1) end + if ctx.pos ~= ctx.start then incomplete(ctx) end +end + +-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). +local function create(code, addr, out) + local ctx = {} + ctx.code = code + ctx.addr = (addr or 0) - 1 + ctx.out = out or io.write + ctx.symtab = {} + ctx.disass = disass_block + ctx.hexdump = 16 + ctx.x64 = false + ctx.map1 = map_opc1_32 + ctx.aregs = map_regs.D + return ctx +end + +local function create64(code, addr, out) + local ctx = create(code, addr, out) + ctx.x64 = true + ctx.map1 = map_opc1_64 + ctx.aregs = map_regs.Q + return ctx +end + +-- Simple API: disassemble code (a string) at address and output via out. +local function disass(code, addr, out) + create(code, addr, out):disass() +end + +local function disass64(code, addr, out) + create64(code, addr, out):disass() +end + +-- Return register name for RID. +local function regname(r) + if r < 8 then return map_regs.D[r+1] end + return map_regs.X[r-7] +end + +local function regname64(r) + if r < 16 then return map_regs.Q[r+1] end + return map_regs.X[r-15] +end + +-- Public module functions. +return { + create = create, + create64 = create64, + disass = disass, + disass64 = disass64, + regname = regname, + regname64 = regname64 +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dump.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dump.lua new file mode 100644 index 00000000..9eda08c4 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/dump.lua @@ -0,0 +1,726 @@ +---------------------------------------------------------------------------- +-- LuaJIT compiler dump module. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module can be used to debug the JIT compiler itself. It dumps the +-- code representations and structures used in various compiler stages. +-- +-- Example usage: +-- +-- luajit -jdump -e "local x=0; for i=1,1e6 do x=x+i end; print(x)" +-- luajit -jdump=im -e "for i=1,1000 do for j=1,1000 do end end" | less -R +-- luajit -jdump=is myapp.lua | less -R +-- luajit -jdump=-b myapp.lua +-- luajit -jdump=+aH,myapp.html myapp.lua +-- luajit -jdump=ixT,myapp.dump myapp.lua +-- +-- The first argument specifies the dump mode. The second argument gives +-- the output file name. Default output is to stdout, unless the environment +-- variable LUAJIT_DUMPFILE is set. The file is overwritten every time the +-- module is started. +-- +-- Different features can be turned on or off with the dump mode. If the +-- mode starts with a '+', the following features are added to the default +-- set of features; a '-' removes them. Otherwise the features are replaced. +-- +-- The following dump features are available (* marks the default): +-- +-- * t Print a line for each started, ended or aborted trace (see also -jv). +-- * b Dump the traced bytecode. +-- * i Dump the IR (intermediate representation). +-- r Augment the IR with register/stack slots. +-- s Dump the snapshot map. +-- * m Dump the generated machine code. +-- x Print each taken trace exit. +-- X Print each taken trace exit and the contents of all registers. +-- a Print the IR of aborted traces, too. +-- +-- The output format can be set with the following characters: +-- +-- T Plain text output. +-- A ANSI-colored text output +-- H Colorized HTML + CSS output. +-- +-- The default output format is plain text. It's set to ANSI-colored text +-- if the COLORTERM variable is set. Note: this is independent of any output +-- redirection, which is actually considered a feature. +-- +-- You probably want to use less -R to enjoy viewing ANSI-colored text from +-- a pipe or a file. Add this to your ~/.bashrc: export LESS="-R" +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local funcinfo, funcbc = jutil.funcinfo, jutil.funcbc +local traceinfo, traceir, tracek = jutil.traceinfo, jutil.traceir, jutil.tracek +local tracemc, tracesnap = jutil.tracemc, jutil.tracesnap +local traceexitstub, ircalladdr = jutil.traceexitstub, jutil.ircalladdr +local bit = require("bit") +local band, shr, tohex = bit.band, bit.rshift, bit.tohex +local sub, gsub, format = string.sub, string.gsub, string.format +local byte, rep = string.byte, string.rep +local type, tostring = type, tostring +local stdout, stderr = io.stdout, io.stderr + +-- Load other modules on-demand. +local bcline, disass + +-- Active flag, output file handle and dump mode. +local active, out, dumpmode + +------------------------------------------------------------------------------ + +local symtabmt = { __index = false } +local symtab = {} +local nexitsym = 0 + +-- Fill nested symbol table with per-trace exit stub addresses. +local function fillsymtab_tr(tr, nexit) + local t = {} + symtabmt.__index = t + if jit.arch:sub(1, 4) == "mips" then + t[traceexitstub(tr, 0)] = "exit" + return + end + for i=0,nexit-1 do + local addr = traceexitstub(tr, i) + if addr < 0 then addr = addr + 2^32 end + t[addr] = tostring(i) + end + local addr = traceexitstub(tr, nexit) + if addr then t[addr] = "stack_check" end +end + +-- Fill symbol table with trace exit stub addresses. +local function fillsymtab(tr, nexit) + local t = symtab + if nexitsym == 0 then + local maskaddr = jit.arch == "arm" and -2 + local ircall = vmdef.ircall + for i=0,#ircall do + local addr = ircalladdr(i) + if addr ~= 0 then + if maskaddr then addr = band(addr, maskaddr) end + if addr < 0 then addr = addr + 2^32 end + t[addr] = ircall[i] + end + end + end + if nexitsym == 1000000 then -- Per-trace exit stubs. + fillsymtab_tr(tr, nexit) + elseif nexit > nexitsym then -- Shared exit stubs. + for i=nexitsym,nexit-1 do + local addr = traceexitstub(i) + if addr == nil then -- Fall back to per-trace exit stubs. + fillsymtab_tr(tr, nexit) + setmetatable(symtab, symtabmt) + nexit = 1000000 + break + end + if addr < 0 then addr = addr + 2^32 end + t[addr] = tostring(i) + end + nexitsym = nexit + end + return t +end + +local function dumpwrite(s) + out:write(s) +end + +-- Disassemble machine code. +local function dump_mcode(tr) + local info = traceinfo(tr) + if not info then return end + local mcode, addr, loop = tracemc(tr) + if not mcode then return end + if not disass then disass = require("jit.dis_"..jit.arch) end + if addr < 0 then addr = addr + 2^32 end + out:write("---- TRACE ", tr, " mcode ", #mcode, "\n") + local ctx = disass.create(mcode, addr, dumpwrite) + ctx.hexdump = 0 + ctx.symtab = fillsymtab(tr, info.nexit) + if loop ~= 0 then + symtab[addr+loop] = "LOOP" + ctx:disass(0, loop) + out:write("->LOOP:\n") + ctx:disass(loop, #mcode-loop) + symtab[addr+loop] = nil + else + ctx:disass(0, #mcode) + end +end + +------------------------------------------------------------------------------ + +local irtype_text = { + [0] = "nil", + "fal", + "tru", + "lud", + "str", + "p32", + "thr", + "pro", + "fun", + "p64", + "cdt", + "tab", + "udt", + "flt", + "num", + "i8 ", + "u8 ", + "i16", + "u16", + "int", + "u32", + "i64", + "u64", + "sfp", +} + +local colortype_ansi = { + [0] = "%s", + "%s", + "%s", + "\027[36m%s\027[m", + "\027[32m%s\027[m", + "%s", + "\027[1m%s\027[m", + "%s", + "\027[1m%s\027[m", + "%s", + "\027[33m%s\027[m", + "\027[31m%s\027[m", + "\027[36m%s\027[m", + "\027[34m%s\027[m", + "\027[34m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", + "\027[35m%s\027[m", +} + +local function colorize_text(s) + return s +end + +local function colorize_ansi(s, t, extra) + local out = format(colortype_ansi[t], s) + if extra then out = "\027[3m"..out end + return out +end + +local irtype_ansi = setmetatable({}, + { __index = function(tab, t) + local s = colorize_ansi(irtype_text[t], t); tab[t] = s; return s; end }) + +local html_escape = { ["<"] = "<", [">"] = ">", ["&"] = "&", } + +local function colorize_html(s, t, extra) + s = gsub(s, "[<>&]", html_escape) + return format('%s', + irtype_text[t], extra and " irt_extra" or "", s) +end + +local irtype_html = setmetatable({}, + { __index = function(tab, t) + local s = colorize_html(irtype_text[t], t); tab[t] = s; return s; end }) + +local header_html = [[ + +]] + +local colorize, irtype + +-- Lookup tables to convert some literals into names. +local litname = { + ["SLOAD "] = setmetatable({}, { __index = function(t, mode) + local s = "" + if band(mode, 1) ~= 0 then s = s.."P" end + if band(mode, 2) ~= 0 then s = s.."F" end + if band(mode, 4) ~= 0 then s = s.."T" end + if band(mode, 8) ~= 0 then s = s.."C" end + if band(mode, 16) ~= 0 then s = s.."R" end + if band(mode, 32) ~= 0 then s = s.."I" end + if band(mode, 64) ~= 0 then s = s.."K" end + t[mode] = s + return s + end}), + ["XLOAD "] = { [0] = "", "R", "V", "RV", "U", "RU", "VU", "RVU", }, + ["CONV "] = setmetatable({}, { __index = function(t, mode) + local s = irtype[band(mode, 31)] + s = irtype[band(shr(mode, 5), 31)].."."..s + if band(mode, 0x800) ~= 0 then s = s.." sext" end + local c = shr(mode, 12) + if c == 1 then s = s.." none" + elseif c == 2 then s = s.." index" + elseif c == 3 then s = s.." check" end + t[mode] = s + return s + end}), + ["FLOAD "] = vmdef.irfield, + ["FREF "] = vmdef.irfield, + ["FPMATH"] = vmdef.irfpm, + ["TMPREF"] = { [0] = "", "IN", "OUT", "INOUT", "", "", "OUT2", "INOUT2" }, + ["BUFHDR"] = { [0] = "RESET", "APPEND", "WRITE" }, + ["TOSTR "] = { [0] = "INT", "NUM", "CHAR" }, +} + +local function ctlsub(c) + if c == "\n" then return "\\n" + elseif c == "\r" then return "\\r" + elseif c == "\t" then return "\\t" + else return format("\\%03d", byte(c)) + end +end + +local function fmtfunc(func, pc) + local fi = funcinfo(func, pc) + if fi.loc then + return fi.loc + elseif fi.ffid then + return vmdef.ffnames[fi.ffid] + elseif fi.addr then + return format("C:%x", fi.addr) + else + return "(?)" + end +end + +local function formatk(tr, idx, sn) + local k, t, slot = tracek(tr, idx) + local tn = type(k) + local s + if tn == "number" then + if t < 12 then + s = k == 0 and "NULL" or format("[0x%08x]", k) + elseif band(sn or 0, 0x30000) ~= 0 then + s = band(sn, 0x20000) ~= 0 and "contpc" or "ftsz" + elseif k == 2^52+2^51 then + s = "bias" + else + s = format(0 < k and k < 0x1p-1026 and "%+a" or "%+.14g", k) + end + elseif tn == "string" then + s = format(#k > 20 and '"%.20s"~' or '"%s"', gsub(k, "%c", ctlsub)) + elseif tn == "function" then + s = fmtfunc(k) + elseif tn == "table" then + s = format("{%p}", k) + elseif tn == "userdata" then + if t == 12 then + s = format("userdata:%p", k) + else + s = format("[%p]", k) + if s == "[NULL]" then s = "NULL" end + end + elseif t == 21 then -- int64_t + s = sub(tostring(k), 1, -3) + if sub(s, 1, 1) ~= "-" then s = "+"..s end + elseif sn == 0x1057fff then -- SNAP(1, SNAP_FRAME | SNAP_NORESTORE, REF_NIL) + return "----" -- Special case for LJ_FR2 slot 1. + else + s = tostring(k) -- For primitives. + end + s = colorize(format("%-4s", s), t, band(sn or 0, 0x100000) ~= 0) + if slot then + s = format("%s @%d", s, slot) + end + return s +end + +local function printsnap(tr, snap) + local n = 2 + for s=0,snap[1]-1 do + local sn = snap[n] + if shr(sn, 24) == s then + n = n + 1 + local ref = band(sn, 0xffff) - 0x8000 -- REF_BIAS + if ref < 0 then + out:write(formatk(tr, ref, sn)) + elseif band(sn, 0x80000) ~= 0 then -- SNAP_SOFTFPNUM + out:write(colorize(format("%04d/%04d", ref, ref+1), 14)) + else + local m, ot, op1, op2 = traceir(tr, ref) + out:write(colorize(format("%04d", ref), band(ot, 31), band(sn, 0x100000) ~= 0)) + end + out:write(band(sn, 0x10000) == 0 and " " or "|") -- SNAP_FRAME + else + out:write("---- ") + end + end + out:write("]\n") +end + +-- Dump snapshots (not interleaved with IR). +local function dump_snap(tr) + out:write("---- TRACE ", tr, " snapshots\n") + for i=0,1000000000 do + local snap = tracesnap(tr, i) + if not snap then break end + out:write(format("#%-3d %04d [ ", i, snap[0])) + printsnap(tr, snap) + end +end + +-- Return a register name or stack slot for a rid/sp location. +local function ridsp_name(ridsp, ins) + if not disass then disass = require("jit.dis_"..jit.arch) end + local rid, slot = band(ridsp, 0xff), shr(ridsp, 8) + if rid == 253 or rid == 254 then + return (slot == 0 or slot == 255) and " {sink" or format(" {%04d", ins-slot) + end + if ridsp > 255 then return format("[%x]", slot*4) end + if rid < 128 then return disass.regname(rid) end + return "" +end + +-- Dump CALL* function ref and return optional ctype. +local function dumpcallfunc(tr, ins) + local ctype + if ins > 0 then + local m, ot, op1, op2 = traceir(tr, ins) + if band(ot, 31) == 0 then -- nil type means CARG(func, ctype). + ins = op1 + ctype = formatk(tr, op2) + end + end + if ins < 0 then + out:write(format("[0x%x](", tonumber((tracek(tr, ins))))) + else + out:write(format("%04d (", ins)) + end + return ctype +end + +-- Recursively gather CALL* args and dump them. +local function dumpcallargs(tr, ins) + if ins < 0 then + out:write(formatk(tr, ins)) + else + local m, ot, op1, op2 = traceir(tr, ins) + local oidx = 6*shr(ot, 8) + local op = sub(vmdef.irnames, oidx+1, oidx+6) + if op == "CARG " then + dumpcallargs(tr, op1) + if op2 < 0 then + out:write(" ", formatk(tr, op2)) + else + out:write(" ", format("%04d", op2)) + end + else + out:write(format("%04d", ins)) + end + end +end + +-- Dump IR and interleaved snapshots. +local function dump_ir(tr, dumpsnap, dumpreg) + local info = traceinfo(tr) + if not info then return end + local nins = info.nins + out:write("---- TRACE ", tr, " IR\n") + local irnames = vmdef.irnames + local snapref = 65536 + local snap, snapno + if dumpsnap then + snap = tracesnap(tr, 0) + snapref = snap[0] + snapno = 0 + end + for ins=1,nins do + if ins >= snapref then + if dumpreg then + out:write(format(".... SNAP #%-3d [ ", snapno)) + else + out:write(format(".... SNAP #%-3d [ ", snapno)) + end + printsnap(tr, snap) + snapno = snapno + 1 + snap = tracesnap(tr, snapno) + snapref = snap and snap[0] or 65536 + end + local m, ot, op1, op2, ridsp = traceir(tr, ins) + local oidx, t = 6*shr(ot, 8), band(ot, 31) + local op = sub(irnames, oidx+1, oidx+6) + if op == "LOOP " then + if dumpreg then + out:write(format("%04d ------------ LOOP ------------\n", ins)) + else + out:write(format("%04d ------ LOOP ------------\n", ins)) + end + elseif op ~= "NOP " and op ~= "CARG " and + (dumpreg or op ~= "RENAME") then + local rid = band(ridsp, 255) + if dumpreg then + out:write(format("%04d %-6s", ins, ridsp_name(ridsp, ins))) + else + out:write(format("%04d ", ins)) + end + out:write(format("%s%s %s %s ", + (rid == 254 or rid == 253) and "}" or + (band(ot, 128) == 0 and " " or ">"), + band(ot, 64) == 0 and " " or "+", + irtype[t], op)) + local m1, m2 = band(m, 3), band(m, 3*4) + if sub(op, 1, 4) == "CALL" then + local ctype + if m2 == 1*4 then -- op2 == IRMlit + out:write(format("%-10s (", vmdef.ircall[op2])) + else + ctype = dumpcallfunc(tr, op2) + end + if op1 ~= -1 then dumpcallargs(tr, op1) end + out:write(")") + if ctype then out:write(" ctype ", ctype) end + elseif op == "CNEW " and op2 == -1 then + out:write(formatk(tr, op1)) + elseif m1 ~= 3 then -- op1 != IRMnone + if op1 < 0 then + out:write(formatk(tr, op1)) + else + out:write(format(m1 == 0 and "%04d" or "#%-3d", op1)) + end + if m2 ~= 3*4 then -- op2 != IRMnone + if m2 == 1*4 then -- op2 == IRMlit + local litn = litname[op] + if litn and litn[op2] then + out:write(" ", litn[op2]) + elseif op == "UREFO " or op == "UREFC " then + out:write(format(" #%-3d", shr(op2, 8))) + else + out:write(format(" #%-3d", op2)) + end + elseif op2 < 0 then + out:write(" ", formatk(tr, op2)) + else + out:write(format(" %04d", op2)) + end + end + end + out:write("\n") + end + end + if snap then + if dumpreg then + out:write(format(".... SNAP #%-3d [ ", snapno)) + else + out:write(format(".... SNAP #%-3d [ ", snapno)) + end + printsnap(tr, snap) + end +end + +------------------------------------------------------------------------------ + +local recprefix = "" +local recdepth = 0 + +-- Format trace error message. +local function fmterr(err, info) + if type(err) == "number" then + if type(info) == "function" then info = fmtfunc(info) end + err = format(vmdef.traceerr[err], info) + end + return err +end + +-- Dump trace states. +local function dump_trace(what, tr, func, pc, otr, oex) + if what == "stop" or (what == "abort" and dumpmode.a) then + if dumpmode.i then dump_ir(tr, dumpmode.s, dumpmode.r and what == "stop") + elseif dumpmode.s then dump_snap(tr) end + if dumpmode.m then dump_mcode(tr) end + end + if what == "start" then + if dumpmode.H then out:write('
\n') end
+    out:write("---- TRACE ", tr, " ", what)
+    if otr then out:write(" ", otr, "/", oex == -1 and "stitch" or oex) end
+    out:write(" ", fmtfunc(func, pc), "\n")
+  elseif what == "stop" or what == "abort" then
+    out:write("---- TRACE ", tr, " ", what)
+    if what == "abort" then
+      out:write(" ", fmtfunc(func, pc), " -- ", fmterr(otr, oex), "\n")
+    else
+      local info = traceinfo(tr)
+      local link, ltype = info.link, info.linktype
+      if link == tr or link == 0 then
+	out:write(" -> ", ltype, "\n")
+      elseif ltype == "root" then
+	out:write(" -> ", link, "\n")
+      else
+	out:write(" -> ", link, " ", ltype, "\n")
+      end
+    end
+    if dumpmode.H then out:write("
\n\n") else out:write("\n") end + else + if what == "flush" then symtab, nexitsym = {}, 0 end + out:write("---- TRACE ", what, "\n\n") + end + out:flush() +end + +-- Dump recorded bytecode. +local function dump_record(tr, func, pc, depth) + if depth ~= recdepth then + recdepth = depth + recprefix = rep(" .", depth) + end + local line + if pc >= 0 then + line = bcline(func, pc, recprefix) + if dumpmode.H then line = gsub(line, "[<>&]", html_escape) end + else + line = "0000 "..recprefix.." FUNCC \n" + end + if pc <= 0 then + out:write(sub(line, 1, -2), " ; ", fmtfunc(func), "\n") + else + out:write(line) + end + if pc >= 0 and band(funcbc(func, pc), 0xff) < 16 then -- ORDER BC + out:write(bcline(func, pc+1, recprefix)) -- Write JMP for cond. + end +end + +------------------------------------------------------------------------------ + +local gpr64 = jit.arch:match("64") +local fprmips32 = jit.arch == "mips" or jit.arch == "mipsel" + +-- Dump taken trace exits. +local function dump_texit(tr, ex, ngpr, nfpr, ...) + out:write("---- TRACE ", tr, " exit ", ex, "\n") + if dumpmode.X then + local regs = {...} + if gpr64 then + for i=1,ngpr do + out:write(format(" %016x", regs[i])) + if i % 4 == 0 then out:write("\n") end + end + else + for i=1,ngpr do + out:write(" ", tohex(regs[i])) + if i % 8 == 0 then out:write("\n") end + end + end + if fprmips32 then + for i=1,nfpr,2 do + out:write(format(" %+17.14g", regs[ngpr+i])) + if i % 8 == 7 then out:write("\n") end + end + else + for i=1,nfpr do + out:write(format(" %+17.14g", regs[ngpr+i])) + if i % 4 == 0 then out:write("\n") end + end + end + end +end + +------------------------------------------------------------------------------ + +-- Detach dump handlers. +local function dumpoff() + if active then + active = false + jit.attach(dump_texit) + jit.attach(dump_record) + jit.attach(dump_trace) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach dump handlers. +local function dumpon(opt, outfile) + if active then dumpoff() end + + local term = os.getenv("TERM") + local colormode = (term and term:match("color") or os.getenv("COLORTERM")) and "A" or "T" + if opt then + opt = gsub(opt, "[TAH]", function(mode) colormode = mode; return ""; end) + end + + local m = { t=true, b=true, i=true, m=true, } + if opt and opt ~= "" then + local o = sub(opt, 1, 1) + if o ~= "+" and o ~= "-" then m = {} end + for i=1,#opt do m[sub(opt, i, i)] = (o ~= "-") end + end + dumpmode = m + + if m.t or m.b or m.i or m.s or m.m then + jit.attach(dump_trace, "trace") + end + if m.b then + jit.attach(dump_record, "record") + if not bcline then bcline = require("jit.bc").line end + end + if m.x or m.X then + jit.attach(dump_texit, "texit") + end + + if not outfile then outfile = os.getenv("LUAJIT_DUMPFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stdout + end + + m[colormode] = true + if colormode == "A" then + colorize = colorize_ansi + irtype = irtype_ansi + elseif colormode == "H" then + colorize = colorize_html + irtype = irtype_html + out:write(header_html) + else + colorize = colorize_text + irtype = irtype_text + end + + active = true +end + +-- Public module functions. +return { + on = dumpon, + off = dumpoff, + start = dumpon -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/p.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/p.lua new file mode 100644 index 00000000..c9ec1d8b --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/p.lua @@ -0,0 +1,312 @@ +---------------------------------------------------------------------------- +-- LuaJIT profiler. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module is a simple command line interface to the built-in +-- low-overhead profiler of LuaJIT. +-- +-- The lower-level API of the profiler is accessible via the "jit.profile" +-- module or the luaJIT_profile_* C API. +-- +-- Example usage: +-- +-- luajit -jp myapp.lua +-- luajit -jp=s myapp.lua +-- luajit -jp=-s myapp.lua +-- luajit -jp=vl myapp.lua +-- luajit -jp=G,profile.txt myapp.lua +-- +-- The following dump features are available: +-- +-- f Stack dump: function name, otherwise module:line. Default mode. +-- F Stack dump: ditto, but always prepend module. +-- l Stack dump: module:line. +-- stack dump depth (callee < caller). Default: 1. +-- - Inverse stack dump depth (caller > callee). +-- s Split stack dump after first stack level. Implies abs(depth) >= 2. +-- p Show full path for module names. +-- v Show VM states. Can be combined with stack dumps, e.g. vf or fv. +-- z Show zones. Can be combined with stack dumps, e.g. zf or fz. +-- r Show raw sample counts. Default: show percentages. +-- a Annotate excerpts from source code files. +-- A Annotate complete source code files. +-- G Produce raw output suitable for graphical tools (e.g. flame graphs). +-- m Minimum sample percentage to be shown. Default: 3. +-- i Sampling interval in milliseconds. Default: 10. +-- +---------------------------------------------------------------------------- + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local profile = require("jit.profile") +local vmdef = require("jit.vmdef") +local math = math +local pairs, ipairs, tonumber, floor = pairs, ipairs, tonumber, math.floor +local sort, format = table.sort, string.format +local stdout = io.stdout +local zone -- Load jit.zone module on demand. + +-- Output file handle. +local out + +------------------------------------------------------------------------------ + +local prof_ud +local prof_states, prof_split, prof_min, prof_raw, prof_fmt, prof_depth +local prof_ann, prof_count1, prof_count2, prof_samples + +local map_vmmode = { + N = "Compiled", + I = "Interpreted", + C = "C code", + G = "Garbage Collector", + J = "JIT Compiler", +} + +-- Profiler callback. +local function prof_cb(th, samples, vmmode) + prof_samples = prof_samples + samples + local key_stack, key_stack2, key_state + -- Collect keys for sample. + if prof_states then + if prof_states == "v" then + key_state = map_vmmode[vmmode] or vmmode + else + key_state = zone:get() or "(none)" + end + end + if prof_fmt then + key_stack = profile.dumpstack(th, prof_fmt, prof_depth) + key_stack = key_stack:gsub("%[builtin#(%d+)%]", function(x) + return vmdef.ffnames[tonumber(x)] + end) + if prof_split == 2 then + local k1, k2 = key_stack:match("(.-) [<>] (.*)") + if k2 then key_stack, key_stack2 = k1, k2 end + elseif prof_split == 3 then + key_stack2 = profile.dumpstack(th, "l", 1) + end + end + -- Order keys. + local k1, k2 + if prof_split == 1 then + if key_state then + k1 = key_state + if key_stack then k2 = key_stack end + end + elseif key_stack then + k1 = key_stack + if key_stack2 then k2 = key_stack2 elseif key_state then k2 = key_state end + end + -- Coalesce samples in one or two levels. + if k1 then + local t1 = prof_count1 + t1[k1] = (t1[k1] or 0) + samples + if k2 then + local t2 = prof_count2 + local t3 = t2[k1] + if not t3 then t3 = {}; t2[k1] = t3 end + t3[k2] = (t3[k2] or 0) + samples + end + end +end + +------------------------------------------------------------------------------ + +-- Show top N list. +local function prof_top(count1, count2, samples, indent) + local t, n = {}, 0 + for k in pairs(count1) do + n = n + 1 + t[n] = k + end + sort(t, function(a, b) return count1[a] > count1[b] end) + for i=1,n do + local k = t[i] + local v = count1[k] + local pct = floor(v*100/samples + 0.5) + if pct < prof_min then break end + if not prof_raw then + out:write(format("%s%2d%% %s\n", indent, pct, k)) + elseif prof_raw == "r" then + out:write(format("%s%5d %s\n", indent, v, k)) + else + out:write(format("%s %d\n", k, v)) + end + if count2 then + local r = count2[k] + if r then + prof_top(r, nil, v, (prof_split == 3 or prof_split == 1) and " -- " or + (prof_depth < 0 and " -> " or " <- ")) + end + end + end +end + +-- Annotate source code +local function prof_annotate(count1, samples) + local files = {} + local ms = 0 + for k, v in pairs(count1) do + local pct = floor(v*100/samples + 0.5) + ms = math.max(ms, v) + if pct >= prof_min then + local file, line = k:match("^(.*):(%d+)$") + if not file then file = k; line = 0 end + local fl = files[file] + if not fl then fl = {}; files[file] = fl; files[#files+1] = file end + line = tonumber(line) + fl[line] = prof_raw and v or pct + end + end + sort(files) + local fmtv, fmtn = " %3d%% | %s\n", " | %s\n" + if prof_raw then + local n = math.max(5, math.ceil(math.log10(ms))) + fmtv = "%"..n.."d | %s\n" + fmtn = (" "):rep(n).." | %s\n" + end + local ann = prof_ann + for _, file in ipairs(files) do + local f0 = file:byte() + if f0 == 40 or f0 == 91 then + out:write(format("\n====== %s ======\n[Cannot annotate non-file]\n", file)) + break + end + local fp, err = io.open(file) + if not fp then + out:write(format("====== ERROR: %s: %s\n", file, err)) + break + end + out:write(format("\n====== %s ======\n", file)) + local fl = files[file] + local n, show = 1, false + if ann ~= 0 then + for i=1,ann do + if fl[i] then show = true; out:write("@@ 1 @@\n"); break end + end + end + for line in fp:lines() do + if line:byte() == 27 then + out:write("[Cannot annotate bytecode file]\n") + break + end + local v = fl[n] + if ann ~= 0 then + local v2 = fl[n+ann] + if show then + if v2 then show = n+ann elseif v then show = n + elseif show+ann < n then show = false end + elseif v2 then + show = n+ann + out:write(format("@@ %d @@\n", n)) + end + if not show then goto next end + end + if v then + out:write(format(fmtv, v, line)) + else + out:write(format(fmtn, line)) + end + ::next:: + n = n + 1 + end + fp:close() + end +end + +------------------------------------------------------------------------------ + +-- Finish profiling and dump result. +local function prof_finish() + if prof_ud then + profile.stop() + local samples = prof_samples + if samples == 0 then + if prof_raw ~= true then out:write("[No samples collected]\n") end + return + end + if prof_ann then + prof_annotate(prof_count1, samples) + else + prof_top(prof_count1, prof_count2, samples, "") + end + prof_count1 = nil + prof_count2 = nil + prof_ud = nil + if out ~= stdout then out:close() end + end +end + +-- Start profiling. +local function prof_start(mode) + local interval = "" + mode = mode:gsub("i%d*", function(s) interval = s; return "" end) + prof_min = 3 + mode = mode:gsub("m(%d+)", function(s) prof_min = tonumber(s); return "" end) + prof_depth = 1 + mode = mode:gsub("%-?%d+", function(s) prof_depth = tonumber(s); return "" end) + local m = {} + for c in mode:gmatch(".") do m[c] = c end + prof_states = m.z or m.v + if prof_states == "z" then zone = require("jit.zone") end + local scope = m.l or m.f or m.F or (prof_states and "" or "f") + local flags = (m.p or "") + prof_raw = m.r + if m.s then + prof_split = 2 + if prof_depth == -1 or m["-"] then prof_depth = -2 + elseif prof_depth == 1 then prof_depth = 2 end + elseif mode:find("[fF].*l") then + scope = "l" + prof_split = 3 + else + prof_split = (scope == "" or mode:find("[zv].*[lfF]")) and 1 or 0 + end + prof_ann = m.A and 0 or (m.a and 3) + if prof_ann then + scope = "l" + prof_fmt = "pl" + prof_split = 0 + prof_depth = 1 + elseif m.G and scope ~= "" then + prof_fmt = flags..scope.."Z;" + prof_depth = -100 + prof_raw = true + prof_min = 0 + elseif scope == "" then + prof_fmt = false + else + local sc = prof_split == 3 and m.f or m.F or scope + prof_fmt = flags..sc..(prof_depth >= 0 and "Z < " or "Z > ") + end + prof_count1 = {} + prof_count2 = {} + prof_samples = 0 + profile.start(scope:lower()..interval, prof_cb) + prof_ud = newproxy(true) + getmetatable(prof_ud).__gc = prof_finish +end + +------------------------------------------------------------------------------ + +local function start(mode, outfile) + if not outfile then outfile = os.getenv("LUAJIT_PROFILEFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stdout + end + prof_start(mode or "f") +end + +-- Public module functions. +return { + start = start, -- For -j command line option. + stop = prof_finish +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/v.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/v.lua new file mode 100644 index 00000000..83589143 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/v.lua @@ -0,0 +1,170 @@ +---------------------------------------------------------------------------- +-- Verbose mode of the LuaJIT compiler. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module shows verbose information about the progress of the +-- JIT compiler. It prints one line for each generated trace. This module +-- is useful to see which code has been compiled or where the compiler +-- punts and falls back to the interpreter. +-- +-- Example usage: +-- +-- luajit -jv -e "for i=1,1000 do for j=1,1000 do end end" +-- luajit -jv=myapp.out myapp.lua +-- +-- Default output is to stderr. To redirect the output to a file, pass a +-- filename as an argument (use '-' for stdout) or set the environment +-- variable LUAJIT_VERBOSEFILE. The file is overwritten every time the +-- module is started. +-- +-- The output from the first example should look like this: +-- +-- [TRACE 1 (command line):1 loop] +-- [TRACE 2 (1/3) (command line):1 -> 1] +-- +-- The first number in each line is the internal trace number. Next are +-- the file name ('(command line)') and the line number (':1') where the +-- trace has started. Side traces also show the parent trace number and +-- the exit number where they are attached to in parentheses ('(1/3)'). +-- An arrow at the end shows where the trace links to ('-> 1'), unless +-- it loops to itself. +-- +-- In this case the inner loop gets hot and is traced first, generating +-- a root trace. Then the last exit from the 1st trace gets hot, too, +-- and triggers generation of the 2nd trace. The side trace follows the +-- path along the outer loop and *around* the inner loop, back to its +-- start, and then links to the 1st trace. Yes, this may seem unusual, +-- if you know how traditional compilers work. Trace compilers are full +-- of surprises like this -- have fun! :-) +-- +-- Aborted traces are shown like this: +-- +-- [TRACE --- foo.lua:44 -- leaving loop in root trace at foo:lua:50] +-- +-- Don't worry -- trace aborts are quite common, even in programs which +-- can be fully compiled. The compiler may retry several times until it +-- finds a suitable trace. +-- +-- Of course this doesn't work with features that are not-yet-implemented +-- (NYI error messages). The VM simply falls back to the interpreter. This +-- may not matter at all if the particular trace is not very high up in +-- the CPU usage profile. Oh, and the interpreter is quite fast, too. +-- +-- Also check out the -jdump module, which prints all the gory details. +-- +------------------------------------------------------------------------------ + +-- Cache some library functions and objects. +local jit = require("jit") +assert(jit.version_num == 20100, "LuaJIT core/library version mismatch") +local jutil = require("jit.util") +local vmdef = require("jit.vmdef") +local funcinfo, traceinfo = jutil.funcinfo, jutil.traceinfo +local type, format = type, string.format +local stdout, stderr = io.stdout, io.stderr + +-- Active flag and output file handle. +local active, out + +------------------------------------------------------------------------------ + +local startloc, startex + +local function fmtfunc(func, pc) + local fi = funcinfo(func, pc) + if fi.loc then + return fi.loc + elseif fi.ffid then + return vmdef.ffnames[fi.ffid] + elseif fi.addr then + return format("C:%x", fi.addr) + else + return "(?)" + end +end + +-- Format trace error message. +local function fmterr(err, info) + if type(err) == "number" then + if type(info) == "function" then info = fmtfunc(info) end + err = format(vmdef.traceerr[err], info) + end + return err +end + +-- Dump trace states. +local function dump_trace(what, tr, func, pc, otr, oex) + if what == "start" then + startloc = fmtfunc(func, pc) + startex = otr and "("..otr.."/"..(oex == -1 and "stitch" or oex)..") " or "" + else + if what == "abort" then + local loc = fmtfunc(func, pc) + if loc ~= startloc then + out:write(format("[TRACE --- %s%s -- %s at %s]\n", + startex, startloc, fmterr(otr, oex), loc)) + else + out:write(format("[TRACE --- %s%s -- %s]\n", + startex, startloc, fmterr(otr, oex))) + end + elseif what == "stop" then + local info = traceinfo(tr) + local link, ltype = info.link, info.linktype + if ltype == "interpreter" then + out:write(format("[TRACE %3s %s%s -- fallback to interpreter]\n", + tr, startex, startloc)) + elseif ltype == "stitch" then + out:write(format("[TRACE %3s %s%s %s %s]\n", + tr, startex, startloc, ltype, fmtfunc(func, pc))) + elseif link == tr or link == 0 then + out:write(format("[TRACE %3s %s%s %s]\n", + tr, startex, startloc, ltype)) + elseif ltype == "root" then + out:write(format("[TRACE %3s %s%s -> %d]\n", + tr, startex, startloc, link)) + else + out:write(format("[TRACE %3s %s%s -> %d %s]\n", + tr, startex, startloc, link, ltype)) + end + else + out:write(format("[TRACE %s]\n", what)) + end + out:flush() + end +end + +------------------------------------------------------------------------------ + +-- Detach dump handlers. +local function dumpoff() + if active then + active = false + jit.attach(dump_trace) + if out and out ~= stdout and out ~= stderr then out:close() end + out = nil + end +end + +-- Open the output file and attach dump handlers. +local function dumpon(outfile) + if active then dumpoff() end + if not outfile then outfile = os.getenv("LUAJIT_VERBOSEFILE") end + if outfile then + out = outfile == "-" and stdout or assert(io.open(outfile, "w")) + else + out = stderr + end + jit.attach(dump_trace, "trace") + active = true +end + +-- Public module functions. +return { + on = dumpon, + off = dumpoff, + start = dumpon -- For -j command line option. +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/vmdef.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/vmdef.lua new file mode 100644 index 00000000..68ff7ba2 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/vmdef.lua @@ -0,0 +1,395 @@ +-- This is a generated file. DO NOT EDIT! + +return { + +bcnames = "ISLT ISGE ISLE ISGT ISEQV ISNEV ISEQS ISNES ISEQN ISNEN ISEQP ISNEP ISTC ISFC IST ISF ISTYPEISNUM MOV NOT UNM LEN ADDVN SUBVN MULVN DIVVN MODVN ADDNV SUBNV MULNV DIVNV MODNV ADDVV SUBVV MULVV DIVVV MODVV POW CAT KSTR KCDATAKSHORTKNUM KPRI KNIL UGET USETV USETS USETN USETP UCLO FNEW TNEW TDUP GGET GSET TGETV TGETS TGETB TGETR TSETV TSETS TSETB TSETM TSETR CALLM CALL CALLMTCALLT ITERC ITERN VARG ISNEXTRETM RET RET0 RET1 FORI JFORI FORL IFORL JFORL ITERL IITERLJITERLLOOP ILOOP JLOOP JMP FUNCF IFUNCFJFUNCFFUNCV IFUNCVJFUNCVFUNCC FUNCCW", + +irnames = "LT GE LE GT ULT UGE ULE UGT EQ NE ABC RETF NOP BASE PVAL GCSTEPHIOP LOOP USE PHI RENAMEPROF KPRI KINT KGC KPTR KKPTR KNULL KNUM KINT64KSLOT BNOT BSWAP BAND BOR BXOR BSHL BSHR BSAR BROL BROR ADD SUB MUL DIV MOD POW NEG ABS LDEXP MIN MAX FPMATHADDOV SUBOV MULOV AREF HREFK HREF NEWREFUREFO UREFC FREF TMPREFSTRREFLREF ALOAD HLOAD ULOAD FLOAD XLOAD SLOAD VLOAD ALEN ASTOREHSTOREUSTOREFSTOREXSTORESNEW XSNEW TNEW TDUP CNEW CNEWI BUFHDRBUFPUTBUFSTRTBAR OBAR XBAR CONV TOBIT TOSTR STRTO CALLN CALLA CALLL CALLS CALLXSCARG ", + +irfpm = { [0]="floor", "ceil", "trunc", "sqrt", "log", "log2", "other", }, + +irfield = { [0]="str.len", "func.env", "func.pc", "func.ffid", "thread.env", "tab.meta", "tab.array", "tab.node", "tab.asize", "tab.hmask", "tab.nomm", "udata.meta", "udata.udtype", "udata.file", "sbuf.w", "sbuf.e", "sbuf.b", "sbuf.l", "sbuf.ref", "sbuf.r", "cdata.ctypeid", "cdata.ptr", "cdata.int", "cdata.int64", "cdata.int64_4", }, + +ircall = { +[0]="lj_str_cmp", +"lj_str_find", +"lj_str_new", +"lj_strscan_num", +"lj_strfmt_int", +"lj_strfmt_num", +"lj_strfmt_char", +"lj_strfmt_putint", +"lj_strfmt_putnum", +"lj_strfmt_putquoted", +"lj_strfmt_putfxint", +"lj_strfmt_putfnum_int", +"lj_strfmt_putfnum_uint", +"lj_strfmt_putfnum", +"lj_strfmt_putfstr", +"lj_strfmt_putfchar", +"lj_buf_putmem", +"lj_buf_putstr", +"lj_buf_putchar", +"lj_buf_putstr_reverse", +"lj_buf_putstr_lower", +"lj_buf_putstr_upper", +"lj_buf_putstr_rep", +"lj_buf_puttab", +"lj_bufx_set", +"lj_bufx_more", +"lj_serialize_put", +"lj_serialize_get", +"lj_serialize_encode", +"lj_serialize_decode", +"lj_buf_tostr", +"lj_tab_new_ah", +"lj_tab_new1", +"lj_tab_dup", +"lj_tab_clear", +"lj_tab_newkey", +"lj_tab_keyindex", +"lj_vm_next", +"lj_tab_len", +"lj_tab_len_hint", +"lj_gc_step_jit", +"lj_gc_barrieruv", +"lj_mem_newgco", +"lj_prng_u64d", +"lj_vm_modi", +"log10", +"exp", +"sin", +"cos", +"tan", +"asin", +"acos", +"atan", +"sinh", +"cosh", +"tanh", +"fputc", +"fwrite", +"fflush", +"lj_vm_floor", +"lj_vm_ceil", +"lj_vm_trunc", +"sqrt", +"log", +"lj_vm_log2", +"lj_vm_powi", +"pow", +"atan2", +"ldexp", +"lj_vm_tobit", +"softfp_add", +"softfp_sub", +"softfp_mul", +"softfp_div", +"softfp_cmp", +"softfp_i2d", +"softfp_d2i", +"lj_vm_sfmin", +"lj_vm_sfmax", +"lj_vm_tointg", +"softfp_ui2d", +"softfp_f2d", +"softfp_d2ui", +"softfp_d2f", +"softfp_i2f", +"softfp_ui2f", +"softfp_f2i", +"softfp_f2ui", +"fp64_l2d", +"fp64_ul2d", +"fp64_l2f", +"fp64_ul2f", +"fp64_d2l", +"fp64_d2ul", +"fp64_f2l", +"fp64_f2ul", +"lj_carith_divi64", +"lj_carith_divu64", +"lj_carith_modi64", +"lj_carith_modu64", +"lj_carith_powi64", +"lj_carith_powu64", +"lj_cdata_newv", +"lj_cdata_setfin", +"strlen", +"memcpy", +"memset", +"lj_vm_errno", +"lj_carith_mul64", +"lj_carith_shl64", +"lj_carith_shr64", +"lj_carith_sar64", +"lj_carith_rol64", +"lj_carith_ror64", +}, + +traceerr = { +[0]="error thrown or hook called during recording", +"trace too short", +"trace too long", +"trace too deep", +"too many snapshots", +"blacklisted", +"retry recording", +"NYI: bytecode %d", +"leaving loop in root trace", +"inner loop in root trace", +"loop unroll limit reached", +"bad argument type", +"JIT compilation disabled for function", +"call unroll limit reached", +"down-recursion, restarting", +"NYI: unsupported variant of FastFunc %s", +"NYI: return to lower frame", +"store with nil or NaN key", +"missing metamethod", +"looping index lookup", +"NYI: mixed sparse/dense table", +"symbol not in cache", +"NYI: unsupported C type conversion", +"NYI: unsupported C function type", +"guard would always fail", +"too many PHIs", +"persistent type instability", +"failed to allocate mcode memory", +"machine code too long", +"hit mcode limit (retrying)", +"too many spill slots", +"inconsistent register allocation", +"NYI: cannot assemble IR instruction %d", +"NYI: PHI shuffling too complex", +"NYI: register coalescing too complex", +}, + +ffnames = { +[0]="Lua", +"C", +"assert", +"type", +"next", +"pairs", +"ipairs_aux", +"ipairs", +"getmetatable", +"setmetatable", +"getfenv", +"setfenv", +"rawget", +"rawset", +"rawequal", +"unpack", +"select", +"tonumber", +"tostring", +"error", +"pcall", +"xpcall", +"loadfile", +"load", +"loadstring", +"dofile", +"gcinfo", +"collectgarbage", +"newproxy", +"print", +"coroutine.status", +"coroutine.running", +"coroutine.isyieldable", +"coroutine.create", +"coroutine.yield", +"coroutine.resume", +"coroutine.wrap_aux", +"coroutine.wrap", +"math.abs", +"math.floor", +"math.ceil", +"math.sqrt", +"math.log10", +"math.exp", +"math.sin", +"math.cos", +"math.tan", +"math.asin", +"math.acos", +"math.atan", +"math.sinh", +"math.cosh", +"math.tanh", +"math.frexp", +"math.modf", +"math.log", +"math.atan2", +"math.pow", +"math.fmod", +"math.ldexp", +"math.min", +"math.max", +"math.random", +"math.randomseed", +"bit.tobit", +"bit.bnot", +"bit.bswap", +"bit.lshift", +"bit.rshift", +"bit.arshift", +"bit.rol", +"bit.ror", +"bit.band", +"bit.bor", +"bit.bxor", +"bit.tohex", +"string.byte", +"string.char", +"string.sub", +"string.rep", +"string.reverse", +"string.lower", +"string.upper", +"string.dump", +"string.find", +"string.match", +"string.gmatch_aux", +"string.gmatch", +"string.gsub", +"string.format", +"table.maxn", +"table.insert", +"table.concat", +"table.sort", +"table.new", +"table.clear", +"io.method.close", +"io.method.read", +"io.method.write", +"io.method.flush", +"io.method.seek", +"io.method.setvbuf", +"io.method.lines", +"io.method.__gc", +"io.method.__tostring", +"io.open", +"io.popen", +"io.tmpfile", +"io.close", +"io.read", +"io.write", +"io.flush", +"io.input", +"io.output", +"io.lines", +"io.type", +"os.execute", +"os.remove", +"os.rename", +"os.tmpname", +"os.getenv", +"os.exit", +"os.clock", +"os.date", +"os.time", +"os.difftime", +"os.setlocale", +"debug.getregistry", +"debug.getmetatable", +"debug.setmetatable", +"debug.getfenv", +"debug.setfenv", +"debug.getinfo", +"debug.getlocal", +"debug.setlocal", +"debug.getupvalue", +"debug.setupvalue", +"debug.upvalueid", +"debug.upvaluejoin", +"debug.sethook", +"debug.gethook", +"debug.debug", +"debug.traceback", +"jit.on", +"jit.off", +"jit.flush", +"jit.status", +"jit.security", +"jit.attach", +"jit.util.funcinfo", +"jit.util.funcbc", +"jit.util.funck", +"jit.util.funcuvname", +"jit.util.traceinfo", +"jit.util.traceir", +"jit.util.tracek", +"jit.util.tracesnap", +"jit.util.tracemc", +"jit.util.traceexitstub", +"jit.util.ircalladdr", +"jit.opt.start", +"jit.profile.start", +"jit.profile.stop", +"jit.profile.dumpstack", +"ffi.meta.__index", +"ffi.meta.__newindex", +"ffi.meta.__eq", +"ffi.meta.__len", +"ffi.meta.__lt", +"ffi.meta.__le", +"ffi.meta.__concat", +"ffi.meta.__call", +"ffi.meta.__add", +"ffi.meta.__sub", +"ffi.meta.__mul", +"ffi.meta.__div", +"ffi.meta.__mod", +"ffi.meta.__pow", +"ffi.meta.__unm", +"ffi.meta.__tostring", +"ffi.meta.__pairs", +"ffi.meta.__ipairs", +"ffi.clib.__index", +"ffi.clib.__newindex", +"ffi.clib.__gc", +"ffi.callback.free", +"ffi.callback.set", +"ffi.cdef", +"ffi.new", +"ffi.cast", +"ffi.typeof", +"ffi.typeinfo", +"ffi.istype", +"ffi.sizeof", +"ffi.alignof", +"ffi.offsetof", +"ffi.errno", +"ffi.string", +"ffi.copy", +"ffi.fill", +"ffi.abi", +"ffi.metatype", +"ffi.gc", +"ffi.load", +"buffer.method.free", +"buffer.method.reset", +"buffer.method.skip", +"buffer.method.set", +"buffer.method.put", +"buffer.method.putf", +"buffer.method.get", +"buffer.method.putcdata", +"buffer.method.reserve", +"buffer.method.commit", +"buffer.method.ref", +"buffer.method.encode", +"buffer.method.decode", +"buffer.method.__gc", +"buffer.method.__tostring", +"buffer.method.__len", +"buffer.new", +"buffer.encode", +"buffer.decode", +}, + +} + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/jit/zone.lua b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/zone.lua new file mode 100644 index 00000000..94357854 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/android/x86_64/jit/zone.lua @@ -0,0 +1,45 @@ +---------------------------------------------------------------------------- +-- LuaJIT profiler zones. +-- +-- Copyright (C) 2005-2021 Mike Pall. All rights reserved. +-- Released under the MIT license. See Copyright Notice in luajit.h +---------------------------------------------------------------------------- +-- +-- This module implements a simple hierarchical zone model. +-- +-- Example usage: +-- +-- local zone = require("jit.zone") +-- zone("AI") +-- ... +-- zone("A*") +-- ... +-- print(zone:get()) --> "A*" +-- ... +-- zone() +-- ... +-- print(zone:get()) --> "AI" +-- ... +-- zone() +-- +---------------------------------------------------------------------------- + +local remove = table.remove + +return setmetatable({ + flush = function(t) + for i=#t,1,-1 do t[i] = nil end + end, + get = function(t) + return t[#t] + end +}, { + __call = function(t, zone) + if zone then + t[#t+1] = zone + else + return (assert(remove(t), "empty zone stack")) + end + end +}) + diff --git a/love/src/jni/LuaJIT-2.1/android/x86_64/libluajit.a b/love/src/jni/LuaJIT-2.1/android/x86_64/libluajit.a index e2089063..3275e35c 100644 Binary files a/love/src/jni/LuaJIT-2.1/android/x86_64/libluajit.a and b/love/src/jni/LuaJIT-2.1/android/x86_64/libluajit.a differ diff --git a/love/src/jni/LuaJIT-2.1/build_instructions.bat b/love/src/jni/LuaJIT-2.1/build_instructions.bat index 8f04ba16..c3003340 100644 --- a/love/src/jni/LuaJIT-2.1/build_instructions.bat +++ b/love/src/jni/LuaJIT-2.1/build_instructions.bat @@ -28,6 +28,8 @@ wsl make HOST_LUA=luajit.exe "HOST_CC=clang.exe -m32" HOST_CFLAGS=-D_CRT_SECURE_ if not "%ERRORLEVEL%" == "0" goto :error copy src\libluajit.a android\armeabi-v7a\libluajit.a if not "%ERRORLEVEL%" == "0" goto :error +xcopy src\jit android\armeabi-v7a\jit /I +del android\armeabi-v7a\jit\.gitignore :x86 rem x86 @@ -48,6 +50,8 @@ wsl make HOST_LUA=luajit.exe "HOST_CC=clang.exe %4" HOST_CFLAGS=-D_CRT_SECURE_NO if not "%ERRORLEVEL%" == "0" goto :error copy src\libluajit.a android\%1\libluajit.a if not "%ERRORLEVEL%" == "0" goto :error +xcopy src\jit android\%1\jit /I +del android\%1\jit\.gitignore goto :done :error diff --git a/love/src/jni/LuaJIT-2.1/doc/contact.html b/love/src/jni/LuaJIT-2.1/doc/contact.html index 0c775f98..c253a08b 100644 --- a/love/src/jni/LuaJIT-2.1/doc/contact.html +++ b/love/src/jni/LuaJIT-2.1/doc/contact.html @@ -1,8 +1,8 @@ - + Contact - + @@ -37,6 +37,8 @@ FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/ext_buffer.html b/love/src/jni/LuaJIT-2.1/doc/ext_buffer.html new file mode 100644 index 00000000..63c2efe3 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/doc/ext_buffer.html @@ -0,0 +1,693 @@ + + + +String Buffer Library + + + + + + + + +
    +Lua +
    + + +
    +

    +The string buffer library allows high-performance manipulation of +string-like data. +

    +

    +Unlike Lua strings, which are constants, string buffers are +mutable sequences of 8-bit (binary-transparent) characters. Data +can be stored, formatted and encoded into a string buffer and later +converted, extracted or decoded. +

    +

    +The convenient string buffer API simplifies common string manipulation +tasks, that would otherwise require creating many intermediate strings. +String buffers improve performance by eliminating redundant memory +copies, object creation, string interning and garbage collection +overhead. In conjunction with the FFI library, they allow zero-copy +operations. +

    +

    +The string buffer libary also includes a high-performance +serializer for Lua objects. +

    + +

    Work in Progress

    +

    +This library is a work in progress. More +functionality will be added soon. +

    + +

    Using the String Buffer Library

    +

    +The string buffer library is built into LuaJIT by default, but it's not +loaded by default. Add this to the start of every Lua file that needs +one of its functions: +

    +
    +local buffer = require("string.buffer")
+
    +

    +The convention for the syntax shown on this page is that buffer +refers to the buffer library and buf refers to an individual +buffer object. +

    +

    +Please note the difference between a Lua function call, e.g. +buffer.new() (with a dot) and a Lua method call, e.g. +buf:reset() (with a colon). +

    + +

    Buffer Objects

    +

    +A buffer object is a garbage-collected Lua object. After creation with +buffer.new(), it can (and should) be reused for many operations. +When the last reference to a buffer object is gone, it will eventually +be freed by the garbage collector, along with the allocated buffer +space. +

    +

    +Buffers operate like a FIFO (first-in first-out) data structure. Data +can be appended (written) to the end of the buffer and consumed (read) +from the front of the buffer. These operations may be freely mixed. +

    +

    +The buffer space that holds the characters is managed automatically +— it grows as needed and already consumed space is recycled. Use +buffer.new(size) and buf:free(), if you need more +control. +

    +

    +The maximum size of a single buffer is the same as the maximum size of a +Lua string, which is slightly below two gigabytes. For huge data sizes, +neither strings nor buffers are the right data structure — use the +FFI library to directly map memory or files up to the virtual memory +limit of your OS. +

    + +

    Buffer Method Overview

    +
      +
    • +The buf:put*()-like methods append (write) characters to the +end of the buffer. +
      • +
    • +The buf:get*()-like methods consume (read) characters from the +front of the buffer. +
      • +
    • +Other methods, like buf:tostring() only read the buffer +contents, but don't change the buffer. +
      • +
    • +The buf:set() method allows zero-copy consumption of a string +or an FFI cdata object as a buffer. +
      • +
    • +The FFI-specific methods allow zero-copy read/write-style operations or +modifying the buffer contents in-place. Please check the +FFI caveats below, too. +
      • +
    • +Methods that don't need to return anything specific, return the buffer +object itself as a convenience. This allows method chaining, e.g.: +buf:reset():encode(obj) or buf:skip(len):get() +
      • +
    + +

    Buffer Creation and Management

    + +

    local buf = buffer.new([size [,options]])
    +local buf = buffer.new([options])

    +

    +Creates a new buffer object. +

    +

    +The optional size argument ensures a minimum initial buffer +size. This is strictly an optimization when the required buffer size is +known beforehand. The buffer space will grow as needed, in any case. +

    +

    +The optional table options sets various +serialization options. +

    + +

    buf = buf:reset()

    +

    +Reset (empty) the buffer. The allocated buffer space is not freed and +may be reused. +

    + +

    buf = buf:free()

    +

    +The buffer space of the buffer object is freed. The object itself +remains intact, empty and may be reused. +

    +

    +Note: you normally don't need to use this method. The garbage collector +automatically frees the buffer space, when the buffer object is +collected. Use this method, if you need to free the associated memory +immediately. +

    + +

    Buffer Writers

    + +

    buf = buf:put([str|num|obj] [,…])

    +

    +Appends a string str, a number num or any object +obj with a __tostring metamethod to the buffer. +Multiple arguments are appended in the given order. +

    +

    +Appending a buffer to a buffer is possible and short-circuited +internally. But it still involves a copy. Better combine the buffer +writes to use a single buffer. +

    + +

    buf = buf:putf(format, …)

    +

    +Appends the formatted arguments to the buffer. The format +string supports the same options as string.format(). +

    + +

    buf = buf:putcdata(cdata, len)FFI

    +

    +Appends the given len number of bytes from the memory pointed +to by the FFI cdata object to the buffer. The object needs to +be convertible to a (constant) pointer. +

    + +

    buf = buf:set(str)
    +buf = buf:set(cdata, len)    FFI

    +

    +This method allows zero-copy consumption of a string or an FFI cdata +object as a buffer. It stores a reference to the passed string +str or the FFI cdata object in the buffer. Any buffer +space originally allocated is freed. This is not an append +operation, unlike the buf:put*() methods. +

    +

    +After calling this method, the buffer behaves as if +buf:free():put(str) or buf:free():put(cdata, len) +had been called. However, the data is only referenced and not copied, as +long as the buffer is only consumed. +

    +

    +In case the buffer is written to later on, the referenced data is copied +and the object reference is removed (copy-on-write semantics). +

    +

    +The stored reference is an anchor for the garbage collector and keeps the +originally passed string or FFI cdata object alive. +

    + +

    ptr, len = buf:reserve(size)FFI
    +buf = buf:commit(used)FFI

    +

    +The reserve method reserves at least size bytes of +write space in the buffer. It returns an uint8_t * FFI +cdata pointer ptr that points to this space. +

    +

    +The available length in bytes is returned in len. This is at +least size bytes, but may be more to facilitate efficient +buffer growth. You can either make use of the additional space or ignore +len and only use size bytes. +

    +

    +The commit method appends the used bytes of the +previously returned write space to the buffer data. +

    +

    +This pair of methods allows zero-copy use of C read-style APIs: +

    +
    +local MIN_SIZE = 65536
+repeat
+  local ptr, len = buf:reserve(MIN_SIZE)
+  local n = C.read(fd, ptr, len)
+  if n == 0 then break end -- EOF.
+  if n < 0 then error("read error") end
+  buf:commit(n)
+until false
+
    +

    +The reserved write space is not initialized. At least the +used bytes must be written to before calling the +commit method. There's no need to call the commit +method, if nothing is added to the buffer (e.g. on error). +

    + +

    Buffer Readers

    + +

    len = #buf

    +

    +Returns the current length of the buffer data in bytes. +

    + +

    res = str|num|buf .. str|num|buf […]

    +

    +The Lua concatenation operator .. also accepts buffers, just +like strings or numbers. It always returns a string and not a buffer. +

    +

    +Note that although this is supported for convenience, this thwarts one +of the main reasons to use buffers, which is to avoid string +allocations. Rewrite it with buf:put() and buf:get(). +

    +

    +Mixing this with unrelated objects that have a __concat +metamethod may not work, since these probably only expect strings. +

    + +

    buf = buf:skip(len)

    +

    +Skips (consumes) len bytes from the buffer up to the current +length of the buffer data. +

    + +

    str, … = buf:get([len|nil] [,…])

    +

    +Consumes the buffer data and returns one or more strings. If called +without arguments, the whole buffer data is consumed. If called with a +number, up to len bytes are consumed. A nil argument +consumes the remaining buffer space (this only makes sense as the last +argument). Multiple arguments consume the buffer data in the given +order. +

    +

    +Note: a zero length or no remaining buffer data returns an empty string +and not nil. +

    + +

    str = buf:tostring()
    +str = tostring(buf)

    +

    +Creates a string from the buffer data, but doesn't consume it. The +buffer remains unchanged. +

    +

    +Buffer objects also define a __tostring metamethod. This means +buffers can be passed to the global tostring() function and +many other functions that accept this in place of strings. The important +internal uses in functions like io.write() are short-circuited +to avoid the creation of an intermediate string object. +

    + +

    ptr, len = buf:ref()FFI

    +

    +Returns an uint8_t * FFI cdata pointer ptr that +points to the buffer data. The length of the buffer data in bytes is +returned in len. +

    +

    +The returned pointer can be directly passed to C functions that expect a +buffer and a length. You can also do bytewise reads +(local x = ptr[i]) or writes +(ptr[i] = 0x40) of the buffer data. +

    +

    +In conjunction with the skip method, this allows zero-copy use +of C write-style APIs: +

    +
    +repeat
+  local ptr, len = buf:ref()
+  if len == 0 then break end
+  local n = C.write(fd, ptr, len)
+  if n < 0 then error("write error") end
+  buf:skip(n)
+until n >= len
+
    +

    +Unlike Lua strings, buffer data is not implicitly +zero-terminated. It's not safe to pass ptr to C functions that +expect zero-terminated strings. If you're not using len, then +you're doing something wrong. +

    + +

    Serialization of Lua Objects

    +

    +The following functions and methods allow high-speed serialization +(encoding) of a Lua object into a string and decoding it back to a Lua +object. This allows convenient storage and transport of structured +data. +

    +

    +The encoded data is in an internal binary +format. The data can be stored in files, binary-transparent +databases or transmitted to other LuaJIT instances across threads, +processes or networks. +

    +

    +Encoding speed can reach up to 1 Gigabyte/second on a modern desktop- or +server-class system, even when serializing many small objects. Decoding +speed is mostly constrained by object creation cost. +

    +

    +The serializer handles most Lua types, common FFI number types and +nested structures. Functions, thread objects, other FFI cdata and full +userdata cannot be serialized (yet). +

    +

    +The encoder serializes nested structures as trees. Multiple references +to a single object will be stored separately and create distinct objects +after decoding. Circular references cause an error. +

    + +

    Serialization Functions and Methods

    + +

    str = buffer.encode(obj)
    +buf = buf:encode(obj)

    +

    +Serializes (encodes) the Lua object obj. The stand-alone +function returns a string str. The buffer method appends the +encoding to the buffer. +

    +

    +obj can be any of the supported Lua types — it doesn't +need to be a Lua table. +

    +

    +This function may throw an error when attempting to serialize +unsupported object types, circular references or deeply nested tables. +

    + +

    obj = buffer.decode(str)
    +obj = buf:decode()

    +

    +The stand-alone function de-serializes (decodes) the string +str, the buffer method de-serializes one object from the +buffer. Both return a Lua object obj. +

    +

    +The returned object may be any of the supported Lua types — +even nil. +

    +

    +This function may throw an error when fed with malformed or incomplete +encoded data. The stand-alone function throws when there's left-over +data after decoding a single top-level object. The buffer method leaves +any left-over data in the buffer. +

    + +

    Serialization Options

    +

    +The options table passed to buffer.new() may contain +the following members (all optional): +

    +
      +
    • +dict is a Lua table holding a dictionary of strings that +commonly occur as table keys of objects you are serializing. These keys +are compactly encoded as indexes during serialization. A well chosen +dictionary saves space and improves serialization performance. +
      • +
    • +metatable is a Lua table holding a dictionary of metatables +for the table objects you are serializing. +
      • +
    +

    +dict needs to be an array of strings and metatable needs +to be an array of tables. Both starting at index 1 and without holes (no +nil inbetween). The tables are anchored in the buffer object and +internally modified into a two-way index (don't do this yourself, just pass +a plain array). The tables must not be modified after they have been passed +to buffer.new(). +

    +

    +The dict and metatable tables used by the encoder and +decoder must be the same. Put the most common entries at the front. Extend +at the end to ensure backwards-compatibility — older encodings can +then still be read. You may also set some indexes to false to +explicitly drop backwards-compatibility. Old encodings that use these +indexes will throw an error when decoded. +

    +

    +Metatables that are not found in the metatable dictionary are +ignored when encoding. Decoding returns a table with a nil +metatable. +

    +

    +Note: parsing and preparation of the options table is somewhat +expensive. Create a buffer object only once and recycle it for multiple +uses. Avoid mixing encoder and decoder buffers, since the +buf:set() method frees the already allocated buffer space: +

    +
    +local options = {
+  dict = { "commonly", "used", "string", "keys" },
+}
+local buf_enc = buffer.new(options)
+local buf_dec = buffer.new(options)
+
+local function encode(obj)
+  return buf_enc:reset():encode(obj):get()
+end
+
+local function decode(str)
+  return buf_dec:set(str):decode()
+end
+
    + +

    Streaming Serialization

    +

    +In some contexts, it's desirable to do piecewise serialization of large +datasets, also known as streaming. +

    +

    +This serialization format can be safely concatenated and supports streaming. +Multiple encodings can simply be appended to a buffer and later decoded +individually: +

    +
    +local buf = buffer.new()
+buf:encode(obj1)
+buf:encode(obj2)
+local copy1 = buf:decode()
+local copy2 = buf:decode()
+
    +

    +Here's how to iterate over a stream: +

    +
    +while #buf ~= 0 do
+  local obj = buf:decode()
+  -- Do something with obj.
+end
+
    +

    +Since the serialization format doesn't prepend a length to its encoding, +network applications may need to transmit the length, too. +

    + +

    Serialization Format Specification

    +

    +This serialization format is designed for internal use by LuaJIT +applications. Serialized data is upwards-compatible and portable across +all supported LuaJIT platforms. +

    +

    +It's an 8-bit binary format and not human-readable. It uses e.g. +embedded zeroes and stores embedded Lua string objects unmodified, which +are 8-bit-clean, too. Encoded data can be safely concatenated for +streaming and later decoded one top-level object at a time. +

    +

    +The encoding is reasonably compact, but tuned for maximum performance, +not for minimum space usage. It compresses well with any of the common +byte-oriented data compression algorithms. +

    +

    +Although documented here for reference, this format is explicitly +not intended to be a 'public standard' for structured data +interchange across computer languages (like JSON or MessagePack). Please +do not use it as such. +

    +

    +The specification is given below as a context-free grammar with a +top-level object as the starting point. Alternatives are +separated by the | symbol and * indicates repeats. +Grouping is implicit or indicated by {…}. Terminals are +either plain hex numbers, encoded as bytes, or have a .format +suffix. +

    +
    +object    → nil | false | true
+          | null | lightud32 | lightud64
+          | int | num | tab | tab_mt
+          | int64 | uint64 | complex
+          | string
+
+nil       → 0x00
+false     → 0x01
+true      → 0x02
+
+null      → 0x03                            // NULL lightuserdata
+lightud32 → 0x04 data.I                   // 32 bit lightuserdata
+lightud64 → 0x05 data.L                   // 64 bit lightuserdata
+
+int       → 0x06 int.I                                 // int32_t
+num       → 0x07 double.L
+
+tab       → 0x08                                   // Empty table
+          | 0x09 h.U h*{object object}          // Key/value hash
+          | 0x0a a.U a*object                    // 0-based array
+          | 0x0b a.U a*object h.U h*{object object}      // Mixed
+          | 0x0c a.U (a-1)*object                // 1-based array
+          | 0x0d a.U (a-1)*object h.U h*{object object}  // Mixed
+tab_mt    → 0x0e (index-1).U tab          // Metatable dict entry
+
+int64     → 0x10 int.L                             // FFI int64_t
+uint64    → 0x11 uint.L                           // FFI uint64_t
+complex   → 0x12 re.L im.L                         // FFI complex
+
+string    → (0x20+len).U len*char.B
+          | 0x0f (index-1).U                 // String dict entry
+
+.B = 8 bit
+.I = 32 bit little-endian
+.L = 64 bit little-endian
+.U = prefix-encoded 32 bit unsigned number n:
+     0x00..0xdf   → n.B
+     0xe0..0x1fdf → (0xe0|(((n-0xe0)>>8)&0x1f)).B ((n-0xe0)&0xff).B
+   0x1fe0..       → 0xff n.I
+
    + +

    Error handling

    +

    +Many of the buffer methods can throw an error. Out-of-memory or usage +errors are best caught with an outer wrapper for larger parts of code. +There's not much one can do after that, anyway. +

    +

    +OTOH you may want to catch some errors individually. Buffer methods need +to receive the buffer object as the first argument. The Lua colon-syntax +obj:method() does that implicitly. But to wrap a method with +pcall(), the arguments need to be passed like this: +

    +
    +local ok, err = pcall(buf.encode, buf, obj)
+if not ok then
+  -- Handle error in err.
+end
+
    + +

    FFI caveats

    +

    +The string buffer library has been designed to work well together with +the FFI library. But due to the low-level nature of the FFI library, +some care needs to be taken: +

    +

    +First, please remember that FFI pointers are zero-indexed. The space +returned by buf:reserve() and buf:ref() starts at the +returned pointer and ends before len bytes after that. +

    +

    +I.e. the first valid index is ptr[0] and the last valid index +is ptr[len-1]. If the returned length is zero, there's no valid +index at all. The returned pointer may even be NULL. +

    +

    +The space pointed to by the returned pointer is only valid as long as +the buffer is not modified in any way (neither append, nor consume, nor +reset, etc.). The pointer is also not a GC anchor for the buffer object +itself. +

    +

    +Buffer data is only guaranteed to be byte-aligned. Casting the returned +pointer to a data type with higher alignment may cause unaligned +accesses. It depends on the CPU architecture whether this is allowed or +not (it's always OK on x86/x64 and mostly OK on other modern +architectures). +

    +

    +FFI pointers or references do not count as GC anchors for an underlying +object. E.g. an array allocated with ffi.new() is +anchored by buf:set(array, len), but not by +buf:set(array+offset, len). The addition of the offset +creates a new pointer, even when the offset is zero. In this case, you +need to make sure there's still a reference to the original array as +long as its contents are in use by the buffer. +

    +

    +Even though each LuaJIT VM instance is single-threaded (but you can +create multiple VMs), FFI data structures can be accessed concurrently. +Be careful when reading/writing FFI cdata from/to buffers to avoid +concurrent accesses or modifications. In particular, the memory +referenced by buf:set(cdata, len) must not be modified +while buffer readers are working on it. Shared, but read-only memory +mappings of files are OK, but only if the file does not change. +

    +
    +
    + + + diff --git a/love/src/jni/LuaJIT-2.1/doc/ext_c_api.html b/love/src/jni/LuaJIT-2.1/doc/ext_c_api.html index e204459a..9f1ad212 100644 --- a/love/src/jni/LuaJIT-2.1/doc/ext_c_api.html +++ b/love/src/jni/LuaJIT-2.1/doc/ext_c_api.html @@ -1,8 +1,8 @@ - + Lua/C API Extensions - + @@ -37,6 +37,8 @@ FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/ext_ffi.html b/love/src/jni/LuaJIT-2.1/doc/ext_ffi.html index 115a77fe..b934dc78 100644 --- a/love/src/jni/LuaJIT-2.1/doc/ext_ffi.html +++ b/love/src/jni/LuaJIT-2.1/doc/ext_ffi.html @@ -1,8 +1,8 @@ - + FFI Library - + @@ -37,6 +37,8 @@ FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/ext_ffi_api.html b/love/src/jni/LuaJIT-2.1/doc/ext_ffi_api.html index 65761aa7..061cc42a 100644 --- a/love/src/jni/LuaJIT-2.1/doc/ext_ffi_api.html +++ b/love/src/jni/LuaJIT-2.1/doc/ext_ffi_api.html @@ -1,8 +1,8 @@ - + ffi.* API Functions - + @@ -42,6 +42,8 @@ td.abiparam { font-weight: bold; width: 6em; } FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/ext_ffi_semantics.html b/love/src/jni/LuaJIT-2.1/doc/ext_ffi_semantics.html index 16f87739..fef39c32 100644 --- a/love/src/jni/LuaJIT-2.1/doc/ext_ffi_semantics.html +++ b/love/src/jni/LuaJIT-2.1/doc/ext_ffi_semantics.html @@ -1,8 +1,8 @@ - + FFI Semantics - + @@ -42,6 +42,8 @@ td.convop { font-style: italic; width: 40%; } FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/ext_ffi_tutorial.html b/love/src/jni/LuaJIT-2.1/doc/ext_ffi_tutorial.html index c7d453ae..ca71be4d 100644 --- a/love/src/jni/LuaJIT-2.1/doc/ext_ffi_tutorial.html +++ b/love/src/jni/LuaJIT-2.1/doc/ext_ffi_tutorial.html @@ -1,8 +1,8 @@ - + FFI Tutorial - + @@ -44,6 +44,8 @@ td.idiomlua b { font-weight: normal; color: #2142bf; } FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/ext_jit.html b/love/src/jni/LuaJIT-2.1/doc/ext_jit.html index 17362d6a..6dd54c70 100644 --- a/love/src/jni/LuaJIT-2.1/doc/ext_jit.html +++ b/love/src/jni/LuaJIT-2.1/doc/ext_jit.html @@ -1,8 +1,8 @@ - + jit.* Library - + @@ -37,6 +37,8 @@ FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/ext_profiler.html b/love/src/jni/LuaJIT-2.1/doc/ext_profiler.html index a2f9cab0..2783abdb 100644 --- a/love/src/jni/LuaJIT-2.1/doc/ext_profiler.html +++ b/love/src/jni/LuaJIT-2.1/doc/ext_profiler.html @@ -1,8 +1,8 @@ - + Profiler - + @@ -37,6 +37,8 @@ FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/extensions.html b/love/src/jni/LuaJIT-2.1/doc/extensions.html index d0c3ca7a..748c1793 100644 --- a/love/src/jni/LuaJIT-2.1/doc/extensions.html +++ b/love/src/jni/LuaJIT-2.1/doc/extensions.html @@ -1,8 +1,8 @@ - + Extensions - + @@ -54,6 +54,8 @@ td.excinterop { FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API @@ -392,29 +394,19 @@ the toolchain used to compile LuaJIT: Interoperability -POSIX/x64, DWARF2 unwinding -GCC 4.3+, Clang +External frame unwinding +GCC, Clang, MSVC Full -ARM -DLUAJIT_UNWIND_EXTERNAL -GCC, Clang -Full - - -Other platforms, DWARF2 unwinding +Internal frame unwinding + DWARF2 GCC, Clang Limited - -Windows/x64 -MSVC -Full - -Windows/x86 -Any -Full +Windows 64 bit +non-MSVC +Limited Other platforms @@ -435,7 +427,9 @@ the toolchain used to compile LuaJIT: on the C stack. The contents of the C++ exception object pass through unmodified.
  • Lua errors can be caught on the C++ side with catch(...). -The corresponding Lua error message can be retrieved from the Lua stack.
    • +The corresponding Lua error message can be retrieved from the Lua stack.
    +For MSVC for Windows 64 bit this requires compilation of your C++ code +with /EHa.
  • Throwing Lua errors across C++ frames is safe. C++ destructors will be called.
    • diff --git a/love/src/jni/LuaJIT-2.1/doc/faq.html b/love/src/jni/LuaJIT-2.1/doc/faq.html index 891a3402..1b7cb371 100644 --- a/love/src/jni/LuaJIT-2.1/doc/faq.html +++ b/love/src/jni/LuaJIT-2.1/doc/faq.html @@ -1,8 +1,8 @@ - + Frequently Asked Questions (FAQ) - + @@ -40,6 +40,8 @@ dd { margin-left: 1.5em; } FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API @@ -74,7 +76,7 @@ about LuaJIT.
    • and links to books and papers about Lua.
  • The community-managed » Lua Wiki has information about diverse topics.
    • - +
    @@ -119,8 +121,7 @@ Consider testing your application with older versions, too.
    Similarly, the Borland/Delphi runtime modifies the FPU control word and enables FP exceptions. Of course this violates the Windows ABI, too. -Please check the Delphi docs for the Set8087CW method. - +Please check the Delphi docs for the Set8087CW method.
    @@ -133,6 +134,19 @@ You have to press Ctrl-C twice to get stop your program. That's similar to when it's stuck running inside a C function under the Lua interpreter.
    +
    +
    Q: Table iteration with pairs() does not result in the same order?
    +
    The order of table iteration is explicitly undefined by +the Lua language standard.
    +Different Lua implementations or versions may use different orders for +otherwise identical tables. Different ways of constructing a table may +result in different orders, too.
    +Due to improved VM security, LuaJIT 2.1 may even use a different order +on separate VM invocations or when string keys are newly interned.

    +If your program relies on a deterministic order, it has a bug. Rewrite it, +so it doesn't rely on the key order. Or sort the table keys, if you must.
    +
    +
    Q: Can Lua code be safely sandboxed?
    diff --git a/love/src/jni/LuaJIT-2.1/doc/install.html b/love/src/jni/LuaJIT-2.1/doc/install.html index b7217537..e4af9dde 100644 --- a/love/src/jni/LuaJIT-2.1/doc/install.html +++ b/love/src/jni/LuaJIT-2.1/doc/install.html @@ -1,8 +1,8 @@ - + Installation - + @@ -65,6 +65,8 @@ td.compatno { FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API @@ -186,14 +188,18 @@ Debian/Ubuntu, install libc6-dev with the package manager.

    The recommended way to fetch the latest version is to do a pull from -the git repository. Alternatively download the latest source package of -LuaJIT (pick the .tar.gz). Move it to a directory of your choice, -open a terminal window and change to this directory. Now unpack the archive -and change to the newly created directory: +the git repository. +

    +

    +Alternatively download the latest source package of LuaJIT (pick the .tar.gz). +Move it to a directory of your choice, open a terminal window and change +to this directory. Now unpack the archive and change to the newly created +directory (replace XX.YY.ZZ with the version you downloaded): 

    -tar zxf LuaJIT-2.1.0-beta3.tar.gz
-cd LuaJIT-2.1.0-beta3
    +tar zxf LuaJIT-XX.YY.ZZ.tar.gz +cd LuaJIT-XX.YY.ZZ +

    Building LuaJIT

    The supplied Makefiles try to auto-detect the settings needed for your @@ -217,9 +223,12 @@ You can add an extra prefix to the search paths by appending the make PREFIX=/home/myself/lj2

    -Note for macOS: you must set the MACOSX_DEPLOYMENT_TARGET -environment variable to a value supported by your toolchain. +Note for macOS: you must set the MACOSX_DEPLOYMENT_TARGET +environment variable to a value supported by your toolchain:

    +
    +MACOSX_DEPLOYMENT_TARGET=XX.YY make
+

    Installing LuaJIT

    The top-level Makefile installs LuaJIT by default under diff --git a/love/src/jni/LuaJIT-2.1/doc/luajit.html b/love/src/jni/LuaJIT-2.1/doc/luajit.html index 0d3bf31c..a25267a6 100644 --- a/love/src/jni/LuaJIT-2.1/doc/luajit.html +++ b/love/src/jni/LuaJIT-2.1/doc/luajit.html @@ -1,8 +1,8 @@ - + LuaJIT - + @@ -122,6 +122,8 @@ table.feature small { FFI Semantics

  • +String Buffers +
  • jit.* Library
  • Lua/C API diff --git a/love/src/jni/LuaJIT-2.1/doc/running.html b/love/src/jni/LuaJIT-2.1/doc/running.html index 6dd3f196..b55b8439 100644 --- a/love/src/jni/LuaJIT-2.1/doc/running.html +++ b/love/src/jni/LuaJIT-2.1/doc/running.html @@ -1,8 +1,8 @@ - + Running LuaJIT - + @@ -59,6 +59,8 @@ td.param_default { FFI Semantics
  • +String Buffers +
  • jit.* Library
  • Lua/C API @@ -183,8 +185,8 @@ itself. For a description of their options and output format, please read the comment block at the start of their source. They can be found in the lib directory of the source distribution or installed under the jit directory. By default -this is /usr/local/share/luajit-2.1.0-beta3/jit on POSIX -systems. +this is /usr/local/share/luajit-XX.YY.ZZ>/jit on POSIX +systems (replace XX.YY.ZZ by the installed version).

    -O[level]
    diff --git a/love/src/jni/LuaJIT-2.1/doc/status.html b/love/src/jni/LuaJIT-2.1/doc/status.html index f607f5cb..1d3ba984 100644 --- a/love/src/jni/LuaJIT-2.1/doc/status.html +++ b/love/src/jni/LuaJIT-2.1/doc/status.html @@ -1,8 +1,8 @@ - + Status - + @@ -40,6 +40,8 @@ ul li { padding-bottom: 0.3em; } FFI Semantics

  • +String Buffers +
  • jit.* Library
  • Lua/C API @@ -90,12 +92,6 @@ The Lua debug API is missing a couple of features (return hooks for non-Lua functions) and shows slightly different behavior in LuaJIT (no per-coroutine hooks, no tail call counting).
    • -
  • -Currently some out-of-memory errors from on-trace code are not -handled correctly. The error may fall through an on-trace -pcall or it may be passed on to the function set with -lua_atpanic on x64. -

    • diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm.h b/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm.h index a7295663..ebcf4ac0 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm.h +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm.h @@ -294,7 +294,7 @@ int dasm_link(Dst_DECL, size_t *szp) { /* Handle globals not defined in this translation unit. */ int idx; - for (idx = 20; idx*sizeof(int) < D->lgsize; idx++) { + for (idx = 10; idx*sizeof(int) < D->lgsize; idx++) { int n = D->lglabels[idx]; /* Undefined label: Collapse rel chain and replace with marker (< 0). */ while (n > 0) { int *pb = DASM_POS2PTR(D, n); n = *pb; *pb = -idx; } @@ -371,7 +371,10 @@ int dasm_encode(Dst_DECL, void *buffer) ins &= 255; while ((((char *)cp - base) & ins)) *cp++ = 0xe1a00000; break; case DASM_REL_LG: - CK(n >= 0, UNDEF_LG); + if (n < 0) { + n = (int)((ptrdiff_t)D->globals[-n] - (ptrdiff_t)cp - 4); + goto patchrel; + } /* fallthrough */ case DASM_REL_PC: CK(n >= 0, UNDEF_PC); diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm.lua b/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm.lua index acc8b48d..0c775ae2 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm.lua +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm.lua @@ -9,9 +9,9 @@ local _info = { arch = "arm", description = "DynASM ARM module", - version = "1.4.0", - vernum = 10400, - release = "2015-10-18", + version = "1.5.0", + vernum = 10500, + release = "2021-05-02", author = "Mike Pall", license = "MIT", } diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm64.h b/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm64.h index 577b54bc..d6da4528 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm64.h +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm64.h @@ -21,8 +21,9 @@ enum { /* The following actions need a buffer position. */ DASM_ALIGN, DASM_REL_LG, DASM_LABEL_LG, /* The following actions also have an argument. */ - DASM_REL_PC, DASM_LABEL_PC, + DASM_REL_PC, DASM_LABEL_PC, DASM_REL_A, DASM_IMM, DASM_IMM6, DASM_IMM12, DASM_IMM13W, DASM_IMM13X, DASM_IMML, + DASM_IMMV, DASM_VREG, DASM__MAX }; @@ -39,6 +40,7 @@ enum { #define DASM_S_RANGE_LG 0x13000000 #define DASM_S_RANGE_PC 0x14000000 #define DASM_S_RANGE_REL 0x15000000 +#define DASM_S_RANGE_VREG 0x16000000 #define DASM_S_UNDEF_LG 0x21000000 #define DASM_S_UNDEF_PC 0x22000000 @@ -247,7 +249,7 @@ void dasm_put(Dst_DECL, int start, ...) n = (ins & 255); CK(n < D->maxsection, RANGE_SEC); D->section = &D->sections[n]; goto stop; case DASM_ESC: p++; ofs += 4; break; - case DASM_REL_EXT: break; + case DASM_REL_EXT: if ((ins & 0x8000)) ofs += 8; break; case DASM_ALIGN: ofs += (ins & 255); b[pos++] = ofs; break; case DASM_REL_LG: n = (ins & 2047) - 10; pl = D->lglabels + n; @@ -268,6 +270,11 @@ void dasm_put(Dst_DECL, int start, ...) *pl = pos; } pos++; + if ((ins & 0x8000)) ofs += 8; + break; + case DASM_REL_A: + b[pos++] = n; + b[pos++] = va_arg(ap, int); break; case DASM_LABEL_LG: pl = D->lglabels + (ins & 2047) - 10; CKPL(lg, LG); goto putlabel; @@ -312,13 +319,21 @@ void dasm_put(Dst_DECL, int start, ...) } case DASM_IMML: { #ifdef DASM_CHECKS - int scale = (p[-2] >> 30); + int scale = (ins & 3); CK((!(n & ((1<>scale) < 4096) || (unsigned int)(n+256) < 512, RANGE_I); #endif b[pos++] = n; break; } + case DASM_IMMV: + ofs += 4; + b[pos++] = n; + break; + case DASM_VREG: + CK(n < 32, RANGE_VREG); + b[pos++] = n; + break; } } } @@ -348,7 +363,7 @@ int dasm_link(Dst_DECL, size_t *szp) { /* Handle globals not defined in this translation unit. */ int idx; - for (idx = 20; idx*sizeof(int) < D->lgsize; idx++) { + for (idx = 10; idx*sizeof(int) < D->lgsize; idx++) { int n = D->lglabels[idx]; /* Undefined label: Collapse rel chain and replace with marker (< 0). */ while (n > 0) { int *pb = DASM_POS2PTR(D, n); n = *pb; *pb = -idx; } @@ -375,8 +390,8 @@ int dasm_link(Dst_DECL, size_t *szp) case DASM_REL_LG: case DASM_REL_PC: pos++; break; case DASM_LABEL_LG: case DASM_LABEL_PC: b[pos++] += ofs; break; case DASM_IMM: case DASM_IMM6: case DASM_IMM12: case DASM_IMM13W: - case DASM_IMML: pos++; break; - case DASM_IMM13X: pos += 2; break; + case DASM_IMML: case DASM_IMMV: case DASM_VREG: pos++; break; + case DASM_IMM13X: case DASM_REL_A: pos += 2; break; } } stop: (void)0; @@ -426,7 +441,12 @@ int dasm_encode(Dst_DECL, void *buffer) ins &= 255; while ((((char *)cp - base) & ins)) *cp++ = 0xe1a00000; break; case DASM_REL_LG: - CK(n >= 0, UNDEF_LG); + if (n < 0) { + ptrdiff_t na = (ptrdiff_t)D->globals[-n] - (ptrdiff_t)cp + 4; + n = (int)na; + CK((ptrdiff_t)n == na, RANGE_REL); + goto patchrel; + } /* fallthrough */ case DASM_REL_PC: CK(n >= 0, UNDEF_PC); @@ -446,8 +466,24 @@ int dasm_encode(Dst_DECL, void *buffer) } else if ((ins & 0x1000)) { /* TBZ, TBNZ */ CK((n & 3) == 0 && ((n+0x00008000) >> 16) == 0, RANGE_REL); cp[-1] |= ((n << 3) & 0x0007ffe0); + } else if ((ins & 0x8000)) { /* absolute */ + cp[0] = (unsigned int)((ptrdiff_t)cp - 4 + n); + cp[1] = (unsigned int)(((ptrdiff_t)cp - 4 + n) >> 32); + cp += 2; } break; + case DASM_REL_A: { + ptrdiff_t na = (((ptrdiff_t)(*b++) << 32) | (unsigned int)n); + if ((ins & 0x3000) == 0x3000) { /* ADRP */ + ins &= ~0x1000; + na = (na >> 12) - (((ptrdiff_t)cp - 4) >> 12); + } else { + na = na - (ptrdiff_t)cp + 4; + } + n = (int)na; + CK((ptrdiff_t)n == na, RANGE_REL); + goto patchrel; + } case DASM_LABEL_LG: ins &= 2047; if (ins >= 20) D->globals[ins-10] = (void *)(base + n); break; @@ -468,11 +504,17 @@ int dasm_encode(Dst_DECL, void *buffer) cp[-1] |= (dasm_imm13(n, *b++) << 10); break; case DASM_IMML: { - int scale = (p[-2] >> 30); + int scale = (ins & 3); cp[-1] |= (!(n & ((1<>scale) < 4096) ? ((n << (10-scale)) | 0x01000000) : ((n & 511) << 12); break; } + case DASM_IMMV: + *cp++ = n; + break; + case DASM_VREG: + cp[-1] |= (n & 0x1f) << (ins & 0x1f); + break; default: *cp++ = ins; break; } } diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm64.lua b/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm64.lua index 861ff2b2..cb82dc4a 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm64.lua +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_arm64.lua @@ -9,9 +9,9 @@ local _info = { arch = "arm", description = "DynASM ARM64 module", - version = "1.4.0", - vernum = 10400, - release = "2015-10-18", + version = "1.5.0", + vernum = 10500, + release = "2021-05-02", author = "Mike Pall", license = "MIT", } @@ -23,12 +23,12 @@ local _M = { _info = _info } local type, tonumber, pairs, ipairs = type, tonumber, pairs, ipairs local assert, setmetatable, rawget = assert, setmetatable, rawget local _s = string -local sub, format, byte, char = _s.sub, _s.format, _s.byte, _s.char +local format, byte, char = _s.format, _s.byte, _s.char local match, gmatch, gsub = _s.match, _s.gmatch, _s.gsub local concat, sort, insert = table.concat, table.sort, table.insert local bit = bit or require("bit") local band, shl, shr, sar = bit.band, bit.lshift, bit.rshift, bit.arshift -local ror, tohex = bit.ror, bit.tohex +local ror, tohex, tobit = bit.ror, bit.tohex, bit.tobit -- Inherited tables and callbacks. local g_opt, g_arch @@ -39,7 +39,9 @@ local wline, werror, wfatal, wwarn local action_names = { "STOP", "SECTION", "ESC", "REL_EXT", "ALIGN", "REL_LG", "LABEL_LG", - "REL_PC", "LABEL_PC", "IMM", "IMM6", "IMM12", "IMM13W", "IMM13X", "IMML", + "REL_PC", "LABEL_PC", "REL_A", + "IMM", "IMM6", "IMM12", "IMM13W", "IMM13X", "IMML", "IMMV", + "VREG", } -- Maximum number of section buffer positions for dasm_put(). @@ -246,9 +248,12 @@ local map_cond = { local parse_reg_type -local function parse_reg(expr) +local function parse_reg(expr, shift) if not expr then werror("expected register name") end local tname, ovreg = match(expr, "^([%w_]+):(@?%l%d+)$") + if not tname then + tname, ovreg = match(expr, "^([%w_]+):(R[xwqdshb]%b())$") + end local tp = map_type[tname or expr] if tp then local reg = ovreg or tp.reg @@ -266,18 +271,28 @@ local function parse_reg(expr) elseif parse_reg_type ~= rt then werror("register size mismatch") end - return r, tp + return shl(r, shift), tp end end + local vrt, vreg = match(expr, "^R([xwqdshb])(%b())$") + if vreg then + if not parse_reg_type then + parse_reg_type = vrt + elseif parse_reg_type ~= vrt then + werror("register size mismatch") + end + if shift then waction("VREG", shift, vreg) end + return 0 + end werror("bad register name `"..expr.."'") end local function parse_reg_base(expr) if expr == "sp" then return 0x3e0 end - local base, tp = parse_reg(expr) + local base, tp = parse_reg(expr, 5) if parse_reg_type ~= "x" then werror("bad register type") end parse_reg_type = false - return shl(base, 5), tp + return base, tp end local parse_ctx = {} @@ -297,7 +312,7 @@ local function parse_number(n) local code = loadenv("return "..n) if code then local ok, y = pcall(code) - if ok then return y end + if ok and type(y) == "number" then return y end end return nil end @@ -403,7 +418,7 @@ local function parse_imm_load(imm, scale) end werror("out of range immediate `"..imm.."'") else - waction("IMML", 0, imm) + waction("IMML", scale, imm) return 0 end end @@ -462,6 +477,7 @@ end local function parse_load(params, nparams, n, op) if params[n+2] then werror("too many operands") end + local scale = shr(op, 30) local pn, p2 = params[n], params[n+1] local p1, wb = match(pn, "^%[%s*(.-)%s*%](!?)$") if not p1 then @@ -470,14 +486,13 @@ local function parse_load(params, nparams, n, op) if reg and tailr ~= "" then local base, tp = parse_reg_base(reg) if tp then - waction("IMML", 0, format(tp.ctypefmt, tailr)) + waction("IMML", scale, format(tp.ctypefmt, tailr)) return op + base end end end werror("expected address operand") end - local scale = shr(op, 30) if p2 then if wb == "!" then werror("bad use of '!'") end op = op + parse_reg_base(p1) + parse_imm(p2, 9, 12, 0, true) + 0x400 @@ -494,7 +509,7 @@ local function parse_load(params, nparams, n, op) op = op + parse_imm_load(imm, scale) else local p2b, p3b, p3s = match(p2a, "^,%s*([^,%s]*)%s*,?%s*(%S*)%s*(.*)$") - op = op + shl(parse_reg(p2b), 16) + 0x00200800 + op = op + parse_reg(p2b, 16) + 0x00200800 if parse_reg_type ~= "x" and parse_reg_type ~= "w" then werror("bad index register type") end @@ -561,14 +576,14 @@ local function parse_load_pair(params, nparams, n, op) end local function parse_label(label, def) - local prefix = sub(label, 1, 2) + local prefix = label:sub(1, 2) -- =>label (pc label reference) if prefix == "=>" then - return "PC", 0, sub(label, 3) + return "PC", 0, label:sub(3) end -- ->name (global label reference) if prefix == "->" then - return "LG", map_global[sub(label, 3)] + return "LG", map_global[label:sub(3)] end if def then -- [1-9] (local label definition) @@ -586,8 +601,11 @@ local function parse_label(label, def) if extname then return "EXT", map_extern[extname] end + -- &expr (pointer) + if label:sub(1, 1) == "&" then + return "A", 0, format("(ptrdiff_t)(%s)", label:sub(2)) + end end - werror("bad label `"..label.."'") end local function branch_type(op) @@ -620,24 +638,24 @@ local function alias_bfx(p) end local function alias_bfiz(p) - parse_reg(p[1]) + parse_reg(p[1], 0) if parse_reg_type == "w" then - p[3] = "#-("..p[3]:sub(2)..")%32" + p[3] = "#(32-("..p[3]:sub(2).."))%32" p[4] = "#("..p[4]:sub(2)..")-1" else - p[3] = "#-("..p[3]:sub(2)..")%64" + p[3] = "#(64-("..p[3]:sub(2).."))%64" p[4] = "#("..p[4]:sub(2)..")-1" end end local alias_lslimm = op_alias("ubfm_4", function(p) - parse_reg(p[1]) + parse_reg(p[1], 0) local sh = p[3]:sub(2) if parse_reg_type == "w" then - p[3] = "#-("..sh..")%32" + p[3] = "#(32-("..sh.."))%32" p[4] = "#31-("..sh..")" else - p[3] = "#-("..sh..")%64" + p[3] = "#(64-("..sh.."))%64" p[4] = "#63-("..sh..")" end end) @@ -881,25 +899,25 @@ end -- Handle opcodes defined with template strings. local function parse_template(params, template, nparams, pos) - local op = tonumber(sub(template, 1, 8), 16) + local op = tonumber(template:sub(1, 8), 16) local n = 1 local rtt = {} parse_reg_type = false -- Process each character. - for p in gmatch(sub(template, 9), ".") do + for p in gmatch(template:sub(9), ".") do local q = params[n] if p == "D" then - op = op + parse_reg(q); n = n + 1 + op = op + parse_reg(q, 0); n = n + 1 elseif p == "N" then - op = op + shl(parse_reg(q), 5); n = n + 1 + op = op + parse_reg(q, 5); n = n + 1 elseif p == "M" then - op = op + shl(parse_reg(q), 16); n = n + 1 + op = op + parse_reg(q, 16); n = n + 1 elseif p == "A" then - op = op + shl(parse_reg(q), 10); n = n + 1 + op = op + parse_reg(q, 10); n = n + 1 elseif p == "m" then - op = op + shl(parse_reg(params[n-1]), 16) + op = op + parse_reg(params[n-1], 16) elseif p == "p" then if q == "sp" then params[n] = "@x31" end @@ -930,8 +948,14 @@ local function parse_template(params, template, nparams, pos) elseif p == "B" then local mode, v, s = parse_label(q, false); n = n + 1 + if not mode then werror("bad label `"..q.."'") end local m = branch_type(op) - waction("REL_"..mode, v+m, s, 1) + if mode == "A" then + waction("REL_"..mode, v+m, format("(unsigned int)(%s)", s)) + actargs[#actargs+1] = format("(unsigned int)((%s)>>32)", s) + else + waction("REL_"..mode, v+m, s, 1) + end elseif p == "I" then op = op + parse_imm12(q); n = n + 1 @@ -977,8 +1001,8 @@ function op_template(params, template, nparams) if not params then return template:gsub("%x%x%x%x%x%x%x%x", "") end -- Limit number of section buffer positions used by a single dasm_put(). - -- A single opcode needs a maximum of 3 positions. - if secpos+3 > maxsecpos then wflush() end + -- A single opcode needs a maximum of 4 positions. + if secpos+4 > maxsecpos then wflush() end local pos = wpos() local lpos, apos, spos = #actlist, #actargs, secpos @@ -990,9 +1014,11 @@ function op_template(params, template, nparams) actlist[lpos+1] = nil actlist[lpos+2] = nil actlist[lpos+3] = nil + actlist[lpos+4] = nil actargs[apos+1] = nil actargs[apos+2] = nil actargs[apos+3] = nil + actargs[apos+4] = nil end error(err, 0) end @@ -1036,23 +1062,50 @@ map_op[".label_1"] = function(params) if not params then return "[1-9] | ->global | =>pcexpr" end if secpos+1 > maxsecpos then wflush() end local mode, n, s = parse_label(params[1], true) - if mode == "EXT" then werror("bad label definition") end + if not mode or mode == "EXT" then werror("bad label definition") end waction("LABEL_"..mode, n, s, 1) end ------------------------------------------------------------------------------ -- Pseudo-opcodes for data storage. -map_op[".long_*"] = function(params) +local function op_data(params) if not params then return "imm..." end + local sz = params.op == ".long" and 4 or 8 for _,p in ipairs(params) do - local n = tonumber(p) - if not n then werror("bad immediate `"..p.."'") end - if n < 0 then n = n + 2^32 end - wputw(n) + local imm = parse_number(p) + if imm then + local n = tobit(imm) + if n == imm or (n < 0 and n + 2^32 == imm) then + wputw(n < 0 and n + 2^32 or n) + if sz == 8 then + wputw(imm < 0 and 0xffffffff or 0) + end + elseif sz == 4 then + werror("bad immediate `"..p.."'") + else + imm = nil + end + end + if not imm then + local mode, v, s = parse_label(p, false) + if sz == 4 then + if mode then werror("label does not fit into .long") end + waction("IMMV", 0, p) + elseif mode and mode ~= "A" then + waction("REL_"..mode, v+0x8000, s, 1) + else + if mode == "A" then p = s end + waction("IMMV", 0, format("(unsigned int)(%s)", p)) + waction("IMMV", 0, format("(unsigned int)((unsigned long long)(%s)>>32)", p)) + end + end if secpos+2 > maxsecpos then wflush() end end end +map_op[".long_*"] = op_data +map_op[".quad_*"] = op_data +map_op[".addr_*"] = op_data -- Alignment pseudo-opcode. map_op[".align_1"] = function(params) diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_mips.h b/love/src/jni/LuaJIT-2.1/dynasm/dasm_mips.h index 2fd7d71f..b99b56b0 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_mips.h +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_mips.h @@ -273,7 +273,7 @@ int dasm_link(Dst_DECL, size_t *szp) { /* Handle globals not defined in this translation unit. */ int idx; - for (idx = 20; idx*sizeof(int) < D->lgsize; idx++) { + for (idx = 10; idx*sizeof(int) < D->lgsize; idx++) { int n = D->lglabels[idx]; /* Undefined label: Collapse rel chain and replace with marker (< 0). */ while (n > 0) { int *pb = DASM_POS2PTR(D, n); n = *pb; *pb = -idx; } @@ -349,7 +349,10 @@ int dasm_encode(Dst_DECL, void *buffer) ins &= 255; while ((((char *)cp - base) & ins)) *cp++ = 0x60000000; break; case DASM_REL_LG: - CK(n >= 0, UNDEF_LG); + if (n < 0) { + n = (int)((ptrdiff_t)D->globals[-n] - (ptrdiff_t)cp); + goto patchrel; + } /* fallthrough */ case DASM_REL_PC: CK(n >= 0, UNDEF_PC); diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_mips.lua b/love/src/jni/LuaJIT-2.1/dynasm/dasm_mips.lua index 45b82755..59147015 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_mips.lua +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_mips.lua @@ -12,9 +12,9 @@ local mipsr6 = _map_def.MIPSR6 local _info = { arch = mips64 and "mips64" or "mips", description = "DynASM MIPS32/MIPS64 module", - version = "1.4.0", - vernum = 10400, - release = "2020-01-20", + version = "1.5.0", + vernum = 10500, + release = "2021-05-02", author = "Mike Pall", license = "MIT", } diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_ppc.h b/love/src/jni/LuaJIT-2.1/dynasm/dasm_ppc.h index ae540e46..35264f2e 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_ppc.h +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_ppc.h @@ -277,7 +277,7 @@ int dasm_link(Dst_DECL, size_t *szp) { /* Handle globals not defined in this translation unit. */ int idx; - for (idx = 20; idx*sizeof(int) < D->lgsize; idx++) { + for (idx = 10; idx*sizeof(int) < D->lgsize; idx++) { int n = D->lglabels[idx]; /* Undefined label: Collapse rel chain and replace with marker (< 0). */ while (n > 0) { int *pb = DASM_POS2PTR(D, n); n = *pb; *pb = -idx; } @@ -353,7 +353,10 @@ int dasm_encode(Dst_DECL, void *buffer) ins &= 255; while ((((char *)cp - base) & ins)) *cp++ = 0x60000000; break; case DASM_REL_LG: - CK(n >= 0, UNDEF_LG); + if (n < 0) { + n = (int)((ptrdiff_t)D->globals[-n] - (ptrdiff_t)cp); + goto patchrel; + } /* fallthrough */ case DASM_REL_PC: CK(n >= 0, UNDEF_PC); diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_ppc.lua b/love/src/jni/LuaJIT-2.1/dynasm/dasm_ppc.lua index d295ef62..ee2afb2e 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_ppc.lua +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_ppc.lua @@ -11,9 +11,9 @@ local _info = { arch = "ppc", description = "DynASM PPC module", - version = "1.4.0", - vernum = 10400, - release = "2015-10-18", + version = "1.5.0", + vernum = 10500, + release = "2021-05-02", author = "Mike Pall", license = "MIT", } diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_proto.h b/love/src/jni/LuaJIT-2.1/dynasm/dasm_proto.h index ea2cc260..8914596a 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_proto.h +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_proto.h @@ -10,8 +10,8 @@ #include #include -#define DASM_IDENT "DynASM 1.4.0" -#define DASM_VERSION 10400 /* 1.4.0 */ +#define DASM_IDENT "DynASM 1.5.0" +#define DASM_VERSION 10500 /* 1.5.0 */ #ifndef Dst_DECL #define Dst_DECL dasm_State **Dst diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_x86.h b/love/src/jni/LuaJIT-2.1/dynasm/dasm_x86.h index 2e2f2334..d8d4928c 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_x86.h +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_x86.h @@ -239,8 +239,11 @@ void dasm_put(Dst_DECL, int start, ...) } pos++; ofs += 4; /* Maximum offset needed. */ - if (action == DASM_REL_LG || action == DASM_REL_PC) + if (action == DASM_REL_LG || action == DASM_REL_PC) { b[pos++] = ofs; /* Store pass1 offset estimate. */ + } else if (sizeof(ptrdiff_t) == 8) { + ofs += 4; + } break; case DASM_LABEL_LG: pl = D->lglabels + *p++; CKPL(lg, LG); goto putlabel; case DASM_LABEL_PC: pl = D->pclabels + va_arg(ap, int); CKPL(pc, PC); @@ -365,10 +368,22 @@ int dasm_link(Dst_DECL, size_t *szp) do { *((unsigned short *)cp) = (unsigned short)(x); cp+=2; } while (0) #define dasmd(x) \ do { *((unsigned int *)cp) = (unsigned int)(x); cp+=4; } while (0) +#define dasmq(x) \ + do { *((unsigned long long *)cp) = (unsigned long long)(x); cp+=8; } while (0) #else #define dasmw(x) do { dasmb(x); dasmb((x)>>8); } while (0) #define dasmd(x) do { dasmw(x); dasmw((x)>>16); } while (0) +#define dasmq(x) do { dasmd(x); dasmd((x)>>32); } while (0) #endif +static unsigned char *dasma_(unsigned char *cp, ptrdiff_t x) +{ + if (sizeof(ptrdiff_t) == 8) + dasmq((unsigned long long)x); + else + dasmd((unsigned int)x); + return cp; +} +#define dasma(x) (cp = dasma_(cp, (x))) /* Pass 3: Encode sections. */ int dasm_encode(Dst_DECL, void *buffer) @@ -443,12 +458,13 @@ int dasm_encode(Dst_DECL, void *buffer) goto wb; } case DASM_IMM_LG: - p++; if (n < 0) { n = (int)(ptrdiff_t)D->globals[-n]; goto wd; } + p++; + if (n < 0) { dasma((ptrdiff_t)D->globals[-n]); break; } /* fallthrough */ case DASM_IMM_PC: { int *pb = DASM_POS2PTR(D, n); - n = *pb < 0 ? pb[1] : (*pb + (int)(ptrdiff_t)base); - goto wd; + dasma(*pb < 0 ? (ptrdiff_t)pb[1] : (*pb + (ptrdiff_t)base)); + break; } case DASM_LABEL_LG: { int idx = *p++; diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dasm_x86.lua b/love/src/jni/LuaJIT-2.1/dynasm/dasm_x86.lua index bc9a0001..fe2cf579 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dasm_x86.lua +++ b/love/src/jni/LuaJIT-2.1/dynasm/dasm_x86.lua @@ -11,9 +11,9 @@ local x64 = x64 local _info = { arch = x64 and "x64" or "x86", description = "DynASM x86/x64 module", - version = "1.4.0", - vernum = 10400, - release = "2015-10-18", + version = "1.5.0", + vernum = 10500, + release = "2021-05-02", author = "Mike Pall", license = "MIT", } @@ -484,6 +484,22 @@ local function wputdarg(n) end end +-- Put signed or unsigned qword or arg. +local function wputqarg(n) + local tn = type(n) + if tn == "number" then -- This is only used for numbers from -2^31..2^32-1. + wputb(band(n, 255)) + wputb(band(shr(n, 8), 255)) + wputb(band(shr(n, 16), 255)) + wputb(shr(n, 24)) + local sign = n < 0 and 255 or 0 + wputb(sign); wputb(sign); wputb(sign); wputb(sign) + else + waction("IMM_D", format("(unsigned int)(%s)", n)) + waction("IMM_D", format("(unsigned int)((unsigned long long)(%s)>>32)", n)) + end +end + -- Put operand-size dependent number or arg (defaults to dword). local function wputszarg(sz, n) if not sz or sz == "d" or sz == "q" then wputdarg(n) @@ -663,10 +679,16 @@ local function opmodestr(op, args) end -- Convert number to valid integer or nil. -local function toint(expr) +local function toint(expr, isqword) local n = tonumber(expr) if n then - if n % 1 ~= 0 or n < -2147483648 or n > 4294967295 then + if n % 1 ~= 0 then + werror("not an integer number `"..expr.."'") + elseif isqword then + if n < -2147483648 or n > 2147483647 then + n = nil -- Handle it as an expression to avoid precision loss. + end + elseif n < -2147483648 or n > 4294967295 then werror("bad integer number `"..expr.."'") end return n @@ -749,7 +771,7 @@ local function rtexpr(expr) end -- Parse operand and return { mode, opsize, reg, xreg, xsc, disp, imm }. -local function parseoperand(param) +local function parseoperand(param, isqword) local t = {} local expr = param @@ -810,7 +832,7 @@ local function parseoperand(param) if t.disp then break end -- [reg+xreg...] - local xreg, tailx = match(tailr, "^+%s*([@%w_:]+)%s*(.*)$") + local xreg, tailx = match(tailr, "^%+%s*([@%w_:]+)%s*(.*)$") xreg, t.xreg, tp = rtexpr(xreg) if not t.xreg then -- [reg+-expr] @@ -837,7 +859,7 @@ local function parseoperand(param) t.disp = dispexpr(tailx) else -- imm or opsize*imm - local imm = toint(expr) + local imm = toint(expr, isqword) if not imm and sub(expr, 1, 1) == "*" and t.opsize then imm = toint(sub(expr, 2)) if imm then @@ -1952,7 +1974,7 @@ local function dopattern(pat, args, sz, op, needrex) local a = args[narg] narg = narg + 1 local mode, imm = a.mode, a.imm - if mode == "iJ" and not match("iIJ", c) then + if mode == "iJ" and not match(x64 and "J" or "iIJ", c) then werror("bad operand size for label") end if c == "S" then @@ -2144,14 +2166,16 @@ end local function op_data(params) if not params then return "imm..." end local sz = sub(params.op, 2, 2) - if sz == "a" then sz = addrsize end + if sz == "l" then sz = "d" elseif sz == "a" then sz = addrsize end for _,p in ipairs(params) do - local a = parseoperand(p) + local a = parseoperand(p, sz == "q") if sub(a.mode, 1, 1) ~= "i" or (a.opsize and a.opsize ~= sz) then werror("bad mode or size in `"..p.."'") end if a.mode == "iJ" then wputlabel("IMM_", a.imm, 1) + elseif sz == "q" then + wputqarg(a.imm) else wputszarg(sz, a.imm) end @@ -2163,7 +2187,11 @@ map_op[".byte_*"] = op_data map_op[".sbyte_*"] = op_data map_op[".word_*"] = op_data map_op[".dword_*"] = op_data +map_op[".qword_*"] = op_data map_op[".aword_*"] = op_data +map_op[".long_*"] = op_data +map_op[".quad_*"] = op_data +map_op[".addr_*"] = op_data ------------------------------------------------------------------------------ diff --git a/love/src/jni/LuaJIT-2.1/dynasm/dynasm.lua b/love/src/jni/LuaJIT-2.1/dynasm/dynasm.lua index 92d109f2..f4e71eca 100644 --- a/love/src/jni/LuaJIT-2.1/dynasm/dynasm.lua +++ b/love/src/jni/LuaJIT-2.1/dynasm/dynasm.lua @@ -10,9 +10,9 @@ local _info = { name = "DynASM", description = "A dynamic assembler for code generation engines", - version = "1.4.0", - vernum = 10400, - release = "2015-10-18", + version = "1.5.0", + vernum = 10500, + release = "2021-05-02", author = "Mike Pall", url = "https://luajit.org/dynasm.html", license = "MIT", diff --git a/love/src/jni/LuaJIT-2.1/src/Makefile b/love/src/jni/LuaJIT-2.1/src/Makefile index 2e1a2888..2538503f 100644 --- a/love/src/jni/LuaJIT-2.1/src/Makefile +++ b/love/src/jni/LuaJIT-2.1/src/Makefile @@ -211,7 +211,7 @@ TARGET_CC= $(STATIC_CC) TARGET_STCC= $(STATIC_CC) TARGET_DYNCC= $(DYNAMIC_CC) TARGET_LD= $(CROSS)$(CC) -TARGET_AR= $(CROSS)ar rcus 2>/dev/null +TARGET_AR= $(CROSS)ar rcus TARGET_STRIP= $(CROSS)strip TARGET_LIBPATH= $(or $(PREFIX),/usr/local)/$(or $(MULTILIB),lib) @@ -311,6 +311,7 @@ ifeq (Windows,$(TARGET_SYS)) TARGET_XSHLDFLAGS= -shared -Wl,--out-implib,$(TARGET_DLLDOTANAME) TARGET_DYNXLDOPTS= else + TARGET_AR+= 2>/dev/null ifeq (,$(shell $(TARGET_CC) -o /dev/null -c -x c /dev/null -fno-stack-protector 2>/dev/null || echo 1)) TARGET_XCFLAGS+= -fno-stack-protector endif @@ -319,13 +320,10 @@ ifeq (Darwin,$(TARGET_SYS)) $(error missing: export MACOSX_DEPLOYMENT_TARGET=XX.YY) endif TARGET_STRIP+= -x + TARGET_XCFLAGS+= -DLUAJIT_UNWIND_EXTERNAL TARGET_XSHLDFLAGS= -dynamiclib -single_module -undefined dynamic_lookup -fPIC TARGET_DYNXLDOPTS= TARGET_XSHLDFLAGS+= -install_name $(TARGET_DYLIBPATH) -compatibility_version $(MAJVER).$(MINVER) -current_version $(MAJVER).$(MINVER).$(RELVER) - ifeq (x64,$(TARGET_LJARCH)) - TARGET_XLDFLAGS+= -pagezero_size 10000 -image_base 100000000 - TARGET_XSHLDFLAGS+= -image_base 7fff04c4a000 - endif else ifeq (iOS,$(TARGET_SYS)) TARGET_STRIP+= -x @@ -336,6 +334,13 @@ ifeq (iOS,$(TARGET_SYS)) TARGET_XCFLAGS+= -fno-omit-frame-pointer endif else + ifeq (,$(findstring LJ_NO_UNWIND 1,$(TARGET_TESTARCH))) + # Find out whether the target toolchain always generates unwind tables. + TARGET_TESTUNWIND=$(shell exec 2>/dev/null; echo 'extern void b(void);int a(void){b();return 0;}' | $(TARGET_CC) -c -x c - -o tmpunwind.o && { grep -qa -e eh_frame -e __unwind_info tmpunwind.o || grep -qU -e eh_frame -e __unwind_info tmpunwind.o; } && echo E; rm -f tmpunwind.o) + ifneq (,$(findstring E,$(TARGET_TESTUNWIND))) + TARGET_XCFLAGS+= -DLUAJIT_UNWIND_EXTERNAL + endif + endif ifneq (SunOS,$(TARGET_SYS)) ifneq (PS3,$(TARGET_SYS)) TARGET_XLDFLAGS+= -Wl,-E @@ -479,13 +484,15 @@ LJVM_BOUT= $(LJVM_S) LJVM_MODE= elfasm LJLIB_O= lib_base.o lib_math.o lib_bit.o lib_string.o lib_table.o \ - lib_io.o lib_os.o lib_package.o lib_debug.o lib_jit.o lib_ffi.o + lib_io.o lib_os.o lib_package.o lib_debug.o lib_jit.o lib_ffi.o \ + lib_buffer.o LJLIB_C= $(LJLIB_O:.o=.c) LJCORE_O= lj_assert.o lj_gc.o lj_err.o lj_char.o lj_bc.o lj_obj.o lj_buf.o \ lj_str.o lj_tab.o lj_func.o lj_udata.o lj_meta.o lj_debug.o \ lj_prng.o lj_state.o lj_dispatch.o lj_vmevent.o lj_vmmath.o \ - lj_strscan.o lj_strfmt.o lj_strfmt_num.o lj_api.o lj_profile.o \ + lj_strscan.o lj_strfmt.o lj_strfmt_num.o lj_serialize.o \ + lj_api.o lj_profile.o \ lj_lex.o lj_parse.o lj_bcread.o lj_bcwrite.o lj_load.o \ lj_ir.o lj_opt_mem.o lj_opt_fold.o lj_opt_narrow.o \ lj_opt_dce.o lj_opt_loop.o lj_opt_split.o lj_opt_sink.o \ diff --git a/love/src/jni/LuaJIT-2.1/src/Makefile.dep b/love/src/jni/LuaJIT-2.1/src/Makefile.dep index 3f26599e..1ad6701a 100644 --- a/love/src/jni/LuaJIT-2.1/src/Makefile.dep +++ b/love/src/jni/LuaJIT-2.1/src/Makefile.dep @@ -2,14 +2,18 @@ lib_aux.o: lib_aux.c lua.h luaconf.h lauxlib.h lj_obj.h lj_def.h \ lj_arch.h lj_err.h lj_errmsg.h lj_state.h lj_trace.h lj_jit.h lj_ir.h \ lj_dispatch.h lj_bc.h lj_traceerr.h lj_lib.h lib_base.o: lib_base.c lua.h luaconf.h lauxlib.h lualib.h lj_obj.h \ - lj_def.h lj_arch.h lj_gc.h lj_err.h lj_errmsg.h lj_debug.h lj_str.h \ - lj_tab.h lj_meta.h lj_state.h lj_frame.h lj_bc.h lj_ctype.h lj_cconv.h \ - lj_ff.h lj_ffdef.h lj_dispatch.h lj_jit.h lj_ir.h lj_char.h lj_strscan.h \ - lj_strfmt.h lj_lib.h lj_libdef.h + lj_def.h lj_arch.h lj_gc.h lj_err.h lj_errmsg.h lj_debug.h lj_buf.h \ + lj_str.h lj_tab.h lj_meta.h lj_state.h lj_frame.h lj_bc.h lj_ctype.h \ + lj_cconv.h lj_ff.h lj_ffdef.h lj_dispatch.h lj_jit.h lj_ir.h lj_char.h \ + lj_strscan.h lj_strfmt.h lj_lib.h lj_libdef.h lib_bit.o: lib_bit.c lua.h luaconf.h lauxlib.h lualib.h lj_obj.h lj_def.h \ lj_arch.h lj_err.h lj_errmsg.h lj_buf.h lj_gc.h lj_str.h lj_strscan.h \ lj_strfmt.h lj_ctype.h lj_cdata.h lj_cconv.h lj_carith.h lj_ff.h \ lj_ffdef.h lj_lib.h lj_libdef.h +lib_buffer.o: lib_buffer.c lua.h luaconf.h lauxlib.h lualib.h lj_obj.h \ + lj_def.h lj_arch.h lj_gc.h lj_err.h lj_errmsg.h lj_buf.h lj_str.h \ + lj_tab.h lj_udata.h lj_meta.h lj_ctype.h lj_cdata.h lj_cconv.h \ + lj_strfmt.h lj_serialize.h lj_lib.h lj_libdef.h lib_debug.o: lib_debug.c lua.h luaconf.h lauxlib.h lualib.h lj_obj.h \ lj_def.h lj_arch.h lj_gc.h lj_err.h lj_errmsg.h lj_debug.h lj_lib.h \ lj_libdef.h @@ -48,10 +52,10 @@ lj_api.o: lj_api.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_gc.h \ lj_meta.h lj_state.h lj_bc.h lj_frame.h lj_trace.h lj_jit.h lj_ir.h \ lj_dispatch.h lj_traceerr.h lj_vm.h lj_strscan.h lj_strfmt.h lj_asm.o: lj_asm.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_gc.h \ - lj_str.h lj_tab.h lj_frame.h lj_bc.h lj_ctype.h lj_ir.h lj_jit.h \ - lj_ircall.h lj_iropt.h lj_mcode.h lj_trace.h lj_dispatch.h lj_traceerr.h \ - lj_snap.h lj_asm.h lj_vm.h lj_target.h lj_target_*.h lj_emit_*.h \ - lj_asm_*.h + lj_buf.h lj_str.h lj_tab.h lj_frame.h lj_bc.h lj_ctype.h lj_ir.h \ + lj_jit.h lj_ircall.h lj_iropt.h lj_mcode.h lj_trace.h lj_dispatch.h \ + lj_traceerr.h lj_snap.h lj_asm.h lj_vm.h lj_target.h lj_target_*.h \ + lj_emit_*.h lj_asm_*.h lj_assert.o: lj_assert.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_bc.o: lj_bc.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_bc.h \ lj_bcdef.h @@ -77,8 +81,8 @@ lj_ccallback.o: lj_ccallback.c lj_obj.h lua.h luaconf.h lj_def.h \ lj_target_*.h lj_mcode.h lj_jit.h lj_ir.h lj_trace.h lj_dispatch.h \ lj_traceerr.h lj_vm.h lj_cconv.o: lj_cconv.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ - lj_err.h lj_errmsg.h lj_tab.h lj_ctype.h lj_gc.h lj_cdata.h lj_cconv.h \ - lj_ccallback.h + lj_err.h lj_errmsg.h lj_buf.h lj_gc.h lj_str.h lj_tab.h lj_ctype.h \ + lj_cdata.h lj_cconv.h lj_ccallback.h lj_cdata.o: lj_cdata.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ lj_gc.h lj_err.h lj_errmsg.h lj_tab.h lj_ctype.h lj_cconv.h lj_cdata.h lj_char.o: lj_char.c lj_char.h lj_def.h lua.h luaconf.h @@ -110,10 +114,10 @@ lj_err.o: lj_err.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_err.h \ lj_ff.h lj_ffdef.h lj_trace.h lj_jit.h lj_ir.h lj_dispatch.h \ lj_traceerr.h lj_vm.h lj_strfmt.h lj_ffrecord.o: lj_ffrecord.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ - lj_err.h lj_errmsg.h lj_str.h lj_tab.h lj_frame.h lj_bc.h lj_ff.h \ - lj_ffdef.h lj_ir.h lj_jit.h lj_ircall.h lj_iropt.h lj_trace.h \ - lj_dispatch.h lj_traceerr.h lj_record.h lj_ffrecord.h lj_crecord.h \ - lj_vm.h lj_strscan.h lj_strfmt.h lj_recdef.h + lj_err.h lj_errmsg.h lj_buf.h lj_gc.h lj_str.h lj_tab.h lj_frame.h \ + lj_bc.h lj_ff.h lj_ffdef.h lj_ir.h lj_jit.h lj_ircall.h lj_iropt.h \ + lj_trace.h lj_dispatch.h lj_traceerr.h lj_record.h lj_ffrecord.h \ + lj_crecord.h lj_vm.h lj_strscan.h lj_strfmt.h lj_serialize.h lj_recdef.h lj_func.o: lj_func.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_gc.h \ lj_func.h lj_trace.h lj_jit.h lj_ir.h lj_dispatch.h lj_bc.h \ lj_traceerr.h lj_vm.h @@ -127,15 +131,15 @@ lj_gdbjit.o: lj_gdbjit.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ lj_ir.o: lj_ir.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_gc.h \ lj_buf.h lj_str.h lj_tab.h lj_ir.h lj_jit.h lj_ircall.h lj_iropt.h \ lj_trace.h lj_dispatch.h lj_bc.h lj_traceerr.h lj_ctype.h lj_cdata.h \ - lj_carith.h lj_vm.h lj_strscan.h lj_strfmt.h lj_prng.h + lj_carith.h lj_vm.h lj_strscan.h lj_serialize.h lj_strfmt.h lj_prng.h lj_lex.o: lj_lex.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_gc.h \ lj_err.h lj_errmsg.h lj_buf.h lj_str.h lj_tab.h lj_ctype.h lj_cdata.h \ lualib.h lj_state.h lj_lex.h lj_parse.h lj_char.h lj_strscan.h \ lj_strfmt.h lj_lib.o: lj_lib.c lauxlib.h lua.h luaconf.h lj_obj.h lj_def.h lj_arch.h \ lj_gc.h lj_err.h lj_errmsg.h lj_str.h lj_tab.h lj_func.h lj_bc.h \ - lj_dispatch.h lj_jit.h lj_ir.h lj_vm.h lj_strscan.h lj_strfmt.h lj_lex.h \ - lj_bcdump.h lj_lib.h + lj_dispatch.h lj_jit.h lj_ir.h lj_ctype.h lj_vm.h lj_strscan.h \ + lj_strfmt.h lj_lex.h lj_bcdump.h lj_lib.h lj_load.o: lj_load.c lua.h luaconf.h lauxlib.h lj_obj.h lj_def.h \ lj_arch.h lj_gc.h lj_err.h lj_errmsg.h lj_buf.h lj_str.h lj_func.h \ lj_frame.h lj_bc.h lj_vm.h lj_lex.h lj_bcdump.h lj_parse.h @@ -170,15 +174,18 @@ lj_parse.o: lj_parse.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ lj_gc.h lj_err.h lj_errmsg.h lj_debug.h lj_buf.h lj_str.h lj_tab.h \ lj_func.h lj_state.h lj_bc.h lj_ctype.h lj_strfmt.h lj_lex.h lj_parse.h \ lj_vm.h lj_vmevent.h +lj_prng.o: lj_prng.c lj_def.h lua.h luaconf.h lj_arch.h lj_prng.h lj_profile.o: lj_profile.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ lj_buf.h lj_gc.h lj_str.h lj_frame.h lj_bc.h lj_debug.h lj_dispatch.h \ lj_jit.h lj_ir.h lj_trace.h lj_traceerr.h lj_profile.h luajit.h -lj_prng.o: lj_prng.c lj_def.h lua.h luaconf.h lj_arch.h lj_prng.h lj_record.o: lj_record.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ lj_err.h lj_errmsg.h lj_str.h lj_tab.h lj_meta.h lj_frame.h lj_bc.h \ lj_ctype.h lj_gc.h lj_ff.h lj_ffdef.h lj_debug.h lj_ir.h lj_jit.h \ lj_ircall.h lj_iropt.h lj_trace.h lj_dispatch.h lj_traceerr.h \ lj_record.h lj_ffrecord.h lj_snap.h lj_vm.h lj_prng.h +lj_serialize.o: lj_serialize.c lj_obj.h lua.h luaconf.h lj_def.h \ + lj_arch.h lj_err.h lj_errmsg.h lj_buf.h lj_gc.h lj_str.h lj_tab.h \ + lj_udata.h lj_ctype.h lj_cdata.h lj_ir.h lj_serialize.h lj_snap.o: lj_snap.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_gc.h \ lj_tab.h lj_state.h lj_frame.h lj_bc.h lj_ir.h lj_jit.h lj_iropt.h \ lj_trace.h lj_dispatch.h lj_traceerr.h lj_snap.h lj_target.h \ @@ -189,9 +196,10 @@ lj_state.o: lj_state.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ lj_ir.h lj_dispatch.h lj_traceerr.h lj_vm.h lj_prng.h lj_lex.h \ lj_alloc.h luajit.h lj_str.o: lj_str.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h lj_gc.h \ - lj_err.h lj_errmsg.h lj_str.h lj_char.h + lj_err.h lj_errmsg.h lj_str.h lj_char.h lj_prng.h lj_strfmt.o: lj_strfmt.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ - lj_buf.h lj_gc.h lj_str.h lj_state.h lj_char.h lj_strfmt.h + lj_err.h lj_errmsg.h lj_buf.h lj_gc.h lj_str.h lj_meta.h lj_state.h \ + lj_char.h lj_strfmt.h lj_ctype.h lj_lib.h lj_strfmt_num.o: lj_strfmt_num.c lj_obj.h lua.h luaconf.h lj_def.h \ lj_arch.h lj_buf.h lj_gc.h lj_str.h lj_strfmt.h lj_strscan.o: lj_strscan.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ @@ -204,7 +212,7 @@ lj_trace.o: lj_trace.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ lj_dispatch.h lj_traceerr.h lj_snap.h lj_gdbjit.h lj_record.h lj_asm.h \ lj_vm.h lj_vmevent.h lj_target.h lj_target_*.h lj_prng.h lj_udata.o: lj_udata.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ - lj_gc.h lj_udata.h + lj_gc.h lj_err.h lj_errmsg.h lj_udata.h lj_vmevent.o: lj_vmevent.c lj_obj.h lua.h luaconf.h lj_def.h lj_arch.h \ lj_str.h lj_tab.h lj_state.h lj_dispatch.h lj_bc.h lj_jit.h lj_ir.h \ lj_vm.h lj_vmevent.h @@ -216,23 +224,23 @@ ljamalg.o: ljamalg.c lua.h luaconf.h lauxlib.h lj_assert.c lj_obj.h \ lj_bc.h lj_ctype.h lj_cdata.h lj_trace.h lj_jit.h lj_ir.h lj_dispatch.h \ lj_traceerr.h lj_vm.h lj_err.c lj_debug.h lj_ff.h lj_ffdef.h lj_strfmt.h \ lj_char.c lj_char.h lj_bc.c lj_bcdef.h lj_obj.c lj_buf.c lj_str.c \ - lj_tab.c lj_func.c lj_udata.c lj_meta.c lj_strscan.h lj_lib.h lj_debug.c \ - lj_prng.c lj_prng.h lj_state.c lj_lex.h lj_alloc.h luajit.h \ + lj_prng.h lj_tab.c lj_func.c lj_udata.c lj_meta.c lj_strscan.h lj_lib.h \ + lj_debug.c lj_prng.c lj_state.c lj_lex.h lj_alloc.h luajit.h \ lj_dispatch.c lj_ccallback.h lj_profile.h lj_vmevent.c lj_vmevent.h \ - lj_vmmath.c lj_strscan.c lj_strfmt.c lj_strfmt_num.c lj_api.c \ - lj_profile.c lj_lex.c lualib.h lj_parse.h lj_parse.c lj_bcread.c \ - lj_bcdump.h lj_bcwrite.c lj_load.c lj_ctype.c lj_cdata.c lj_cconv.h \ - lj_cconv.c lj_ccall.c lj_ccall.h lj_ccallback.c lj_target.h \ - lj_target_*.h lj_mcode.h lj_carith.c lj_carith.h lj_clib.c lj_clib.h \ - lj_cparse.c lj_cparse.h lj_lib.c lj_ir.c lj_ircall.h lj_iropt.h \ - lj_opt_mem.c lj_opt_fold.c lj_folddef.h lj_opt_narrow.c lj_opt_dce.c \ - lj_opt_loop.c lj_snap.h lj_opt_split.c lj_opt_sink.c lj_mcode.c \ - lj_snap.c lj_record.c lj_record.h lj_ffrecord.h lj_crecord.c \ + lj_vmmath.c lj_strscan.c lj_strfmt.c lj_strfmt_num.c lj_serialize.c \ + lj_serialize.h lj_api.c lj_profile.c lj_lex.c lualib.h lj_parse.h \ + lj_parse.c lj_bcread.c lj_bcdump.h lj_bcwrite.c lj_load.c lj_ctype.c \ + lj_cdata.c lj_cconv.h lj_cconv.c lj_ccall.c lj_ccall.h lj_ccallback.c \ + lj_target.h lj_target_*.h lj_mcode.h lj_carith.c lj_carith.h lj_clib.c \ + lj_clib.h lj_cparse.c lj_cparse.h lj_lib.c lj_ir.c lj_ircall.h \ + lj_iropt.h lj_opt_mem.c lj_opt_fold.c lj_folddef.h lj_opt_narrow.c \ + lj_opt_dce.c lj_opt_loop.c lj_snap.h lj_opt_split.c lj_opt_sink.c \ + lj_mcode.c lj_snap.c lj_record.c lj_record.h lj_ffrecord.h lj_crecord.c \ lj_crecord.h lj_ffrecord.c lj_recdef.h lj_asm.c lj_asm.h lj_emit_*.h \ lj_asm_*.h lj_trace.c lj_gdbjit.h lj_gdbjit.c lj_alloc.c lib_aux.c \ lib_base.c lj_libdef.h lib_math.c lib_string.c lib_table.c lib_io.c \ lib_os.c lib_package.c lib_debug.c lib_bit.c lib_jit.c lib_ffi.c \ - lib_init.c + lib_buffer.c lib_init.c luajit.o: luajit.c lua.h luaconf.h lauxlib.h lualib.h luajit.h lj_arch.h host/buildvm.o: host/buildvm.c host/buildvm.h lj_def.h lua.h luaconf.h \ lj_arch.h lj_obj.h lj_def.h lj_arch.h lj_gc.h lj_obj.h lj_bc.h lj_ir.h \ diff --git a/love/src/jni/LuaJIT-2.1/src/host/buildvm_lib.c b/love/src/jni/LuaJIT-2.1/src/host/buildvm_lib.c index 17e0111c..20bb77cd 100644 --- a/love/src/jni/LuaJIT-2.1/src/host/buildvm_lib.c +++ b/love/src/jni/LuaJIT-2.1/src/host/buildvm_lib.c @@ -385,6 +385,8 @@ void emit_lib(BuildCtx *ctx) ok = LJ_HASJIT; else if (!strcmp(buf, "#if LJ_HASFFI\n")) ok = LJ_HASFFI; + else if (!strcmp(buf, "#if LJ_HASBUFFER\n")) + ok = LJ_HASBUFFER; if (!ok) { int lvl = 1; while (fgets(buf, sizeof(buf), fp) != NULL) { diff --git a/love/src/jni/LuaJIT-2.1/src/host/minilua.c b/love/src/jni/LuaJIT-2.1/src/host/minilua.c index 79150286..cfc7491d 100644 --- a/love/src/jni/LuaJIT-2.1/src/host/minilua.c +++ b/love/src/jni/LuaJIT-2.1/src/host/minilua.c @@ -1134,7 +1134,7 @@ if(!cl->isC){ CallInfo*ci; StkId st,base; Proto*p=cl->p; -luaD_checkstack(L,p->maxstacksize); +luaD_checkstack(L,p->maxstacksize+p->numparams); func=restorestack(L,funcr); if(!p->is_vararg){ base=func+1; diff --git a/love/src/jni/LuaJIT-2.1/src/jit/dis_arm64.lua b/love/src/jni/LuaJIT-2.1/src/jit/dis_arm64.lua index ea7ca828..ad909fbd 100644 --- a/love/src/jni/LuaJIT-2.1/src/jit/dis_arm64.lua +++ b/love/src/jni/LuaJIT-2.1/src/jit/dis_arm64.lua @@ -1089,7 +1089,7 @@ local function disass_ins(ctx) last = "#"..(sf+32 - immr) operands[#operands] = last x = x + 1 - elseif x >= immr then + else name = a2 x = x - immr + 1 end diff --git a/love/src/jni/LuaJIT-2.1/src/jit/dump.lua b/love/src/jni/LuaJIT-2.1/src/jit/dump.lua index 4800a6b8..9eda08c4 100644 --- a/love/src/jni/LuaJIT-2.1/src/jit/dump.lua +++ b/love/src/jni/LuaJIT-2.1/src/jit/dump.lua @@ -102,10 +102,12 @@ end local function fillsymtab(tr, nexit) local t = symtab if nexitsym == 0 then + local maskaddr = jit.arch == "arm" and -2 local ircall = vmdef.ircall for i=0,#ircall do local addr = ircalladdr(i) if addr ~= 0 then + if maskaddr then addr = band(addr, maskaddr) end if addr < 0 then addr = addr + 2^32 end t[addr] = ircall[i] end @@ -217,8 +219,10 @@ local function colorize_text(s) return s end -local function colorize_ansi(s, t) - return format(colortype_ansi[t], s) +local function colorize_ansi(s, t, extra) + local out = format(colortype_ansi[t], s) + if extra then out = "\027[3m"..out end + return out end local irtype_ansi = setmetatable({}, @@ -227,9 +231,10 @@ local irtype_ansi = setmetatable({}, local html_escape = { ["<"] = "<", [">"] = ">", ["&"] = "&", } -local function colorize_html(s, t) +local function colorize_html(s, t, extra) s = gsub(s, "[<>&]", html_escape) - return format('%s', irtype_text[t], s) + return format('%s', + irtype_text[t], extra and " irt_extra" or "", s) end local irtype_html = setmetatable({}, @@ -254,6 +259,7 @@ span.irt_tab { color: #c00000; } span.irt_udt, span.irt_lud { color: #00c0c0; } span.irt_num { color: #4040c0; } span.irt_int, span.irt_i8, span.irt_u8, span.irt_i16, span.irt_u16 { color: #b040b0; } +span.irt_extra { font-style: italic; } ]] @@ -269,6 +275,7 @@ local litname = { if band(mode, 8) ~= 0 then s = s.."C" end if band(mode, 16) ~= 0 then s = s.."R" end if band(mode, 32) ~= 0 then s = s.."I" end + if band(mode, 64) ~= 0 then s = s.."K" end t[mode] = s return s end}), @@ -277,15 +284,18 @@ local litname = { local s = irtype[band(mode, 31)] s = irtype[band(shr(mode, 5), 31)].."."..s if band(mode, 0x800) ~= 0 then s = s.." sext" end - local c = shr(mode, 14) - if c == 2 then s = s.." index" elseif c == 3 then s = s.." check" end + local c = shr(mode, 12) + if c == 1 then s = s.." none" + elseif c == 2 then s = s.." index" + elseif c == 3 then s = s.." check" end t[mode] = s return s end}), ["FLOAD "] = vmdef.irfield, ["FREF "] = vmdef.irfield, ["FPMATH"] = vmdef.irfpm, - ["BUFHDR"] = { [0] = "RESET", "APPEND" }, + ["TMPREF"] = { [0] = "", "IN", "OUT", "INOUT", "", "", "OUT2", "INOUT2" }, + ["BUFHDR"] = { [0] = "RESET", "APPEND", "WRITE" }, ["TOSTR "] = { [0] = "INT", "NUM", "CHAR" }, } @@ -345,7 +355,7 @@ local function formatk(tr, idx, sn) else s = tostring(k) -- For primitives. end - s = colorize(format("%-4s", s), t) + s = colorize(format("%-4s", s), t, band(sn or 0, 0x100000) ~= 0) if slot then s = format("%s @%d", s, slot) end @@ -365,7 +375,7 @@ local function printsnap(tr, snap) out:write(colorize(format("%04d/%04d", ref, ref+1), 14)) else local m, ot, op1, op2 = traceir(tr, ref) - out:write(colorize(format("%04d", ref), band(ot, 31))) + out:write(colorize(format("%04d", ref), band(ot, 31), band(sn, 0x100000) ~= 0)) end out:write(band(sn, 0x10000) == 0 and " " or "|") -- SNAP_FRAME else diff --git a/love/src/jni/LuaJIT-2.1/src/jit/p.lua b/love/src/jni/LuaJIT-2.1/src/jit/p.lua index c0ad6c0e..c9ec1d8b 100644 --- a/love/src/jni/LuaJIT-2.1/src/jit/p.lua +++ b/love/src/jni/LuaJIT-2.1/src/jit/p.lua @@ -238,6 +238,7 @@ local function prof_finish() prof_count1 = nil prof_count2 = nil prof_ud = nil + if out ~= stdout then out:close() end end end diff --git a/love/src/jni/LuaJIT-2.1/src/lib_base.c b/love/src/jni/LuaJIT-2.1/src/lib_base.c index cb2e244e..55e3c6b8 100644 --- a/love/src/jni/LuaJIT-2.1/src/lib_base.c +++ b/love/src/jni/LuaJIT-2.1/src/lib_base.c @@ -19,6 +19,7 @@ #include "lj_gc.h" #include "lj_err.h" #include "lj_debug.h" +#include "lj_buf.h" #include "lj_str.h" #include "lj_tab.h" #include "lj_meta.h" @@ -75,9 +76,10 @@ LJLIB_ASM_(type) LJLIB_REC(.) /* This solves a circular dependency problem -- change FF_next_N as needed. */ LJ_STATIC_ASSERT((int)FF_next == FF_next_N); -LJLIB_ASM(next) +LJLIB_ASM(next) LJLIB_REC(.) { lj_lib_checktab(L, 1); + lj_err_msg(L, LJ_ERR_NEXTIDX); return FFH_UNREACHABLE; } @@ -406,10 +408,22 @@ LJLIB_CF(load) GCstr *name = lj_lib_optstr(L, 2); GCstr *mode = lj_lib_optstr(L, 3); int status; - if (L->base < L->top && (tvisstr(L->base) || tvisnumber(L->base))) { - GCstr *s = lj_lib_checkstr(L, 1); + if (L->base < L->top && + (tvisstr(L->base) || tvisnumber(L->base) || tvisbuf(L->base))) { + const char *s; + MSize len; + if (tvisbuf(L->base)) { + SBufExt *sbx = bufV(L->base); + s = sbx->r; + len = sbufxlen(sbx); + if (!name) name = &G(L)->strempty; /* Buffers are not NUL-terminated. */ + } else { + GCstr *str = lj_lib_checkstr(L, 1); + s = strdata(str); + len = str->len; + } lua_settop(L, 4); /* Ensure env arg exists. */ - status = luaL_loadbufferx(L, strdata(s), s->len, strdata(name ? name : s), + status = luaL_loadbufferx(L, s, len, name ? strdata(name) : s, mode ? strdata(mode) : NULL); } else { lj_lib_checkfunc(L, 1); diff --git a/love/src/jni/LuaJIT-2.1/src/lib_buffer.c b/love/src/jni/LuaJIT-2.1/src/lib_buffer.c new file mode 100644 index 00000000..2e364861 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/src/lib_buffer.c @@ -0,0 +1,356 @@ +/* +** Buffer library. +** Copyright (C) 2005-2021 Mike Pall. See Copyright Notice in luajit.h +*/ + +#define lib_buffer_c +#define LUA_LIB + +#include "lua.h" +#include "lauxlib.h" +#include "lualib.h" + +#include "lj_obj.h" + +#if LJ_HASBUFFER +#include "lj_gc.h" +#include "lj_err.h" +#include "lj_buf.h" +#include "lj_str.h" +#include "lj_tab.h" +#include "lj_udata.h" +#include "lj_meta.h" +#if LJ_HASFFI +#include "lj_ctype.h" +#include "lj_cdata.h" +#include "lj_cconv.h" +#endif +#include "lj_strfmt.h" +#include "lj_serialize.h" +#include "lj_lib.h" + +/* -- Helper functions ---------------------------------------------------- */ + +/* Check that the first argument is a string buffer. */ +static SBufExt *buffer_tobuf(lua_State *L) +{ + if (!(L->base < L->top && tvisbuf(L->base))) + lj_err_argtype(L, 1, "buffer"); + return bufV(L->base); +} + +/* Ditto, but for writers. */ +static LJ_AINLINE SBufExt *buffer_tobufw(lua_State *L) +{ + SBufExt *sbx = buffer_tobuf(L); + setsbufXL_(sbx, L); + return sbx; +} + +#define buffer_toudata(sbx) ((GCudata *)(sbx)-1) + +/* -- Buffer methods ------------------------------------------------------ */ + +#define LJLIB_MODULE_buffer_method + +LJLIB_CF(buffer_method_free) +{ + SBufExt *sbx = buffer_tobuf(L); + lj_bufx_free(L, sbx); + L->top = L->base+1; /* Chain buffer object. */ + return 1; +} + +LJLIB_CF(buffer_method_reset) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobuf(L); + lj_bufx_reset(sbx); + L->top = L->base+1; /* Chain buffer object. */ + return 1; +} + +LJLIB_CF(buffer_method_skip) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobuf(L); + MSize n = (MSize)lj_lib_checkintrange(L, 2, 0, LJ_MAX_BUF); + MSize len = sbufxlen(sbx); + if (n < len) { + sbx->r += n; + } else { + sbx->r = sbx->w = sbx->b; + } + L->top = L->base+1; /* Chain buffer object. */ + return 1; +} + +LJLIB_CF(buffer_method_set) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobuf(L); + GCobj *ref; + const char *p; + MSize len; +#if LJ_HASFFI + if (tviscdata(L->base+1)) { + CTState *cts = ctype_cts(L); + lj_cconv_ct_tv(cts, ctype_get(cts, CTID_P_CVOID), (uint8_t *)&p, + L->base+1, CCF_ARG(2)); + len = (MSize)lj_lib_checkintrange(L, 3, 0, LJ_MAX_BUF); + } else +#endif + { + GCstr *str = lj_lib_checkstrx(L, 2); + p = strdata(str); + len = str->len; + } + lj_bufx_free(L, sbx); + lj_bufx_set_cow(L, sbx, p, len); + ref = gcV(L->base+1); + setgcref(sbx->cowref, ref); + lj_gc_objbarrier(L, buffer_toudata(sbx), ref); + L->top = L->base+1; /* Chain buffer object. */ + return 1; +} + +LJLIB_CF(buffer_method_put) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobufw(L); + ptrdiff_t arg, narg = L->top - L->base; + for (arg = 1; arg < narg; arg++) { + cTValue *o = &L->base[arg], *mo = NULL; + retry: + if (tvisstr(o)) { + lj_buf_putstr((SBuf *)sbx, strV(o)); + } else if (tvisint(o)) { + lj_strfmt_putint((SBuf *)sbx, intV(o)); + } else if (tvisnum(o)) { + lj_strfmt_putfnum((SBuf *)sbx, STRFMT_G14, numV(o)); + } else if (tvisbuf(o)) { + SBufExt *sbx2 = bufV(o); + if (sbx2 == sbx) lj_err_arg(L, arg+1, LJ_ERR_BUFFER_SELF); + lj_buf_putmem((SBuf *)sbx, sbx2->r, sbufxlen(sbx2)); + } else if (!mo && !tvisnil(mo = lj_meta_lookup(L, o, MM_tostring))) { + /* Call __tostring metamethod inline. */ + copyTV(L, L->top++, mo); + copyTV(L, L->top++, o); + lua_call(L, 1, 1); + o = &L->base[arg]; /* The stack may have been reallocated. */ + copyTV(L, &L->base[arg], L->top-1); + L->top = L->base + narg; + goto retry; /* Retry with the result. */ + } else { + lj_err_argtype(L, arg+1, "string/number/__tostring"); + } + /* Probably not useful to inline other __tostring MMs, e.g. FFI numbers. */ + } + L->top = L->base+1; /* Chain buffer object. */ + lj_gc_check(L); + return 1; +} + +LJLIB_CF(buffer_method_putf) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobufw(L); + lj_strfmt_putarg(L, (SBuf *)sbx, 2, 2); + L->top = L->base+1; /* Chain buffer object. */ + lj_gc_check(L); + return 1; +} + +LJLIB_CF(buffer_method_get) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobuf(L); + ptrdiff_t arg, narg = L->top - L->base; + if (narg == 1) { + narg++; + setnilV(L->top++); /* get() is the same as get(nil). */ + } + for (arg = 1; arg < narg; arg++) { + TValue *o = &L->base[arg]; + MSize n = tvisnil(o) ? LJ_MAX_BUF : + (MSize) lj_lib_checkintrange(L, arg+1, 0, LJ_MAX_BUF); + MSize len = sbufxlen(sbx); + if (n > len) n = len; + setstrV(L, o, lj_str_new(L, sbx->r, n)); + sbx->r += n; + } + if (sbx->r == sbx->w) sbx->r = sbx->w = sbx->b; + lj_gc_check(L); + return narg-1; +} + +#if LJ_HASFFI +LJLIB_CF(buffer_method_putcdata) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobufw(L); + const char *p; + MSize len; + if (tviscdata(L->base+1)) { + CTState *cts = ctype_cts(L); + lj_cconv_ct_tv(cts, ctype_get(cts, CTID_P_CVOID), (uint8_t *)&p, + L->base+1, CCF_ARG(2)); + } else { + lj_err_argtype(L, 2, "cdata"); + } + len = (MSize)lj_lib_checkintrange(L, 3, 0, LJ_MAX_BUF); + lj_buf_putmem((SBuf *)sbx, p, len); + L->top = L->base+1; /* Chain buffer object. */ + return 1; +} + +LJLIB_CF(buffer_method_reserve) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobufw(L); + MSize sz = (MSize)lj_lib_checkintrange(L, 2, 0, LJ_MAX_BUF); + GCcdata *cd; + lj_buf_more((SBuf *)sbx, sz); + ctype_loadffi(L); + cd = lj_cdata_new_(L, CTID_P_UINT8, CTSIZE_PTR); + *(void **)cdataptr(cd) = sbx->w; + setcdataV(L, L->top++, cd); + setintV(L->top++, sbufleft(sbx)); + return 2; +} + +LJLIB_CF(buffer_method_commit) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobuf(L); + MSize len = (MSize)lj_lib_checkintrange(L, 2, 0, LJ_MAX_BUF); + if (len > sbufleft(sbx)) lj_err_arg(L, 2, LJ_ERR_NUMRNG); + sbx->w += len; + L->top = L->base+1; /* Chain buffer object. */ + return 1; +} + +LJLIB_CF(buffer_method_ref) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobuf(L); + GCcdata *cd; + ctype_loadffi(L); + cd = lj_cdata_new_(L, CTID_P_UINT8, CTSIZE_PTR); + *(void **)cdataptr(cd) = sbx->r; + setcdataV(L, L->top++, cd); + setintV(L->top++, sbufxlen(sbx)); + return 2; +} +#endif + +LJLIB_CF(buffer_method_encode) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobufw(L); + cTValue *o = lj_lib_checkany(L, 2); + lj_serialize_put(sbx, o); + lj_gc_check(L); + L->top = L->base+1; /* Chain buffer object. */ + return 1; +} + +LJLIB_CF(buffer_method_decode) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobufw(L); + setnilV(L->top++); + sbx->r = lj_serialize_get(sbx, L->top-1); + lj_gc_check(L); + return 1; +} + +LJLIB_CF(buffer_method___gc) +{ + SBufExt *sbx = buffer_tobuf(L); + lj_bufx_free(L, sbx); + return 0; +} + +LJLIB_CF(buffer_method___tostring) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobuf(L); + setstrV(L, L->top-1, lj_str_new(L, sbx->r, sbufxlen(sbx))); + lj_gc_check(L); + return 1; +} + +LJLIB_CF(buffer_method___len) LJLIB_REC(.) +{ + SBufExt *sbx = buffer_tobuf(L); + setintV(L->top-1, (int32_t)sbufxlen(sbx)); + return 1; +} + +LJLIB_PUSH("buffer") LJLIB_SET(__metatable) +LJLIB_PUSH(top-1) LJLIB_SET(__index) + +/* -- Buffer library functions -------------------------------------------- */ + +#define LJLIB_MODULE_buffer + +LJLIB_PUSH(top-2) LJLIB_SET(!) /* Set environment. */ + +LJLIB_CF(buffer_new) +{ + MSize sz = 0; + int targ = 1; + GCtab *env, *dict_str = NULL, *dict_mt = NULL; + GCudata *ud; + SBufExt *sbx; + if (L->base < L->top && !tvistab(L->base)) { + targ = 2; + if (!tvisnil(L->base)) + sz = (MSize)lj_lib_checkintrange(L, 1, 0, LJ_MAX_BUF); + } + if (L->base+targ-1 < L->top) { + GCtab *options = lj_lib_checktab(L, targ); + cTValue *opt_dict, *opt_mt; + opt_dict = lj_tab_getstr(options, lj_str_newlit(L, "dict")); + if (opt_dict && tvistab(opt_dict)) { + dict_str = tabV(opt_dict); + lj_serialize_dict_prep_str(L, dict_str); + } + opt_mt = lj_tab_getstr(options, lj_str_newlit(L, "metatable")); + if (opt_mt && tvistab(opt_mt)) { + dict_mt = tabV(opt_mt); + lj_serialize_dict_prep_mt(L, dict_mt); + } + } + env = tabref(curr_func(L)->c.env); + ud = lj_udata_new(L, sizeof(SBufExt), env); + ud->udtype = UDTYPE_BUFFER; + /* NOBARRIER: The GCudata is new (marked white). */ + setgcref(ud->metatable, obj2gco(env)); + setudataV(L, L->top++, ud); + sbx = (SBufExt *)uddata(ud); + lj_bufx_init(L, sbx); + setgcref(sbx->dict_str, obj2gco(dict_str)); + setgcref(sbx->dict_mt, obj2gco(dict_mt)); + if (sz > 0) lj_buf_need2((SBuf *)sbx, sz); + return 1; +} + +LJLIB_CF(buffer_encode) LJLIB_REC(.) +{ + cTValue *o = lj_lib_checkany(L, 1); + setstrV(L, L->top++, lj_serialize_encode(L, o)); + lj_gc_check(L); + return 1; +} + +LJLIB_CF(buffer_decode) LJLIB_REC(.) +{ + GCstr *str = lj_lib_checkstrx(L, 1); + setnilV(L->top++); + lj_serialize_decode(L, L->top-1, str); + return 1; +} + +/* ------------------------------------------------------------------------ */ + +#include "lj_libdef.h" + +int luaopen_string_buffer(lua_State *L) +{ + LJ_LIB_REG(L, NULL, buffer_method); + lua_getfield(L, -1, "__tostring"); + lua_setfield(L, -2, "tostring"); + LJ_LIB_REG(L, NULL, buffer); + return 1; +} + +#endif diff --git a/love/src/jni/LuaJIT-2.1/src/lib_ffi.c b/love/src/jni/LuaJIT-2.1/src/lib_ffi.c index a29014e5..b4321048 100644 --- a/love/src/jni/LuaJIT-2.1/src/lib_ffi.c +++ b/love/src/jni/LuaJIT-2.1/src/lib_ffi.c @@ -573,6 +573,7 @@ LJLIB_CF(ffi_typeinfo) setintV(lj_tab_setstr(L, t, lj_str_newlit(L, "sib")), (int32_t)ct->sib); if (gcref(ct->name)) { GCstr *s = gco2str(gcref(ct->name)); + if (isdead(G(L), obj2gco(s))) flipwhite(obj2gco(s)); setstrV(L, lj_tab_setstr(L, t, lj_str_newlit(L, "name")), s); } lj_gc_check(L); diff --git a/love/src/jni/LuaJIT-2.1/src/lib_io.c b/love/src/jni/LuaJIT-2.1/src/lib_io.c index a5163e01..b9d8cc75 100644 --- a/love/src/jni/LuaJIT-2.1/src/lib_io.c +++ b/love/src/jni/LuaJIT-2.1/src/lib_io.c @@ -60,12 +60,12 @@ static IOFileUD *io_tofile(lua_State *L) return iof; } -static FILE *io_stdfile(lua_State *L, ptrdiff_t id) +static IOFileUD *io_stdfile(lua_State *L, ptrdiff_t id) { IOFileUD *iof = IOSTDF_IOF(L, id); if (iof->fp == NULL) lj_err_caller(L, LJ_ERR_IOSTDCL); - return iof->fp; + return iof; } static IOFileUD *io_file_new(lua_State *L) @@ -178,7 +178,7 @@ static int io_file_readlen(lua_State *L, FILE *fp, MSize m) MSize n = (MSize)fread(buf, 1, m, fp); setstrV(L, L->top++, lj_str_new(L, buf, (size_t)n)); lj_gc_check(L); - return (n > 0 || m == 0); + return n > 0; } else { int c = getc(fp); ungetc(c, fp); @@ -187,8 +187,9 @@ static int io_file_readlen(lua_State *L, FILE *fp, MSize m) } } -static int io_file_read(lua_State *L, FILE *fp, int start) +static int io_file_read(lua_State *L, IOFileUD *iof, int start) { + FILE *fp = iof->fp; int ok, n, nargs = (int)(L->top - L->base) - start; clearerr(fp); if (nargs == 0) { @@ -224,8 +225,9 @@ static int io_file_read(lua_State *L, FILE *fp, int start) return n - start; } -static int io_file_write(lua_State *L, FILE *fp, int start) +static int io_file_write(lua_State *L, IOFileUD *iof, int start) { + FILE *fp = iof->fp; cTValue *tv; int status = 1; for (tv = L->base+start; tv < L->top; tv++) { @@ -253,13 +255,11 @@ static int io_file_iter(lua_State *L) lj_err_caller(L, LJ_ERR_IOCLFL); L->top = L->base; if (n) { /* Copy upvalues with options to stack. */ - if (n > LUAI_MAXCSTACK) - lj_err_caller(L, LJ_ERR_STKOV); lj_state_checkstack(L, (MSize)n); memcpy(L->top, &fn->c.upvalue[1], n*sizeof(TValue)); L->top += n; } - n = io_file_read(L, iof->fp, 0); + n = io_file_read(L, iof, 0); if (ferror(iof->fp)) lj_err_callermsg(L, strVdata(L->top-2)); if (tvisnil(L->base) && (iof->type & IOFILE_FLAG_CLOSE)) { @@ -284,19 +284,25 @@ static int io_file_lines(lua_State *L) LJLIB_CF(io_method_close) { - IOFileUD *iof = L->base < L->top ? io_tofile(L) : - IOSTDF_IOF(L, GCROOT_IO_OUTPUT); + IOFileUD *iof; + if (L->base < L->top) { + iof = io_tofile(L); + } else { + iof = IOSTDF_IOF(L, GCROOT_IO_OUTPUT); + if (iof->fp == NULL) + lj_err_caller(L, LJ_ERR_IOCLFL); + } return io_file_close(L, iof); } LJLIB_CF(io_method_read) { - return io_file_read(L, io_tofile(L)->fp, 1); + return io_file_read(L, io_tofile(L), 1); } LJLIB_CF(io_method_write) LJLIB_REC(io_write 0) { - return io_file_write(L, io_tofile(L)->fp, 1); + return io_file_write(L, io_tofile(L), 1); } LJLIB_CF(io_method_flush) LJLIB_REC(io_flush 0) @@ -458,7 +464,7 @@ LJLIB_CF(io_write) LJLIB_REC(io_write GCROOT_IO_OUTPUT) LJLIB_CF(io_flush) LJLIB_REC(io_flush GCROOT_IO_OUTPUT) { - return luaL_fileresult(L, fflush(io_stdfile(L, GCROOT_IO_OUTPUT)) == 0, NULL); + return luaL_fileresult(L, fflush(io_stdfile(L, GCROOT_IO_OUTPUT)->fp) == 0, NULL); } static int io_std_getset(lua_State *L, ptrdiff_t id, const char *mode) diff --git a/love/src/jni/LuaJIT-2.1/src/lib_jit.c b/love/src/jni/LuaJIT-2.1/src/lib_jit.c index 12f51d75..817c2967 100644 --- a/love/src/jni/LuaJIT-2.1/src/lib_jit.c +++ b/love/src/jni/LuaJIT-2.1/src/lib_jit.c @@ -346,11 +346,7 @@ LJLIB_CF(jit_util_tracek) ir = &T->ir[ir->op1]; } #if LJ_HASFFI - if (ir->o == IR_KINT64 && !ctype_ctsG(G(L))) { - ptrdiff_t oldtop = savestack(L, L->top); - luaopen_ffi(L); /* Load FFI library on-demand. */ - L->top = restorestack(L, oldtop); - } + if (ir->o == IR_KINT64) ctype_loadffi(L); #endif lj_ir_kvalue(L, L->top-2, ir); setintV(L->top-1, (int32_t)irt_type(ir->t)); diff --git a/love/src/jni/LuaJIT-2.1/src/lib_string.c b/love/src/jni/LuaJIT-2.1/src/lib_string.c index 51d1c4b0..75d855d6 100644 --- a/love/src/jni/LuaJIT-2.1/src/lib_string.c +++ b/love/src/jni/LuaJIT-2.1/src/lib_string.c @@ -640,89 +640,14 @@ LJLIB_CF(string_gsub) /* ------------------------------------------------------------------------ */ -/* Emulate tostring() inline. */ -static GCstr *string_fmt_tostring(lua_State *L, int arg, int retry) -{ - TValue *o = L->base+arg-1; - cTValue *mo; - lj_assertL(o < L->top, "bad usage"); /* Caller already checks for existence. */ - if (LJ_LIKELY(tvisstr(o))) - return strV(o); - if (retry != 2 && !tvisnil(mo = lj_meta_lookup(L, o, MM_tostring))) { - copyTV(L, L->top++, mo); - copyTV(L, L->top++, o); - lua_call(L, 1, 1); - copyTV(L, L->base+arg-1, --L->top); - return NULL; /* Buffer may be overwritten, retry. */ - } - return lj_strfmt_obj(L, o); -} - LJLIB_CF(string_format) LJLIB_REC(.) { - int arg, top = (int)(L->top - L->base); - GCstr *fmt; - SBuf *sb; - FormatState fs; - SFormat sf; int retry = 0; -again: - arg = 1; - sb = lj_buf_tmp_(L); - fmt = lj_lib_checkstr(L, arg); - lj_strfmt_init(&fs, strdata(fmt), fmt->len); - while ((sf = lj_strfmt_parse(&fs)) != STRFMT_EOF) { - if (sf == STRFMT_LIT) { - lj_buf_putmem(sb, fs.str, fs.len); - } else if (sf == STRFMT_ERR) { - lj_err_callerv(L, LJ_ERR_STRFMT, strdata(lj_str_new(L, fs.str, fs.len))); - } else { - if (++arg > top) - luaL_argerror(L, arg, lj_obj_typename[0]); - switch (STRFMT_TYPE(sf)) { - case STRFMT_INT: - if (tvisint(L->base+arg-1)) { - int32_t k = intV(L->base+arg-1); - if (sf == STRFMT_INT) - lj_strfmt_putint(sb, k); /* Shortcut for plain %d. */ - else - lj_strfmt_putfxint(sb, sf, k); - } else { - lj_strfmt_putfnum_int(sb, sf, lj_lib_checknum(L, arg)); - } - break; - case STRFMT_UINT: - if (tvisint(L->base+arg-1)) - lj_strfmt_putfxint(sb, sf, intV(L->base+arg-1)); - else - lj_strfmt_putfnum_uint(sb, sf, lj_lib_checknum(L, arg)); - break; - case STRFMT_NUM: - lj_strfmt_putfnum(sb, sf, lj_lib_checknum(L, arg)); - break; - case STRFMT_STR: { - GCstr *str = string_fmt_tostring(L, arg, retry); - if (str == NULL) - retry = 1; - else if ((sf & STRFMT_T_QUOTED)) - lj_strfmt_putquoted(sb, str); /* No formatting. */ - else - lj_strfmt_putfstr(sb, sf, str); - break; - } - case STRFMT_CHAR: - lj_strfmt_putfchar(sb, sf, lj_lib_checkint(L, arg)); - break; - case STRFMT_PTR: /* No formatting. */ - lj_strfmt_putptr(sb, lj_obj_ptr(G(L), L->base+arg-1)); - break; - default: - lj_assertL(0, "bad string format type"); - break; - } - } - } - if (retry++ == 1) goto again; + SBuf *sb; + do { + sb = lj_buf_tmp_(L); + retry = lj_strfmt_putarg(L, sb, 1, -retry); + } while (retry > 0); setstrV(L, L->top-1, lj_buf_str(L, sb)); lj_gc_check(L); return 1; @@ -743,6 +668,9 @@ LUALIB_API int luaopen_string(lua_State *L) setgcref(basemt_it(g, LJ_TSTR), obj2gco(mt)); settabV(L, lj_tab_setstr(L, mt, mmname_str(g, MM_index)), tabV(L->top-1)); mt->nomm = (uint8_t)(~(1u<top-1)); +#endif return 1; } diff --git a/love/src/jni/LuaJIT-2.1/src/lib_table.c b/love/src/jni/LuaJIT-2.1/src/lib_table.c index 63a6a4c6..0214bb40 100644 --- a/love/src/jni/LuaJIT-2.1/src/lib_table.c +++ b/love/src/jni/LuaJIT-2.1/src/lib_table.c @@ -159,7 +159,7 @@ LJLIB_CF(table_concat) LJLIB_REC(.) SBuf *sb = lj_buf_tmp_(L); SBuf *sbx = lj_buf_puttab(sb, t, sep, i, e); if (LJ_UNLIKELY(!sbx)) { /* Error: bad element type. */ - int32_t idx = (int32_t)(intptr_t)sbufP(sb); + int32_t idx = (int32_t)(intptr_t)sb->w; cTValue *o = lj_tab_getint(t, idx); lj_err_callerv(L, LJ_ERR_TABCAT, lj_obj_itypename[o ? itypemap(o) : ~LJ_TNIL], idx); diff --git a/love/src/jni/LuaJIT-2.1/src/lj_alloc.c b/love/src/jni/LuaJIT-2.1/src/lj_alloc.c index 21fb7af5..165203fa 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_alloc.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_alloc.c @@ -365,7 +365,7 @@ static void *CALL_MREMAP_(void *ptr, size_t osz, size_t nsz, int flags) #define CALL_MREMAP(addr, osz, nsz, mv) CALL_MREMAP_((addr), (osz), (nsz), (mv)) #define CALL_MREMAP_NOMOVE 0 #define CALL_MREMAP_MAYMOVE 1 -#if LJ_64 && !LJ_GC64 +#if LJ_64 && (!LJ_GC64 || LJ_TARGET_ARM64) #define CALL_MREMAP_MV CALL_MREMAP_NOMOVE #else #define CALL_MREMAP_MV CALL_MREMAP_MAYMOVE diff --git a/love/src/jni/LuaJIT-2.1/src/lj_api.c b/love/src/jni/LuaJIT-2.1/src/lj_api.c index 8b00436d..8c60c058 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_api.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_api.c @@ -707,36 +707,10 @@ LUA_API void lua_pushboolean(lua_State *L, int b) incr_top(L); } -#if LJ_64 -static void *lightud_intern(lua_State *L, void *p) -{ - global_State *g = G(L); - uint64_t u = (uint64_t)p; - uint32_t up = lightudup(u); - uint32_t *segmap = mref(g->gc.lightudseg, uint32_t); - MSize segnum = g->gc.lightudnum; - if (segmap) { - MSize seg; - for (seg = 0; seg <= segnum; seg++) - if (segmap[seg] == up) /* Fast path. */ - return (void *)(((uint64_t)seg << LJ_LIGHTUD_BITS_LO) | lightudlo(u)); - segnum++; - } - if (!((segnum-1) & segnum) && segnum != 1) { - if (segnum >= (1 << LJ_LIGHTUD_BITS_SEG)) lj_err_msg(L, LJ_ERR_BADLU); - lj_mem_reallocvec(L, segmap, segnum, segnum ? 2*segnum : 2u, uint32_t); - setmref(g->gc.lightudseg, segmap); - } - g->gc.lightudnum = segnum; - segmap[segnum] = up; - return (void *)(((uint64_t)segnum << LJ_LIGHTUD_BITS_LO) | lightudlo(u)); -} -#endif - LUA_API void lua_pushlightuserdata(lua_State *L, void *p) { #if LJ_64 - p = lightud_intern(L, p); + p = lj_lightud_intern(L, p); #endif setrawlightudV(L->top, p); incr_top(L); @@ -919,11 +893,13 @@ LUA_API int lua_next(lua_State *L, int idx) cTValue *t = index2adr(L, idx); int more; lj_checkapi(tvistab(t), "stack slot %d is not a table", idx); - more = lj_tab_next(L, tabV(t), L->top-1); - if (more) { + more = lj_tab_next(tabV(t), L->top-1, L->top-1); + if (more > 0) { incr_top(L); /* Return new key and value slot. */ - } else { /* End of traversal. */ + } else if (!more) { /* End of traversal. */ L->top--; /* Remove key slot. */ + } else { + lj_err_msg(L, LJ_ERR_NEXTIDX); } return more; } @@ -1179,7 +1155,7 @@ static TValue *cpcall(lua_State *L, lua_CFunction func, void *ud) setfuncV(L, top++, fn); if (LJ_FR2) setnilV(top++); #if LJ_64 - ud = lightud_intern(L, ud); + ud = lj_lightud_intern(L, ud); #endif setrawlightudV(top++, ud); cframe_nres(L->cframe) = 1+0; /* Zero results. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_arch.h b/love/src/jni/LuaJIT-2.1/src/lj_arch.h index d4fd9c9d..ae999467 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_arch.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_arch.h @@ -170,14 +170,10 @@ #define LJ_ARCH_NAME "x86" #define LJ_ARCH_BITS 32 #define LJ_ARCH_ENDIAN LUAJIT_LE -#if LJ_TARGET_WINDOWS || LJ_TARGET_CYGWIN -#define LJ_ABI_WIN 1 -#else -#define LJ_ABI_WIN 0 -#endif #define LJ_TARGET_X86 1 #define LJ_TARGET_X86ORX64 1 #define LJ_TARGET_EHRETREG 0 +#define LJ_TARGET_EHRAREG 8 #define LJ_TARGET_MASKSHIFT 1 #define LJ_TARGET_MASKROT 1 #define LJ_TARGET_UNALIGNED 1 @@ -188,14 +184,10 @@ #define LJ_ARCH_NAME "x64" #define LJ_ARCH_BITS 64 #define LJ_ARCH_ENDIAN LUAJIT_LE -#if LJ_TARGET_WINDOWS || LJ_TARGET_CYGWIN -#define LJ_ABI_WIN 1 -#else -#define LJ_ABI_WIN 0 -#endif #define LJ_TARGET_X64 1 #define LJ_TARGET_X86ORX64 1 #define LJ_TARGET_EHRETREG 0 +#define LJ_TARGET_EHRAREG 16 #define LJ_TARGET_JUMPRANGE 31 /* +-2^31 = +-2GB */ #define LJ_TARGET_MASKSHIFT 1 #define LJ_TARGET_MASKROT 1 @@ -203,6 +195,8 @@ #define LJ_ARCH_NUMMODE LJ_NUMMODE_SINGLE_DUAL #ifndef LUAJIT_DISABLE_GC64 #define LJ_TARGET_GC64 1 +#elif LJ_TARGET_OSX +#error "macOS requires GC64 -- don't disable it" #endif #elif LUAJIT_TARGET == LUAJIT_ARCH_ARM @@ -219,6 +213,7 @@ #define LJ_ABI_EABI 1 #define LJ_TARGET_ARM 1 #define LJ_TARGET_EHRETREG 0 +#define LJ_TARGET_EHRAREG 14 #define LJ_TARGET_JUMPRANGE 25 /* +-2^25 = +-32MB */ #define LJ_TARGET_MASKSHIFT 0 #define LJ_TARGET_MASKROT 1 @@ -249,6 +244,7 @@ #endif #define LJ_TARGET_ARM64 1 #define LJ_TARGET_EHRETREG 0 +#define LJ_TARGET_EHRAREG 30 #define LJ_TARGET_JUMPRANGE 27 /* +-2^27 = +-128MB */ #define LJ_TARGET_MASKSHIFT 1 #define LJ_TARGET_MASKROT 1 @@ -304,6 +300,7 @@ #define LJ_TARGET_PPC 1 #define LJ_TARGET_EHRETREG 3 +#define LJ_TARGET_EHRAREG 65 #define LJ_TARGET_JUMPRANGE 25 /* +-2^25 = +-32MB */ #define LJ_TARGET_MASKSHIFT 0 #define LJ_TARGET_MASKROT 1 @@ -406,6 +403,7 @@ #endif #define LJ_TARGET_MIPS 1 #define LJ_TARGET_EHRETREG 4 +#define LJ_TARGET_EHRAREG 31 #define LJ_TARGET_JUMPRANGE 27 /* 2*2^27 = 256MB-aligned region */ #define LJ_TARGET_MASKSHIFT 1 #define LJ_TARGET_MASKROT 1 @@ -551,6 +549,13 @@ #define LJ_HASFFI 1 #endif +/* Disable or enable the string buffer extension. */ +#if defined(LUAJIT_DISABLE_BUFFER) +#define LJ_HASBUFFER 0 +#else +#define LJ_HASBUFFER 1 +#endif + #if defined(LUAJIT_DISABLE_PROFILE) #define LJ_HASPROFILE 0 #elif LJ_TARGET_POSIX @@ -611,13 +616,10 @@ #define LJ_NO_SYSTEM 1 #endif -#if !defined(LUAJIT_NO_UNWIND) && __GNU_COMPACT_EH__ -/* NYI: no support for compact unwind specification, yet. */ -#define LUAJIT_NO_UNWIND 1 -#endif - -#if defined(LUAJIT_NO_UNWIND) || defined(__symbian__) || LJ_TARGET_IOS || LJ_TARGET_PS3 || LJ_TARGET_PS4 -#define LJ_NO_UNWIND 1 +#if LJ_TARGET_WINDOWS || LJ_TARGET_CYGWIN +#define LJ_ABI_WIN 1 +#else +#define LJ_ABI_WIN 0 #endif #if LJ_TARGET_WINDOWS @@ -632,6 +634,22 @@ extern void *LJ_WIN_LOADLIBA(const char *path); #endif #endif +#if defined(LUAJIT_NO_UNWIND) || __GNU_COMPACT_EH__ || defined(__symbian__) || LJ_TARGET_IOS || LJ_TARGET_PS3 || LJ_TARGET_PS4 +#define LJ_NO_UNWIND 1 +#endif + +#if !LJ_NO_UNWIND && !defined(LUAJIT_UNWIND_INTERNAL) && (LJ_ABI_WIN || (defined(LUAJIT_UNWIND_EXTERNAL) && (defined(__GNUC__) || defined(__clang__)))) +#define LJ_UNWIND_EXT 1 +#else +#define LJ_UNWIND_EXT 0 +#endif + +#if LJ_UNWIND_EXT && LJ_HASJIT && !LJ_TARGET_ARM && !(LJ_ABI_WIN && LJ_TARGET_X86) +#define LJ_UNWIND_JIT 1 +#else +#define LJ_UNWIND_JIT 0 +#endif + /* Compatibility with Lua 5.1 vs. 5.2. */ #ifdef LUAJIT_ENABLE_LUA52COMPAT #define LJ_52 1 diff --git a/love/src/jni/LuaJIT-2.1/src/lj_asm.c b/love/src/jni/LuaJIT-2.1/src/lj_asm.c index 7c9a4237..5968c5e3 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_asm.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_asm.c @@ -11,6 +11,7 @@ #if LJ_HASJIT #include "lj_gc.h" +#include "lj_buf.h" #include "lj_str.h" #include "lj_tab.h" #include "lj_frame.h" @@ -71,6 +72,8 @@ typedef struct ASMState { IRRef snaprename; /* Rename highwater mark for snapshot check. */ SnapNo snapno; /* Current snapshot number. */ SnapNo loopsnapno; /* Loop snapshot number. */ + int snapalloc; /* Current snapshot needs allocation. */ + BloomFilter snapfilt1, snapfilt2; /* Filled with snapshot refs. */ IRRef fuseref; /* Fusion limit (loopref, 0 or FUSE_DISABLED). */ IRRef sectref; /* Section base reference (loopref or 0). */ @@ -84,6 +87,7 @@ typedef struct ASMState { MCode *mcbot; /* Bottom of reserved MCode. */ MCode *mctop; /* Top of generated MCode. */ + MCode *mctoporig; /* Original top of generated MCode. */ MCode *mcloop; /* Pointer to loop MCode (or NULL). */ MCode *invmcp; /* Points to invertible loop branch (or NULL). */ MCode *flagmcp; /* Pending opportunity to merge flag setting ins. */ @@ -694,7 +698,14 @@ static void ra_rename(ASMState *as, Reg down, Reg up) RA_DBGX((as, "rename $f $r $r", regcost_ref(as->cost[up]), down, up)); emit_movrr(as, ir, down, up); /* Backwards codegen needs inverse move. */ if (!ra_hasspill(IR(ref)->s)) { /* Add the rename to the IR. */ - ra_addrename(as, down, ref, as->snapno); + /* + ** The rename is effective at the subsequent (already emitted) exit + ** branch. This is for the current snapshot (as->snapno). Except if we + ** haven't yet allocated any refs for the snapshot (as->snapalloc == 1), + ** then it belongs to the next snapshot. + ** See also the discussion at asm_snap_checkrename(). + */ + ra_addrename(as, down, ref, as->snapno + as->snapalloc); } } @@ -807,11 +818,11 @@ static void ra_leftov(ASMState *as, Reg dest, IRRef lref) } #endif -#if !LJ_64 /* Force a RID_RETLO/RID_RETHI destination register pair (marked as free). */ static void ra_destpair(ASMState *as, IRIns *ir) { Reg destlo = ir->r, desthi = (ir+1)->r; + IRIns *irx = (LJ_64 && !irt_is64(ir->t)) ? ir+1 : ir; /* First spill unrelated refs blocking the destination registers. */ if (!rset_test(as->freeset, RID_RETLO) && destlo != RID_RETLO && desthi != RID_RETLO) @@ -835,29 +846,28 @@ static void ra_destpair(ASMState *as, IRIns *ir) /* Check for conflicts and shuffle the registers as needed. */ if (destlo == RID_RETHI) { if (desthi == RID_RETLO) { -#if LJ_TARGET_X86 - *--as->mcp = XI_XCHGa + RID_RETHI; +#if LJ_TARGET_X86ORX64 + *--as->mcp = REX_64IR(irx, XI_XCHGa + RID_RETHI); #else - emit_movrr(as, ir, RID_RETHI, RID_TMP); - emit_movrr(as, ir, RID_RETLO, RID_RETHI); - emit_movrr(as, ir, RID_TMP, RID_RETLO); + emit_movrr(as, irx, RID_RETHI, RID_TMP); + emit_movrr(as, irx, RID_RETLO, RID_RETHI); + emit_movrr(as, irx, RID_TMP, RID_RETLO); #endif } else { - emit_movrr(as, ir, RID_RETHI, RID_RETLO); - if (desthi != RID_RETHI) emit_movrr(as, ir, desthi, RID_RETHI); + emit_movrr(as, irx, RID_RETHI, RID_RETLO); + if (desthi != RID_RETHI) emit_movrr(as, irx, desthi, RID_RETHI); } } else if (desthi == RID_RETLO) { - emit_movrr(as, ir, RID_RETLO, RID_RETHI); - if (destlo != RID_RETLO) emit_movrr(as, ir, destlo, RID_RETLO); + emit_movrr(as, irx, RID_RETLO, RID_RETHI); + if (destlo != RID_RETLO) emit_movrr(as, irx, destlo, RID_RETLO); } else { - if (desthi != RID_RETHI) emit_movrr(as, ir, desthi, RID_RETHI); - if (destlo != RID_RETLO) emit_movrr(as, ir, destlo, RID_RETLO); + if (desthi != RID_RETHI) emit_movrr(as, irx, desthi, RID_RETHI); + if (destlo != RID_RETLO) emit_movrr(as, irx, destlo, RID_RETLO); } /* Restore spill slots (if any). */ if (ra_hasspill((ir+1)->s)) ra_save(as, ir+1, RID_RETHI); if (ra_hasspill(ir->s)) ra_save(as, ir, RID_RETLO); } -#endif /* -- Snapshot handling --------- ----------------------------------------- */ @@ -892,7 +902,10 @@ static int asm_sunk_store(ASMState *as, IRIns *ira, IRIns *irs) static void asm_snap_alloc1(ASMState *as, IRRef ref) { IRIns *ir = IR(ref); - if (!irref_isk(ref) && (!(ra_used(ir) || ir->r == RID_SUNK))) { + if (!irref_isk(ref) && ir->r != RID_SUNK) { + bloomset(as->snapfilt1, ref); + bloomset(as->snapfilt2, hashrot(ref, ref + HASH_BIAS)); + if (ra_used(ir)) return; if (ir->r == RID_SINK) { ir->r = RID_SUNK; #if LJ_HASFFI @@ -947,11 +960,12 @@ static void asm_snap_alloc1(ASMState *as, IRRef ref) } /* Allocate refs escaping to a snapshot. */ -static void asm_snap_alloc(ASMState *as) +static void asm_snap_alloc(ASMState *as, int snapno) { - SnapShot *snap = &as->T->snap[as->snapno]; + SnapShot *snap = &as->T->snap[snapno]; SnapEntry *map = &as->T->snapmap[snap->mapofs]; MSize n, nent = snap->nent; + as->snapfilt1 = as->snapfilt2 = 0; for (n = 0; n < nent; n++) { SnapEntry sn = map[n]; IRRef ref = snap_ref(sn); @@ -960,7 +974,7 @@ static void asm_snap_alloc(ASMState *as) if (LJ_SOFTFP && (sn & SNAP_SOFTFPNUM)) { lj_assertA(irt_type(IR(ref+1)->t) == IRT_SOFTFP, "snap %d[%d] points to bad SOFTFP IR %04d", - as->snapno, n, ref - REF_BIAS); + snapno, n, ref - REF_BIAS); asm_snap_alloc1(as, ref+1); } } @@ -976,35 +990,26 @@ static void asm_snap_alloc(ASMState *as) */ static int asm_snap_checkrename(ASMState *as, IRRef ren) { - SnapShot *snap = &as->T->snap[as->snapno]; - SnapEntry *map = &as->T->snapmap[snap->mapofs]; - MSize n, nent = snap->nent; - for (n = 0; n < nent; n++) { - SnapEntry sn = map[n]; - IRRef ref = snap_ref(sn); - if (ref == ren || (LJ_SOFTFP && (sn & SNAP_SOFTFPNUM) && ++ref == ren)) { - IRIns *ir = IR(ref); - ra_spill(as, ir); /* Register renamed, so force a spill slot. */ - RA_DBGX((as, "snaprensp $f $s", ref, ir->s)); - return 1; /* Found. */ - } + if (bloomtest(as->snapfilt1, ren) && + bloomtest(as->snapfilt2, hashrot(ren, ren + HASH_BIAS))) { + IRIns *ir = IR(ren); + ra_spill(as, ir); /* Register renamed, so force a spill slot. */ + RA_DBGX((as, "snaprensp $f $s", ren, ir->s)); + return 1; /* Found. */ } return 0; /* Not found. */ } -/* Prepare snapshot for next guard instruction. */ +/* Prepare snapshot for next guard or throwing instruction. */ static void asm_snap_prep(ASMState *as) { - if (as->curins < as->snapref) { - do { - if (as->snapno == 0) return; /* Called by sunk stores before snap #0. */ - as->snapno--; - as->snapref = as->T->snap[as->snapno].ref; - } while (as->curins < as->snapref); - asm_snap_alloc(as); + if (as->snapalloc) { + /* Alloc on first invocation for each snapshot. */ + as->snapalloc = 0; + asm_snap_alloc(as, as->snapno); as->snaprename = as->T->nins; } else { - /* Process any renames above the highwater mark. */ + /* Check any renames above the highwater mark. */ for (; as->snaprename < as->T->nins; as->snaprename++) { IRIns *ir = &as->T->ir[as->snaprename]; if (asm_snap_checkrename(as, ir->op1)) @@ -1013,6 +1018,35 @@ static void asm_snap_prep(ASMState *as) } } +/* Move to previous snapshot when we cross the current snapshot ref. */ +static void asm_snap_prev(ASMState *as) +{ + if (as->curins < as->snapref) { + uintptr_t ofs = (uintptr_t)(as->mctoporig - as->mcp); + if (ofs >= 0x10000) lj_trace_err(as->J, LJ_TRERR_MCODEOV); + do { + if (as->snapno == 0) return; + as->snapno--; + as->snapref = as->T->snap[as->snapno].ref; + as->T->snap[as->snapno].mcofs = (uint16_t)ofs; /* Remember mcode ofs. */ + } while (as->curins < as->snapref); /* May have no ins inbetween. */ + as->snapalloc = 1; + } +} + +/* Fixup snapshot mcode offsetst. */ +static void asm_snap_fixup_mcofs(ASMState *as) +{ + uint32_t sz = (uint32_t)(as->mctoporig - as->mcp); + SnapShot *snap = as->T->snap; + SnapNo i; + for (i = as->T->nsnap-1; i > 0; i--) { + /* Compute offset from mcode start and store in correct snapshot. */ + snap[i].mcofs = (uint16_t)(sz - snap[i-1].mcofs); + } + snap[0].mcofs = 0; +} + /* -- Miscellaneous helpers ----------------------------------------------- */ /* Calculate stack adjustment. */ @@ -1057,6 +1091,7 @@ static void asm_snew(ASMState *as, IRIns *ir) { const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_str_new]; IRRef args[3]; + asm_snap_prep(as); args[0] = ASMREF_L; /* lua_State *L */ args[1] = ir->op1; /* const char *str */ args[2] = ir->op2; /* size_t len */ @@ -1069,6 +1104,7 @@ static void asm_tnew(ASMState *as, IRIns *ir) { const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_tab_new1]; IRRef args[2]; + asm_snap_prep(as); args[0] = ASMREF_L; /* lua_State *L */ args[1] = ASMREF_TMP1; /* uint32_t ahsize */ as->gcsteps++; @@ -1081,6 +1117,7 @@ static void asm_tdup(ASMState *as, IRIns *ir) { const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_tab_dup]; IRRef args[2]; + asm_snap_prep(as); args[0] = ASMREF_L; /* lua_State *L */ args[1] = ir->op1; /* const GCtab *kt */ as->gcsteps++; @@ -1106,28 +1143,43 @@ static void asm_gcstep(ASMState *as, IRIns *ir) /* -- Buffer operations --------------------------------------------------- */ -static void asm_tvptr(ASMState *as, Reg dest, IRRef ref); +static void asm_tvptr(ASMState *as, Reg dest, IRRef ref, MSize mode); +#if LJ_HASBUFFER +static void asm_bufhdr_write(ASMState *as, Reg sb); +#endif static void asm_bufhdr(ASMState *as, IRIns *ir) { Reg sb = ra_dest(as, ir, RSET_GPR); - if ((ir->op2 & IRBUFHDR_APPEND)) { + switch (ir->op2) { + case IRBUFHDR_RESET: { + Reg tmp = ra_scratch(as, rset_exclude(RSET_GPR, sb)); + IRIns irbp; + irbp.ot = IRT(0, IRT_PTR); /* Buffer data pointer type. */ + emit_storeofs(as, &irbp, tmp, sb, offsetof(SBuf, w)); + emit_loadofs(as, &irbp, tmp, sb, offsetof(SBuf, b)); + break; + } + case IRBUFHDR_APPEND: { /* Rematerialize const buffer pointer instead of likely spill. */ IRIns *irp = IR(ir->op1); if (!(ra_hasreg(irp->r) || irp == ir-1 || (irp == ir-2 && !ra_used(ir-1)))) { - while (!(irp->o == IR_BUFHDR && !(irp->op2 & IRBUFHDR_APPEND))) + while (!(irp->o == IR_BUFHDR && irp->op2 == IRBUFHDR_RESET)) irp = IR(irp->op1); if (irref_isk(irp->op1)) { ra_weak(as, ra_allocref(as, ir->op1, RSET_GPR)); ir = irp; } } - } else { - Reg tmp = ra_scratch(as, rset_exclude(RSET_GPR, sb)); - /* Passing ir isn't strictly correct, but it's an IRT_PGC, too. */ - emit_storeofs(as, ir, tmp, sb, offsetof(SBuf, p)); - emit_loadofs(as, ir, tmp, sb, offsetof(SBuf, b)); + break; + } +#if LJ_HASBUFFER + case IRBUFHDR_WRITE: + asm_bufhdr_write(as, sb); + break; +#endif + default: lj_assertA(0, "bad BUFHDR op2 %d", ir->op2); break; } #if LJ_TARGET_X86ORX64 ra_left(as, sb, ir->op1); @@ -1179,7 +1231,7 @@ static void asm_bufput(ASMState *as, IRIns *ir) if (args[1] == ASMREF_TMP1) { Reg tmp = ra_releasetmp(as, ASMREF_TMP1); if (kchar == -129) - asm_tvptr(as, tmp, irs->op1); + asm_tvptr(as, tmp, irs->op1, IRTMPREF_IN1); else ra_allockreg(as, kchar, tmp); } @@ -1201,6 +1253,7 @@ static void asm_tostr(ASMState *as, IRIns *ir) { const CCallInfo *ci; IRRef args[2]; + asm_snap_prep(as); args[0] = ASMREF_L; as->gcsteps++; if (ir->op2 == IRTOSTR_NUM) { @@ -1216,7 +1269,7 @@ static void asm_tostr(ASMState *as, IRIns *ir) asm_setupresult(as, ir, ci); /* GCstr * */ asm_gencall(as, ci, args); if (ir->op2 == IRTOSTR_NUM) - asm_tvptr(as, ra_releasetmp(as, ASMREF_TMP1), ir->op1); + asm_tvptr(as, ra_releasetmp(as, ASMREF_TMP1), ir->op1, IRTMPREF_IN1); } #if LJ_32 && LJ_HASFFI && !LJ_SOFTFP && !LJ_TARGET_X86 @@ -1257,12 +1310,19 @@ static void asm_newref(ASMState *as, IRIns *ir) IRRef args[3]; if (ir->r == RID_SINK) return; + asm_snap_prep(as); args[0] = ASMREF_L; /* lua_State *L */ args[1] = ir->op1; /* GCtab *t */ args[2] = ASMREF_TMP1; /* cTValue *key */ asm_setupresult(as, ir, ci); /* TValue * */ asm_gencall(as, ci, args); - asm_tvptr(as, ra_releasetmp(as, ASMREF_TMP1), ir->op2); + asm_tvptr(as, ra_releasetmp(as, ASMREF_TMP1), ir->op2, IRTMPREF_IN1); +} + +static void asm_tmpref(ASMState *as, IRIns *ir) +{ + Reg r = ra_dest(as, ir, RSET_GPR); + asm_tvptr(as, r, ir->op1, ir->op2); } static void asm_lref(ASMState *as, IRIns *ir) @@ -1744,6 +1804,7 @@ static void asm_ir(ASMState *as, IRIns *ir) case IR_NEWREF: asm_newref(as, ir); break; case IR_UREFO: case IR_UREFC: asm_uref(as, ir); break; case IR_FREF: asm_fref(as, ir); break; + case IR_TMPREF: asm_tmpref(as, ir); break; case IR_STRREF: asm_strref(as, ir); break; case IR_LREF: asm_lref(as, ir); break; @@ -1838,8 +1899,7 @@ static void asm_head_side(ASMState *as) if (as->snapno && as->topslot > as->parent->topslot) { /* Force snap #0 alloc to prevent register overwrite in stack check. */ - as->snapno = 0; - asm_snap_alloc(as); + asm_snap_alloc(as, 0); } allow = asm_head_side_base(as, irp, allow); @@ -2078,6 +2138,9 @@ static void asm_setup_regsp(ASMState *as) #endif ra_setup(as); +#if LJ_TARGET_ARM64 + ra_setkref(as, RID_GL, (intptr_t)J2G(as->J)); +#endif /* Clear reg/sp for constants. */ for (ir = IR(T->nk), lastir = IR(REF_BASE); ir < lastir; ir++) { @@ -2100,6 +2163,7 @@ static void asm_setup_regsp(ASMState *as) as->snaprename = nins; as->snapref = nins; as->snapno = T->nsnap; + as->snapalloc = 0; as->stopins = REF_BASE; as->orignins = nins; @@ -2148,6 +2212,10 @@ static void asm_setup_regsp(ASMState *as) ir->prev = (uint16_t)REGSP_HINT((rload & 15)); rload = lj_ror(rload, 4); continue; + case IR_TMPREF: + if ((ir->op2 & IRTMPREF_OUT2) && as->evenspill < 4) + as->evenspill = 4; /* TMPREF OUT2 needs two TValues on the stack. */ + break; #endif case IR_CALLXS: { CCallInfo ci; @@ -2157,7 +2225,17 @@ static void asm_setup_regsp(ASMState *as) as->modset |= RSET_SCRATCH; continue; } - case IR_CALLN: case IR_CALLA: case IR_CALLL: case IR_CALLS: { + case IR_CALLL: + /* lj_vm_next needs two TValues on the stack. */ +#if LJ_TARGET_X64 && LJ_ABI_WIN + if (ir->op2 == IRCALL_lj_vm_next && as->evenspill < SPS_FIRST + 4) + as->evenspill = SPS_FIRST + 4; +#else + if (SPS_FIRST < 4 && ir->op2 == IRCALL_lj_vm_next && as->evenspill < 4) + as->evenspill = 4; +#endif + /* fallthrough */ + case IR_CALLN: case IR_CALLA: case IR_CALLS: { const CCallInfo *ci = &lj_ir_callinfo[ir->op2]; ir->prev = asm_setup_call_slots(as, ir, ci); if (inloop) @@ -2165,7 +2243,6 @@ static void asm_setup_regsp(ASMState *as) (RSET_SCRATCH & ~RSET_FPR) : RSET_SCRATCH; continue; } -#if LJ_SOFTFP || (LJ_32 && LJ_HASFFI) case IR_HIOP: switch ((ir-1)->o) { #if LJ_SOFTFP && LJ_TARGET_ARM @@ -2176,7 +2253,7 @@ static void asm_setup_regsp(ASMState *as) } break; #endif -#if !LJ_SOFTFP && LJ_NEED_FP64 +#if !LJ_SOFTFP && LJ_NEED_FP64 && LJ_32 && LJ_HASFFI case IR_CONV: if (irt_isfp((ir-1)->t)) { ir->prev = REGSP_HINT(RID_FPRET); @@ -2184,7 +2261,7 @@ static void asm_setup_regsp(ASMState *as) } #endif /* fallthrough */ - case IR_CALLN: case IR_CALLXS: + case IR_CALLN: case IR_CALLL: case IR_CALLS: case IR_CALLXS: #if LJ_SOFTFP case IR_MIN: case IR_MAX: #endif @@ -2195,7 +2272,6 @@ static void asm_setup_regsp(ASMState *as) break; } break; -#endif #if LJ_SOFTFP case IR_MIN: case IR_MAX: if ((ir+1)->o != IR_HIOP) break; @@ -2250,13 +2326,23 @@ static void asm_setup_regsp(ASMState *as) } /* fallthrough */ /* for integer POW */ case IR_DIV: case IR_MOD: - if (!irt_isnum(ir->t)) { + if ((LJ_64 && LJ_SOFTFP) || !irt_isnum(ir->t)) { ir->prev = REGSP_HINT(RID_RET); if (inloop) as->modset |= (RSET_SCRATCH & RSET_GPR); continue; } break; +#if LJ_64 && LJ_SOFTFP + case IR_ADD: case IR_SUB: case IR_MUL: + if (irt_isnum(ir->t)) { + ir->prev = REGSP_HINT(RID_RET); + if (inloop) + as->modset |= (RSET_SCRATCH & RSET_GPR); + continue; + } + break; +#endif case IR_FPMATH: #if LJ_TARGET_X86ORX64 if (ir->op2 <= IRFPM_TRUNC) { @@ -2327,7 +2413,6 @@ void lj_asm_trace(jit_State *J, GCtrace *T) { ASMState as_; ASMState *as = &as_; - MCode *origtop; /* Remove nops/renames left over from ASM restart due to LJ_TRERR_MCODELM. */ { @@ -2355,7 +2440,7 @@ void lj_asm_trace(jit_State *J, GCtrace *T) as->parent = J->parent ? traceref(J, J->parent) : NULL; /* Reserve MCode memory. */ - as->mctop = origtop = lj_mcode_reserve(J, &as->mcbot); + as->mctop = as->mctoporig = lj_mcode_reserve(J, &as->mcbot); as->mcp = as->mctop; as->mclim = as->mcbot + MCLIM_REDZONE; asm_setup_target(as); @@ -2417,6 +2502,7 @@ void lj_asm_trace(jit_State *J, GCtrace *T) lj_assertA(!(LJ_32 && irt_isint64(ir->t)), "IR %04d has unsplit 64 bit type", (int)(ir - as->ir) - REF_BIAS); + asm_snap_prev(as); if (!ra_used(ir) && !ir_sideeff(ir) && (as->flags & JIT_F_OPT_DCE)) continue; /* Dead-code elimination can be soooo easy. */ if (irt_isguard(ir->t)) @@ -2450,6 +2536,9 @@ void lj_asm_trace(jit_State *J, GCtrace *T) memcpy(J->curfinal->ir + as->orignins, T->ir + as->orignins, (T->nins - as->orignins) * sizeof(IRIns)); /* Copy RENAMEs. */ T->nins = J->curfinal->nins; + /* Fill mcofs of any unprocessed snapshots. */ + as->curins = REF_FIRST; + asm_snap_prev(as); break; /* Done. */ } @@ -2468,13 +2557,16 @@ void lj_asm_trace(jit_State *J, GCtrace *T) /* Set trace entry point before fixing up tail to allow link to self. */ T->mcode = as->mcp; T->mcloop = as->mcloop ? (MSize)((char *)as->mcloop - (char *)as->mcp) : 0; - if (!as->loopref) + if (as->loopref) + asm_loop_tail_fixup(as); + else asm_tail_fixup(as, T->link); /* Note: this may change as->mctop! */ T->szmcode = (MSize)((char *)as->mctop - (char *)as->mcp); + asm_snap_fixup_mcofs(as); #if LJ_TARGET_MCODE_FIXUP asm_mcode_fixup(T->mcode, T->szmcode); #endif - lj_mcode_sync(T->mcode, origtop); + lj_mcode_sync(T->mcode, as->mctoporig); } #undef IR diff --git a/love/src/jni/LuaJIT-2.1/src/lj_asm_arm.h b/love/src/jni/LuaJIT-2.1/src/lj_asm_arm.h index c93c99c9..cc608c0d 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_asm_arm.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_asm_arm.h @@ -185,6 +185,9 @@ static Reg asm_fuseahuref(ASMState *as, IRRef ref, int32_t *ofsp, RegSet allow, *ofsp = (ofs & 255); /* Mask out less bits to allow LDRD. */ return ra_allock(as, (ofs & ~255), allow); } + } else if (ir->o == IR_TMPREF) { + *ofsp = 0; + return RID_SP; } } *ofsp = 0; @@ -498,6 +501,30 @@ static void asm_retf(ASMState *as, IRIns *ir) emit_lso(as, ARMI_LDR, RID_TMP, base, -4); } +/* -- Buffer operations --------------------------------------------------- */ + +#if LJ_HASBUFFER +static void asm_bufhdr_write(ASMState *as, Reg sb) +{ + Reg tmp = ra_scratch(as, rset_exclude(RSET_GPR, sb)); + IRIns irgc; + int32_t addr = i32ptr((void *)&J2G(as->J)->cur_L); + irgc.ot = IRT(0, IRT_PGC); /* GC type. */ + emit_storeofs(as, &irgc, RID_TMP, sb, offsetof(SBuf, L)); + if ((as->flags & JIT_F_ARMV6T2)) { + emit_dnm(as, ARMI_BFI, RID_TMP, lj_fls(SBUF_MASK_FLAG), tmp); + } else { + emit_dnm(as, ARMI_ORR, RID_TMP, RID_TMP, tmp); + emit_dn(as, ARMI_AND|ARMI_K12|SBUF_MASK_FLAG, tmp, tmp); + } + emit_lso(as, ARMI_LDR, RID_TMP, + ra_allock(as, (addr & ~4095), + rset_exclude(rset_exclude(RSET_GPR, sb), tmp)), + (addr & 4095)); + emit_loadofs(as, &irgc, tmp, sb, offsetof(SBuf, L)); +} +#endif + /* -- Type conversions ---------------------------------------------------- */ #if !LJ_SOFTFP @@ -666,35 +693,55 @@ static void asm_strto(ASMState *as, IRIns *ir) /* -- Memory references --------------------------------------------------- */ /* Get pointer to TValue. */ -static void asm_tvptr(ASMState *as, Reg dest, IRRef ref) +static void asm_tvptr(ASMState *as, Reg dest, IRRef ref, MSize mode) { - IRIns *ir = IR(ref); - if (irt_isnum(ir->t)) { - if (irref_isk(ref)) { - /* Use the number constant itself as a TValue. */ - ra_allockreg(as, i32ptr(ir_knum(ir)), dest); - } else { + if ((mode & IRTMPREF_IN1)) { + IRIns *ir = IR(ref); + if (irt_isnum(ir->t)) { + if ((mode & IRTMPREF_OUT1)) { #if LJ_SOFTFP - lj_assertA(0, "unsplit FP op"); + lj_assertA(irref_isk(ref), "unsplit FP op"); + emit_dm(as, ARMI_MOV, dest, RID_SP); + emit_lso(as, ARMI_STR, + ra_allock(as, (int32_t)ir_knum(ir)->u32.lo, RSET_GPR), + RID_SP, 0); + emit_lso(as, ARMI_STR, + ra_allock(as, (int32_t)ir_knum(ir)->u32.hi, RSET_GPR), + RID_SP, 4); #else - /* Otherwise force a spill and use the spill slot. */ - emit_opk(as, ARMI_ADD, dest, RID_SP, ra_spill(as, ir), RSET_GPR); + Reg src = ra_alloc1(as, ref, RSET_FPR); + emit_dm(as, ARMI_MOV, dest, RID_SP); + emit_vlso(as, ARMI_VSTR_D, src, RID_SP, 0); #endif + } else if (irref_isk(ref)) { + /* Use the number constant itself as a TValue. */ + ra_allockreg(as, i32ptr(ir_knum(ir)), dest); + } else { +#if LJ_SOFTFP + lj_assertA(0, "unsplit FP op"); +#else + /* Otherwise force a spill and use the spill slot. */ + emit_opk(as, ARMI_ADD, dest, RID_SP, ra_spill(as, ir), RSET_GPR); +#endif + } + } else { + /* Otherwise use [sp] and [sp+4] to hold the TValue. + ** This assumes the following call has max. 4 args. + */ + Reg type; + emit_dm(as, ARMI_MOV, dest, RID_SP); + if (!irt_ispri(ir->t)) { + Reg src = ra_alloc1(as, ref, RSET_GPR); + emit_lso(as, ARMI_STR, src, RID_SP, 0); + } + if (LJ_SOFTFP && (ir+1)->o == IR_HIOP && !irt_isnil((ir+1)->t)) + type = ra_alloc1(as, ref+1, RSET_GPR); + else + type = ra_allock(as, irt_toitype(ir->t), RSET_GPR); + emit_lso(as, ARMI_STR, type, RID_SP, 4); } } else { - /* Otherwise use [sp] and [sp+4] to hold the TValue. */ - RegSet allow = rset_exclude(RSET_GPR, dest); - Reg type; emit_dm(as, ARMI_MOV, dest, RID_SP); - if (!irt_ispri(ir->t)) { - Reg src = ra_alloc1(as, ref, allow); - emit_lso(as, ARMI_STR, src, RID_SP, 0); - } - if (LJ_SOFTFP && (ir+1)->o == IR_HIOP) - type = ra_alloc1(as, ref+1, allow); - else - type = ra_allock(as, irt_toitype(ir->t), allow); - emit_lso(as, ARMI_STR, type, RID_SP, 4); } } @@ -1086,6 +1133,7 @@ static void asm_ahuvload(ASMState *as, IRIns *ir) } idx = asm_fuseahuref(as, ir->op1, &ofs, allow, (!LJ_SOFTFP && t == IRT_NUM) ? 1024 : 4096); + if (ir->o == IR_VLOAD) ofs += 8 * ir->op2; if (!hiop || type == RID_NONE) { rset_clear(allow, idx); if (ofs < 256 && ra_hasreg(dest) && (dest & 1) == 0 && @@ -1837,15 +1885,15 @@ static void asm_int64comp(ASMState *as, IRIns *ir) } #endif -/* -- Support for 64 bit ops in 32 bit mode ------------------------------- */ +/* -- Split register ops -------------------------------------------------- */ -/* Hiword op of a split 64 bit op. Previous op must be the loword op. */ +/* Hiword op of a split 32/32 bit op. Previous op is the loword op. */ static void asm_hiop(ASMState *as, IRIns *ir) { -#if LJ_HASFFI || LJ_SOFTFP /* HIOP is marked as a store because it needs its own DCE logic. */ int uselo = ra_used(ir-1), usehi = ra_used(ir); /* Loword/hiword used? */ if (LJ_UNLIKELY(!(as->flags & JIT_F_OPT_DCE))) uselo = usehi = 1; +#if LJ_HASFFI || LJ_SOFTFP if ((ir-1)->o <= IR_NE) { /* 64 bit integer or FP comparisons. ORDER IR. */ as->curins--; /* Always skip the loword comparison. */ #if LJ_SOFTFP @@ -1876,6 +1924,7 @@ static void asm_hiop(ASMState *as, IRIns *ir) asm_xstore_(as, ir, 4); return; } +#endif if (!usehi) return; /* Skip unused hiword op for all remaining ops. */ switch ((ir-1)->o) { #if LJ_HASFFI @@ -1894,6 +1943,9 @@ static void asm_hiop(ASMState *as, IRIns *ir) asm_intneg(as, ir, ARMI_RSC); asm_intneg(as, ir-1, ARMI_RSB|ARMI_S); break; + case IR_CNEWI: + /* Nothing to do here. Handled by lo op itself. */ + break; #endif #if LJ_SOFTFP case IR_SLOAD: case IR_ALOAD: case IR_HLOAD: case IR_ULOAD: case IR_VLOAD: @@ -1901,25 +1953,16 @@ static void asm_hiop(ASMState *as, IRIns *ir) if (!uselo) ra_allocref(as, ir->op1, RSET_GPR); /* Mark lo op as used. */ break; + case IR_ASTORE: case IR_HSTORE: case IR_USTORE: case IR_TOSTR: case IR_TMPREF: + /* Nothing to do here. Handled by lo op itself. */ + break; #endif - case IR_CALLN: - case IR_CALLS: - case IR_CALLXS: + case IR_CALLN: case IR_CALLL: case IR_CALLS: case IR_CALLXS: if (!uselo) ra_allocref(as, ir->op1, RID2RSET(RID_RETLO)); /* Mark lo op as used. */ break; -#if LJ_SOFTFP - case IR_ASTORE: case IR_HSTORE: case IR_USTORE: case IR_TOSTR: -#endif - case IR_CNEWI: - /* Nothing to do here. Handled by lo op itself. */ - break; default: lj_assertA(0, "bad HIOP for op %d", (ir-1)->o); break; } -#else - /* Unused without SOFTFP or FFI. */ - UNUSED(as); UNUSED(ir); lj_assertA(0, "unexpected HIOP"); -#endif } /* -- Profiling ----------------------------------------------------------- */ @@ -2021,6 +2064,8 @@ static void asm_stack_restore(ASMState *as, SnapShot *snap) } else if ((sn & SNAP_SOFTFPNUM)) { type = ra_alloc1(as, ref+1, rset_exclude(RSET_GPRODD, RID_BASE)); #endif + } else if ((sn & SNAP_KEYINDEX)) { + type = ra_allock(as, (int32_t)LJ_KEYINDEX, odd); } else { type = ra_allock(as, (int32_t)irt_toitype(ir->t), odd); } @@ -2082,6 +2127,12 @@ static void asm_loop_fixup(ASMState *as) } } +/* Fixup the tail of the loop. */ +static void asm_loop_tail_fixup(ASMState *as) +{ + UNUSED(as); /* Nothing to do. */ +} + /* -- Head of trace ------------------------------------------------------- */ /* Reload L register from g->cur_L. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_asm_arm64.h b/love/src/jni/LuaJIT-2.1/src/lj_asm_arm64.h index b346b034..67c53ee2 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_asm_arm64.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_asm_arm64.h @@ -198,6 +198,9 @@ static Reg asm_fuseahuref(ASMState *as, IRRef ref, int32_t *ofsp, RegSet allow, return RID_GL; } } + } else if (ir->o == IR_TMPREF) { + *ofsp = (int32_t)glofs(as, &J2G(as->J)->tmptv); + return RID_GL; } } *ofsp = 0; @@ -457,8 +460,11 @@ static void asm_gencall(ASMState *as, const CCallInfo *ci, IRRef *args) static void asm_setupresult(ASMState *as, IRIns *ir, const CCallInfo *ci) { RegSet drop = RSET_SCRATCH; + int hiop = ((ir+1)->o == IR_HIOP && !irt_isnil((ir+1)->t)); if (ra_hasreg(ir->r)) rset_clear(drop, ir->r); /* Dest reg handled below. */ + if (hiop && ra_hasreg((ir+1)->r)) + rset_clear(drop, (ir+1)->r); /* Dest reg handled below. */ ra_evictset(as, drop); /* Evictions must be performed first. */ if (ra_used(ir)) { lj_assertA(!irt_ispri(ir->t), "PRI dest"); @@ -470,6 +476,8 @@ static void asm_setupresult(ASMState *as, IRIns *ir, const CCallInfo *ci) } else { ra_destreg(as, ir, RID_FPRET); } + } else if (hiop) { + ra_destpair(as, ir); } else { ra_destreg(as, ir, RID_RET); } @@ -519,6 +527,21 @@ static void asm_retf(ASMState *as, IRIns *ir) emit_lso(as, A64I_LDRx, RID_TMP, base, -8); } +/* -- Buffer operations --------------------------------------------------- */ + +#if LJ_HASBUFFER +static void asm_bufhdr_write(ASMState *as, Reg sb) +{ + Reg tmp = ra_scratch(as, rset_exclude(RSET_GPR, sb)); + IRIns irgc; + irgc.ot = IRT(0, IRT_PGC); /* GC type. */ + emit_storeofs(as, &irgc, RID_TMP, sb, offsetof(SBuf, L)); + emit_dn(as, A64I_BFMx | A64F_IMMS(lj_fls(SBUF_MASK_FLAG)) | A64F_IMMR(0), RID_TMP, tmp); + emit_getgl(as, RID_TMP, cur_L); + emit_loadofs(as, &irgc, tmp, sb, offsetof(SBuf, L)); +} +#endif + /* -- Type conversions ---------------------------------------------------- */ static void asm_tointg(ASMState *as, IRIns *ir, Reg left) @@ -602,7 +625,7 @@ static void asm_conv(ASMState *as, IRIns *ir) emit_dn(as, A64I_SXTW, dest, left); } } else { - if (st64) { + if (st64 && !(ir->op2 & IRCONV_NONE)) { /* This is either a 32 bit reg/reg mov which zeroes the hiword ** or a load of the loword from a 64 bit address. */ @@ -675,22 +698,23 @@ static void asm_tvstore64(ASMState *as, Reg base, int32_t ofs, IRRef ref) } /* Get pointer to TValue. */ -static void asm_tvptr(ASMState *as, Reg dest, IRRef ref) +static void asm_tvptr(ASMState *as, Reg dest, IRRef ref, MSize mode) { - IRIns *ir = IR(ref); - if (irt_isnum(ir->t)) { - if (irref_isk(ref)) { - /* Use the number constant itself as a TValue. */ - ra_allockreg(as, i64ptr(ir_knum(ir)), dest); + if ((mode & IRTMPREF_IN1)) { + IRIns *ir = IR(ref); + if (irt_isnum(ir->t)) { + if (irref_isk(ref) && !(mode & IRTMPREF_OUT1)) { + /* Use the number constant itself as a TValue. */ + ra_allockreg(as, i64ptr(ir_knum(ir)), dest); + return; + } + emit_lso(as, A64I_STRd, (ra_alloc1(as, ref, RSET_FPR) & 31), dest, 0); } else { - /* Otherwise force a spill and use the spill slot. */ - emit_opk(as, A64I_ADDx, dest, RID_SP, ra_spill(as, ir), RSET_GPR); + asm_tvstore64(as, dest, 0, ref); } - } else { - /* Otherwise use g->tmptv to hold the TValue. */ - asm_tvstore64(as, dest, 0, ref); - ra_allockreg(as, i64ptr(&J2G(as->J)->tmptv), dest); } + /* g->tmptv holds the TValue(s). */ + emit_dn(as, A64I_ADDx^emit_isk12(glofs(as, &J2G(as->J)->tmptv)), dest, RID_GL); } static void asm_aref(ASMState *as, IRIns *ir) @@ -767,7 +791,7 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge) } } else if (irt_isaddr(kt)) { if (isk) { - int64_t kk = ((int64_t)irt_toitype(irkey->t) << 47) | irkey[1].tv.u64; + int64_t kk = ((int64_t)irt_toitype(kt) << 47) | irkey[1].tv.u64; scr = ra_allock(as, kk, allow); } else { scr = ra_scratch(as, allow); @@ -775,7 +799,7 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge) rset_clear(allow, scr); } else { lj_assertA(irt_ispri(kt) && !irt_isnil(kt), "bad HREF key type"); - type = ra_allock(as, ~((int64_t)~irt_toitype(ir->t) << 47), allow); + type = ra_allock(as, ~((int64_t)~irt_toitype(kt) << 47), allow); scr = ra_scratch(as, rset_clear(allow, type)); rset_clear(allow, scr); } @@ -824,7 +848,7 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge) emit_lso(as, A64I_LDRx, scr, dest, offsetof(Node, key.u64)); } } else { - emit_nm(as, A64I_CMPw, scr, type); + emit_nm(as, A64I_CMPx, scr, type); emit_lso(as, A64I_LDRx, scr, dest, offsetof(Node, key)); } @@ -1059,6 +1083,7 @@ static void asm_ahuvload(ASMState *as, IRIns *ir) } type = ra_scratch(as, rset_clear(gpr, tmp)); idx = asm_fuseahuref(as, ir->op1, &ofs, rset_clear(gpr, type), A64I_LDRx); + if (ir->o == IR_VLOAD) ofs += 8 * ir->op2; /* Always do the type check, even if the load result is unused. */ asm_guardcc(as, irt_isnum(ir->t) ? CC_LS : CC_NE); if (irt_type(ir->t) >= IRT_NUM) { @@ -1261,17 +1286,13 @@ static void asm_tbar(ASMState *as, IRIns *ir) { Reg tab = ra_alloc1(as, ir->op1, RSET_GPR); Reg link = ra_scratch(as, rset_exclude(RSET_GPR, tab)); - Reg gr = ra_allock(as, i64ptr(J2G(as->J)), - rset_exclude(rset_exclude(RSET_GPR, tab), link)); Reg mark = RID_TMP; MCLabel l_end = emit_label(as); emit_lso(as, A64I_STRx, link, tab, (int32_t)offsetof(GCtab, gclist)); emit_lso(as, A64I_STRB, mark, tab, (int32_t)offsetof(GCtab, marked)); - emit_lso(as, A64I_STRx, tab, gr, - (int32_t)offsetof(global_State, gc.grayagain)); + emit_setgl(as, tab, gc.grayagain); emit_dn(as, A64I_ANDw^emit_isk13(~LJ_GC_BLACK, 0), mark, mark); - emit_lso(as, A64I_LDRx, link, gr, - (int32_t)offsetof(global_State, gc.grayagain)); + emit_getgl(as, link, gc.grayagain); emit_cond_branch(as, CC_EQ, l_end); emit_n(as, A64I_TSTw^emit_isk13(LJ_GC_BLACK, 0), mark); emit_lso(as, A64I_LDRB, mark, tab, (int32_t)offsetof(GCtab, marked)); @@ -1291,7 +1312,7 @@ static void asm_obar(ASMState *as, IRIns *ir) args[0] = ASMREF_TMP1; /* global_State *g */ args[1] = ir->op1; /* TValue *tv */ asm_gencall(as, ci, args); - ra_allockreg(as, i64ptr(J2G(as->J)), ra_releasetmp(as, ASMREF_TMP1) ); + emit_dm(as, A64I_MOVx, ra_releasetmp(as, ASMREF_TMP1), RID_GL); obj = IR(ir->op1)->r; tmp = ra_scratch(as, rset_exclude(allow, obj)); emit_cond_branch(as, CC_EQ, l_end); @@ -1704,13 +1725,25 @@ static void asm_comp(ASMState *as, IRIns *ir) #define asm_equal(as, ir) asm_comp(as, ir) -/* -- Support for 64 bit ops in 32 bit mode ------------------------------- */ +/* -- Split register ops -------------------------------------------------- */ -/* Hiword op of a split 64 bit op. Previous op must be the loword op. */ +/* Hiword op of a split 64/64 bit op. Previous op is the loword op. */ static void asm_hiop(ASMState *as, IRIns *ir) { - UNUSED(as); UNUSED(ir); - lj_assertA(0, "unexpected HIOP"); /* Unused on 64 bit. */ + /* HIOP is marked as a store because it needs its own DCE logic. */ + int uselo = ra_used(ir-1), usehi = ra_used(ir); /* Loword/hiword used? */ + if (LJ_UNLIKELY(!(as->flags & JIT_F_OPT_DCE))) uselo = usehi = 1; + if (!usehi) return; /* Skip unused hiword op for all remaining ops. */ + switch ((ir-1)->o) { + case IR_CALLN: + case IR_CALLL: + case IR_CALLS: + case IR_CALLXS: + if (!uselo) + ra_allocref(as, ir->op1, RID2RSET(RID_RETLO)); /* Mark lo op as used. */ + break; + default: lj_assertA(0, "bad HIOP for op %d", (ir-1)->o); break; + } } /* -- Profiling ----------------------------------------------------------- */ @@ -1781,7 +1814,14 @@ static void asm_stack_restore(ASMState *as, SnapShot *snap) IRIns *ir = IR(ref); if ((sn & SNAP_NORESTORE)) continue; - if (irt_isnum(ir->t)) { + if ((sn & SNAP_KEYINDEX)) { + RegSet allow = rset_exclude(RSET_GPR, RID_BASE); + Reg r = irref_isk(ref) ? ra_allock(as, ir->i, allow) : + ra_alloc1(as, ref, allow); + rset_clear(allow, r); + emit_lso(as, A64I_STRw, r, RID_BASE, ofs); + emit_lso(as, A64I_STRw, ra_allock(as, LJ_KEYINDEX, allow), RID_BASE, ofs+4); + } else if (irt_isnum(ir->t)) { Reg src = ra_alloc1(as, ref, RSET_FPR); emit_lso(as, A64I_STRd, (src & 31), RID_BASE, ofs); } else { @@ -1804,7 +1844,7 @@ static void asm_gc_check(ASMState *as) const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_gc_step_jit]; IRRef args[2]; MCLabel l_end; - Reg tmp1, tmp2; + Reg tmp2; ra_evictset(as, RSET_SCRATCH); l_end = emit_label(as); /* Exit trace if in GCSatomic or GCSfinalize. Avoids syncing GC objects. */ @@ -1813,17 +1853,14 @@ static void asm_gc_check(ASMState *as) args[0] = ASMREF_TMP1; /* global_State *g */ args[1] = ASMREF_TMP2; /* MSize steps */ asm_gencall(as, ci, args); - tmp1 = ra_releasetmp(as, ASMREF_TMP1); + emit_dm(as, A64I_MOVx, ra_releasetmp(as, ASMREF_TMP1), RID_GL); tmp2 = ra_releasetmp(as, ASMREF_TMP2); emit_loadi(as, tmp2, as->gcsteps); /* Jump around GC step if GC total < GC threshold. */ emit_cond_branch(as, CC_LS, l_end); emit_nm(as, A64I_CMPx, RID_TMP, tmp2); - emit_lso(as, A64I_LDRx, tmp2, tmp1, - (int32_t)offsetof(global_State, gc.threshold)); - emit_lso(as, A64I_LDRx, RID_TMP, tmp1, - (int32_t)offsetof(global_State, gc.total)); - ra_allockreg(as, i64ptr(J2G(as->J)), tmp1); + emit_getgl(as, tmp2, gc.threshold); + emit_getgl(as, RID_TMP, gc.total); as->gcsteps = 0; checkmclim(as); } @@ -1846,6 +1883,12 @@ static void asm_loop_fixup(ASMState *as) } } +/* Fixup the tail of the loop. */ +static void asm_loop_tail_fixup(ASMState *as) +{ + UNUSED(as); /* Nothing to do. */ +} + /* -- Head of trace ------------------------------------------------------- */ /* Reload L register from g->cur_L. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_asm_mips.h b/love/src/jni/LuaJIT-2.1/src/lj_asm_mips.h index 321d63c0..1497ad9c 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_asm_mips.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_asm_mips.h @@ -64,19 +64,31 @@ static Reg ra_alloc2(ASMState *as, IRIns *ir, RegSet allow) /* Setup spare long-range jump slots per mcarea. */ static void asm_sparejump_setup(ASMState *as) { - MCode *mxp = as->mcbot; - if (((uintptr_t)mxp & (LJ_PAGESIZE-1)) == sizeof(MCLink)) { + MCode *mxp = as->mctop; + if ((char *)mxp == (char *)as->J->mcarea + as->J->szmcarea) { + mxp -= MIPS_SPAREJUMP*2; lj_assertA(MIPSI_NOP == 0, "bad NOP"); memset(mxp, 0, MIPS_SPAREJUMP*2*sizeof(MCode)); - mxp += MIPS_SPAREJUMP*2; - lj_assertA(mxp < as->mctop, "MIPS_SPAREJUMP too big"); - lj_mcode_sync(as->mcbot, mxp); - lj_mcode_commitbot(as->J, mxp); - as->mcbot = mxp; - as->mclim = as->mcbot + MCLIM_REDZONE; + as->mctop = mxp; } } +static MCode *asm_sparejump_use(MCode *mcarea, MCode tjump) +{ + MCode *mxp = (MCode *)((char *)mcarea + ((MCLink *)mcarea)->size); + int slot = MIPS_SPAREJUMP; + while (slot--) { + mxp -= 2; + if (*mxp == tjump) { + return mxp; + } else if (*mxp == MIPSI_NOP) { + *mxp = tjump; + return mxp; + } + } + return NULL; +} + /* Setup exit stub after the end of each trace. */ static void asm_exitstub_setup(ASMState *as) { @@ -181,6 +193,9 @@ static Reg asm_fuseahuref(ASMState *as, IRRef ref, int32_t *ofsp, RegSet allow) return ra_allock(as, ofs-(int16_t)ofs, allow); } } + } else if (ir->o == IR_TMPREF) { + *ofsp = (int32_t)(offsetof(global_State, tmptv)-32768); + return RID_JGL; } } *ofsp = 0; @@ -336,19 +351,15 @@ static void asm_gencall(ASMState *as, const CCallInfo *ci, IRRef *args) static void asm_setupresult(ASMState *as, IRIns *ir, const CCallInfo *ci) { RegSet drop = RSET_SCRATCH; -#if LJ_32 int hiop = ((ir+1)->o == IR_HIOP && !irt_isnil((ir+1)->t)); -#endif #if !LJ_SOFTFP if ((ci->flags & CCI_NOFPRCLOBBER)) drop &= ~RSET_FPR; #endif if (ra_hasreg(ir->r)) rset_clear(drop, ir->r); /* Dest reg handled below. */ -#if LJ_32 if (hiop && ra_hasreg((ir+1)->r)) rset_clear(drop, (ir+1)->r); /* Dest reg handled below. */ -#endif ra_evictset(as, drop); /* Evictions must be performed first. */ if (ra_used(ir)) { lj_assertA(!irt_ispri(ir->t), "PRI dest"); @@ -377,10 +388,8 @@ static void asm_setupresult(ASMState *as, IRIns *ir, const CCallInfo *ci) } else { ra_destreg(as, ir, RID_FPRET); } -#if LJ_32 } else if (hiop) { ra_destpair(as, ir); -#endif } else { ra_destreg(as, ir, RID_RET); } @@ -450,6 +459,27 @@ static void asm_retf(ASMState *as, IRIns *ir) emit_tsi(as, MIPSI_AL, RID_TMP, base, -8); } +/* -- Buffer operations --------------------------------------------------- */ + +#if LJ_HASBUFFER +static void asm_bufhdr_write(ASMState *as, Reg sb) +{ + Reg tmp = ra_scratch(as, rset_exclude(RSET_GPR, sb)); + IRIns irgc; + irgc.ot = IRT(0, IRT_PGC); /* GC type. */ + emit_storeofs(as, &irgc, RID_TMP, sb, offsetof(SBuf, L)); + if ((as->flags & JIT_F_MIPSXXR2)) { + emit_tsml(as, LJ_64 ? MIPSI_DINS : MIPSI_INS, RID_TMP, tmp, + lj_fls(SBUF_MASK_FLAG), 0); + } else { + emit_dst(as, MIPSI_OR, RID_TMP, RID_TMP, tmp); + emit_tsi(as, MIPSI_ANDI, tmp, tmp, SBUF_MASK_FLAG); + } + emit_getgl(as, RID_TMP, cur_L); + emit_loadofs(as, &irgc, tmp, sb, offsetof(SBuf, L)); +} +#endif + /* -- Type conversions ---------------------------------------------------- */ #if !LJ_SOFTFP @@ -739,7 +769,7 @@ static void asm_conv(ASMState *as, IRIns *ir) } } } else { - if (st64) { + if (st64 && !(ir->op2 & IRCONV_NONE)) { /* This is either a 32 bit reg/reg mov which zeroes the hiword ** or a load of the loword from a 64 bit address. */ @@ -827,34 +857,63 @@ static void asm_tvstore64(ASMState *as, Reg base, int32_t ofs, IRRef ref) #endif /* Get pointer to TValue. */ -static void asm_tvptr(ASMState *as, Reg dest, IRRef ref) +static void asm_tvptr(ASMState *as, Reg dest, IRRef ref, MSize mode) { - IRIns *ir = IR(ref); - if (irt_isnum(ir->t)) { - if (irref_isk(ref)) /* Use the number constant itself as a TValue. */ - ra_allockreg(as, igcptr(ir_knum(ir)), dest); - else /* Otherwise force a spill and use the spill slot. */ - emit_tsi(as, MIPSI_AADDIU, dest, RID_SP, ra_spill(as, ir)); - } else { - /* Otherwise use g->tmptv to hold the TValue. */ -#if LJ_32 - RegSet allow = rset_exclude(RSET_GPR, dest); - Reg type; - emit_tsi(as, MIPSI_ADDIU, dest, RID_JGL, (int32_t)(offsetof(global_State, tmptv)-32768)); - if (!irt_ispri(ir->t)) { - Reg src = ra_alloc1(as, ref, allow); - emit_setgl(as, src, tmptv.gcr); - } - if (LJ_SOFTFP && (ir+1)->o == IR_HIOP) - type = ra_alloc1(as, ref+1, allow); - else - type = ra_allock(as, (int32_t)irt_toitype(ir->t), allow); - emit_setgl(as, type, tmptv.it); + int32_t tmpofs = (int32_t)(offsetof(global_State, tmptv)-32768); + if ((mode & IRTMPREF_IN1)) { + IRIns *ir = IR(ref); + if (irt_isnum(ir->t)) { + if ((mode & IRTMPREF_OUT1)) { +#if LJ_SOFTFP + emit_tsi(as, MIPSI_AADDIU, dest, RID_JGL, tmpofs); +#if LJ_64 + emit_setgl(as, ra_alloc1(as, ref, RSET_GPR), tmptv.u64); #else - asm_tvstore64(as, dest, 0, ref); - emit_tsi(as, MIPSI_DADDIU, dest, RID_JGL, - (int32_t)(offsetof(global_State, tmptv)-32768)); + lj_assertA(irref_isk(ref), "unsplit FP op"); + emit_setgl(as, + ra_allock(as, (int32_t)ir_knum(ir)->u32.lo, RSET_GPR), + tmptv.u32.lo); + emit_setgl(as, + ra_allock(as, (int32_t)ir_knum(ir)->u32.hi, RSET_GPR), + tmptv.u32.hi); #endif +#else + Reg src = ra_alloc1(as, ref, RSET_FPR); + emit_tsi(as, MIPSI_AADDIU, dest, RID_JGL, tmpofs); + emit_tsi(as, MIPSI_SDC1, (src & 31), RID_JGL, tmpofs); +#endif + } else if (irref_isk(ref)) { + /* Use the number constant itself as a TValue. */ + ra_allockreg(as, igcptr(ir_knum(ir)), dest); + } else { +#if LJ_SOFTFP32 + lj_assertA(0, "unsplit FP op"); +#else + /* Otherwise force a spill and use the spill slot. */ + emit_tsi(as, MIPSI_AADDIU, dest, RID_SP, ra_spill(as, ir)); +#endif + } + } else { + /* Otherwise use g->tmptv to hold the TValue. */ +#if LJ_32 + Reg type; + emit_tsi(as, MIPSI_ADDIU, dest, RID_JGL, tmpofs); + if (!irt_ispri(ir->t)) { + Reg src = ra_alloc1(as, ref, RSET_GPR); + emit_setgl(as, src, tmptv.gcr); + } + if (LJ_SOFTFP && (ir+1)->o == IR_HIOP && !irt_isnil((ir+1)->t)) + type = ra_alloc1(as, ref+1, RSET_GPR); + else + type = ra_allock(as, (int32_t)irt_toitype(ir->t), RSET_GPR); + emit_setgl(as, type, tmptv.it); +#else + asm_tvstore64(as, dest, 0, ref); + emit_tsi(as, MIPSI_DADDIU, dest, RID_JGL, tmpofs); +#endif + } + } else { + emit_tsi(as, MIPSI_AADDIU, dest, RID_JGL, tmpofs); } } @@ -909,11 +968,16 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge) MCLabel l_end, l_loop, l_next; rset_clear(allow, tab); -#if LJ_SOFTFP32 - if (!isk) { - key = ra_alloc1(as, refkey, allow); - rset_clear(allow, key); - if (irkey[1].o == IR_HIOP) { + if (!LJ_SOFTFP && irt_isnum(kt)) { + key = ra_alloc1(as, refkey, RSET_FPR); + tmpnum = ra_scratch(as, rset_exclude(RSET_FPR, key)); + } else { + if (!irt_ispri(kt)) { + key = ra_alloc1(as, refkey, allow); + rset_clear(allow, key); + } +#if LJ_32 + if (LJ_SOFTFP && irkey[1].o == IR_HIOP) { if (ra_hasreg((irkey+1)->r)) { type = tmpnum = (irkey+1)->r; tmp1 = ra_scratch(as, allow); @@ -924,23 +988,11 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge) } rset_clear(allow, tmpnum); } else { - type = ra_allock(as, (int32_t)irt_toitype(irkey->t), allow); + type = ra_allock(as, (int32_t)irt_toitype(kt), allow); rset_clear(allow, type); } - } -#else - if (!LJ_SOFTFP && irt_isnum(kt)) { - key = ra_alloc1(as, refkey, RSET_FPR); - tmpnum = ra_scratch(as, rset_exclude(RSET_FPR, key)); - } else if (!irt_ispri(kt)) { - key = ra_alloc1(as, refkey, allow); - rset_clear(allow, key); -#if LJ_32 - type = ra_allock(as, (int32_t)irt_toitype(irkey->t), allow); - rset_clear(allow, type); #endif } -#endif tmp2 = ra_scratch(as, allow); rset_clear(allow, tmp2); #if LJ_64 @@ -953,10 +1005,10 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge) } else { int64_t k; if (isk && irt_isaddr(kt)) { - k = ((int64_t)irt_toitype(irkey->t) << 47) | irkey[1].tv.u64; + k = ((int64_t)irt_toitype(kt) << 47) | irkey[1].tv.u64; } else { lj_assertA(irt_ispri(kt) && !irt_isnil(kt), "bad HREF key type"); - k = ~((int64_t)~irt_toitype(ir->t) << 47); + k = ~((int64_t)~irt_toitype(kt) << 47); } cmp64 = ra_allock(as, k, allow); rset_clear(allow, cmp64); @@ -1352,6 +1404,7 @@ static void asm_ahuvload(ASMState *as, IRIns *ir) #endif } idx = asm_fuseahuref(as, ir->op1, &ofs, allow); + if (ir->o == IR_VLOAD) ofs += 8 * ir->op2; rset_clear(allow, idx); if (irt_isnum(t)) { asm_guard(as, MIPSI_BEQ, RID_TMP, RID_ZERO); @@ -2327,15 +2380,15 @@ static void asm_comp64eq(ASMState *as, IRIns *ir) } #endif -/* -- Support for 64 bit ops in 32 bit mode ------------------------------- */ +/* -- Split register ops -------------------------------------------------- */ -/* Hiword op of a split 64 bit op. Previous op must be the loword op. */ +/* Hiword op of a split 32/32 or 64/64 bit op. Previous op is the loword op. */ static void asm_hiop(ASMState *as, IRIns *ir) { -#if LJ_32 && (LJ_HASFFI || LJ_SOFTFP) /* HIOP is marked as a store because it needs its own DCE logic. */ int uselo = ra_used(ir-1), usehi = ra_used(ir); /* Loword/hiword used? */ if (LJ_UNLIKELY(!(as->flags & JIT_F_OPT_DCE))) uselo = usehi = 1; +#if LJ_32 && (LJ_HASFFI || LJ_SOFTFP) if ((ir-1)->o == IR_CONV) { /* Conversions to/from 64 bit. */ as->curins--; /* Always skip the CONV. */ #if LJ_HASFFI && !LJ_SOFTFP @@ -2382,38 +2435,33 @@ static void asm_hiop(ASMState *as, IRIns *ir) } return; } +#endif if (!usehi) return; /* Skip unused hiword op for all remaining ops. */ switch ((ir-1)->o) { -#if LJ_HASFFI +#if LJ_32 && LJ_HASFFI case IR_ADD: as->curins--; asm_add64(as, ir); break; case IR_SUB: as->curins--; asm_sub64(as, ir); break; case IR_NEG: as->curins--; asm_neg64(as, ir); break; + case IR_CNEWI: + /* Nothing to do here. Handled by lo op itself. */ + break; #endif -#if LJ_SOFTFP +#if LJ_32 && LJ_SOFTFP case IR_SLOAD: case IR_ALOAD: case IR_HLOAD: case IR_ULOAD: case IR_VLOAD: case IR_STRTO: if (!uselo) ra_allocref(as, ir->op1, RSET_GPR); /* Mark lo op as used. */ break; + case IR_ASTORE: case IR_HSTORE: case IR_USTORE: case IR_TOSTR: case IR_TMPREF: + /* Nothing to do here. Handled by lo op itself. */ + break; #endif - case IR_CALLN: - case IR_CALLS: - case IR_CALLXS: + case IR_CALLN: case IR_CALLL: case IR_CALLS: case IR_CALLXS: if (!uselo) ra_allocref(as, ir->op1, RID2RSET(RID_RETLO)); /* Mark lo op as used. */ break; -#if LJ_SOFTFP - case IR_ASTORE: case IR_HSTORE: case IR_USTORE: case IR_TOSTR: -#endif - case IR_CNEWI: - /* Nothing to do here. Handled by lo op itself. */ - break; default: lj_assertA(0, "bad HIOP for op %d", (ir-1)->o); break; } -#else - /* Unused on MIPS64 or without SOFTFP or FFI. */ - UNUSED(as); UNUSED(ir); lj_assertA(0, "unexpected HIOP"); -#endif } /* -- Profiling ----------------------------------------------------------- */ @@ -2513,6 +2561,8 @@ static void asm_stack_restore(ASMState *as, SnapShot *snap) } else if ((sn & SNAP_SOFTFPNUM)) { type = ra_alloc1(as, ref+1, rset_exclude(RSET_GPR, RID_BASE)); #endif + } else if ((sn & SNAP_KEYINDEX)) { + type = ra_allock(as, (int32_t)LJ_KEYINDEX, allow); } else { type = ra_allock(as, (int32_t)irt_toitype(ir->t), allow); } @@ -2575,6 +2625,12 @@ static void asm_loop_fixup(ASMState *as) } } +/* Fixup the tail of the loop. */ +static void asm_loop_tail_fixup(ASMState *as) +{ + if (as->loopinv) as->mctop--; +} + /* -- Head of trace ------------------------------------------------------- */ /* Coalesce BASE register for a root trace. */ @@ -2582,7 +2638,6 @@ static void asm_head_root_base(ASMState *as) { IRIns *ir = IR(REF_BASE); Reg r = ir->r; - if (as->loopinv) as->mctop--; if (ra_hasreg(r)) { ra_free(as, r); if (rset_test(as->modset, r) || irt_ismarked(ir->t)) @@ -2597,7 +2652,6 @@ static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow) { IRIns *ir = IR(REF_BASE); Reg r = ir->r; - if (as->loopinv) as->mctop--; if (ra_hasreg(r)) { ra_free(as, r); if (rset_test(as->modset, r) || irt_ismarked(ir->t)) @@ -2704,21 +2758,17 @@ void lj_asm_patchexit(jit_State *J, GCtrace *T, ExitNo exitno, MCode *target) patchbranch: p[-1] = (p[-1] & 0xffff0000u) | (delta & 0xffffu); *p = MIPSI_NOP; /* Replace the load of the exit number. */ - cstop = p; + cstop = p+1; if (!cstart) cstart = p-1; } else { /* Branch out of range. Use spare jump slot in mcarea. */ - int i; - for (i = (int)(sizeof(MCLink)/sizeof(MCode)); - i < (int)(sizeof(MCLink)/sizeof(MCode)+MIPS_SPAREJUMP*2); - i += 2) { - if (mcarea[i] == tjump) { - delta = mcarea+i - p; - goto patchbranch; - } else if (mcarea[i] == MIPSI_NOP) { - mcarea[i] = tjump; - cstart = mcarea+i; - delta = mcarea+i - p; + MCode *mcjump = asm_sparejump_use(mcarea, tjump); + if (mcjump) { + lj_mcode_sync(mcjump, mcjump+1); + delta = mcjump - p; + if (((delta + 0x8000) >> 16) == 0) { goto patchbranch; + } else { + lj_assertJ(0, "spare jump out of range: -Osizemcode too big"); } } /* Ignore jump slot overflow. Child trace is simply not attached. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_asm_ppc.h b/love/src/jni/LuaJIT-2.1/src/lj_asm_ppc.h index 579f8e2a..ac5d88ce 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_asm_ppc.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_asm_ppc.h @@ -156,6 +156,9 @@ static Reg asm_fuseahuref(ASMState *as, IRRef ref, int32_t *ofsp, RegSet allow) return ra_allock(as, ofs-(int16_t)ofs, allow); } } + } else if (ir->o == IR_TMPREF) { + *ofsp = (int32_t)(offsetof(global_State, tmptv)-32768); + return RID_JGL; } } *ofsp = 0; @@ -337,10 +340,8 @@ static void asm_setupresult(ASMState *as, IRIns *ir, const CCallInfo *ci) } else { ra_destreg(as, ir, RID_FPRET); } -#if LJ_32 } else if (hiop) { ra_destpair(as, ir); -#endif } else { ra_destreg(as, ir, RID_RET); } @@ -389,6 +390,21 @@ static void asm_retf(ASMState *as, IRIns *ir) emit_tai(as, PPCI_LWZ, RID_TMP, base, -8); } +/* -- Buffer operations --------------------------------------------------- */ + +#if LJ_HASBUFFER +static void asm_bufhdr_write(ASMState *as, Reg sb) +{ + Reg tmp = ra_scratch(as, rset_exclude(RSET_GPR, sb)); + IRIns irgc; + irgc.ot = IRT(0, IRT_PGC); /* GC type. */ + emit_storeofs(as, &irgc, RID_TMP, sb, offsetof(SBuf, L)); + emit_rot(as, PPCI_RLWIMI, RID_TMP, tmp, 0, 31-lj_fls(SBUF_MASK_FLAG), 31); + emit_getgl(as, RID_TMP, cur_L); + emit_loadofs(as, &irgc, tmp, sb, offsetof(SBuf, L)); +} +#endif + /* -- Type conversions ---------------------------------------------------- */ #if !LJ_SOFTFP @@ -567,28 +583,54 @@ static void asm_strto(ASMState *as, IRIns *ir) /* -- Memory references --------------------------------------------------- */ /* Get pointer to TValue. */ -static void asm_tvptr(ASMState *as, Reg dest, IRRef ref) +static void asm_tvptr(ASMState *as, Reg dest, IRRef ref, MSize mode) { - IRIns *ir = IR(ref); - if (irt_isnum(ir->t)) { - if (irref_isk(ref)) /* Use the number constant itself as a TValue. */ - ra_allockreg(as, i32ptr(ir_knum(ir)), dest); - else /* Otherwise force a spill and use the spill slot. */ - emit_tai(as, PPCI_ADDI, dest, RID_SP, ra_spill(as, ir)); - } else { - /* Otherwise use g->tmptv to hold the TValue. */ - RegSet allow = rset_exclude(RSET_GPR, dest); - Reg type; - emit_tai(as, PPCI_ADDI, dest, RID_JGL, (int32_t)offsetof(global_State, tmptv)-32768); - if (!irt_ispri(ir->t)) { - Reg src = ra_alloc1(as, ref, allow); - emit_setgl(as, src, tmptv.gcr); + int32_t tmpofs = (int32_t)(offsetof(global_State, tmptv)-32768); + if ((mode & IRTMPREF_IN1)) { + IRIns *ir = IR(ref); + if (irt_isnum(ir->t)) { + if ((mode & IRTMPREF_OUT1)) { +#if LJ_SOFTFP + lj_assertA(irref_isk(ref), "unsplit FP op"); + emit_tai(as, PPCI_ADDI, dest, RID_JGL, tmpofs); + emit_setgl(as, + ra_allock(as, (int32_t)ir_knum(ir)->u32.lo, RSET_GPR), + tmptv.u32.lo); + emit_setgl(as, + ra_allock(as, (int32_t)ir_knum(ir)->u32.hi, RSET_GPR), + tmptv.u32.hi); +#else + Reg src = ra_alloc1(as, ref, RSET_FPR); + emit_tai(as, PPCI_ADDI, dest, RID_JGL, tmpofs); + emit_fai(as, PPCI_STFD, src, RID_JGL, tmpofs); +#endif + } else if (irref_isk(ref)) { + /* Use the number constant itself as a TValue. */ + ra_allockreg(as, i32ptr(ir_knum(ir)), dest); + } else { +#if LJ_SOFTFP + lj_assertA(0, "unsplit FP op"); +#else + /* Otherwise force a spill and use the spill slot. */ + emit_tai(as, PPCI_ADDI, dest, RID_SP, ra_spill(as, ir)); +#endif + } + } else { + /* Otherwise use g->tmptv to hold the TValue. */ + Reg type; + emit_tai(as, PPCI_ADDI, dest, RID_JGL, tmpofs); + if (!irt_ispri(ir->t)) { + Reg src = ra_alloc1(as, ref, RSET_GPR); + emit_setgl(as, src, tmptv.gcr); + } + if (LJ_SOFTFP && (ir+1)->o == IR_HIOP && !irt_isnil((ir+1)->t)) + type = ra_alloc1(as, ref+1, RSET_GPR); + else + type = ra_allock(as, irt_toitype(ir->t), RSET_GPR); + emit_setgl(as, type, tmptv.it); } - if (LJ_SOFTFP && (ir+1)->o == IR_HIOP) - type = ra_alloc1(as, ref+1, allow); - else - type = ra_allock(as, irt_toitype(ir->t), allow); - emit_setgl(as, type, tmptv.it); + } else { + emit_tai(as, PPCI_ADDI, dest, RID_JGL, tmpofs); } } @@ -894,7 +936,7 @@ static void asm_fload(ASMState *as, IRIns *ir) int32_t ofs; if (ir->op1 == REF_NIL) { /* FLOAD from GG_State with offset. */ idx = RID_JGL; - ofs = (ir->op2 << 2) - 32768; + ofs = (ir->op2 << 2) - 32768 - GG_OFS(g); } else { idx = ra_alloc1(as, ir->op1, RSET_GPR); if (ir->op2 == IRFL_TAB_ARRAY) { @@ -975,6 +1017,10 @@ static void asm_ahuvload(ASMState *as, IRIns *ir) rset_clear(allow, dest); } idx = asm_fuseahuref(as, ir->op1, &ofs, allow); + if (ir->o == IR_VLOAD) { + ofs = ofs != AHUREF_LSX ? ofs + 8 * ir->op2 : + ir->op2 ? 8 * ir->op2 : AHUREF_LSX; + } if (irt_isnum(t)) { Reg tisnum = ra_allock(as, (int32_t)LJ_TISNUM, rset_exclude(allow, idx)); asm_guardcc(as, CC_GE); @@ -1057,7 +1103,8 @@ static void asm_sload(ASMState *as, IRIns *ir) lj_assertA(irt_isguard(ir->t) || !(ir->op2 & IRSLOAD_TYPECHECK), "inconsistent SLOAD variant"); lj_assertA(LJ_DUALNUM || - !irt_isint(t) || (ir->op2 & (IRSLOAD_CONVERT|IRSLOAD_FRAME)), + !irt_isint(t) || + (ir->op2 & (IRSLOAD_CONVERT|IRSLOAD_FRAME|IRSLOAD_KEYINDEX)), "bad SLOAD type"); #if LJ_SOFTFP lj_assertA(!(ir->op2 & IRSLOAD_CONVERT), @@ -1894,15 +1941,15 @@ static void asm_comp64(ASMState *as, IRIns *ir) } #endif -/* -- Support for 64 bit ops in 32 bit mode ------------------------------- */ +/* -- Split register ops -------------------------------------------------- */ -/* Hiword op of a split 64 bit op. Previous op must be the loword op. */ +/* Hiword op of a split 32/32 bit op. Previous op is be the loword op. */ static void asm_hiop(ASMState *as, IRIns *ir) { -#if LJ_HASFFI || LJ_SOFTFP /* HIOP is marked as a store because it needs its own DCE logic. */ int uselo = ra_used(ir-1), usehi = ra_used(ir); /* Loword/hiword used? */ if (LJ_UNLIKELY(!(as->flags & JIT_F_OPT_DCE))) uselo = usehi = 1; +#if LJ_HASFFI || LJ_SOFTFP if ((ir-1)->o == IR_CONV) { /* Conversions to/from 64 bit. */ as->curins--; /* Always skip the CONV. */ #if LJ_HASFFI && !LJ_SOFTFP @@ -1937,12 +1984,16 @@ static void asm_hiop(ASMState *as, IRIns *ir) } return; } +#endif if (!usehi) return; /* Skip unused hiword op for all remaining ops. */ switch ((ir-1)->o) { #if LJ_HASFFI case IR_ADD: as->curins--; asm_add64(as, ir); break; case IR_SUB: as->curins--; asm_sub64(as, ir); break; case IR_NEG: as->curins--; asm_neg64(as, ir); break; + case IR_CNEWI: + /* Nothing to do here. Handled by lo op itself. */ + break; #endif #if LJ_SOFTFP case IR_SLOAD: case IR_ALOAD: case IR_HLOAD: case IR_ULOAD: case IR_VLOAD: @@ -1950,25 +2001,16 @@ static void asm_hiop(ASMState *as, IRIns *ir) if (!uselo) ra_allocref(as, ir->op1, RSET_GPR); /* Mark lo op as used. */ break; + case IR_ASTORE: case IR_HSTORE: case IR_USTORE: case IR_TOSTR: case IR_TMPREF: + /* Nothing to do here. Handled by lo op itself. */ + break; #endif - case IR_CALLN: - case IR_CALLS: - case IR_CALLXS: + case IR_CALLN: case IR_CALLL: case IR_CALLS: case IR_CALLXS: if (!uselo) ra_allocref(as, ir->op1, RID2RSET(RID_RETLO)); /* Mark lo op as used. */ break; -#if LJ_SOFTFP - case IR_ASTORE: case IR_HSTORE: case IR_USTORE: case IR_TOSTR: -#endif - case IR_CNEWI: - /* Nothing to do here. Handled by lo op itself. */ - break; default: lj_assertA(0, "bad HIOP for op %d", (ir-1)->o); break; } -#else - /* Unused without SOFTFP or FFI. */ - UNUSED(as); UNUSED(ir); lj_assertA(0, "unexpected HIOP"); -#endif } /* -- Profiling ----------------------------------------------------------- */ @@ -2055,6 +2097,8 @@ static void asm_stack_restore(ASMState *as, SnapShot *snap) } else if ((sn & SNAP_SOFTFPNUM)) { type = ra_alloc1(as, ref+1, rset_exclude(RSET_GPR, RID_BASE)); #endif + } else if ((sn & SNAP_KEYINDEX)) { + type = ra_allock(as, (int32_t)LJ_KEYINDEX, allow); } else { type = ra_allock(as, (int32_t)irt_toitype(ir->t), allow); } @@ -2113,6 +2157,12 @@ static void asm_loop_fixup(ASMState *as) } } +/* Fixup the tail of the loop. */ +static void asm_loop_tail_fixup(ASMState *as) +{ + UNUSED(as); /* Nothing to do. */ +} + /* -- Head of trace ------------------------------------------------------- */ /* Coalesce BASE register for a root trace. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_asm_x86.h b/love/src/jni/LuaJIT-2.1/src/lj_asm_x86.h index 060ed974..5eb18365 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_asm_x86.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_asm_x86.h @@ -216,10 +216,17 @@ static void asm_fuseahuref(ASMState *as, IRRef ref, RegSet allow) #endif } break; + case IR_TMPREF: +#if LJ_GC64 + as->mrm.ofs = (int32_t)dispofs(as, &J2G(as->J)->tmptv); + as->mrm.base = RID_DISPATCH; + as->mrm.idx = RID_NONE; +#else + as->mrm.ofs = igcptr(&J2G(as->J)->tmptv); + as->mrm.base = as->mrm.idx = RID_NONE; +#endif + return; default: - lj_assertA(ir->o == IR_HREF || ir->o == IR_NEWREF || ir->o == IR_UREFO || - ir->o == IR_KKPTR, - "bad IR op %d", ir->o); break; } } @@ -480,6 +487,7 @@ static Reg asm_fuseload(ASMState *as, IRRef ref, RegSet allow) } } else if (ir->o == IR_VLOAD && !(LJ_GC64 && irt_isaddr(ir->t))) { asm_fuseahuref(as, ir->op1, xallow); + as->mrm.ofs += 8 * ir->op2; return RID_MRM; } } @@ -651,7 +659,7 @@ static void asm_gencall(ASMState *as, const CCallInfo *ci, IRRef *args) static void asm_setupresult(ASMState *as, IRIns *ir, const CCallInfo *ci) { RegSet drop = RSET_SCRATCH; - int hiop = (LJ_32 && (ir+1)->o == IR_HIOP && !irt_isnil((ir+1)->t)); + int hiop = ((ir+1)->o == IR_HIOP && !irt_isnil((ir+1)->t)); if ((ci->flags & CCI_NOFPRCLOBBER)) drop &= ~RSET_FPR; if (ra_hasreg(ir->r)) @@ -691,10 +699,8 @@ static void asm_setupresult(ASMState *as, IRIns *ir, const CCallInfo *ci) irt_isnum(ir->t) ? XOg_FSTPq : XOg_FSTPd, RID_ESP, ofs); } #endif -#if LJ_32 } else if (hiop) { ra_destpair(as, ir); -#endif } else { lj_assertA(!irt_ispri(ir->t), "PRI dest"); ra_destreg(as, ir, RID_RET); @@ -781,6 +787,21 @@ static void asm_retf(ASMState *as, IRIns *ir) #endif } +/* -- Buffer operations --------------------------------------------------- */ + +#if LJ_HASBUFFER +static void asm_bufhdr_write(ASMState *as, Reg sb) +{ + Reg tmp = ra_scratch(as, rset_exclude(RSET_GPR, sb)); + IRIns irgc; + irgc.ot = IRT(0, IRT_PGC); /* GC type. */ + emit_storeofs(as, &irgc, tmp, sb, offsetof(SBuf, L)); + emit_opgl(as, XO_ARITH(XOg_OR), tmp|REX_GC64, cur_L); + emit_gri(as, XG_ARITHi(XOg_AND), tmp, SBUF_MASK_FLAG); + emit_loadofs(as, &irgc, tmp, sb, offsetof(SBuf, L)); +} +#endif + /* -- Type conversions ---------------------------------------------------- */ static void asm_tointg(ASMState *as, IRIns *ir, Reg left) @@ -924,7 +945,7 @@ static void asm_conv(ASMState *as, IRIns *ir) } } else { Reg dest = ra_dest(as, ir, RSET_GPR); - if (st64) { + if (st64 && !(ir->op2 & IRCONV_NONE)) { Reg left = asm_fuseload(as, lref, RSET_GPR); /* This is either a 32 bit reg/reg mov which zeroes the hiword ** or a load of the loword from a 64 bit address. @@ -1050,47 +1071,48 @@ static void asm_strto(ASMState *as, IRIns *ir) /* -- Memory references --------------------------------------------------- */ /* Get pointer to TValue. */ -static void asm_tvptr(ASMState *as, Reg dest, IRRef ref) +static void asm_tvptr(ASMState *as, Reg dest, IRRef ref, MSize mode) { - IRIns *ir = IR(ref); - if (irt_isnum(ir->t)) { - /* For numbers use the constant itself or a spill slot as a TValue. */ - if (irref_isk(ref)) - emit_loada(as, dest, ir_knum(ir)); - else - emit_rmro(as, XO_LEA, dest|REX_64, RID_ESP, ra_spill(as, ir)); - } else { - /* Otherwise use g->tmptv to hold the TValue. */ -#if LJ_GC64 - if (irref_isk(ref)) { - TValue k; - lj_ir_kvalue(as->J->L, &k, ir); - emit_movmroi(as, dest, 4, k.u32.hi); - emit_movmroi(as, dest, 0, k.u32.lo); - } else { - /* TODO: 64 bit store + 32 bit load-modify-store is suboptimal. */ - Reg src = ra_alloc1(as, ref, rset_exclude(RSET_GPR, dest)); - if (irt_is64(ir->t)) { - emit_u32(as, irt_toitype(ir->t) << 15); - emit_rmro(as, XO_ARITHi, XOg_OR, dest, 4); - } else { - /* Currently, no caller passes integers that might end up here. */ - emit_movmroi(as, dest, 4, (irt_toitype(ir->t) << 15)); + if ((mode & IRTMPREF_IN1)) { + IRIns *ir = IR(ref); + if (irt_isnum(ir->t)) { + if (irref_isk(ref) && !(mode & IRTMPREF_OUT1)) { + /* Use the number constant itself as a TValue. */ + emit_loada(as, dest, ir_knum(ir)); + return; + } + emit_rmro(as, XO_MOVSDto, ra_alloc1(as, ref, RSET_FPR), dest, 0); + } else { +#if LJ_GC64 + if (irref_isk(ref)) { + TValue k; + lj_ir_kvalue(as->J->L, &k, ir); + emit_movmroi(as, dest, 4, k.u32.hi); + emit_movmroi(as, dest, 0, k.u32.lo); + } else { + /* TODO: 64 bit store + 32 bit load-modify-store is suboptimal. */ + Reg src = ra_alloc1(as, ref, rset_exclude(RSET_GPR, dest)); + if (irt_is64(ir->t)) { + emit_u32(as, irt_toitype(ir->t) << 15); + emit_rmro(as, XO_ARITHi, XOg_OR, dest, 4); + } else { + emit_movmroi(as, dest, 4, (irt_toitype(ir->t) << 15)); + } + emit_movtomro(as, REX_64IR(ir, src), dest, 0); } - emit_movtomro(as, REX_64IR(ir, src), dest, 0); - } #else - if (!irref_isk(ref)) { - Reg src = ra_alloc1(as, ref, rset_exclude(RSET_GPR, dest)); - emit_movtomro(as, REX_64IR(ir, src), dest, 0); - } else if (!irt_ispri(ir->t)) { - emit_movmroi(as, dest, 0, ir->i); - } - if (!(LJ_64 && irt_islightud(ir->t))) - emit_movmroi(as, dest, 4, irt_toitype(ir->t)); + if (!irref_isk(ref)) { + Reg src = ra_alloc1(as, ref, rset_exclude(RSET_GPR, dest)); + emit_movtomro(as, REX_64IR(ir, src), dest, 0); + } else if (!irt_ispri(ir->t)) { + emit_movmroi(as, dest, 0, ir->i); + } + if (!(LJ_64 && irt_islightud(ir->t))) + emit_movmroi(as, dest, 4, irt_toitype(ir->t)); #endif - emit_loada(as, dest, &J2G(as->J)->tmptv); + } } + emit_loada(as, dest, &J2G(as->J)->tmptv); /* g->tmptv holds the TValue(s). */ } static void asm_aref(ASMState *as, IRIns *ir) @@ -1524,6 +1546,7 @@ static void asm_ahuvload(ASMState *as, IRIns *ir) Reg dest = asm_load_lightud64(as, ir, 1); if (ra_hasreg(dest)) { asm_fuseahuref(as, ir->op1, RSET_GPR); + if (ir->o == IR_VLOAD) as->mrm.ofs += 8 * ir->op2; emit_mrm(as, XO_MOV, dest|REX_64, RID_MRM); } return; @@ -1533,6 +1556,7 @@ static void asm_ahuvload(ASMState *as, IRIns *ir) RegSet allow = irt_isnum(ir->t) ? RSET_FPR : RSET_GPR; Reg dest = ra_dest(as, ir, allow); asm_fuseahuref(as, ir->op1, RSET_GPR); + if (ir->o == IR_VLOAD) as->mrm.ofs += 8 * ir->op2; #if LJ_GC64 if (irt_isaddr(ir->t)) { emit_shifti(as, XOg_SHR|REX_64, dest, 17); @@ -1560,6 +1584,7 @@ static void asm_ahuvload(ASMState *as, IRIns *ir) } #endif asm_fuseahuref(as, ir->op1, gpr); + if (ir->o == IR_VLOAD) as->mrm.ofs += 8 * ir->op2; } /* Always do the type check, even if the load result is unused. */ as->mrm.ofs += 4; @@ -1675,7 +1700,8 @@ static void asm_sload(ASMState *as, IRIns *ir) lj_assertA(irt_isguard(t) || !(ir->op2 & IRSLOAD_TYPECHECK), "inconsistent SLOAD variant"); lj_assertA(LJ_DUALNUM || - !irt_isint(t) || (ir->op2 & (IRSLOAD_CONVERT|IRSLOAD_FRAME)), + !irt_isint(t) || + (ir->op2 & (IRSLOAD_CONVERT|IRSLOAD_FRAME|IRSLOAD_KEYINDEX)), "bad SLOAD type"); if ((ir->op2 & IRSLOAD_CONVERT) && irt_isguard(t) && irt_isint(t)) { Reg left = ra_scratch(as, RSET_FPR); @@ -2584,15 +2610,15 @@ static void asm_comp_int64(ASMState *as, IRIns *ir) } #endif -/* -- Support for 64 bit ops in 32 bit mode ------------------------------- */ +/* -- Split register ops -------------------------------------------------- */ -/* Hiword op of a split 64 bit op. Previous op must be the loword op. */ +/* Hiword op of a split 32/32 or 64/64 bit op. Previous op is the loword op. */ static void asm_hiop(ASMState *as, IRIns *ir) { -#if LJ_32 && LJ_HASFFI /* HIOP is marked as a store because it needs its own DCE logic. */ int uselo = ra_used(ir-1), usehi = ra_used(ir); /* Loword/hiword used? */ if (LJ_UNLIKELY(!(as->flags & JIT_F_OPT_DCE))) uselo = usehi = 1; +#if LJ_32 && LJ_HASFFI if ((ir-1)->o == IR_CONV) { /* Conversions to/from 64 bit. */ as->curins--; /* Always skip the CONV. */ if (usehi || uselo) @@ -2606,8 +2632,10 @@ static void asm_hiop(ASMState *as, IRIns *ir) asm_fxstore(as, ir); return; } +#endif if (!usehi) return; /* Skip unused hiword op for all remaining ops. */ switch ((ir-1)->o) { +#if LJ_32 && LJ_HASFFI case IR_ADD: as->flagmcp = NULL; as->curins--; @@ -2630,20 +2658,16 @@ static void asm_hiop(ASMState *as, IRIns *ir) asm_neg_not(as, ir-1, XOg_NEG); break; } - case IR_CALLN: - case IR_CALLXS: - if (!uselo) - ra_allocref(as, ir->op1, RID2RSET(RID_RETLO)); /* Mark lo op as used. */ - break; case IR_CNEWI: /* Nothing to do here. Handled by CNEWI itself. */ break; +#endif + case IR_CALLN: case IR_CALLL: case IR_CALLS: case IR_CALLXS: + if (!uselo) + ra_allocref(as, ir->op1, RID2RSET(RID_RETLO)); /* Mark lo op as used. */ + break; default: lj_assertA(0, "bad HIOP for op %d", (ir-1)->o); break; } -#else - /* Unused on x64 or without FFI. */ - UNUSED(as); UNUSED(ir); lj_assertA(0, "unexpected HIOP"); -#endif } /* -- Profiling ----------------------------------------------------------- */ @@ -2704,7 +2728,15 @@ static void asm_stack_restore(ASMState *as, SnapShot *snap) IRIns *ir = IR(ref); if ((sn & SNAP_NORESTORE)) continue; - if (irt_isnum(ir->t)) { + if ((sn & SNAP_KEYINDEX)) { + emit_movmroi(as, RID_BASE, ofs+4, LJ_KEYINDEX); + if (irref_isk(ref)) { + emit_movmroi(as, RID_BASE, ofs, ir->i); + } else { + Reg src = ra_alloc1(as, ref, rset_exclude(RSET_GPR, RID_BASE)); + emit_movtomro(as, src, RID_BASE, ofs); + } + } else if (irt_isnum(ir->t)) { Reg src = ra_alloc1(as, ref, RSET_FPR); emit_rmro(as, XO_MOVSDto, src, RID_BASE, ofs); } else { @@ -2837,6 +2869,12 @@ static void asm_loop_fixup(ASMState *as) } } +/* Fixup the tail of the loop. */ +static void asm_loop_tail_fixup(ASMState *as) +{ + UNUSED(as); /* Nothing to do. */ +} + /* -- Head of trace ------------------------------------------------------- */ /* Coalesce BASE register for a root trace. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_bcread.c b/love/src/jni/LuaJIT-2.1/src/lj_bcread.c index 047f98df..298e6c45 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_bcread.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_bcread.c @@ -53,11 +53,11 @@ static LJ_NOINLINE void bcread_fill(LexState *ls, MSize len, int need) do { const char *buf; size_t sz; - char *p = sbufB(&ls->sb); + char *p = ls->sb.b; MSize n = (MSize)(ls->pe - ls->p); if (n) { /* Copy remainder to buffer. */ if (sbuflen(&ls->sb)) { /* Move down in buffer. */ - lj_assertLS(ls->pe == sbufP(&ls->sb), "bad buffer pointer"); + lj_assertLS(ls->pe == ls->sb.w, "bad buffer pointer"); if (ls->p != p) memmove(p, ls->p, n); } else { /* Copy from buffer provided by reader. */ p = lj_buf_need(&ls->sb, len); @@ -66,7 +66,7 @@ static LJ_NOINLINE void bcread_fill(LexState *ls, MSize len, int need) ls->p = p; ls->pe = p + n; } - setsbufP(&ls->sb, p + n); + ls->sb.w = p + n; buf = ls->rfunc(ls->L, ls->rdata, &sz); /* Get more data from reader. */ if (buf == NULL || sz == 0) { /* EOF? */ if (need) bcread_error(ls, LJ_ERR_BCBAD); @@ -77,8 +77,8 @@ static LJ_NOINLINE void bcread_fill(LexState *ls, MSize len, int need) if (n) { /* Append to buffer. */ n += (MSize)sz; p = lj_buf_need(&ls->sb, n < len ? len : n); - memcpy(sbufP(&ls->sb), buf, sz); - setsbufP(&ls->sb, p + n); + memcpy(ls->sb.w, buf, sz); + ls->sb.w = p + n; ls->p = p; ls->pe = p + n; } else { /* Return buffer provided by reader. */ @@ -399,11 +399,7 @@ static int bcread_header(LexState *ls) if ((flags & BCDUMP_F_FFI)) { #if LJ_HASFFI lua_State *L = ls->L; - if (!ctype_ctsG(G(L))) { - ptrdiff_t oldtop = savestack(L, L->top); - luaopen_ffi(L); /* Load FFI library on-demand. */ - L->top = restorestack(L, oldtop); - } + ctype_loadffi(L); #else return 0; #endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_bcwrite.c b/love/src/jni/LuaJIT-2.1/src/lj_bcwrite.c index 7580a8ea..c5c042e0 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_bcwrite.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_bcwrite.c @@ -62,7 +62,7 @@ static void bcwrite_ktabk(BCWriteCtx *ctx, cTValue *o, int narrow) if (num == (lua_Number)k) { /* -0 is never a constant. */ *p++ = BCDUMP_KTAB_INT; p = lj_strfmt_wuleb128(p, k); - setsbufP(&ctx->sb, p); + ctx->sb.w = p; return; } } @@ -73,7 +73,7 @@ static void bcwrite_ktabk(BCWriteCtx *ctx, cTValue *o, int narrow) lj_assertBCW(tvispri(o), "unhandled type %d", itype(o)); *p++ = BCDUMP_KTAB_NIL+~itype(o); } - setsbufP(&ctx->sb, p); + ctx->sb.w = p; } /* Write a template table. */ @@ -97,7 +97,7 @@ static void bcwrite_ktab(BCWriteCtx *ctx, char *p, const GCtab *t) /* Write number of array slots and hash slots. */ p = lj_strfmt_wuleb128(p, narray); p = lj_strfmt_wuleb128(p, nhash); - setsbufP(&ctx->sb, p); + ctx->sb.w = p; if (narray) { /* Write array entries (may contain nil). */ MSize i; TValue *o = tvref(t->array); @@ -172,7 +172,7 @@ static void bcwrite_kgc(BCWriteCtx *ctx, GCproto *pt) } #endif } - setsbufP(&ctx->sb, p); + ctx->sb.w = p; } } @@ -206,7 +206,7 @@ static void bcwrite_knum(BCWriteCtx *ctx, GCproto *pt) p = lj_strfmt_wuleb128(p, o->u32.hi); } } - setsbufP(&ctx->sb, p); + ctx->sb.w = p; } /* Write bytecode instructions. */ @@ -281,7 +281,7 @@ static void bcwrite_proto(BCWriteCtx *ctx, GCproto *pt) /* Write bytecode instructions and upvalue refs. */ p = bcwrite_bytecode(ctx, p, pt); p = lj_buf_wmem(p, proto_uv(pt), pt->sizeuv*2); - setsbufP(&ctx->sb, p); + ctx->sb.w = p; /* Write constants. */ bcwrite_kgc(ctx, pt); @@ -291,16 +291,16 @@ static void bcwrite_proto(BCWriteCtx *ctx, GCproto *pt) if (sizedbg) { p = lj_buf_more(&ctx->sb, sizedbg); p = lj_buf_wmem(p, proto_lineinfo(pt), sizedbg); - setsbufP(&ctx->sb, p); + ctx->sb.w = p; } /* Pass buffer to writer function. */ if (ctx->status == 0) { MSize n = sbuflen(&ctx->sb) - 5; MSize nn = (lj_fls(n)+8)*9 >> 6; - char *q = sbufB(&ctx->sb) + (5 - nn); + char *q = ctx->sb.b + (5 - nn); p = lj_strfmt_wuleb128(q, n); /* Fill in final size. */ - lj_assertBCW(p == sbufB(&ctx->sb) + 5, "bad ULEB128 write"); + lj_assertBCW(p == ctx->sb.b + 5, "bad ULEB128 write"); ctx->status = ctx->wfunc(sbufL(&ctx->sb), q, nn+n, ctx->wdata); } } @@ -324,8 +324,8 @@ static void bcwrite_header(BCWriteCtx *ctx) p = lj_strfmt_wuleb128(p, len); p = lj_buf_wmem(p, name, len); } - ctx->status = ctx->wfunc(sbufL(&ctx->sb), sbufB(&ctx->sb), - (MSize)(p - sbufB(&ctx->sb)), ctx->wdata); + ctx->status = ctx->wfunc(sbufL(&ctx->sb), ctx->sb.b, + (MSize)(p - ctx->sb.b), ctx->wdata); } /* Write footer of bytecode dump. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_buf.c b/love/src/jni/LuaJIT-2.1/src/lj_buf.c index 731470f4..d31bd99e 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_buf.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_buf.c @@ -20,12 +20,32 @@ static void buf_grow(SBuf *sb, MSize sz) { MSize osz = sbufsz(sb), len = sbuflen(sb), nsz = osz; char *b; + GCSize flag; if (nsz < LJ_MIN_SBUF) nsz = LJ_MIN_SBUF; while (nsz < sz) nsz += nsz; - b = (char *)lj_mem_realloc(sbufL(sb), sbufB(sb), osz, nsz); - setmref(sb->b, b); - setmref(sb->p, b + len); - setmref(sb->e, b + nsz); + flag = sbufflag(sb); + if ((flag & SBUF_FLAG_COW)) { /* Copy-on-write semantics. */ + lj_assertG_(G(sbufL(sb)), sb->w == sb->e, "bad SBuf COW"); + b = (char *)lj_mem_new(sbufL(sb), nsz); + setsbufflag(sb, flag & ~(GCSize)SBUF_FLAG_COW); + setgcrefnull(sbufX(sb)->cowref); + memcpy(b, sb->b, osz); + } else { + b = (char *)lj_mem_realloc(sbufL(sb), sb->b, osz, nsz); + } + if ((flag & SBUF_FLAG_EXT)) { + sbufX(sb)->r = sbufX(sb)->r - sb->b + b; /* Adjust read pointer, too. */ + } + /* Adjust buffer pointers. */ + sb->b = b; + sb->w = b + len; + sb->e = b + nsz; + if ((flag & SBUF_FLAG_BORROW)) { /* Adjust borrowed buffer pointers. */ + SBuf *bsb = mref(sbufX(sb)->bsb, SBuf); + bsb->b = b; + bsb->w = b + len; + bsb->e = b + nsz; + } } LJ_NOINLINE char *LJ_FASTCALL lj_buf_need2(SBuf *sb, MSize sz) @@ -34,30 +54,51 @@ LJ_NOINLINE char *LJ_FASTCALL lj_buf_need2(SBuf *sb, MSize sz) if (LJ_UNLIKELY(sz > LJ_MAX_BUF)) lj_err_mem(sbufL(sb)); buf_grow(sb, sz); - return sbufB(sb); + return sb->b; } LJ_NOINLINE char *LJ_FASTCALL lj_buf_more2(SBuf *sb, MSize sz) { - MSize len = sbuflen(sb); - lj_assertG_(G(sbufL(sb)), sz > sbufleft(sb), "SBuf overflow"); - if (LJ_UNLIKELY(sz > LJ_MAX_BUF || len + sz > LJ_MAX_BUF)) - lj_err_mem(sbufL(sb)); - buf_grow(sb, len + sz); - return sbufP(sb); + if (sbufisext(sb)) { + SBufExt *sbx = (SBufExt *)sb; + MSize len = sbufxlen(sbx); + if (LJ_UNLIKELY(sz > LJ_MAX_BUF || len + sz > LJ_MAX_BUF)) + lj_err_mem(sbufL(sbx)); + if (len + sz > sbufsz(sbx)) { /* Must grow. */ + buf_grow((SBuf *)sbx, len + sz); + } else if (sbufxslack(sbx) < (sbufsz(sbx) >> 3)) { + /* Also grow to avoid excessive compactions, if slack < size/8. */ + buf_grow((SBuf *)sbx, sbuflen(sbx) + sz); /* Not sbufxlen! */ + return sbx->w; + } + if (sbx->r != sbx->b) { /* Compact by moving down. */ + memmove(sbx->b, sbx->r, len); + sbx->r = sbx->b; + sbx->w = sbx->b + len; + lj_assertG_(G(sbufL(sbx)), len + sz <= sbufsz(sbx), "bad SBuf compact"); + } + } else { + MSize len = sbuflen(sb); + lj_assertG_(G(sbufL(sb)), sz > sbufleft(sb), "SBuf overflow"); + if (LJ_UNLIKELY(sz > LJ_MAX_BUF || len + sz > LJ_MAX_BUF)) + lj_err_mem(sbufL(sb)); + buf_grow(sb, len + sz); + } + return sb->w; } void LJ_FASTCALL lj_buf_shrink(lua_State *L, SBuf *sb) { - char *b = sbufB(sb); - MSize osz = (MSize)(sbufE(sb) - b); + char *b = sb->b; + MSize osz = (MSize)(sb->e - b); if (osz > 2*LJ_MIN_SBUF) { - MSize n = (MSize)(sbufP(sb) - b); + MSize n = (MSize)(sb->w - b); b = lj_mem_realloc(L, b, osz, (osz >> 1)); - setmref(sb->b, b); - setmref(sb->p, b + n); - setmref(sb->e, b + (osz >> 1)); + sb->b = b; + sb->w = b + n; + sb->e = b + (osz >> 1); } + lj_assertG_(G(sbufL(sb)), !sbufisext(sb), "YAGNI shrink SBufExt"); } char * LJ_FASTCALL lj_buf_tmp(lua_State *L, MSize sz) @@ -67,30 +108,62 @@ char * LJ_FASTCALL lj_buf_tmp(lua_State *L, MSize sz) return lj_buf_need(sb, sz); } +#if LJ_HASBUFFER && LJ_HASJIT +void lj_bufx_set(SBufExt *sbx, const char *p, MSize len, GCobj *ref) +{ + lua_State *L = sbufL(sbx); + lj_bufx_free(L, sbx); + lj_bufx_set_cow(L, sbx, p, len); + setgcref(sbx->cowref, ref); + lj_gc_objbarrier(L, (GCudata *)sbx - 1, ref); +} + +#if LJ_HASFFI +MSize LJ_FASTCALL lj_bufx_more(SBufExt *sbx, MSize sz) +{ + lj_buf_more((SBuf *)sbx, sz); + return sbufleft(sbx); +} +#endif +#endif + /* -- Low-level buffer put operations ------------------------------------- */ SBuf *lj_buf_putmem(SBuf *sb, const void *q, MSize len) { - char *p = lj_buf_more(sb, len); - p = lj_buf_wmem(p, q, len); - setsbufP(sb, p); + char *w = lj_buf_more(sb, len); + w = lj_buf_wmem(w, q, len); + sb->w = w; + return sb; +} + +#if LJ_HASJIT || LJ_HASFFI +static LJ_NOINLINE SBuf * LJ_FASTCALL lj_buf_putchar2(SBuf *sb, int c) +{ + char *w = lj_buf_more2(sb, 1); + *w++ = (char)c; + sb->w = w; return sb; } SBuf * LJ_FASTCALL lj_buf_putchar(SBuf *sb, int c) { - char *p = lj_buf_more(sb, 1); - *p++ = (char)c; - setsbufP(sb, p); - return sb; + char *w = sb->w; + if (LJ_LIKELY(w < sb->e)) { + *w++ = (char)c; + sb->w = w; + return sb; + } + return lj_buf_putchar2(sb, c); } +#endif SBuf * LJ_FASTCALL lj_buf_putstr(SBuf *sb, GCstr *s) { MSize len = s->len; - char *p = lj_buf_more(sb, len); - p = lj_buf_wmem(p, strdata(s), len); - setsbufP(sb, p); + char *w = lj_buf_more(sb, len); + w = lj_buf_wmem(w, strdata(s), len); + sb->w = w; return sb; } @@ -99,47 +172,47 @@ SBuf * LJ_FASTCALL lj_buf_putstr(SBuf *sb, GCstr *s) SBuf * LJ_FASTCALL lj_buf_putstr_reverse(SBuf *sb, GCstr *s) { MSize len = s->len; - char *p = lj_buf_more(sb, len), *e = p+len; + char *w = lj_buf_more(sb, len), *e = w+len; const char *q = strdata(s)+len-1; - while (p < e) - *p++ = *q--; - setsbufP(sb, p); + while (w < e) + *w++ = *q--; + sb->w = w; return sb; } SBuf * LJ_FASTCALL lj_buf_putstr_lower(SBuf *sb, GCstr *s) { MSize len = s->len; - char *p = lj_buf_more(sb, len), *e = p+len; + char *w = lj_buf_more(sb, len), *e = w+len; const char *q = strdata(s); - for (; p < e; p++, q++) { + for (; w < e; w++, q++) { uint32_t c = *(unsigned char *)q; #if LJ_TARGET_PPC - *p = c + ((c >= 'A' && c <= 'Z') << 5); + *w = c + ((c >= 'A' && c <= 'Z') << 5); #else if (c >= 'A' && c <= 'Z') c += 0x20; - *p = c; + *w = c; #endif } - setsbufP(sb, p); + sb->w = w; return sb; } SBuf * LJ_FASTCALL lj_buf_putstr_upper(SBuf *sb, GCstr *s) { MSize len = s->len; - char *p = lj_buf_more(sb, len), *e = p+len; + char *w = lj_buf_more(sb, len), *e = w+len; const char *q = strdata(s); - for (; p < e; p++, q++) { + for (; w < e; w++, q++) { uint32_t c = *(unsigned char *)q; #if LJ_TARGET_PPC - *p = c - ((c >= 'a' && c <= 'z') << 5); + *w = c - ((c >= 'a' && c <= 'z') << 5); #else if (c >= 'a' && c <= 'z') c -= 0x20; - *p = c; + *w = c; #endif } - setsbufP(sb, p); + sb->w = w; return sb; } @@ -148,21 +221,21 @@ SBuf *lj_buf_putstr_rep(SBuf *sb, GCstr *s, int32_t rep) MSize len = s->len; if (rep > 0 && len) { uint64_t tlen = (uint64_t)rep * len; - char *p; + char *w; if (LJ_UNLIKELY(tlen > LJ_MAX_STR)) lj_err_mem(sbufL(sb)); - p = lj_buf_more(sb, (MSize)tlen); + w = lj_buf_more(sb, (MSize)tlen); if (len == 1) { /* Optimize a common case. */ uint32_t c = strdata(s)[0]; - do { *p++ = c; } while (--rep > 0); + do { *w++ = c; } while (--rep > 0); } else { const char *e = strdata(s) + len; do { const char *q = strdata(s); - do { *p++ = *q++; } while (q < e); + do { *w++ = *q++; } while (q < e); } while (--rep > 0); } - setsbufP(sb, p); + sb->w = w; } return sb; } @@ -173,27 +246,27 @@ SBuf *lj_buf_puttab(SBuf *sb, GCtab *t, GCstr *sep, int32_t i, int32_t e) if (i <= e) { for (;;) { cTValue *o = lj_tab_getint(t, i); - char *p; + char *w; if (!o) { badtype: /* Error: bad element type. */ - setsbufP(sb, (void *)(intptr_t)i); /* Store failing index. */ + sb->w = (char *)(intptr_t)i; /* Store failing index. */ return NULL; } else if (tvisstr(o)) { MSize len = strV(o)->len; - p = lj_buf_wmem(lj_buf_more(sb, len + seplen), strVdata(o), len); + w = lj_buf_wmem(lj_buf_more(sb, len + seplen), strVdata(o), len); } else if (tvisint(o)) { - p = lj_strfmt_wint(lj_buf_more(sb, STRFMT_MAXBUF_INT+seplen), intV(o)); + w = lj_strfmt_wint(lj_buf_more(sb, STRFMT_MAXBUF_INT+seplen), intV(o)); } else if (tvisnum(o)) { - p = lj_buf_more(lj_strfmt_putfnum(sb, STRFMT_G14, numV(o)), seplen); + w = lj_buf_more(lj_strfmt_putfnum(sb, STRFMT_G14, numV(o)), seplen); } else { goto badtype; } if (i++ == e) { - setsbufP(sb, p); + sb->w = w; break; } - if (seplen) p = lj_buf_wmem(p, strdata(sep), seplen); - setsbufP(sb, p); + if (seplen) w = lj_buf_wmem(w, strdata(sep), seplen); + sb->w = w; } } return sb; @@ -203,7 +276,7 @@ SBuf *lj_buf_puttab(SBuf *sb, GCtab *t, GCstr *sep, int32_t i, int32_t e) GCstr * LJ_FASTCALL lj_buf_tostr(SBuf *sb) { - return lj_str_new(sbufL(sb), sbufB(sb), sbuflen(sb)); + return lj_str_new(sbufL(sb), sb->b, sbuflen(sb)); } /* Concatenate two strings. */ @@ -219,14 +292,14 @@ GCstr *lj_buf_cat2str(lua_State *L, GCstr *s1, GCstr *s2) /* Read ULEB128 from buffer. */ uint32_t LJ_FASTCALL lj_buf_ruleb128(const char **pp) { - const uint8_t *p = (const uint8_t *)*pp; - uint32_t v = *p++; + const uint8_t *w = (const uint8_t *)*pp; + uint32_t v = *w++; if (LJ_UNLIKELY(v >= 0x80)) { int sh = 0; v &= 0x7f; - do { v |= ((*p & 0x7f) << (sh += 7)); } while (*p++ >= 0x80); + do { v |= ((*w & 0x7f) << (sh += 7)); } while (*w++ >= 0x80); } - *pp = (const char *)p; + *pp = (const char *)w; return v; } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_buf.h b/love/src/jni/LuaJIT-2.1/src/lj_buf.h index ae875298..e2ac922e 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_buf.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_buf.h @@ -10,16 +10,60 @@ #include "lj_gc.h" #include "lj_str.h" -/* Resizable string buffers. Struct definition in lj_obj.h. */ -#define sbufB(sb) (mref((sb)->b, char)) -#define sbufP(sb) (mref((sb)->p, char)) -#define sbufE(sb) (mref((sb)->e, char)) -#define sbufL(sb) (mref((sb)->L, lua_State)) -#define sbufsz(sb) ((MSize)(sbufE((sb)) - sbufB((sb)))) -#define sbuflen(sb) ((MSize)(sbufP((sb)) - sbufB((sb)))) -#define sbufleft(sb) ((MSize)(sbufE((sb)) - sbufP((sb)))) -#define setsbufP(sb, q) (setmref((sb)->p, (q))) -#define setsbufL(sb, l) (setmref((sb)->L, (l))) +/* Resizable string buffers. */ + +/* The SBuf struct definition is in lj_obj.h: +** char *w; Write pointer. +** char *e; End pointer. +** char *b; Base pointer. +** MRef L; lua_State, used for buffer resizing. Extension bits in 3 LSB. +*/ + +/* Extended string buffer. */ +typedef struct SBufExt { + SBufHeader; + union { + GCRef cowref; /* Copy-on-write object reference. */ + MRef bsb; /* Borrowed string buffer. */ + }; + char *r; /* Read pointer. */ + GCRef dict_str; /* Serialization string dictionary table. */ + GCRef dict_mt; /* Serialization metatable dictionary table. */ + int depth; /* Remaining recursion depth. */ +} SBufExt; + +#define sbufsz(sb) ((MSize)((sb)->e - (sb)->b)) +#define sbuflen(sb) ((MSize)((sb)->w - (sb)->b)) +#define sbufleft(sb) ((MSize)((sb)->e - (sb)->w)) +#define sbufxlen(sbx) ((MSize)((sbx)->w - (sbx)->r)) +#define sbufxslack(sbx) ((MSize)((sbx)->r - (sbx)->b)) + +#define SBUF_MASK_FLAG (7) +#define SBUF_MASK_L (~(GCSize)SBUF_MASK_FLAG) +#define SBUF_FLAG_EXT 1 /* Extended string buffer. */ +#define SBUF_FLAG_COW 2 /* Copy-on-write buffer. */ +#define SBUF_FLAG_BORROW 4 /* Borrowed string buffer. */ + +#define sbufL(sb) \ + ((lua_State *)(void *)(uintptr_t)(mrefu((sb)->L) & SBUF_MASK_L)) +#define setsbufL(sb, l) (setmref((sb)->L, (l))) +#define setsbufXL(sb, l, flag) \ + (setmrefu((sb)->L, (GCSize)(uintptr_t)(void *)(l) + (flag))) +#define setsbufXL_(sb, l) \ + (setmrefu((sb)->L, (GCSize)(uintptr_t)(void *)(l) | (mrefu((sb)->L) & SBUF_MASK_FLAG))) + +#define sbufflag(sb) (mrefu((sb)->L)) +#define sbufisext(sb) (sbufflag((sb)) & SBUF_FLAG_EXT) +#define sbufiscow(sb) (sbufflag((sb)) & SBUF_FLAG_COW) +#define sbufisborrow(sb) (sbufflag((sb)) & SBUF_FLAG_BORROW) +#define sbufiscoworborrow(sb) (sbufflag((sb)) & (SBUF_FLAG_COW|SBUF_FLAG_BORROW)) +#define sbufX(sb) \ + (lj_assertG_(G(sbufL(sb)), sbufisext(sb), "not an SBufExt"), (SBufExt *)(sb)) +#define setsbufflag(sb, flag) (setmrefu((sb)->L, (flag))) + +#define tvisbuf(o) \ + (LJ_HASBUFFER && tvisudata(o) && udataV(o)->udtype == UDTYPE_BUFFER) +#define bufV(o) check_exp(tvisbuf(o), ((SBufExt *)uddata(udataV(o)))) /* Buffer management */ LJ_FUNC char *LJ_FASTCALL lj_buf_need2(SBuf *sb, MSize sz); @@ -30,12 +74,12 @@ LJ_FUNC char * LJ_FASTCALL lj_buf_tmp(lua_State *L, MSize sz); static LJ_AINLINE void lj_buf_init(lua_State *L, SBuf *sb) { setsbufL(sb, L); - setmref(sb->p, NULL); setmref(sb->e, NULL); setmref(sb->b, NULL); + sb->w = sb->e = sb->b = NULL; } static LJ_AINLINE void lj_buf_reset(SBuf *sb) { - setmrefr(sb->p, sb->b); + sb->w = sb->b; } static LJ_AINLINE SBuf *lj_buf_tmp_(lua_State *L) @@ -48,26 +92,77 @@ static LJ_AINLINE SBuf *lj_buf_tmp_(lua_State *L) static LJ_AINLINE void lj_buf_free(global_State *g, SBuf *sb) { - lj_mem_free(g, sbufB(sb), sbufsz(sb)); + lj_assertG(!sbufisext(sb), "bad free of SBufExt"); + lj_mem_free(g, sb->b, sbufsz(sb)); } static LJ_AINLINE char *lj_buf_need(SBuf *sb, MSize sz) { if (LJ_UNLIKELY(sz > sbufsz(sb))) return lj_buf_need2(sb, sz); - return sbufB(sb); + return sb->b; } static LJ_AINLINE char *lj_buf_more(SBuf *sb, MSize sz) { if (LJ_UNLIKELY(sz > sbufleft(sb))) return lj_buf_more2(sb, sz); - return sbufP(sb); + return sb->w; } +/* Extended buffer management */ +static LJ_AINLINE void lj_bufx_init(lua_State *L, SBufExt *sbx) +{ + memset(sbx, 0, sizeof(SBufExt)); + setsbufXL(sbx, L, SBUF_FLAG_EXT); +} + +static LJ_AINLINE void lj_bufx_set_borrow(lua_State *L, SBufExt *sbx, SBuf *sb) +{ + setsbufXL(sbx, L, SBUF_FLAG_EXT | SBUF_FLAG_BORROW); + setmref(sbx->bsb, sb); + sbx->r = sbx->w = sbx->b = sb->b; + sbx->e = sb->e; +} + +static LJ_AINLINE void lj_bufx_set_cow(lua_State *L, SBufExt *sbx, + const char *p, MSize len) +{ + setsbufXL(sbx, L, SBUF_FLAG_EXT | SBUF_FLAG_COW); + sbx->r = sbx->b = (char *)p; + sbx->w = sbx->e = (char *)p + len; +} + +static LJ_AINLINE void lj_bufx_reset(SBufExt *sbx) +{ + if (sbufiscow(sbx)) { + setmrefu(sbx->L, (mrefu(sbx->L) & ~(GCSize)SBUF_FLAG_COW)); + setgcrefnull(sbx->cowref); + sbx->b = sbx->e = NULL; + } + sbx->r = sbx->w = sbx->b; +} + +static LJ_AINLINE void lj_bufx_free(lua_State *L, SBufExt *sbx) +{ + if (!sbufiscoworborrow(sbx)) lj_mem_free(G(L), sbx->b, sbufsz(sbx)); + setsbufXL(sbx, L, SBUF_FLAG_EXT); + setgcrefnull(sbx->cowref); + sbx->r = sbx->w = sbx->b = sbx->e = NULL; +} + +#if LJ_HASBUFFER && LJ_HASJIT +LJ_FUNC void lj_bufx_set(SBufExt *sbx, const char *p, MSize len, GCobj *o); +#if LJ_HASFFI +LJ_FUNC MSize LJ_FASTCALL lj_bufx_more(SBufExt *sbx, MSize sz); +#endif +#endif + /* Low-level buffer put operations */ LJ_FUNC SBuf *lj_buf_putmem(SBuf *sb, const void *q, MSize len); +#if LJ_HASJIT || LJ_HASFFI LJ_FUNC SBuf * LJ_FASTCALL lj_buf_putchar(SBuf *sb, int c); +#endif LJ_FUNC SBuf * LJ_FASTCALL lj_buf_putstr(SBuf *sb, GCstr *s); static LJ_AINLINE char *lj_buf_wmem(char *p, const void *q, MSize len) @@ -77,9 +172,9 @@ static LJ_AINLINE char *lj_buf_wmem(char *p, const void *q, MSize len) static LJ_AINLINE void lj_buf_putb(SBuf *sb, int c) { - char *p = lj_buf_more(sb, 1); - *p++ = (char)c; - setsbufP(sb, p); + char *w = lj_buf_more(sb, 1); + *w++ = (char)c; + sb->w = w; } /* High-level buffer put operations */ @@ -97,7 +192,7 @@ LJ_FUNC uint32_t LJ_FASTCALL lj_buf_ruleb128(const char **pp); static LJ_AINLINE GCstr *lj_buf_str(lua_State *L, SBuf *sb) { - return lj_str_new(L, sbufB(sb), sbuflen(sb)); + return lj_str_new(L, sb->b, sbuflen(sb)); } #endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_ccall.c b/love/src/jni/LuaJIT-2.1/src/lj_ccall.c index a91ffc7e..3c029823 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_ccall.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_ccall.c @@ -334,7 +334,7 @@ isfp = sz == 2*sizeof(float) ? 2 : 1; #define CCALL_HANDLE_REGARG \ - if (LJ_TARGET_IOS && isva) { \ + if (LJ_TARGET_OSX && isva) { \ /* IOS: All variadic arguments are on the stack. */ \ } else if (isfp) { /* Try to pass argument in FPRs. */ \ int n2 = ctype_isvector(d->info) ? 1 : \ @@ -345,10 +345,10 @@ goto done; \ } else { \ nfpr = CCALL_NARG_FPR; /* Prevent reordering. */ \ - if (LJ_TARGET_IOS && d->size < 8) goto err_nyi; \ + if (LJ_TARGET_OSX && d->size < 8) goto err_nyi; \ } \ } else { /* Try to pass argument in GPRs. */ \ - if (!LJ_TARGET_IOS && (d->info & CTF_ALIGN) > CTALIGN_PTR) \ + if (!LJ_TARGET_OSX && (d->info & CTF_ALIGN) > CTALIGN_PTR) \ ngpr = (ngpr + 1u) & ~1u; /* Align to regpair. */ \ if (ngpr + n <= maxgpr) { \ dp = &cc->gpr[ngpr]; \ @@ -356,7 +356,7 @@ goto done; \ } else { \ ngpr = maxgpr; /* Prevent reordering. */ \ - if (LJ_TARGET_IOS && d->size < 8) goto err_nyi; \ + if (LJ_TARGET_OSX && d->size < 8) goto err_nyi; \ } \ } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_ccallback.c b/love/src/jni/LuaJIT-2.1/src/lj_ccallback.c index 9de86258..80d738c6 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_ccallback.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_ccallback.c @@ -256,6 +256,11 @@ static void *callback_mcode_init(global_State *g, uint32_t *page) #ifndef MAP_ANONYMOUS #define MAP_ANONYMOUS MAP_ANON #endif +#ifdef PROT_MPROTECT +#define CCPROT_CREATE (PROT_MPROTECT(PROT_EXEC)) +#else +#define CCPROT_CREATE 0 +#endif #endif @@ -271,7 +276,7 @@ static void callback_mcode_new(CTState *cts) if (!p) lj_err_caller(cts->L, LJ_ERR_FFI_CBACKOV); #elif LJ_TARGET_POSIX - p = mmap(NULL, sz, (PROT_READ|PROT_WRITE), MAP_PRIVATE|MAP_ANONYMOUS, + p = mmap(NULL, sz, (PROT_READ|PROT_WRITE|CCPROT_CREATE), MAP_PRIVATE|MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) lj_err_caller(cts->L, LJ_ERR_FFI_CBACKOV); @@ -409,7 +414,7 @@ void lj_ccallback_mcode_free(CTState *cts) nfpr = CCALL_NARG_FPR; /* Prevent reordering. */ \ } \ } else { \ - if (!LJ_TARGET_IOS && n > 1) \ + if (!LJ_TARGET_OSX && n > 1) \ ngpr = (ngpr + 1u) & ~1u; /* Align to regpair. */ \ if (ngpr + n <= maxgpr) { \ sp = &cts->cb.gpr[ngpr]; \ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_cconv.c b/love/src/jni/LuaJIT-2.1/src/lj_cconv.c index 353786b9..613f66e2 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_cconv.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_cconv.c @@ -8,6 +8,7 @@ #if LJ_HASFFI #include "lj_err.h" +#include "lj_buf.h" #include "lj_tab.h" #include "lj_ctype.h" #include "lj_cdata.h" @@ -568,7 +569,9 @@ void lj_cconv_ct_tv(CTState *cts, CType *d, } s = ctype_raw(cts, sid); if (ctype_isfunc(s->info)) { + CTypeID did = ctype_typeid(cts, d); sid = lj_ctype_intern(cts, CTINFO(CT_PTR, CTALIGN_PTR|sid), CTSIZE_PTR); + d = ctype_get(cts, did); /* cts->tab may have been reallocated. */ } else { if (ctype_isenum(s->info)) s = ctype_child(cts, s); goto doconv; @@ -619,6 +622,8 @@ void lj_cconv_ct_tv(CTState *cts, CType *d, tmpptr = uddata(ud); if (ud->udtype == UDTYPE_IO_FILE) tmpptr = *(void **)tmpptr; + else if (ud->udtype == UDTYPE_BUFFER) + tmpptr = ((SBufExt *)tmpptr)->r; } else if (tvislightud(o)) { tmpptr = lightudV(cts->g, o); } else if (tvisfunc(o)) { diff --git a/love/src/jni/LuaJIT-2.1/src/lj_cparse.c b/love/src/jni/LuaJIT-2.1/src/lj_cparse.c index 16a1d7b7..efe80759 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_cparse.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_cparse.c @@ -133,9 +133,9 @@ LJ_NORET static void cp_errmsg(CPState *cp, CPToken tok, ErrMsg em, ...) tokstr = NULL; } else if (tok == CTOK_IDENT || tok == CTOK_INTEGER || tok == CTOK_STRING || tok >= CTOK_FIRSTDECL) { - if (sbufP(&cp->sb) == sbufB(&cp->sb)) cp_save(cp, '$'); + if (cp->sb.w == cp->sb.b) cp_save(cp, '$'); cp_save(cp, '\0'); - tokstr = sbufB(&cp->sb); + tokstr = cp->sb.b; } else { tokstr = cp_tok2str(cp, tok); } @@ -175,7 +175,7 @@ static CPToken cp_number(CPState *cp) TValue o; do { cp_save(cp, cp->c); } while (lj_char_isident(cp_get(cp))); cp_save(cp, '\0'); - fmt = lj_strscan_scan((const uint8_t *)sbufB(&cp->sb), sbuflen(&cp->sb)-1, + fmt = lj_strscan_scan((const uint8_t *)(cp->sb.b), sbuflen(&cp->sb)-1, &o, STRSCAN_OPT_C); if (fmt == STRSCAN_INT) cp->val.id = CTID_INT32; else if (fmt == STRSCAN_U32) cp->val.id = CTID_UINT32; @@ -279,7 +279,7 @@ static CPToken cp_string(CPState *cp) return CTOK_STRING; } else { if (sbuflen(&cp->sb) != 1) cp_err_token(cp, '\''); - cp->val.i32 = (int32_t)(char)*sbufB(&cp->sb); + cp->val.i32 = (int32_t)(char)*cp->sb.b; cp->val.id = CTID_INT32; return CTOK_INTEGER; } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_crecord.c b/love/src/jni/LuaJIT-2.1/src/lj_crecord.c index c7b60b3a..165f95d9 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_crecord.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_crecord.c @@ -78,7 +78,7 @@ static CTypeID argv2ctype(jit_State *J, TRef tr, cTValue *o) /* Specialize to the string containing the C type declaration. */ emitir(IRTG(IR_EQ, IRT_STR), tr, lj_ir_kstr(J, s)); cp.L = J->L; - cp.cts = ctype_ctsG(J2G(J)); + cp.cts = ctype_cts(J->L); oldtop = cp.cts->top; cp.srcname = strdata(s); cp.p = strdata(s); @@ -616,10 +616,12 @@ static TRef crec_ct_tv(jit_State *J, CType *d, TRef dp, TRef sp, cTValue *sval) sp = lj_ir_kptr(J, NULL); } else if (tref_isudata(sp)) { GCudata *ud = udataV(sval); - if (ud->udtype == UDTYPE_IO_FILE) { + if (ud->udtype == UDTYPE_IO_FILE || ud->udtype == UDTYPE_BUFFER) { TRef tr = emitir(IRT(IR_FLOAD, IRT_U8), sp, IRFL_UDATA_UDTYPE); - emitir(IRTGI(IR_EQ), tr, lj_ir_kint(J, UDTYPE_IO_FILE)); - sp = emitir(IRT(IR_FLOAD, IRT_PTR), sp, IRFL_UDATA_FILE); + emitir(IRTGI(IR_EQ), tr, lj_ir_kint(J, ud->udtype)); + sp = emitir(IRT(IR_FLOAD, IRT_PTR), sp, + ud->udtype == UDTYPE_IO_FILE ? IRFL_UDATA_FILE : + IRFL_SBUF_R); } else { sp = emitir(IRT(IR_ADD, IRT_PTR), sp, lj_ir_kintp(J, sizeof(GCudata))); } @@ -1024,8 +1026,26 @@ static void crec_alloc(jit_State *J, RecordFFData *rd, CTypeID id) crec_ct_tv(J, dc, dp, sp, sval); } } else if (ctype_isstruct(d->info)) { - CTypeID fid = d->sib; + CTypeID fid; MSize i = 1; + if (!J->base[1]) { /* Handle zero-fill of struct-of-NYI. */ + fid = d->sib; + while (fid) { + CType *df = ctype_get(cts, fid); + fid = df->sib; + if (ctype_isfield(df->info)) { + CType *dc; + if (!gcref(df->name)) continue; /* Ignore unnamed fields. */ + dc = ctype_rawchild(cts, df); /* Field type. */ + if (!(ctype_isnum(dc->info) || ctype_isptr(dc->info) || + ctype_isenum(dc->info))) + goto special; + } else if (!ctype_isconstval(df->info)) { + goto special; + } + } + } + fid = d->sib; while (fid) { CType *df = ctype_get(cts, fid); fid = df->sib; @@ -1893,6 +1913,30 @@ void LJ_FASTCALL lj_crecord_tonumber(jit_State *J, RecordFFData *rd) } } +TRef lj_crecord_loadiu64(jit_State *J, TRef tr, cTValue *o) +{ + CTypeID id = argv2cdata(J, tr, o)->ctypeid; + if (!(id == CTID_INT64 || id == CTID_UINT64)) + lj_trace_err(J, LJ_TRERR_BADTYPE); + lj_needsplit(J); + return emitir(IRT(IR_FLOAD, id == CTID_INT64 ? IRT_I64 : IRT_U64), tr, + IRFL_CDATA_INT64); +} + +#if LJ_HASBUFFER +TRef lj_crecord_topcvoid(jit_State *J, TRef tr, cTValue *o) +{ + CTState *cts = ctype_ctsG(J2G(J)); + if (!tref_iscdata(tr)) lj_trace_err(J, LJ_TRERR_BADTYPE); + return crec_ct_tv(J, ctype_get(cts, CTID_P_CVOID), 0, tr, o); +} + +TRef lj_crecord_topuint8(jit_State *J, TRef tr) +{ + return emitir(IRTG(IR_CNEWI, IRT_CDATA), lj_ir_kint(J, CTID_P_UINT8), tr); +} +#endif + #undef IR #undef emitir #undef emitconv diff --git a/love/src/jni/LuaJIT-2.1/src/lj_crecord.h b/love/src/jni/LuaJIT-2.1/src/lj_crecord.h index c6f39a67..e1a2d9c0 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_crecord.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_crecord.h @@ -33,6 +33,11 @@ LJ_FUNC int LJ_FASTCALL recff_bit64_shift(jit_State *J, RecordFFData *rd); LJ_FUNC TRef recff_bit64_tohex(jit_State *J, RecordFFData *rd, TRef hdr); LJ_FUNC void LJ_FASTCALL lj_crecord_tonumber(jit_State *J, RecordFFData *rd); +LJ_FUNC TRef lj_crecord_loadiu64(jit_State *J, TRef tr, cTValue *o); +#if LJ_HASBUFFER +LJ_FUNC TRef lj_crecord_topcvoid(jit_State *J, TRef tr, cTValue *o); +LJ_FUNC TRef lj_crecord_topuint8(jit_State *J, TRef tr); +#endif #endif #endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_ctype.c b/love/src/jni/LuaJIT-2.1/src/lj_ctype.c index 4e913556..6741437c 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_ctype.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_ctype.c @@ -583,7 +583,7 @@ GCstr *lj_ctype_repr_complex(lua_State *L, void *sp, CTSize size) lj_strfmt_putfnum(sb, STRFMT_G14, re.n); if (!(im.u32.hi & 0x80000000u) || im.n != im.n) lj_buf_putchar(sb, '+'); lj_strfmt_putfnum(sb, STRFMT_G14, im.n); - lj_buf_putchar(sb, sbufP(sb)[-1] >= 'a' ? 'I' : 'i'); + lj_buf_putchar(sb, sb->w[-1] >= 'a' ? 'I' : 'i'); return lj_buf_str(L, sb); } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_ctype.h b/love/src/jni/LuaJIT-2.1/src/lj_ctype.h index 49ec71da..700250df 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_ctype.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_ctype.h @@ -298,6 +298,7 @@ typedef struct CTState { _(P_VOID, CTSIZE_PTR, CT_PTR, CTALIGN_PTR|CTID_VOID) \ _(P_CVOID, CTSIZE_PTR, CT_PTR, CTALIGN_PTR|CTID_CVOID) \ _(P_CCHAR, CTSIZE_PTR, CT_PTR, CTALIGN_PTR|CTID_CCHAR) \ + _(P_UINT8, CTSIZE_PTR, CT_PTR, CTALIGN_PTR|CTID_UINT8) \ _(A_CCHAR, -1, CT_ARRAY, CTF_CONST|CTALIGN(0)|CTID_CCHAR) \ _(CTYPEID, 4, CT_ENUM, CTALIGN(2)|CTID_INT32) \ CTTYDEFP(_) \ @@ -389,6 +390,16 @@ static LJ_AINLINE CTState *ctype_cts(lua_State *L) return cts; } +/* Load FFI library on-demand. */ +#define ctype_loadffi(L) \ + do { \ + if (!ctype_ctsG(G(L))) { \ + ptrdiff_t oldtop = (char *)L->top - mref(L->stack, char); \ + luaopen_ffi(L); \ + L->top = (TValue *)(mref(L->stack, char) + oldtop); \ + } \ + } while (0) + /* Save and restore state of C type table. */ #define LJ_CTYPE_SAVE(cts) CTState savects_ = *(cts) #define LJ_CTYPE_RESTORE(cts) \ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_debug.c b/love/src/jni/LuaJIT-2.1/src/lj_debug.c index 0b4e2c87..3dffad90 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_debug.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_debug.c @@ -648,7 +648,7 @@ void lj_debug_dumpstack(lua_State *L, SBuf *sb, const char *fmt, int depth) level += dir; } if (lastlen) - setsbufP(sb, sbufB(sb) + lastlen); /* Zap trailing separator. */ + sb->w = sb->b + lastlen; /* Zap trailing separator. */ } #endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_dispatch.c b/love/src/jni/LuaJIT-2.1/src/lj_dispatch.c index bf8d8812..7b73d3dd 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_dispatch.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_dispatch.c @@ -68,6 +68,8 @@ void lj_dispatch_init(GG_State *GG) /* The JIT engine is off by default. luaopen_jit() turns it on. */ disp[BC_FORL] = disp[BC_IFORL]; disp[BC_ITERL] = disp[BC_IITERL]; + /* Workaround for stable v2.1 bytecode. TODO: Replace with BC_IITERN. */ + disp[BC_ITERN] = &lj_vm_IITERN; disp[BC_LOOP] = disp[BC_ILOOP]; disp[BC_FUNCF] = disp[BC_IFUNCF]; disp[BC_FUNCV] = disp[BC_IFUNCV]; @@ -118,19 +120,21 @@ void lj_dispatch_update(global_State *g) mode |= (g->hookmask & LUA_MASKRET) ? DISPMODE_RET : 0; if (oldmode != mode) { /* Mode changed? */ ASMFunction *disp = G2GG(g)->dispatch; - ASMFunction f_forl, f_iterl, f_loop, f_funcf, f_funcv; + ASMFunction f_forl, f_iterl, f_itern, f_loop, f_funcf, f_funcv; g->dispatchmode = mode; /* Hotcount if JIT is on, but not while recording. */ if ((mode & (DISPMODE_JIT|DISPMODE_REC)) == DISPMODE_JIT) { f_forl = makeasmfunc(lj_bc_ofs[BC_FORL]); f_iterl = makeasmfunc(lj_bc_ofs[BC_ITERL]); + f_itern = makeasmfunc(lj_bc_ofs[BC_ITERN]); f_loop = makeasmfunc(lj_bc_ofs[BC_LOOP]); f_funcf = makeasmfunc(lj_bc_ofs[BC_FUNCF]); f_funcv = makeasmfunc(lj_bc_ofs[BC_FUNCV]); } else { /* Otherwise use the non-hotcounting instructions. */ f_forl = disp[GG_LEN_DDISP+BC_IFORL]; f_iterl = disp[GG_LEN_DDISP+BC_IITERL]; + f_itern = &lj_vm_IITERN; f_loop = disp[GG_LEN_DDISP+BC_ILOOP]; f_funcf = makeasmfunc(lj_bc_ofs[BC_IFUNCF]); f_funcv = makeasmfunc(lj_bc_ofs[BC_IFUNCV]); @@ -138,6 +142,7 @@ void lj_dispatch_update(global_State *g) /* Init static counting instruction dispatch first (may be copied below). */ disp[GG_LEN_DDISP+BC_FORL] = f_forl; disp[GG_LEN_DDISP+BC_ITERL] = f_iterl; + disp[GG_LEN_DDISP+BC_ITERN] = f_itern; disp[GG_LEN_DDISP+BC_LOOP] = f_loop; /* Set dynamic instruction dispatch. */ @@ -165,6 +170,7 @@ void lj_dispatch_update(global_State *g) /* Otherwise set dynamic counting ins. */ disp[BC_FORL] = f_forl; disp[BC_ITERL] = f_iterl; + disp[BC_ITERN] = f_itern; disp[BC_LOOP] = f_loop; /* Set dynamic return dispatch. */ if ((mode & DISPMODE_RET)) { diff --git a/love/src/jni/LuaJIT-2.1/src/lj_dispatch.h b/love/src/jni/LuaJIT-2.1/src/lj_dispatch.h index be7c410a..2331bd42 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_dispatch.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_dispatch.h @@ -31,7 +31,7 @@ extern double __divdf3(double a, double b); #define SFGOTDEF(_) #endif #if LJ_HASJIT -#define JITGOTDEF(_) _(lj_trace_exit) _(lj_trace_hot) +#define JITGOTDEF(_) _(lj_err_trace) _(lj_trace_exit) _(lj_trace_hot) #else #define JITGOTDEF(_) #endif @@ -46,7 +46,7 @@ extern double __divdf3(double a, double b); _(asin) _(acos) _(atan) _(sinh) _(cosh) _(tanh) _(frexp) _(modf) _(atan2) \ _(pow) _(fmod) _(ldexp) _(lj_vm_modi) \ _(lj_dispatch_call) _(lj_dispatch_ins) _(lj_dispatch_stitch) \ - _(lj_dispatch_profile) _(lj_err_throw) _(lj_err_run) \ + _(lj_dispatch_profile) _(lj_err_throw) \ _(lj_ffh_coroutine_wrap_err) _(lj_func_closeuv) _(lj_func_newL_gc) \ _(lj_gc_barrieruv) _(lj_gc_step) _(lj_gc_step_fixtop) _(lj_meta_arith) \ _(lj_meta_call) _(lj_meta_cat) _(lj_meta_comp) _(lj_meta_equal) \ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_emit_arm64.h b/love/src/jni/LuaJIT-2.1/src/lj_emit_arm64.h index abc32be7..00086e8a 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_emit_arm64.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_emit_arm64.h @@ -163,7 +163,7 @@ nopair: /* Try to find an N-step delta relative to other consts with N < lim. */ static int emit_kdelta(ASMState *as, Reg rd, uint64_t k, int lim) { - RegSet work = ~as->freeset & RSET_GPR; + RegSet work = (~as->freeset & RSET_GPR) | RID2RSET(RID_GL); if (lim <= 1) return 0; /* Can't beat that. */ while (work) { Reg r = rset_picktop(work); @@ -194,39 +194,41 @@ static int emit_kdelta(ASMState *as, Reg rd, uint64_t k, int lim) static void emit_loadk(ASMState *as, Reg rd, uint64_t u64, int is64) { - uint32_t k13 = emit_isk13(u64, is64); - if (k13) { /* Can the constant be represented as a bitmask immediate? */ - emit_dn(as, (is64|A64I_ORRw)^k13, rd, RID_ZERO); - } else { - int i, zeros = 0, ones = 0, neg; - if (!is64) u64 = (int64_t)(int32_t)u64; /* Sign-extend. */ - /* Count homogeneous 16 bit fragments. */ - for (i = 0; i < 4; i++) { - uint64_t frag = (u64 >> i*16) & 0xffff; - zeros += (frag == 0); - ones += (frag == 0xffff); + int i, zeros = 0, ones = 0, neg; + if (!is64) u64 = (int64_t)(int32_t)u64; /* Sign-extend. */ + /* Count homogeneous 16 bit fragments. */ + for (i = 0; i < 4; i++) { + uint64_t frag = (u64 >> i*16) & 0xffff; + zeros += (frag == 0); + ones += (frag == 0xffff); + } + neg = ones > zeros; /* Use MOVN if it pays off. */ + if ((neg ? ones : zeros) < 3) { /* Need 2+ ins. Try shorter K13 encoding. */ + uint32_t k13 = emit_isk13(u64, is64); + if (k13) { + emit_dn(as, (is64|A64I_ORRw)^k13, rd, RID_ZERO); + return; } - neg = ones > zeros; /* Use MOVN if it pays off. */ - if (!emit_kdelta(as, rd, u64, 4 - (neg ? ones : zeros))) { - int shift = 0, lshift = 0; - uint64_t n64 = neg ? ~u64 : u64; - if (n64 != 0) { - /* Find first/last fragment to be filled. */ - shift = (63-emit_clz64(n64)) & ~15; - lshift = emit_ctz64(n64) & ~15; - } - /* MOVK requires the original value (u64). */ - while (shift > lshift) { - uint32_t u16 = (u64 >> shift) & 0xffff; - /* Skip fragments that are correctly filled by MOVN/MOVZ. */ - if (u16 != (neg ? 0xffff : 0)) - emit_d(as, is64 | A64I_MOVKw | A64F_U16(u16) | A64F_LSL16(shift), rd); - shift -= 16; - } - /* But MOVN needs an inverted value (n64). */ - emit_d(as, (neg ? A64I_MOVNx : A64I_MOVZx) | - A64F_U16((n64 >> lshift) & 0xffff) | A64F_LSL16(lshift), rd); + } + if (!emit_kdelta(as, rd, u64, 4 - (neg ? ones : zeros))) { + int shift = 0, lshift = 0; + uint64_t n64 = neg ? ~u64 : u64; + if (n64 != 0) { + /* Find first/last fragment to be filled. */ + shift = (63-emit_clz64(n64)) & ~15; + lshift = emit_ctz64(n64) & ~15; } + /* MOVK requires the original value (u64). */ + while (shift > lshift) { + uint32_t u16 = (u64 >> shift) & 0xffff; + /* Skip fragments that are correctly filled by MOVN/MOVZ. */ + if (u16 != (neg ? 0xffff : 0)) + emit_d(as, is64 | A64I_MOVKw | A64F_U16(u16) | A64F_LSL16(shift), rd); + shift -= 16; + } + /* But MOVN needs an inverted value (n64). */ + emit_d(as, (neg ? A64I_MOVNx : A64I_MOVZx) | + A64F_U16((n64 >> lshift) & 0xffff) | A64F_LSL16(lshift), rd); } } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_emit_mips.h b/love/src/jni/LuaJIT-2.1/src/lj_emit_mips.h index 7825eabb..c13615dd 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_emit_mips.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_emit_mips.h @@ -70,7 +70,7 @@ static void emit_rotr(ASMState *as, Reg dest, Reg src, Reg tmp, uint32_t shift) } } -#if LJ_64 +#if LJ_64 || LJ_HASBUFFER static void emit_tsml(ASMState *as, MIPSIns mi, Reg rt, Reg rs, uint32_t msb, uint32_t lsb) { diff --git a/love/src/jni/LuaJIT-2.1/src/lj_err.c b/love/src/jni/LuaJIT-2.1/src/lj_err.c index c0bf38c2..fda4a59c 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_err.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_err.c @@ -29,12 +29,18 @@ ** Pros and Cons: ** ** - EXT requires unwind tables for *all* functions on the C stack between -** the pcall/catch and the error/throw. This is the default on x64, -** but needs to be manually enabled on x86/PPC for non-C++ code. +** the pcall/catch and the error/throw. C modules used by Lua code can +** throw errors, so these need to have unwind tables, too. Transitively +** this applies to all system libraries used by C modules -- at least +** when they have callbacks which may throw an error. ** -** - INT is faster when actually throwing errors (but this happens rarely). +** - INT is faster when actually throwing errors, but this happens rarely. ** Setting up error handlers is zero-cost in any case. ** +** - INT needs to save *all* callee-saved registers when entering the +** interpreter. EXT only needs to save those actually used inside the +** interpreter. JIT-compiled code may need to save some more. +** ** - EXT provides full interoperability with C++ exceptions. You can throw ** Lua errors or C++ exceptions through a mix of Lua frames and C++ frames. ** C++ destructors are called as needed. C++ exceptions caught by pcall @@ -46,27 +52,38 @@ ** the wrapper function feature. Lua errors thrown through C++ frames ** cannot be caught by C++ code and C++ destructors are not run. ** -** EXT is the default on x64 systems and on Windows, INT is the default on all -** other systems. +** - EXT can handle errors from internal helper functions that are called +** from JIT-compiled code (except for Windows/x86 and 32 bit ARM). +** INT has no choice but to call the panic handler, if this happens. +** Note: this is mainly relevant for out-of-memory errors. ** -** EXT can be manually enabled on POSIX systems using GCC and DWARF2 stack -** unwinding with -DLUAJIT_UNWIND_EXTERNAL. *All* C code must be compiled -** with -funwind-tables (or -fexceptions). This includes LuaJIT itself (set -** TARGET_CFLAGS), all of your C/Lua binding code, all loadable C modules -** and all C libraries that have callbacks which may be used to call back -** into Lua. C++ code must *not* be compiled with -fno-exceptions. +** EXT is the default on all systems where the toolchain produces unwind +** tables by default (*). This is hard-coded and/or detected in src/Makefile. +** You can thwart the detection with: TARGET_XCFLAGS=-DLUAJIT_UNWIND_INTERNAL ** -** EXT is mandatory on WIN64 since the calling convention has an abundance -** of callee-saved registers (rbx, rbp, rsi, rdi, r12-r15, xmm6-xmm15). -** The POSIX/x64 interpreter only saves r12/r13 for INT (e.g. PS4). +** INT is the default on all other systems. +** +** EXT can be manually enabled for toolchains that are able to produce +** conforming unwind tables: +** "TARGET_XCFLAGS=-funwind-tables -DLUAJIT_UNWIND_EXTERNAL" +** As explained above, *all* C code used directly or indirectly by LuaJIT +** must be compiled with -funwind-tables (or -fexceptions). C++ code must +** *not* be compiled with -fno-exceptions. +** +** If you're unsure whether error handling inside the VM works correctly, +** try running this and check whether it prints "OK": +** +** luajit -e "print(select(2, load('OK')):match('OK'))" +** +** (*) Originally, toolchains only generated unwind tables for C++ code. For +** interoperability reasons, this can be manually enabled for plain C code, +** too (with -funwind-tables). With the introduction of the x64 architecture, +** the corresponding POSIX and Windows ABIs mandated unwind tables for all +** code. Over the following years most desktop and server platforms have +** enabled unwind tables by default on all architectures. OTOH mobile and +** embedded platforms do not consistently mandate unwind tables. */ -#if (defined(__GNUC__) || defined(__clang__)) && (LJ_TARGET_X64 || defined(LUAJIT_UNWIND_EXTERNAL)) && !LJ_NO_UNWIND -#define LJ_UNWIND_EXT 1 -#elif LJ_TARGET_WINDOWS -#define LJ_UNWIND_EXT 1 -#endif - /* -- Error messages ------------------------------------------------------ */ /* Error message strings. */ @@ -184,7 +201,172 @@ static void *err_unwind(lua_State *L, void *stopcf, int errcode) /* -- External frame unwinding -------------------------------------------- */ -#if (defined(__GNUC__) || defined(__clang__)) && !LJ_NO_UNWIND && !LJ_ABI_WIN +#if LJ_ABI_WIN + +/* +** Someone in Redmond owes me several days of my life. A lot of this is +** undocumented or just plain wrong on MSDN. Some of it can be gathered +** from 3rd party docs or must be found by trial-and-error. They really +** don't want you to write your own language-specific exception handler +** or to interact gracefully with MSVC. :-( +** +** Apparently MSVC doesn't call C++ destructors for foreign exceptions +** unless you compile your C++ code with /EHa. Unfortunately this means +** catch (...) also catches things like access violations. The use of +** _set_se_translator doesn't really help, because it requires /EHa, too. +*/ + +#define WIN32_LEAN_AND_MEAN +#include + +#if LJ_TARGET_X86 +typedef void *UndocumentedDispatcherContext; /* Unused on x86. */ +#else +/* Taken from: http://www.nynaeve.net/?p=99 */ +typedef struct UndocumentedDispatcherContext { + ULONG64 ControlPc; + ULONG64 ImageBase; + PRUNTIME_FUNCTION FunctionEntry; + ULONG64 EstablisherFrame; + ULONG64 TargetIp; + PCONTEXT ContextRecord; + void (*LanguageHandler)(void); + PVOID HandlerData; + PUNWIND_HISTORY_TABLE HistoryTable; + ULONG ScopeIndex; + ULONG Fill0; +} UndocumentedDispatcherContext; +#endif + +/* Another wild guess. */ +extern void __DestructExceptionObject(EXCEPTION_RECORD *rec, int nothrow); + +#if LJ_TARGET_X64 && defined(MINGW_SDK_INIT) +/* Workaround for broken MinGW64 declaration. */ +VOID RtlUnwindEx_FIXED(PVOID,PVOID,PVOID,PVOID,PVOID,PVOID) asm("RtlUnwindEx"); +#define RtlUnwindEx RtlUnwindEx_FIXED +#endif + +#define LJ_MSVC_EXCODE ((DWORD)0xe06d7363) +#define LJ_GCC_EXCODE ((DWORD)0x20474343) + +#define LJ_EXCODE ((DWORD)0xe24c4a00) +#define LJ_EXCODE_MAKE(c) (LJ_EXCODE | (DWORD)(c)) +#define LJ_EXCODE_CHECK(cl) (((cl) ^ LJ_EXCODE) <= 0xff) +#define LJ_EXCODE_ERRCODE(cl) ((int)((cl) & 0xff)) + +/* Windows exception handler for interpreter frame. */ +LJ_FUNCA int lj_err_unwind_win(EXCEPTION_RECORD *rec, + void *f, CONTEXT *ctx, UndocumentedDispatcherContext *dispatch) +{ +#if LJ_TARGET_X86 + void *cf = (char *)f - CFRAME_OFS_SEH; +#else + void *cf = f; +#endif + lua_State *L = cframe_L(cf); + int errcode = LJ_EXCODE_CHECK(rec->ExceptionCode) ? + LJ_EXCODE_ERRCODE(rec->ExceptionCode) : LUA_ERRRUN; + if ((rec->ExceptionFlags & 6)) { /* EH_UNWINDING|EH_EXIT_UNWIND */ + /* Unwind internal frames. */ + err_unwind(L, cf, errcode); + } else { + void *cf2 = err_unwind(L, cf, 0); + if (cf2) { /* We catch it, so start unwinding the upper frames. */ + if (rec->ExceptionCode == LJ_MSVC_EXCODE || + rec->ExceptionCode == LJ_GCC_EXCODE) { +#if !LJ_TARGET_CYGWIN + __DestructExceptionObject(rec, 1); +#endif + setstrV(L, L->top++, lj_err_str(L, LJ_ERR_ERRCPP)); + } else if (!LJ_EXCODE_CHECK(rec->ExceptionCode)) { + /* Don't catch access violations etc. */ + return 1; /* ExceptionContinueSearch */ + } +#if LJ_TARGET_X86 + UNUSED(ctx); + UNUSED(dispatch); + /* Call all handlers for all lower C frames (including ourselves) again + ** with EH_UNWINDING set. Then call the specified function, passing cf + ** and errcode. + */ + lj_vm_rtlunwind(cf, (void *)rec, + (cframe_unwind_ff(cf2) && errcode != LUA_YIELD) ? + (void *)lj_vm_unwind_ff : (void *)lj_vm_unwind_c, errcode); + /* lj_vm_rtlunwind does not return. */ +#else + /* Unwind the stack and call all handlers for all lower C frames + ** (including ourselves) again with EH_UNWINDING set. Then set + ** stack pointer = cf, result = errcode and jump to the specified target. + */ + RtlUnwindEx(cf, (void *)((cframe_unwind_ff(cf2) && errcode != LUA_YIELD) ? + lj_vm_unwind_ff_eh : + lj_vm_unwind_c_eh), + rec, (void *)(uintptr_t)errcode, ctx, dispatch->HistoryTable); + /* RtlUnwindEx should never return. */ +#endif + } + } + return 1; /* ExceptionContinueSearch */ +} + +#if LJ_UNWIND_JIT + +#if LJ_TARGET_X64 +#define CONTEXT_REG_PC Rip +#elif LJ_TARGET_ARM64 +#define CONTEXT_REG_PC Pc +#else +#error "NYI: Windows arch-specific unwinder for JIT-compiled code" +#endif + +/* Windows unwinder for JIT-compiled code. */ +static void err_unwind_win_jit(global_State *g, int errcode) +{ + CONTEXT ctx; + UNWIND_HISTORY_TABLE hist; + + memset(&hist, 0, sizeof(hist)); + RtlCaptureContext(&ctx); + while (1) { + uintptr_t frame, base, addr = ctx.CONTEXT_REG_PC; + void *hdata; + PRUNTIME_FUNCTION func = RtlLookupFunctionEntry(addr, &base, &hist); + if (!func) { /* Found frame without .pdata: must be JIT-compiled code. */ + ExitNo exitno; + uintptr_t stub = lj_trace_unwind(G2J(g), addr - sizeof(MCode), &exitno); + if (stub) { /* Jump to side exit to unwind the trace. */ + ctx.CONTEXT_REG_PC = stub; + G2J(g)->exitcode = errcode; + RtlRestoreContext(&ctx, NULL); /* Does not return. */ + } + break; + } + RtlVirtualUnwind(UNW_FLAG_NHANDLER, base, addr, func, + &ctx, &hdata, &frame, NULL); + if (!addr) break; + } + /* Unwinding failed, if we end up here. */ +} +#endif + +/* Raise Windows exception. */ +static void err_raise_ext(global_State *g, int errcode) +{ +#if LJ_UNWIND_JIT + if (tvref(g->jit_base)) { + err_unwind_win_jit(g, errcode); + return; /* Unwinding failed. */ + } +#elif LJ_HASJIT + /* Cannot catch on-trace errors for Windows/x86 SEH. Unwind to interpreter. */ + setmref(g->jit_base, NULL); +#endif + UNUSED(g); + RaiseException(LJ_EXCODE_MAKE(errcode), 1 /* EH_NONCONTINUABLE */, 0, NULL); +} + +#elif !LJ_NO_UNWIND && (defined(__GNUC__) || defined(__clang__)) /* ** We have to use our own definitions instead of the mandatory (!) unwind.h, @@ -194,6 +376,7 @@ static void *err_unwind(lua_State *L, void *stopcf, int errcode) typedef struct _Unwind_Context _Unwind_Context; #define _URC_OK 0 +#define _URC_FATAL_PHASE2_ERROR 2 #define _URC_FATAL_PHASE1_ERROR 3 #define _URC_HANDLER_FOUND 6 #define _URC_INSTALL_CONTEXT 7 @@ -213,9 +396,11 @@ typedef struct _Unwind_Exception void (*excleanup)(int, struct _Unwind_Exception *); uintptr_t p1, p2; } __attribute__((__aligned__)) _Unwind_Exception; +#define UNWIND_EXCEPTION_TYPE _Unwind_Exception extern uintptr_t _Unwind_GetCFA(_Unwind_Context *); extern void _Unwind_SetGR(_Unwind_Context *, int, uintptr_t); +extern uintptr_t _Unwind_GetIP(_Unwind_Context *); extern void _Unwind_SetIP(_Unwind_Context *, uintptr_t); extern void _Unwind_DeleteException(_Unwind_Exception *); extern int _Unwind_RaiseException(_Unwind_Exception *); @@ -233,7 +418,6 @@ LJ_FUNCA int lj_err_unwind_dwarf(int version, int actions, lua_State *L; if (version != 1) return _URC_FATAL_PHASE1_ERROR; - UNUSED(uexclass); cf = (void *)_Unwind_GetCFA(ctx); L = cframe_L(cf); if ((actions & _UA_SEARCH_PHASE)) { @@ -281,20 +465,139 @@ LJ_FUNCA int lj_err_unwind_dwarf(int version, int actions, ** it on non-x64 because the interpreter restores all callee-saved regs. */ lj_err_throw(L, errcode); +#if LJ_TARGET_X64 +#error "Broken build system -- only use the provided Makefiles!" +#endif #endif } return _URC_CONTINUE_UNWIND; } -#if LJ_UNWIND_EXT -static __thread _Unwind_Exception static_uex; +#if LJ_UNWIND_EXT && defined(LUA_USE_ASSERT) +struct dwarf_eh_bases { void *tbase, *dbase, *func; }; +extern const void *_Unwind_Find_FDE(void *pc, struct dwarf_eh_bases *bases); -/* Raise DWARF2 exception. */ -static void err_raise_ext(int errcode) +/* Verify that external error handling actually has a chance to work. */ +void lj_err_verify(void) { - static_uex.exclass = LJ_UEXCLASS_MAKE(errcode); - static_uex.excleanup = NULL; - _Unwind_RaiseException(&static_uex); + struct dwarf_eh_bases ehb; + lj_assertX(_Unwind_Find_FDE((void *)lj_err_throw, &ehb), "broken build: external frame unwinding enabled, but missing -funwind-tables"); + /* Check disabled, because of broken Fedora/ARM64. See #722. + lj_assertX(_Unwind_Find_FDE((void *)_Unwind_RaiseException, &ehb), "broken build: external frame unwinding enabled, but system libraries have no unwind tables"); + */ +} +#endif + +#if LJ_UNWIND_JIT +/* DWARF2 personality handler for JIT-compiled code. */ +static int err_unwind_jit(int version, int actions, + uint64_t uexclass, _Unwind_Exception *uex, _Unwind_Context *ctx) +{ + /* NYI: FFI C++ exception interoperability. */ + if (version != 1 || !LJ_UEXCLASS_CHECK(uexclass)) + return _URC_FATAL_PHASE1_ERROR; + if ((actions & _UA_SEARCH_PHASE)) { + return _URC_HANDLER_FOUND; + } + if ((actions & _UA_CLEANUP_PHASE)) { + global_State *g = *(global_State **)(uex+1); + ExitNo exitno; + uintptr_t addr = _Unwind_GetIP(ctx); /* Return address _after_ call. */ + uintptr_t stub = lj_trace_unwind(G2J(g), addr - sizeof(MCode), &exitno); + lj_assertG(tvref(g->jit_base), "unexpected throw across mcode frame"); + if (stub) { /* Jump to side exit to unwind the trace. */ + G2J(g)->exitcode = LJ_UEXCLASS_ERRCODE(uexclass); +#ifdef LJ_TARGET_MIPS + _Unwind_SetGR(ctx, 4, stub); + _Unwind_SetGR(ctx, 5, exitno); + _Unwind_SetIP(ctx, (uintptr_t)(void *)lj_vm_unwind_stub); +#else + _Unwind_SetIP(ctx, stub); +#endif + return _URC_INSTALL_CONTEXT; + } + return _URC_FATAL_PHASE2_ERROR; + } + return _URC_FATAL_PHASE1_ERROR; +} + +/* DWARF2 template frame info for JIT-compiled code. +** +** After copying the template to the start of the mcode segment, +** the frame handler function and the code size is patched. +** The frame handler always installs a new context to jump to the exit, +** so don't bother to add any unwind opcodes. +*/ +static const uint8_t err_frame_jit_template[] = { +#if LJ_BE + 0,0,0, +#endif + LJ_64 ? 0x1c : 0x14, /* CIE length. */ +#if LJ_LE + 0,0,0, +#endif + 0,0,0,0, 1, 'z','P','R',0, /* CIE mark, CIE version, augmentation. */ + 1, LJ_64 ? 0x78 : 0x7c, LJ_TARGET_EHRAREG, /* Code/data align, RA. */ +#if LJ_64 + 10, 0, 0,0,0,0,0,0,0,0, 0x1b, /* Aug. data ABS handler, PCREL|SDATA4 code. */ + 0,0,0,0,0, /* Alignment. */ +#else + 6, 0, 0,0,0,0, 0x1b, /* Aug. data ABS handler, PCREL|SDATA4 code. */ + 0, /* Alignment. */ +#endif +#if LJ_BE + 0,0,0, +#endif + LJ_64 ? 0x14 : 0x10, /* FDE length. */ + 0,0,0, + LJ_64 ? 0x24 : 0x1c, /* CIE offset. */ + 0,0,0, + LJ_64 ? 0x14 : 0x10, /* Code offset. After Final FDE. */ +#if LJ_LE + 0,0,0, +#endif + 0,0,0,0, 0, 0,0,0, /* Code size, augmentation length, alignment. */ +#if LJ_64 + 0,0,0,0, /* Alignment. */ +#endif + 0,0,0,0 /* Final FDE. */ +}; + +#define ERR_FRAME_JIT_OFS_HANDLER 0x12 +#define ERR_FRAME_JIT_OFS_FDE (LJ_64 ? 0x20 : 0x18) +#define ERR_FRAME_JIT_OFS_CODE_SIZE (LJ_64 ? 0x2c : 0x24) +#if LJ_TARGET_OSX +#define ERR_FRAME_JIT_OFS_REGISTER ERR_FRAME_JIT_OFS_FDE +#else +#define ERR_FRAME_JIT_OFS_REGISTER 0 +#endif + +extern void __register_frame(const void *); +extern void __deregister_frame(const void *); + +uint8_t *lj_err_register_mcode(void *base, size_t sz, uint8_t *info) +{ + void **handler; + memcpy(info, err_frame_jit_template, sizeof(err_frame_jit_template)); + handler = (void *)err_unwind_jit; + memcpy(info + ERR_FRAME_JIT_OFS_HANDLER, &handler, sizeof(handler)); + *(uint32_t *)(info + ERR_FRAME_JIT_OFS_CODE_SIZE) = + (uint32_t)(sz - sizeof(err_frame_jit_template) - (info - (uint8_t *)base)); + __register_frame(info + ERR_FRAME_JIT_OFS_REGISTER); +#ifdef LUA_USE_ASSERT + { + struct dwarf_eh_bases ehb; + lj_assertX(_Unwind_Find_FDE(info + sizeof(err_frame_jit_template)+1, &ehb), + "bad JIT unwind table registration"); + } +#endif + return info + sizeof(err_frame_jit_template); +} + +void lj_err_deregister_mcode(void *base, size_t sz, uint8_t *info) +{ + UNUSED(base); UNUSED(sz); + __deregister_frame(info + ERR_FRAME_JIT_OFS_REGISTER); } #endif @@ -306,6 +609,7 @@ static void err_raise_ext(int errcode) #define _US_FORCE_UNWIND 8 typedef struct _Unwind_Control_Block _Unwind_Control_Block; +#define UNWIND_EXCEPTION_TYPE _Unwind_Control_Block struct _Unwind_Control_Block { uint64_t exclass; @@ -364,136 +668,63 @@ LJ_FUNCA int lj_err_unwind_arm(int state, _Unwind_Control_Block *ucb, } if (__gnu_unwind_frame(ucb, ctx) != _URC_OK) return _URC_FAILURE; +#ifdef LUA_USE_ASSERT + /* We should never get here unless this is a forced unwind aka backtrace. */ + if (_Unwind_GetGR(ctx, 0) == 0xff33aa77) { + _Unwind_SetGR(ctx, 0, 0xff33aa88); + } +#endif return _URC_CONTINUE_UNWIND; } -#if LJ_UNWIND_EXT -static __thread _Unwind_Control_Block static_uex; +#if LJ_UNWIND_EXT && defined(LUA_USE_ASSERT) +typedef int (*_Unwind_Trace_Fn)(_Unwind_Context *, void *); +extern int _Unwind_Backtrace(_Unwind_Trace_Fn, void *); -static void err_raise_ext(int errcode) +static int err_verify_bt(_Unwind_Context *ctx, int *got) { - memset(&static_uex, 0, sizeof(static_uex)); - static_uex.exclass = LJ_UEXCLASS_MAKE(errcode); - _Unwind_RaiseException(&static_uex); + if (_Unwind_GetGR(ctx, 0) == 0xff33aa88) { *got = 2; } + else if (*got == 0) { *got = 1; _Unwind_SetGR(ctx, 0, 0xff33aa77); } + return _URC_OK; +} + +/* Verify that external error handling actually has a chance to work. */ +void lj_err_verify(void) +{ + int got = 0; + _Unwind_Backtrace((_Unwind_Trace_Fn)err_verify_bt, &got); + lj_assertX(got == 2, "broken build: external frame unwinding enabled, but missing -funwind-tables"); } #endif +/* +** Note: LJ_UNWIND_JIT is not implemented for 32 bit ARM. +** +** The quirky ARM unwind API doesn't have __register_frame(). +** A potential workaround might involve _Unwind_Backtrace. +** But most 32 bit ARM targets don't qualify for LJ_UNWIND_EXT, anyway, +** since they are built without unwind tables by default. +*/ + #endif /* LJ_TARGET_ARM */ -#elif LJ_ABI_WIN -/* -** Someone in Redmond owes me several days of my life. A lot of this is -** undocumented or just plain wrong on MSDN. Some of it can be gathered -** from 3rd party docs or must be found by trial-and-error. They really -** don't want you to write your own language-specific exception handler -** or to interact gracefully with MSVC. :-( -** -** Apparently MSVC doesn't call C++ destructors for foreign exceptions -** unless you compile your C++ code with /EHa. Unfortunately this means -** catch (...) also catches things like access violations. The use of -** _set_se_translator doesn't really help, because it requires /EHa, too. -*/ +#if LJ_UNWIND_EXT +static __thread struct { + UNWIND_EXCEPTION_TYPE ex; + global_State *g; +} static_uex; -#define WIN32_LEAN_AND_MEAN -#include - -#if LJ_TARGET_X64 -/* Taken from: http://www.nynaeve.net/?p=99 */ -typedef struct UndocumentedDispatcherContext { - ULONG64 ControlPc; - ULONG64 ImageBase; - PRUNTIME_FUNCTION FunctionEntry; - ULONG64 EstablisherFrame; - ULONG64 TargetIp; - PCONTEXT ContextRecord; - void (*LanguageHandler)(void); - PVOID HandlerData; - PUNWIND_HISTORY_TABLE HistoryTable; - ULONG ScopeIndex; - ULONG Fill0; -} UndocumentedDispatcherContext; -#else -typedef void *UndocumentedDispatcherContext; -#endif - -/* Another wild guess. */ -extern void __DestructExceptionObject(EXCEPTION_RECORD *rec, int nothrow); - -#if LJ_TARGET_X64 && defined(MINGW_SDK_INIT) -/* Workaround for broken MinGW64 declaration. */ -VOID RtlUnwindEx_FIXED(PVOID,PVOID,PVOID,PVOID,PVOID,PVOID) asm("RtlUnwindEx"); -#define RtlUnwindEx RtlUnwindEx_FIXED -#endif - -#define LJ_MSVC_EXCODE ((DWORD)0xe06d7363) -#define LJ_GCC_EXCODE ((DWORD)0x20474343) - -#define LJ_EXCODE ((DWORD)0xe24c4a00) -#define LJ_EXCODE_MAKE(c) (LJ_EXCODE | (DWORD)(c)) -#define LJ_EXCODE_CHECK(cl) (((cl) ^ LJ_EXCODE) <= 0xff) -#define LJ_EXCODE_ERRCODE(cl) ((int)((cl) & 0xff)) - -/* Windows exception handler for interpreter frame. */ -LJ_FUNCA int lj_err_unwind_win(EXCEPTION_RECORD *rec, - void *f, CONTEXT *ctx, UndocumentedDispatcherContext *dispatch) +/* Raise external exception. */ +static void err_raise_ext(global_State *g, int errcode) { -#if LJ_TARGET_X64 - void *cf = f; -#else - void *cf = (char *)f - CFRAME_OFS_SEH; -#endif - lua_State *L = cframe_L(cf); - int errcode = LJ_EXCODE_CHECK(rec->ExceptionCode) ? - LJ_EXCODE_ERRCODE(rec->ExceptionCode) : LUA_ERRRUN; - if ((rec->ExceptionFlags & 6)) { /* EH_UNWINDING|EH_EXIT_UNWIND */ - /* Unwind internal frames. */ - err_unwind(L, cf, errcode); - } else { - void *cf2 = err_unwind(L, cf, 0); - if (cf2) { /* We catch it, so start unwinding the upper frames. */ - if (rec->ExceptionCode == LJ_MSVC_EXCODE || - rec->ExceptionCode == LJ_GCC_EXCODE) { -#if LJ_TARGET_WINDOWS - __DestructExceptionObject(rec, 1); -#endif - setstrV(L, L->top++, lj_err_str(L, LJ_ERR_ERRCPP)); - } else if (!LJ_EXCODE_CHECK(rec->ExceptionCode)) { - /* Don't catch access violations etc. */ - return 1; /* ExceptionContinueSearch */ - } -#if LJ_TARGET_X64 - /* Unwind the stack and call all handlers for all lower C frames - ** (including ourselves) again with EH_UNWINDING set. Then set - ** rsp = cf, rax = errcode and jump to the specified target. - */ - RtlUnwindEx(cf, (void *)((cframe_unwind_ff(cf2) && errcode != LUA_YIELD) ? - lj_vm_unwind_ff_eh : - lj_vm_unwind_c_eh), - rec, (void *)(uintptr_t)errcode, ctx, dispatch->HistoryTable); - /* RtlUnwindEx should never return. */ -#else - UNUSED(ctx); - UNUSED(dispatch); - /* Call all handlers for all lower C frames (including ourselves) again - ** with EH_UNWINDING set. Then call the specified function, passing cf - ** and errcode. - */ - lj_vm_rtlunwind(cf, (void *)rec, - (cframe_unwind_ff(cf2) && errcode != LUA_YIELD) ? - (void *)lj_vm_unwind_ff : (void *)lj_vm_unwind_c, errcode); - /* lj_vm_rtlunwind does not return. */ -#endif - } - } - return 1; /* ExceptionContinueSearch */ + memset(&static_uex, 0, sizeof(static_uex)); + static_uex.ex.exclass = LJ_UEXCLASS_MAKE(errcode); + static_uex.g = g; + _Unwind_RaiseException(&static_uex.ex); } -/* Raise Windows exception. */ -static void err_raise_ext(int errcode) -{ - RaiseException(LJ_EXCODE_MAKE(errcode), 1 /* EH_NONCONTINUABLE */, 0, NULL); -} +#endif #endif @@ -504,22 +735,23 @@ LJ_NOINLINE void LJ_FASTCALL lj_err_throw(lua_State *L, int errcode) { global_State *g = G(L); lj_trace_abort(g); - setmref(g->jit_base, NULL); L->status = LUA_OK; #if LJ_UNWIND_EXT - err_raise_ext(errcode); + err_raise_ext(g, errcode); /* ** A return from this function signals a corrupt C stack that cannot be ** unwound. We have no choice but to call the panic function and exit. ** ** Usually this is caused by a C function without unwind information. - ** This should never happen on x64, but may happen if you've manually - ** enabled LUAJIT_UNWIND_EXTERNAL and forgot to recompile *every* - ** non-C++ file with -funwind-tables. + ** This may happen if you've manually enabled LUAJIT_UNWIND_EXTERNAL + ** and forgot to recompile *every* non-C++ file with -funwind-tables. */ if (G(L)->panic) G(L)->panic(L); #else +#if LJ_HASJIT + setmref(g->jit_base, NULL); +#endif { void *cf = err_unwind(L, NULL, errcode); if (cframe_unwind_ff(cf)) @@ -600,7 +832,7 @@ static ptrdiff_t finderrfunc(lua_State *L) /* Runtime error. */ LJ_NOINLINE void LJ_FASTCALL lj_err_run(lua_State *L) { - ptrdiff_t ef = finderrfunc(L); + ptrdiff_t ef = (LJ_HASJIT && tvref(G(L)->jit_base)) ? 0 : finderrfunc(L); if (ef) { TValue *errfunc = restorestack(L, ef); TValue *top = L->top; @@ -619,6 +851,16 @@ LJ_NOINLINE void LJ_FASTCALL lj_err_run(lua_State *L) lj_err_throw(L, LUA_ERRRUN); } +#if LJ_HASJIT +LJ_NOINLINE void LJ_FASTCALL lj_err_trace(lua_State *L, int errcode) +{ + if (errcode == LUA_ERRRUN) + lj_err_run(L); + else + lj_err_throw(L, errcode); +} +#endif + /* Formatted runtime error message. */ LJ_NORET LJ_NOINLINE static void err_msgv(lua_State *L, ErrMsg em, ...) { @@ -699,25 +941,27 @@ LJ_NOINLINE void lj_err_optype_call(lua_State *L, TValue *o) /* Error in context of caller. */ LJ_NOINLINE void lj_err_callermsg(lua_State *L, const char *msg) { - TValue *frame = L->base-1; - TValue *pframe = NULL; - if (frame_islua(frame)) { - pframe = frame_prevl(frame); - } else if (frame_iscont(frame)) { - if (frame_iscont_fficb(frame)) { - pframe = frame; - frame = NULL; - } else { - pframe = frame_prevd(frame); + TValue *frame = NULL, *pframe = NULL; + if (!(LJ_HASJIT && tvref(G(L)->jit_base))) { + frame = L->base-1; + if (frame_islua(frame)) { + pframe = frame_prevl(frame); + } else if (frame_iscont(frame)) { + if (frame_iscont_fficb(frame)) { + pframe = frame; + frame = NULL; + } else { + pframe = frame_prevd(frame); #if LJ_HASFFI - /* Remove frame for FFI metamethods. */ - if (frame_func(frame)->c.ffid >= FF_ffi_meta___index && - frame_func(frame)->c.ffid <= FF_ffi_meta___tostring) { - L->base = pframe+1; - L->top = frame; - setcframe_pc(cframe_raw(L->cframe), frame_contpc(frame)); - } + /* Remove frame for FFI metamethods. */ + if (frame_func(frame)->c.ffid >= FF_ffi_meta___index && + frame_func(frame)->c.ffid <= FF_ffi_meta___tostring) { + L->base = pframe+1; + L->top = frame; + setcframe_pc(cframe_raw(L->cframe), frame_contpc(frame)); + } #endif + } } } lj_debug_addloc(L, msg, pframe, frame); diff --git a/love/src/jni/LuaJIT-2.1/src/lj_err.h b/love/src/jni/LuaJIT-2.1/src/lj_err.h index 770553fc..2e8a251f 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_err.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_err.h @@ -23,7 +23,10 @@ LJ_DATA const char *lj_err_allmsg; LJ_FUNC GCstr *lj_err_str(lua_State *L, ErrMsg em); LJ_FUNCA_NORET void LJ_FASTCALL lj_err_throw(lua_State *L, int errcode); LJ_FUNC_NORET void lj_err_mem(lua_State *L); -LJ_FUNCA_NORET void LJ_FASTCALL lj_err_run(lua_State *L); +LJ_FUNC_NORET void LJ_FASTCALL lj_err_run(lua_State *L); +#if LJ_HASJIT +LJ_FUNCA_NORET void LJ_FASTCALL lj_err_trace(lua_State *L, int errcode); +#endif LJ_FUNC_NORET void lj_err_msg(lua_State *L, ErrMsg em); LJ_FUNC_NORET void lj_err_lex(lua_State *L, GCstr *src, const char *tok, BCLine line, ErrMsg em, va_list argp); @@ -38,4 +41,18 @@ LJ_FUNC_NORET void lj_err_argv(lua_State *L, int narg, ErrMsg em, ...); LJ_FUNC_NORET void lj_err_argtype(lua_State *L, int narg, const char *xname); LJ_FUNC_NORET void lj_err_argt(lua_State *L, int narg, int tt); +#if LJ_UNWIND_JIT && !LJ_ABI_WIN +LJ_FUNC uint8_t *lj_err_register_mcode(void *base, size_t sz, uint8_t *info); +LJ_FUNC void lj_err_deregister_mcode(void *base, size_t sz, uint8_t *info); +#else +#define lj_err_register_mcode(base, sz, info) (info) +#define lj_err_deregister_mcode(base, sz, info) UNUSED(base) +#endif + +#if LJ_UNWIND_EXT && !LJ_ABI_WIN && defined(LUA_USE_ASSERT) +LJ_FUNC void lj_err_verify(void); +#else +#define lj_err_verify() ((void)0) +#endif + #endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_errmsg.h b/love/src/jni/LuaJIT-2.1/src/lj_errmsg.h index 9ff4553d..89e67496 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_errmsg.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_errmsg.h @@ -67,6 +67,7 @@ ERRDEF(PROTMT, "cannot change a protected metatable") ERRDEF(UNPACK, "too many results to unpack") ERRDEF(RDRSTR, "reader function must return a string") ERRDEF(PRTOSTR, LUA_QL("tostring") " must return a string to " LUA_QL("print")) +ERRDEF(NUMRNG, "number out of range") ERRDEF(IDXRNG, "index out of range") ERRDEF(BASERNG, "base out of range") ERRDEF(LVLRNG, "level out of range") @@ -179,6 +180,19 @@ ERRDEF(FFI_NYIPACKBIT, "NYI: packed bit fields") ERRDEF(FFI_NYICALL, "NYI: cannot call this C function (yet)") #endif +#if LJ_HASBUFFER +/* String buffer errors. */ +ERRDEF(BUFFER_SELF, "cannot put buffer into itself") +ERRDEF(BUFFER_BADOPT, "bad options table") +ERRDEF(BUFFER_BADENC, "cannot serialize " LUA_QS) +ERRDEF(BUFFER_BADDEC, "cannot deserialize tag 0x%02x") +ERRDEF(BUFFER_BADDICTX, "cannot deserialize dictionary index %d") +ERRDEF(BUFFER_DEPTH, "too deep to serialize") +ERRDEF(BUFFER_DUPKEY, "duplicate table key") +ERRDEF(BUFFER_EOB, "unexpected end of buffer") +ERRDEF(BUFFER_LEFTOV, "left-over data in buffer") +#endif + #undef ERRDEF /* Detecting unused error messages: diff --git a/love/src/jni/LuaJIT-2.1/src/lj_ffrecord.c b/love/src/jni/LuaJIT-2.1/src/lj_ffrecord.c index baf7279d..97b40f98 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_ffrecord.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_ffrecord.c @@ -11,6 +11,7 @@ #if LJ_HASJIT #include "lj_err.h" +#include "lj_buf.h" #include "lj_str.h" #include "lj_tab.h" #include "lj_frame.h" @@ -28,6 +29,7 @@ #include "lj_vm.h" #include "lj_strscan.h" #include "lj_strfmt.h" +#include "lj_serialize.h" /* Some local macros to save typing. Undef'd at the end. */ #define IR(ref) (&J->cur.ir[(ref)]) @@ -107,6 +109,10 @@ static void recff_stitch(jit_State *J) const BCIns *pc = frame_pc(base-1); TValue *pframe = frame_prevl(base-1); + /* Check for this now. Throwing in lj_record_stop messes up the stack. */ + if (J->cur.nsnap >= (MSize)J->param[JIT_P_maxsnap]) + lj_trace_err(J, LJ_TRERR_SNAPOV); + /* Move func + args up in Lua stack and insert continuation. */ memmove(&base[1], &base[-1-LJ_FR2], sizeof(TValue)*nslot); setframe_ftsz(nframe, ((char *)nframe - (char *)pframe) + FRAME_CONT); @@ -182,6 +188,14 @@ static TRef recff_bufhdr(jit_State *J) lj_ir_kptr(J, &J2G(J)->tmpbuf), IRBUFHDR_RESET); } +/* Emit TMPREF. */ +static TRef recff_tmpref(jit_State *J, TRef tr, int mode) +{ + if (!LJ_DUALNUM && tref_isinteger(tr)) + tr = emitir(IRTN(IR_CONV), tr, IRCONV_NUM_INT); + return emitir(IRT(IR_TMPREF, IRT_PGC), tr, mode); +} + /* -- Base library fast functions ----------------------------------------- */ static void LJ_FASTCALL recff_assert(jit_State *J, RecordFFData *rd) @@ -296,7 +310,7 @@ int32_t lj_ffrecord_select_mode(jit_State *J, TRef tr, TValue *tv) } else { TRef trptr = emitir(IRT(IR_STRREF, IRT_PGC), tr, lj_ir_kint(J, 0)); TRef trchar = emitir(IRT(IR_XLOAD, IRT_U8), trptr, IRXLOAD_READONLY); - emitir(IRTG(IR_EQ, IRT_INT), trchar, lj_ir_kint(J, '#')); + emitir(IRTGI(IR_EQ), trchar, lj_ir_kint(J, '#')); } return 0; } else { /* select(n, ...) */ @@ -317,9 +331,9 @@ static void LJ_FASTCALL recff_select(jit_State *J, RecordFFData *rd) ptrdiff_t n = (ptrdiff_t)J->maxslot; if (start < 0) start += n; else if (start > n) start = n; - rd->nres = n - start; if (start >= 1) { ptrdiff_t i; + rd->nres = n - start; for (i = 0; i < n - start; i++) J->base[i] = J->base[start+i]; } /* else: Interpreter will throw. */ @@ -455,6 +469,7 @@ static void LJ_FASTCALL recff_pcall(jit_State *J, RecordFFData *rd) #endif lj_record_call(J, 0, J->maxslot - 1); rd->nres = -1; /* Pending call. */ + J->needsnap = 1; /* Start catching on-trace errors. */ } /* else: Interpreter will throw. */ } @@ -490,6 +505,7 @@ static void LJ_FASTCALL recff_xpcall(jit_State *J, RecordFFData *rd) if (errcode) lj_err_throw(J->L, errcode); /* Propagate errors. */ rd->nres = -1; /* Pending call. */ + J->needsnap = 1; /* Start catching on-trace errors. */ } /* else: Interpreter will throw. */ } @@ -505,6 +521,40 @@ static void LJ_FASTCALL recff_getfenv(jit_State *J, RecordFFData *rd) recff_nyiu(J, rd); } +static void LJ_FASTCALL recff_next(jit_State *J, RecordFFData *rd) +{ +#if LJ_BE + /* YAGNI: Disabled on big-endian due to issues with lj_vm_next, + ** IR_HIOP, RID_RETLO/RID_RETHI and ra_destpair. + */ + recff_nyi(J, rd); +#else + TRef tab = J->base[0]; + if (tref_istab(tab)) { + RecordIndex ix; + cTValue *keyv; + ix.tab = tab; + if (tref_isnil(J->base[1])) { /* Shortcut for start of traversal. */ + ix.key = lj_ir_kint(J, 0); + keyv = niltvg(J2G(J)); + } else { + TRef tmp = recff_tmpref(J, J->base[1], IRTMPREF_IN1); + ix.key = lj_ir_call(J, IRCALL_lj_tab_keyindex, tab, tmp); + keyv = &rd->argv[1]; + } + copyTV(J->L, &ix.tabv, &rd->argv[0]); + ix.keyv.u32.lo = lj_tab_keyindex(tabV(&ix.tabv), keyv); + /* Omit the value, if not used by the caller. */ + ix.idxchain = (J->framedepth && frame_islua(J->L->base-1) && + bc_b(frame_pc(J->L->base-1)[-1])-1 < 2); + ix.mobj = 0; /* We don't need the next index. */ + rd->nres = lj_record_next(J, &ix); + J->base[0] = ix.key; + J->base[1] = ix.val; + } /* else: Interpreter will throw. */ +#endif +} + /* -- Math library fast functions ----------------------------------------- */ static void LJ_FASTCALL recff_math_abs(jit_State *J, RecordFFData *rd) @@ -707,7 +757,7 @@ static void LJ_FASTCALL recff_bit_tohex(jit_State *J, RecordFFData *rd) #if LJ_HASFFI TRef hdr = recff_bufhdr(J); TRef tr = recff_bit64_tohex(J, rd, hdr); - J->base[0] = emitir(IRT(IR_BUFSTR, IRT_STR), tr, hdr); + J->base[0] = emitir(IRTG(IR_BUFSTR, IRT_STR), tr, hdr); #else recff_nyiu(J, rd); /* Don't bother working around this NYI. */ #endif @@ -833,8 +883,8 @@ static void LJ_FASTCALL recff_string_char(jit_State *J, RecordFFData *rd) if (i > 1) { /* Concatenate the strings, if there's more than one. */ TRef hdr = recff_bufhdr(J), tr = hdr; for (i = 0; J->base[i] != 0; i++) - tr = emitir(IRT(IR_BUFPUT, IRT_PGC), tr, J->base[i]); - J->base[0] = emitir(IRT(IR_BUFSTR, IRT_STR), tr, hdr); + tr = emitir(IRTG(IR_BUFPUT, IRT_PGC), tr, J->base[i]); + J->base[0] = emitir(IRTG(IR_BUFSTR, IRT_STR), tr, hdr); } else if (i == 0) { J->base[0] = lj_ir_kstr(J, &J2G(J)->strempty); } @@ -852,19 +902,19 @@ static void LJ_FASTCALL recff_string_rep(jit_State *J, RecordFFData *rd) emitir(IRTGI(vrep > 1 ? IR_GT : IR_LE), rep, lj_ir_kint(J, 1)); if (vrep > 1) { TRef hdr2 = recff_bufhdr(J); - TRef tr2 = emitir(IRT(IR_BUFPUT, IRT_PGC), hdr2, sep); - tr2 = emitir(IRT(IR_BUFPUT, IRT_PGC), tr2, str); - str2 = emitir(IRT(IR_BUFSTR, IRT_STR), tr2, hdr2); + TRef tr2 = emitir(IRTG(IR_BUFPUT, IRT_PGC), hdr2, sep); + tr2 = emitir(IRTG(IR_BUFPUT, IRT_PGC), tr2, str); + str2 = emitir(IRTG(IR_BUFSTR, IRT_STR), tr2, hdr2); } } tr = hdr = recff_bufhdr(J); if (str2) { - tr = emitir(IRT(IR_BUFPUT, IRT_PGC), tr, str); + tr = emitir(IRTG(IR_BUFPUT, IRT_PGC), tr, str); str = str2; rep = emitir(IRTI(IR_ADD), rep, lj_ir_kint(J, -1)); } tr = lj_ir_call(J, IRCALL_lj_buf_putstr_rep, tr, str, rep); - J->base[0] = emitir(IRT(IR_BUFSTR, IRT_STR), tr, hdr); + J->base[0] = emitir(IRTG(IR_BUFSTR, IRT_STR), tr, hdr); } static void LJ_FASTCALL recff_string_op(jit_State *J, RecordFFData *rd) @@ -872,7 +922,7 @@ static void LJ_FASTCALL recff_string_op(jit_State *J, RecordFFData *rd) TRef str = lj_ir_tostr(J, J->base[0]); TRef hdr = recff_bufhdr(J); TRef tr = lj_ir_call(J, rd->data, hdr, str); - J->base[0] = emitir(IRT(IR_BUFSTR, IRT_STR), tr, hdr); + J->base[0] = emitir(IRTG(IR_BUFSTR, IRT_STR), tr, hdr); } static void LJ_FASTCALL recff_string_find(jit_State *J, RecordFFData *rd) @@ -935,34 +985,40 @@ static void LJ_FASTCALL recff_string_find(jit_State *J, RecordFFData *rd) } } -static void LJ_FASTCALL recff_string_format(jit_State *J, RecordFFData *rd) +static void recff_format(jit_State *J, RecordFFData *rd, TRef hdr, int sbufx) { - TRef trfmt = lj_ir_tostr(J, J->base[0]); - GCstr *fmt = argv2str(J, &rd->argv[0]); - int arg = 1; - TRef hdr, tr; + ptrdiff_t arg = sbufx; + TRef tr = hdr, trfmt = lj_ir_tostr(J, J->base[arg]); + GCstr *fmt = argv2str(J, &rd->argv[arg]); FormatState fs; SFormat sf; /* Specialize to the format string. */ emitir(IRTG(IR_EQ, IRT_STR), trfmt, lj_ir_kstr(J, fmt)); - tr = hdr = recff_bufhdr(J); lj_strfmt_init(&fs, strdata(fmt), fmt->len); while ((sf = lj_strfmt_parse(&fs)) != STRFMT_EOF) { /* Parse format. */ - TRef tra = sf == STRFMT_LIT ? 0 : J->base[arg++]; + TRef tra = sf == STRFMT_LIT ? 0 : J->base[++arg]; TRef trsf = lj_ir_kint(J, (int32_t)sf); IRCallID id; switch (STRFMT_TYPE(sf)) { case STRFMT_LIT: - tr = emitir(IRT(IR_BUFPUT, IRT_PGC), tr, + tr = emitir(IRTG(IR_BUFPUT, IRT_PGC), tr, lj_ir_kstr(J, lj_str_new(J->L, fs.str, fs.len))); break; case STRFMT_INT: id = IRCALL_lj_strfmt_putfnum_int; handle_int: - if (!tref_isinteger(tra)) + if (!tref_isinteger(tra)) { +#if LJ_HASFFI + if (tref_iscdata(tra)) { + tra = lj_crecord_loadiu64(J, tra, &rd->argv[arg]); + tr = lj_ir_call(J, IRCALL_lj_strfmt_putfxint, tr, trsf, tra); + break; + } +#endif goto handle_num; + } if (sf == STRFMT_INT) { /* Shortcut for plain %d. */ - tr = emitir(IRT(IR_BUFPUT, IRT_PGC), tr, + tr = emitir(IRTG(IR_BUFPUT, IRT_PGC), tr, emitir(IRT(IR_TOSTR, IRT_STR), tra, IRTOSTR_INT)); } else { #if LJ_HASFFI @@ -989,10 +1045,11 @@ static void LJ_FASTCALL recff_string_format(jit_State *J, RecordFFData *rd) case STRFMT_STR: if (!tref_isstr(tra)) { recff_nyiu(J, rd); /* NYI: __tostring and non-string types for %s. */ + /* NYI: also buffers. */ return; } if (sf == STRFMT_STR) /* Shortcut for plain %s. */ - tr = emitir(IRT(IR_BUFPUT, IRT_PGC), tr, tra); + tr = emitir(IRTG(IR_BUFPUT, IRT_PGC), tr, tra); else if ((sf & STRFMT_T_QUOTED)) tr = lj_ir_call(J, IRCALL_lj_strfmt_putquoted, tr, tra); else @@ -1001,7 +1058,7 @@ static void LJ_FASTCALL recff_string_format(jit_State *J, RecordFFData *rd) case STRFMT_CHAR: tra = lj_opt_narrow_toint(J, tra); if (sf == STRFMT_CHAR) /* Shortcut for plain %c. */ - tr = emitir(IRT(IR_BUFPUT, IRT_PGC), tr, + tr = emitir(IRTG(IR_BUFPUT, IRT_PGC), tr, emitir(IRT(IR_TOSTR, IRT_STR), tra, IRTOSTR_CHAR)); else tr = lj_ir_call(J, IRCALL_lj_strfmt_putfchar, tr, trsf, tra); @@ -1013,9 +1070,326 @@ static void LJ_FASTCALL recff_string_format(jit_State *J, RecordFFData *rd) return; } } - J->base[0] = emitir(IRT(IR_BUFSTR, IRT_STR), tr, hdr); + if (sbufx) { + emitir(IRT(IR_USE, IRT_NIL), tr, 0); + } else { + J->base[0] = emitir(IRTG(IR_BUFSTR, IRT_STR), tr, hdr); + } } +static void LJ_FASTCALL recff_string_format(jit_State *J, RecordFFData *rd) +{ + recff_format(J, rd, recff_bufhdr(J), 0); +} + +/* -- Buffer library fast functions --------------------------------------- */ + +#if LJ_HASBUFFER + +static LJ_AINLINE TRef recff_sbufx_get_L(jit_State *J, TRef ud) +{ + return emitir(IRT(IR_FLOAD, IRT_PGC), ud, IRFL_SBUF_L); +} + +static LJ_AINLINE void recff_sbufx_set_L(jit_State *J, TRef ud, TRef val) +{ + TRef fref = emitir(IRT(IR_FREF, IRT_PGC), ud, IRFL_SBUF_L); + emitir(IRT(IR_FSTORE, IRT_PGC), fref, val); +} + +static LJ_AINLINE TRef recff_sbufx_get_ptr(jit_State *J, TRef ud, IRFieldID fl) +{ + return emitir(IRT(IR_FLOAD, IRT_PTR), ud, fl); +} + +static LJ_AINLINE void recff_sbufx_set_ptr(jit_State *J, TRef ud, IRFieldID fl, TRef val) +{ + TRef fref = emitir(IRT(IR_FREF, IRT_PTR), ud, fl); + emitir(IRT(IR_FSTORE, IRT_PTR), fref, val); +} + +static LJ_AINLINE TRef recff_sbufx_len(jit_State *J, TRef trr, TRef trw) +{ + TRef len = emitir(IRT(IR_SUB, IRT_INTP), trw, trr); + if (LJ_64) + len = emitir(IRTI(IR_CONV), len, (IRT_INT<<5)|IRT_INTP|IRCONV_NONE); + return len; +} + +/* Emit typecheck for string buffer. */ +static TRef recff_sbufx_check(jit_State *J, RecordFFData *rd, int arg) +{ + TRef trtype, ud = J->base[arg]; + if (!tvisbuf(&rd->argv[arg])) lj_trace_err(J, LJ_TRERR_BADTYPE); + trtype = emitir(IRT(IR_FLOAD, IRT_U8), ud, IRFL_UDATA_UDTYPE); + emitir(IRTGI(IR_EQ), trtype, lj_ir_kint(J, UDTYPE_BUFFER)); + J->needsnap = 1; + return ud; +} + +/* Emit BUFHDR for write to extended string buffer. */ +static TRef recff_sbufx_write(jit_State *J, TRef ud) +{ + TRef trbuf = emitir(IRT(IR_ADD, IRT_PGC), ud, lj_ir_kint(J, sizeof(GCudata))); + return emitir(IRT(IR_BUFHDR, IRT_PGC), trbuf, IRBUFHDR_WRITE); +} + +/* Check for integer in range for the buffer API. */ +static TRef recff_sbufx_checkint(jit_State *J, RecordFFData *rd, int arg) +{ + TRef tr = J->base[arg]; + TRef trlim = lj_ir_kint(J, LJ_MAX_BUF); + if (tref_isinteger(tr)) { + emitir(IRTGI(IR_ULE), tr, trlim); + } else if (tref_isnum(tr)) { + tr = emitir(IRTI(IR_CONV), tr, IRCONV_INT_NUM|IRCONV_ANY); + emitir(IRTGI(IR_ULE), tr, trlim); +#if LJ_HASFFI + } else if (tref_iscdata(tr)) { + tr = lj_crecord_loadiu64(J, tr, &rd->argv[arg]); + emitir(IRTG(IR_ULE, IRT_U64), tr, lj_ir_kint64(J, LJ_MAX_BUF)); + tr = emitir(IRTI(IR_CONV), tr, (IRT_INT<<5)|IRT_I64|IRCONV_NONE); +#else + UNUSED(rd); +#endif + } else { + lj_trace_err(J, LJ_TRERR_BADTYPE); + } + return tr; +} + +static void LJ_FASTCALL recff_buffer_method_reset(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + SBufExt *sbx = bufV(&rd->argv[0]); + int iscow = (int)sbufiscow(sbx); + TRef trl = recff_sbufx_get_L(J, ud); + TRef trcow = emitir(IRT(IR_BAND, IRT_IGC), trl, lj_ir_kint(J, SBUF_FLAG_COW)); + TRef zero = lj_ir_kint(J, 0); + emitir(IRTG(iscow ? IR_NE : IR_EQ, IRT_IGC), trcow, zero); + if (iscow) { + trl = emitir(IRT(IR_BXOR, IRT_IGC), trl, + LJ_GC64 ? lj_ir_kint64(J, SBUF_FLAG_COW) : + lj_ir_kint(J, SBUF_FLAG_COW)); + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_W, zero); + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_E, zero); + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_B, zero); + recff_sbufx_set_L(J, ud, trl); + emitir(IRT(IR_FSTORE, IRT_PGC), + emitir(IRT(IR_FREF, IRT_PGC), ud, IRFL_SBUF_REF), zero); + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_R, zero); + } else { + TRef trb = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_B); + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_W, trb); + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_R, trb); + } +} + +static void LJ_FASTCALL recff_buffer_method_skip(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trr = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_R); + TRef trw = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_W); + TRef len = recff_sbufx_len(J, trr, trw); + TRef trn = recff_sbufx_checkint(J, rd, 1); + len = emitir(IRTI(IR_MIN), len, trn); + trr = emitir(IRT(IR_ADD, IRT_PTR), trr, len); + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_R, trr); +} + +static void LJ_FASTCALL recff_buffer_method_set(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trbuf = recff_sbufx_write(J, ud); + TRef tr = J->base[1]; + if (tref_isstr(tr)) { + TRef trp = emitir(IRT(IR_STRREF, IRT_PGC), tr, lj_ir_kint(J, 0)); + TRef len = emitir(IRTI(IR_FLOAD), tr, IRFL_STR_LEN); + lj_ir_call(J, IRCALL_lj_bufx_set, trbuf, trp, len, tr); +#if LJ_HASFFI + } else if (tref_iscdata(tr)) { + TRef trp = lj_crecord_topcvoid(J, tr, &rd->argv[1]); + TRef len = recff_sbufx_checkint(J, rd, 2); + lj_ir_call(J, IRCALL_lj_bufx_set, trbuf, trp, len, tr); +#endif + } /* else: Interpreter will throw. */ +} + +static void LJ_FASTCALL recff_buffer_method_put(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trbuf = recff_sbufx_write(J, ud); + TRef tr; + ptrdiff_t arg; + if (!J->base[1]) return; + for (arg = 1; (tr = J->base[arg]); arg++) { + if (tref_isstr(tr)) { + trbuf = emitir(IRTG(IR_BUFPUT, IRT_PGC), trbuf, tr); + } else if (tref_isnumber(tr)) { + trbuf = emitir(IRTG(IR_BUFPUT, IRT_PGC), trbuf, + emitir(IRT(IR_TOSTR, IRT_STR), tr, + tref_isnum(tr) ? IRTOSTR_NUM : IRTOSTR_INT)); + } else if (tref_isudata(tr)) { + TRef ud2 = recff_sbufx_check(J, rd, arg); + TRef trr = recff_sbufx_get_ptr(J, ud2, IRFL_SBUF_R); + TRef trw = recff_sbufx_get_ptr(J, ud2, IRFL_SBUF_W); + TRef len = recff_sbufx_len(J, trr, trw); + emitir(IRTG(IR_NE, IRT_PGC), ud, ud2); + trbuf = lj_ir_call(J, IRCALL_lj_buf_putmem, trbuf, trr, len); + } else { + recff_nyiu(J, rd); + } + } + emitir(IRT(IR_USE, IRT_NIL), trbuf, 0); +} + +static void LJ_FASTCALL recff_buffer_method_putf(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trbuf = recff_sbufx_write(J, ud); + recff_format(J, rd, trbuf, 1); +} + +static void LJ_FASTCALL recff_buffer_method_get(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trr = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_R); + TRef trw = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_W); + TRef tr; + ptrdiff_t arg; + if (!J->base[1]) { J->base[1] = TREF_NIL; J->base[2] = 0; } + for (arg = 0; (tr = J->base[arg+1]); arg++) { + TRef len = recff_sbufx_len(J, trr, trw); + if (tref_isnil(tr)) { + J->base[arg] = emitir(IRT(IR_XSNEW, IRT_STR), trr, len); + trr = trw; + } else { + TRef trn = recff_sbufx_checkint(J, rd, arg+1); + TRef tru; + len = emitir(IRTI(IR_MIN), len, trn); + tru = emitir(IRT(IR_ADD, IRT_PTR), trr, len); + J->base[arg] = emitir(IRT(IR_XSNEW, IRT_STR), trr, len); + trr = tru; /* Doing the ADD before the SNEW generates better code. */ + } + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_R, trr); + } + rd->nres = arg; +} + +static void LJ_FASTCALL recff_buffer_method___tostring(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trr = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_R); + TRef trw = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_W); + J->base[0] = emitir(IRT(IR_XSNEW, IRT_STR), trr, recff_sbufx_len(J, trr, trw)); +} + +static void LJ_FASTCALL recff_buffer_method___len(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trr = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_R); + TRef trw = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_W); + J->base[0] = recff_sbufx_len(J, trr, trw); +} + +#if LJ_HASFFI +static void LJ_FASTCALL recff_buffer_method_putcdata(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trbuf = recff_sbufx_write(J, ud); + TRef tr = lj_crecord_topcvoid(J, J->base[1], &rd->argv[1]); + TRef len = recff_sbufx_checkint(J, rd, 2); + trbuf = lj_ir_call(J, IRCALL_lj_buf_putmem, trbuf, tr, len); + emitir(IRT(IR_USE, IRT_NIL), trbuf, 0); +} + +static void LJ_FASTCALL recff_buffer_method_reserve(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trbuf = recff_sbufx_write(J, ud); + TRef trsz = recff_sbufx_checkint(J, rd, 1); + J->base[1] = lj_ir_call(J, IRCALL_lj_bufx_more, trbuf, trsz); + J->base[0] = lj_crecord_topuint8(J, recff_sbufx_get_ptr(J, ud, IRFL_SBUF_W)); + rd->nres = 2; +} + +static void LJ_FASTCALL recff_buffer_method_commit(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef len = recff_sbufx_checkint(J, rd, 1); + TRef trw = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_W); + TRef tre = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_E); + TRef left = emitir(IRT(IR_SUB, IRT_INTP), tre, trw); + if (LJ_64) + left = emitir(IRTI(IR_CONV), left, (IRT_INT<<5)|IRT_INTP|IRCONV_NONE); + emitir(IRTGI(IR_ULE), len, left); + trw = emitir(IRT(IR_ADD, IRT_PTR), trw, len); + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_W, trw); +} + +static void LJ_FASTCALL recff_buffer_method_ref(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trr = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_R); + TRef trw = recff_sbufx_get_ptr(J, ud, IRFL_SBUF_W); + J->base[0] = lj_crecord_topuint8(J, trr); + J->base[1] = recff_sbufx_len(J, trr, trw); + rd->nres = 2; +} +#endif + +static void LJ_FASTCALL recff_buffer_method_encode(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trbuf = recff_sbufx_write(J, ud); + TRef tmp = recff_tmpref(J, J->base[1], IRTMPREF_IN1); + lj_ir_call(J, IRCALL_lj_serialize_put, trbuf, tmp); + /* No IR_USE needed, since the call is a store. */ +} + +static void LJ_FASTCALL recff_buffer_method_decode(jit_State *J, RecordFFData *rd) +{ + TRef ud = recff_sbufx_check(J, rd, 0); + TRef trbuf = recff_sbufx_write(J, ud); + TRef tmp = recff_tmpref(J, TREF_NIL, IRTMPREF_OUT1); + TRef trr = lj_ir_call(J, IRCALL_lj_serialize_get, trbuf, tmp); + IRType t = (IRType)lj_serialize_peektype(bufV(&rd->argv[0])); + /* No IR_USE needed, since the call is a store. */ + J->base[0] = lj_record_vload(J, tmp, 0, t); + /* The sbx->r store must be after the VLOAD type check, in case it fails. */ + recff_sbufx_set_ptr(J, ud, IRFL_SBUF_R, trr); +} + +static void LJ_FASTCALL recff_buffer_encode(jit_State *J, RecordFFData *rd) +{ + TRef tmp = recff_tmpref(J, J->base[0], IRTMPREF_IN1); + J->base[0] = lj_ir_call(J, IRCALL_lj_serialize_encode, tmp); + /* IR_USE needed for IR_CALLA, because the encoder may throw non-OOM. */ + emitir(IRT(IR_USE, IRT_NIL), J->base[0], 0); + UNUSED(rd); +} + +static void LJ_FASTCALL recff_buffer_decode(jit_State *J, RecordFFData *rd) +{ + if (tvisstr(&rd->argv[0])) { + GCstr *str = strV(&rd->argv[0]); + SBufExt sbx; + IRType t; + TRef tmp = recff_tmpref(J, TREF_NIL, IRTMPREF_OUT1); + TRef tr = lj_ir_call(J, IRCALL_lj_serialize_decode, tmp, J->base[0]); + /* IR_USE needed for IR_CALLA, because the decoder may throw non-OOM. + ** That's why IRCALL_lj_serialize_decode needs a fake INT result. + */ + emitir(IRT(IR_USE, IRT_NIL), tr, 0); + memset(&sbx, 0, sizeof(SBufExt)); + lj_bufx_set_cow(J->L, &sbx, strdata(str), str->len); + t = (IRType)lj_serialize_peektype(&sbx); + J->base[0] = lj_record_vload(J, tmp, 0, t); + } /* else: Interpreter will throw. */ +} + +#endif + /* -- Table library fast functions ---------------------------------------- */ static void LJ_FASTCALL recff_table_insert(jit_State *J, RecordFFData *rd) @@ -1054,7 +1428,7 @@ static void LJ_FASTCALL recff_table_concat(jit_State *J, RecordFFData *rd) TRef hdr = recff_bufhdr(J); TRef tr = lj_ir_call(J, IRCALL_lj_buf_puttab, hdr, tab, sep, tri, tre); emitir(IRTG(IR_NE, IRT_PTR), tr, lj_ir_kptr(J, NULL)); - J->base[0] = emitir(IRT(IR_BUFSTR, IRT_STR), tr, hdr); + J->base[0] = emitir(IRTG(IR_BUFSTR, IRT_STR), tr, hdr); } /* else: Interpreter will throw. */ UNUSED(rd); } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_frame.h b/love/src/jni/LuaJIT-2.1/src/lj_frame.h index f67a2332..b6146454 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_frame.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_frame.h @@ -192,12 +192,12 @@ enum { LJ_CONT_TAILCALL, LJ_CONT_FFI_CALLBACK }; /* Special continuations. */ #endif #define CFRAME_SHIFT_MULTRES 3 #elif LJ_TARGET_ARM64 -#define CFRAME_OFS_ERRF 196 -#define CFRAME_OFS_NRES 200 -#define CFRAME_OFS_PREV 160 -#define CFRAME_OFS_L 176 -#define CFRAME_OFS_PC 168 -#define CFRAME_OFS_MULTRES 192 +#define CFRAME_OFS_ERRF 36 +#define CFRAME_OFS_NRES 40 +#define CFRAME_OFS_PREV 0 +#define CFRAME_OFS_L 16 +#define CFRAME_OFS_PC 8 +#define CFRAME_OFS_MULTRES 32 #define CFRAME_SIZE 208 #define CFRAME_SHIFT_MULTRES 3 #elif LJ_TARGET_PPC diff --git a/love/src/jni/LuaJIT-2.1/src/lj_gc.c b/love/src/jni/LuaJIT-2.1/src/lj_gc.c index cfbce037..5a238542 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_gc.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_gc.c @@ -65,6 +65,15 @@ static void gc_mark(global_State *g, GCobj *o) gray2black(o); /* Userdata are never gray. */ if (mt) gc_markobj(g, mt); gc_markobj(g, tabref(gco2ud(o)->env)); + if (LJ_HASBUFFER && gco2ud(o)->udtype == UDTYPE_BUFFER) { + SBufExt *sbx = (SBufExt *)uddata(gco2ud(o)); + if (sbufiscow(sbx) && gcref(sbx->cowref)) + gc_markobj(g, gcref(sbx->cowref)); + if (gcref(sbx->dict_str)) + gc_markobj(g, gcref(sbx->dict_str)); + if (gcref(sbx->dict_mt)) + gc_markobj(g, gcref(sbx->dict_mt)); + } } else if (LJ_UNLIKELY(gct == ~LJ_TUPVAL)) { GCupval *uv = gco2uv(o); gc_marktv(g, uvval(uv)); diff --git a/love/src/jni/LuaJIT-2.1/src/lj_ir.c b/love/src/jni/LuaJIT-2.1/src/lj_ir.c index 4be11a61..71bf8855 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_ir.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_ir.c @@ -30,6 +30,7 @@ #endif #include "lj_vm.h" #include "lj_strscan.h" +#include "lj_serialize.h" #include "lj_strfmt.h" #include "lj_prng.h" @@ -147,7 +148,7 @@ TRef lj_ir_call(jit_State *J, IRCallID id, ...) } /* Load field of type t from GG_State + offset. Must be 32 bit aligned. */ -LJ_FUNC TRef lj_ir_ggfload(jit_State *J, IRType t, uintptr_t ofs) +TRef lj_ir_ggfload(jit_State *J, IRType t, uintptr_t ofs) { lj_assertJ((ofs & 3) == 0, "unaligned GG_State field offset"); ofs >>= 2; diff --git a/love/src/jni/LuaJIT-2.1/src/lj_ir.h b/love/src/jni/LuaJIT-2.1/src/lj_ir.h index 1aedfd0e..2b127f6c 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_ir.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_ir.h @@ -95,6 +95,7 @@ _(UREFO, LW, ref, lit) \ _(UREFC, LW, ref, lit) \ _(FREF, R , ref, lit) \ + _(TMPREF, S , ref, lit) \ _(STRREF, N , ref, ref) \ _(LREF, L , ___, ___) \ \ @@ -105,7 +106,7 @@ _(FLOAD, L , ref, lit) \ _(XLOAD, L , ref, lit) \ _(SLOAD, L , lit, lit) \ - _(VLOAD, L , ref, ___) \ + _(VLOAD, L , ref, lit) \ _(ALEN, L , ref, ref) \ \ _(ASTORE, S , ref, ref) \ @@ -124,8 +125,8 @@ \ /* Buffer operations. */ \ _(BUFHDR, L , ref, lit) \ - _(BUFPUT, L , ref, ref) \ - _(BUFSTR, A , ref, ref) \ + _(BUFPUT, LW, ref, ref) \ + _(BUFSTR, AW, ref, ref) \ \ /* Barriers. */ \ _(TBAR, S , ref, ___) \ @@ -139,9 +140,9 @@ _(STRTO, N , ref, ___) \ \ /* Calls. */ \ - _(CALLN, N , ref, lit) \ - _(CALLA, A , ref, lit) \ - _(CALLL, L , ref, lit) \ + _(CALLN, NW, ref, lit) \ + _(CALLA, AW, ref, lit) \ + _(CALLL, LW, ref, lit) \ _(CALLS, S , ref, lit) \ _(CALLXS, S , ref, ref) \ _(CARG, N , ref, ref) \ @@ -204,9 +205,15 @@ IRFPMDEF(FPMENUM) _(UDATA_META, offsetof(GCudata, metatable)) \ _(UDATA_UDTYPE, offsetof(GCudata, udtype)) \ _(UDATA_FILE, sizeof(GCudata)) \ + _(SBUF_W, sizeof(GCudata) + offsetof(SBufExt, w)) \ + _(SBUF_E, sizeof(GCudata) + offsetof(SBufExt, e)) \ + _(SBUF_B, sizeof(GCudata) + offsetof(SBufExt, b)) \ + _(SBUF_L, sizeof(GCudata) + offsetof(SBufExt, L)) \ + _(SBUF_REF, sizeof(GCudata) + offsetof(SBufExt, cowref)) \ + _(SBUF_R, sizeof(GCudata) + offsetof(SBufExt, r)) \ _(CDATA_CTYPEID, offsetof(GCcdata, ctypeid)) \ _(CDATA_PTR, sizeof(GCcdata)) \ - _(CDATA_INT, sizeof(GCcdata)) \ + _(CDATA_INT, sizeof(GCcdata)) \ _(CDATA_INT64, sizeof(GCcdata)) \ _(CDATA_INT64_4, sizeof(GCcdata) + 4) @@ -217,6 +224,11 @@ IRFLDEF(FLENUM) IRFL__MAX } IRFieldID; +/* TMPREF mode bits, stored in op2. */ +#define IRTMPREF_IN1 0x01 /* First input value. */ +#define IRTMPREF_OUT1 0x02 /* First output value. */ +#define IRTMPREF_OUT2 0x04 /* Second output value. */ + /* SLOAD mode bits, stored in op2. */ #define IRSLOAD_PARENT 0x01 /* Coalesce with parent trace. */ #define IRSLOAD_FRAME 0x02 /* Load 32 bits of ftsz. */ @@ -224,15 +236,17 @@ IRFLDEF(FLENUM) #define IRSLOAD_CONVERT 0x08 /* Number to integer conversion. */ #define IRSLOAD_READONLY 0x10 /* Read-only, omit slot store. */ #define IRSLOAD_INHERIT 0x20 /* Inherited by exits/side traces. */ +#define IRSLOAD_KEYINDEX 0x40 /* Table traversal key index. */ -/* XLOAD mode, stored in op2. */ -#define IRXLOAD_READONLY 1 /* Load from read-only data. */ -#define IRXLOAD_VOLATILE 2 /* Load from volatile data. */ -#define IRXLOAD_UNALIGNED 4 /* Unaligned load. */ +/* XLOAD mode bits, stored in op2. */ +#define IRXLOAD_READONLY 0x01 /* Load from read-only data. */ +#define IRXLOAD_VOLATILE 0x02 /* Load from volatile data. */ +#define IRXLOAD_UNALIGNED 0x04 /* Unaligned load. */ /* BUFHDR mode, stored in op2. */ #define IRBUFHDR_RESET 0 /* Reset buffer. */ #define IRBUFHDR_APPEND 1 /* Append to buffer. */ +#define IRBUFHDR_WRITE 2 /* Write to string buffer. */ /* CONV mode, stored in op2. */ #define IRCONV_SRCMASK 0x001f /* Source IRType. */ @@ -249,6 +263,7 @@ IRFLDEF(FLENUM) #define IRCONV_ANY (1<sb), sbuflen(&ls->sb)-1, tv, + fmt = lj_strscan_scan((const uint8_t *)ls->sb.b, sbuflen(&ls->sb)-1, tv, (LJ_DUALNUM ? STRSCAN_OPT_TOINT : STRSCAN_OPT_TONUM) | (LJ_HASFFI ? (STRSCAN_OPT_LL|STRSCAN_OPT_IMAG) : 0)); if (LJ_DUALNUM && fmt == STRSCAN_INT) { @@ -118,11 +118,7 @@ static void lex_number(LexState *ls, TValue *tv) GCcdata *cd; lj_assertLS(fmt == STRSCAN_I64 || fmt == STRSCAN_U64 || fmt == STRSCAN_IMAG, "unexpected number format %d", fmt); - if (!ctype_ctsG(G(L))) { - ptrdiff_t oldtop = savestack(L, L->top); - luaopen_ffi(L); /* Load FFI library on-demand. */ - L->top = restorestack(L, oldtop); - } + ctype_loadffi(L); if (fmt == STRSCAN_IMAG) { cd = lj_cdata_new_(L, CTID_COMPLEX_DOUBLE, 2*sizeof(double)); ((double *)cdataptr(cd))[0] = 0; @@ -180,7 +176,7 @@ static void lex_longstring(LexState *ls, TValue *tv, int sep) } } endloop: if (tv) { - GCstr *str = lj_parse_keepstr(ls, sbufB(&ls->sb) + (2 + (MSize)sep), + GCstr *str = lj_parse_keepstr(ls, ls->sb.b + (2 + (MSize)sep), sbuflen(&ls->sb) - 2*(2 + (MSize)sep)); setstrV(ls->L, tv, str); } @@ -286,7 +282,7 @@ static void lex_string(LexState *ls, TValue *tv) } lex_savenext(ls); /* Skip trailing delimiter. */ setstrV(ls->L, tv, - lj_parse_keepstr(ls, sbufB(&ls->sb)+1, sbuflen(&ls->sb)-2)); + lj_parse_keepstr(ls, ls->sb.b+1, sbuflen(&ls->sb)-2)); } /* -- Main lexical scanner ------------------------------------------------ */ @@ -306,7 +302,7 @@ static LexToken lex_scan(LexState *ls, TValue *tv) do { lex_savenext(ls); } while (lj_char_isident(ls->c)); - s = lj_parse_keepstr(ls, sbufB(&ls->sb), sbuflen(&ls->sb)); + s = lj_parse_keepstr(ls, ls->sb.b, sbuflen(&ls->sb)); setstrV(ls->L, tv, s); if (s->reserved > 0) /* Reserved word? */ return TK_OFS + s->reserved; @@ -496,7 +492,7 @@ void lj_lex_error(LexState *ls, LexToken tok, ErrMsg em, ...) tokstr = NULL; } else if (tok == TK_name || tok == TK_string || tok == TK_number) { lex_save(ls, '\0'); - tokstr = sbufB(&ls->sb); + tokstr = ls->sb.b; } else { tokstr = lj_lex_token2str(ls, tok); } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_lib.c b/love/src/jni/LuaJIT-2.1/src/lj_lib.c index a962ddc1..438056d8 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_lib.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_lib.c @@ -16,6 +16,9 @@ #include "lj_func.h" #include "lj_bc.h" #include "lj_dispatch.h" +#if LJ_HASFFI +#include "lj_ctype.h" +#endif #include "lj_vm.h" #include "lj_strscan.h" #include "lj_strfmt.h" @@ -301,3 +304,56 @@ int lj_lib_checkopt(lua_State *L, int narg, int def, const char *lst) return def; } +/* -- Strict type checks -------------------------------------------------- */ + +/* The following type checks do not coerce between strings and numbers. +** And they handle plain int64_t/uint64_t FFI numbers, too. +*/ + +#if LJ_HASBUFFER +GCstr *lj_lib_checkstrx(lua_State *L, int narg) +{ + TValue *o = L->base + narg-1; + if (!(o < L->top && tvisstr(o))) lj_err_argt(L, narg, LUA_TSTRING); + return strV(o); +} + +int32_t lj_lib_checkintrange(lua_State *L, int narg, int32_t a, int32_t b) +{ + TValue *o = L->base + narg-1; + lj_assertL(b >= 0, "expected range must be non-negative"); + if (o < L->top) { + if (LJ_LIKELY(tvisint(o))) { + int32_t i = intV(o); + if (i >= a && i <= b) return i; + } else if (LJ_LIKELY(tvisnum(o))) { + /* For performance reasons, this doesn't check for integerness or + ** integer overflow. Overflow detection still works, since all FPUs + ** return either MININT or MAXINT, which is then out of range. + */ + int32_t i = (int32_t)numV(o); + if (i >= a && i <= b) return i; +#if LJ_HASFFI + } else if (tviscdata(o)) { + GCcdata *cd = cdataV(o); + if (cd->ctypeid == CTID_INT64) { + int64_t i = *(int64_t *)cdataptr(cd); + if (i >= (int64_t)a && i <= (int64_t)b) return (int32_t)i; + } else if (cd->ctypeid == CTID_UINT64) { + uint64_t i = *(uint64_t *)cdataptr(cd); + if ((a < 0 || i >= (uint64_t)a) && i <= (uint64_t)b) return (int32_t)i; + } else { + goto badtype; + } +#endif + } else { + goto badtype; + } + lj_err_arg(L, narg, LJ_ERR_NUMRNG); + } +badtype: + lj_err_argt(L, narg, LUA_TNUMBER); + return 0; /* unreachable */ +} +#endif + diff --git a/love/src/jni/LuaJIT-2.1/src/lj_lib.h b/love/src/jni/LuaJIT-2.1/src/lj_lib.h index 718d8eb4..f59e9ea2 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_lib.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_lib.h @@ -46,6 +46,12 @@ LJ_FUNC GCtab *lj_lib_checktab(lua_State *L, int narg); LJ_FUNC GCtab *lj_lib_checktabornil(lua_State *L, int narg); LJ_FUNC int lj_lib_checkopt(lua_State *L, int narg, int def, const char *lst); +#if LJ_HASBUFFER +LJ_FUNC GCstr *lj_lib_checkstrx(lua_State *L, int narg); +LJ_FUNC int32_t lj_lib_checkintrange(lua_State *L, int narg, + int32_t a, int32_t b); +#endif + /* Avoid including lj_frame.h. */ #if LJ_GC64 #define lj_lib_upvalue(L, n) \ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_mcode.c b/love/src/jni/LuaJIT-2.1/src/lj_mcode.c index 9ce1deec..b3efbc55 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_mcode.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_mcode.c @@ -97,10 +97,15 @@ static int mcode_setprot(void *p, size_t sz, DWORD prot) #define MCPROT_RW (PROT_READ|PROT_WRITE) #define MCPROT_RX (PROT_READ|PROT_EXEC) #define MCPROT_RWX (PROT_READ|PROT_WRITE|PROT_EXEC) +#ifdef PROT_MPROTECT +#define MCPROT_CREATE (PROT_MPROTECT(MCPROT_RWX)) +#else +#define MCPROT_CREATE 0 +#endif static void *mcode_alloc_at(jit_State *J, uintptr_t hint, size_t sz, int prot) { - void *p = mmap((void *)hint, sz, prot, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0); + void *p = mmap((void *)hint, sz, prot|MCPROT_CREATE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0); if (p == MAP_FAILED) { if (!hint) lj_trace_err(J, LJ_TRERR_MCODEAL); p = NULL; @@ -238,7 +243,7 @@ static void *mcode_alloc(jit_State *J, size_t sz) /* All memory addresses are reachable by relative jumps. */ static void *mcode_alloc(jit_State *J, size_t sz) { -#if defined(__OpenBSD__) || LJ_TARGET_UWP +#if defined(__OpenBSD__) || defined(__NetBSD__) || LJ_TARGET_UWP /* Allow better executable memory allocation for OpenBSD W^X mode. */ void *p = mcode_alloc_at(J, 0, sz, MCPROT_RUN); if (p && mcode_setprot(p, sz, MCPROT_GEN)) { @@ -269,6 +274,7 @@ static void mcode_allocarea(jit_State *J) ((MCLink *)J->mcarea)->next = oldarea; ((MCLink *)J->mcarea)->size = sz; J->szallmcarea += sz; + J->mcbot = (MCode *)lj_err_register_mcode(J->mcarea, sz, (uint8_t *)J->mcbot); } /* Free all MCode areas. */ @@ -279,7 +285,9 @@ void lj_mcode_free(jit_State *J) J->szallmcarea = 0; while (mc) { MCode *next = ((MCLink *)mc)->next; - mcode_free(J, mc, ((MCLink *)mc)->size); + size_t sz = ((MCLink *)mc)->size; + lj_err_deregister_mcode(mc, sz, (uint8_t *)mc + sizeof(MCLink)); + mcode_free(J, mc, sz); mc = next; } } @@ -314,21 +322,21 @@ void lj_mcode_abort(jit_State *J) /* Set/reset protection to allow patching of MCode areas. */ MCode *lj_mcode_patch(jit_State *J, MCode *ptr, int finish) { -#if LUAJIT_SECURITY_MCODE == 0 - UNUSED(J); UNUSED(ptr); UNUSED(finish); - return NULL; -#else if (finish) { +#if LUAJIT_SECURITY_MCODE if (J->mcarea == ptr) mcode_protect(J, MCPROT_RUN); else if (LJ_UNLIKELY(mcode_setprot(ptr, ((MCLink *)ptr)->size, MCPROT_RUN))) mcode_protfail(J); +#endif return NULL; } else { MCode *mc = J->mcarea; /* Try current area first to use the protection cache. */ if (ptr >= mc && ptr < (MCode *)((char *)mc + J->szmcarea)) { +#if LUAJIT_SECURITY_MCODE mcode_protect(J, MCPROT_GEN); +#endif return mc; } /* Otherwise search through the list of MCode areas. */ @@ -336,13 +344,14 @@ MCode *lj_mcode_patch(jit_State *J, MCode *ptr, int finish) mc = ((MCLink *)mc)->next; lj_assertJ(mc != NULL, "broken MCode area chain"); if (ptr >= mc && ptr < (MCode *)((char *)mc + ((MCLink *)mc)->size)) { +#if LUAJIT_SECURITY_MCODE if (LJ_UNLIKELY(mcode_setprot(mc, ((MCLink *)mc)->size, MCPROT_GEN))) mcode_protfail(J); +#endif return mc; } } } -#endif } /* Limit of MCode reservation reached. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_meta.c b/love/src/jni/LuaJIT-2.1/src/lj_meta.c index 07defa55..660dfec0 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_meta.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_meta.c @@ -240,8 +240,8 @@ TValue *lj_meta_cat(lua_State *L, TValue *top, int left) int fromc = 0; if (left < 0) { left = -left; fromc = 1; } do { - if (!(tvisstr(top) || tvisnumber(top)) || - !(tvisstr(top-1) || tvisnumber(top-1))) { + if (!(tvisstr(top) || tvisnumber(top) || tvisbuf(top)) || + !(tvisstr(top-1) || tvisnumber(top-1) || tvisbuf(top-1))) { cTValue *mo = lj_meta_lookup(L, top-1, MM_concat); if (tvisnil(mo)) { mo = lj_meta_lookup(L, top, MM_concat); @@ -277,10 +277,12 @@ TValue *lj_meta_cat(lua_State *L, TValue *top, int left) ** next step: [...][CAT stack ............] */ TValue *e, *o = top; - uint64_t tlen = tvisstr(o) ? strV(o)->len : STRFMT_MAXBUF_NUM; + uint64_t tlen = tvisstr(o) ? strV(o)->len : + tvisbuf(o) ? sbufxlen(bufV(o)) : STRFMT_MAXBUF_NUM; SBuf *sb; do { - o--; tlen += tvisstr(o) ? strV(o)->len : STRFMT_MAXBUF_NUM; + o--; tlen += tvisstr(o) ? strV(o)->len : + tvisbuf(o) ? sbufxlen(bufV(o)) : STRFMT_MAXBUF_NUM; } while (--left > 0 && (tvisstr(o-1) || tvisnumber(o-1))); if (tlen >= LJ_MAX_STR) lj_err_msg(L, LJ_ERR_STROV); sb = lj_buf_tmp_(L); @@ -290,6 +292,9 @@ TValue *lj_meta_cat(lua_State *L, TValue *top, int left) GCstr *s = strV(o); MSize len = s->len; lj_buf_putmem(sb, strdata(s), len); + } else if (tvisbuf(o)) { + SBufExt *sbx = bufV(o); + lj_buf_putmem(sb, sbx->r, sbufxlen(sbx)); } else if (tvisint(o)) { lj_strfmt_putint(sb, intV(o)); } else { diff --git a/love/src/jni/LuaJIT-2.1/src/lj_obj.h b/love/src/jni/LuaJIT-2.1/src/lj_obj.h index 11434ce2..1a6445fc 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_obj.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_obj.h @@ -34,13 +34,17 @@ typedef struct MRef { #if LJ_GC64 #define mref(r, t) ((t *)(void *)(r).ptr64) +#define mrefu(r) ((r).ptr64) #define setmref(r, p) ((r).ptr64 = (uint64_t)(void *)(p)) +#define setmrefu(r, u) ((r).ptr64 = (uint64_t)(u)) #define setmrefr(r, v) ((r).ptr64 = (v).ptr64) #else #define mref(r, t) ((t *)(void *)(uintptr_t)(r).ptr32) +#define mrefu(r) ((r).ptr32) #define setmref(r, p) ((r).ptr32 = (uint32_t)(uintptr_t)(void *)(p)) +#define setmrefu(r, u) ((r).ptr32 = (uint32_t)(u)) #define setmrefr(r, v) ((r).ptr32 = (v).ptr32) #endif @@ -153,11 +157,9 @@ typedef int32_t BCLine; /* Bytecode line number. */ typedef void (*ASMFunction)(void); /* Resizable string buffer. Need this here, details in lj_buf.h. */ +#define SBufHeader char *w, *e, *b; MRef L typedef struct SBuf { - MRef p; /* String buffer pointer. */ - MRef e; /* String buffer end pointer. */ - MRef b; /* String buffer base. */ - MRef L; /* lua_State, used for buffer resizing. */ + SBufHeader; } SBuf; /* -- Tags and values ----------------------------------------------------- */ @@ -282,6 +284,9 @@ typedef const TValue cTValue; #define LJ_TISGCV (LJ_TSTR+1) #define LJ_TISTABUD LJ_TTAB +/* Type marker for slot holding a traversal index. Must be lightuserdata. */ +#define LJ_KEYINDEX 0xfffe7fffu + #if LJ_GC64 #define LJ_GCVMASK (((uint64_t)1 << 47) - 1) #endif @@ -330,6 +335,7 @@ enum { UDTYPE_USERDATA, /* Regular userdata. */ UDTYPE_IO_FILE, /* I/O library FILE. */ UDTYPE_FFI_CLIB, /* FFI C library namespace. */ + UDTYPE_BUFFER, /* String buffer. */ UDTYPE__MAX }; @@ -920,7 +926,7 @@ static LJ_AINLINE void setgcV(lua_State *L, TValue *o, GCobj *v, uint32_t it) } #define define_setV(name, type, tag) \ -static LJ_AINLINE void name(lua_State *L, TValue *o, type *v) \ +static LJ_AINLINE void name(lua_State *L, TValue *o, const type *v) \ { \ setgcV(L, o, obj2gco(v), tag); \ } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_opt_fold.c b/love/src/jni/LuaJIT-2.1/src/lj_opt_fold.c index 3cf448d6..2f903e27 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_opt_fold.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_opt_fold.c @@ -514,6 +514,7 @@ LJFOLDF(kfold_snew_kptr) } LJFOLD(SNEW any KINT) +LJFOLD(XSNEW any KINT) LJFOLDF(kfold_snew_empty) { if (fright->i == 0) @@ -577,22 +578,49 @@ LJFOLDF(kfold_strcmp) ** The compromise is to declare them as loads, emit them like stores and ** CSE whole chains manually when the BUFSTR is to be emitted. Any chain ** fragments left over from CSE are eliminated by DCE. +** +** The string buffer methods emit a USE instead of a BUFSTR to keep the +** chain alive. */ -/* BUFHDR is emitted like a store, see below. */ - -LJFOLD(BUFPUT BUFHDR BUFSTR) -LJFOLDF(bufput_append) +LJFOLD(BUFHDR any any) +LJFOLDF(bufhdr_merge) { - /* New buffer, no other buffer op inbetween and same buffer? */ - if ((J->flags & JIT_F_OPT_FWD) && - !(fleft->op2 & IRBUFHDR_APPEND) && - fleft->prev == fright->op2 && - fleft->op1 == IR(fright->op2)->op1) { - IRRef ref = fins->op1; - IR(ref)->op2 = (fleft->op2 | IRBUFHDR_APPEND); /* Modify BUFHDR. */ - IR(ref)->op1 = fright->op1; - return ref; + return fins->op2 == IRBUFHDR_WRITE ? CSEFOLD : EMITFOLD; +} + +LJFOLD(BUFPUT any BUFSTR) +LJFOLDF(bufput_bufstr) +{ + if ((J->flags & JIT_F_OPT_FWD)) { + IRRef hdr = fright->op2; + /* New buffer, no other buffer op inbetween and same buffer? */ + if (fleft->o == IR_BUFHDR && fleft->op2 == IRBUFHDR_RESET && + fleft->prev == hdr && + fleft->op1 == IR(hdr)->op1) { + IRRef ref = fins->op1; + IR(ref)->op2 = IRBUFHDR_APPEND; /* Modify BUFHDR. */ + IR(ref)->op1 = fright->op1; + return ref; + } + /* Replay puts to global temporary buffer. */ + if (IR(hdr)->op2 == IRBUFHDR_RESET) { + IRIns *ir = IR(fright->op1); + /* For now only handle single string.reverse .lower .upper .rep. */ + if (ir->o == IR_CALLL && + ir->op2 >= IRCALL_lj_buf_putstr_reverse && + ir->op2 <= IRCALL_lj_buf_putstr_rep) { + IRIns *carg1 = IR(ir->op1); + if (ir->op2 == IRCALL_lj_buf_putstr_rep) { + IRIns *carg2 = IR(carg1->op1); + if (carg2->op1 == hdr) { + return lj_ir_call(J, ir->op2, fins->op1, carg2->op2, carg1->op2); + } + } else if (carg1->op1 == hdr) { + return lj_ir_call(J, ir->op2, fins->op1, carg1->op2); + } + } + } } return EMITFOLD; /* Always emit, CSE later. */ } @@ -626,14 +654,14 @@ LJFOLDF(bufstr_kfold_cse) "bad buffer constructor IR op %d", fleft->o); if (LJ_LIKELY(J->flags & JIT_F_OPT_FOLD)) { if (fleft->o == IR_BUFHDR) { /* No put operations? */ - if (!(fleft->op2 & IRBUFHDR_APPEND)) /* Empty buffer? */ + if (fleft->op2 == IRBUFHDR_RESET) /* Empty buffer? */ return lj_ir_kstr(J, &J2G(J)->strempty); fins->op1 = fleft->op1; fins->op2 = fleft->prev; /* Relies on checks in bufput_append. */ return CSEFOLD; } else if (fleft->o == IR_BUFPUT) { IRIns *irb = IR(fleft->op1); - if (irb->o == IR_BUFHDR && !(irb->op2 & IRBUFHDR_APPEND)) + if (irb->o == IR_BUFHDR && irb->op2 == IRBUFHDR_RESET) return fleft->op2; /* Shortcut for a single put operation. */ } } @@ -646,7 +674,7 @@ LJFOLDF(bufstr_kfold_cse) lj_assertJ(ira->o == IR_BUFHDR || ira->o == IR_BUFPUT || ira->o == IR_CALLL || ira->o == IR_CARG, "bad buffer constructor IR op %d", ira->o); - if (ira->o == IR_BUFHDR && !(ira->op2 & IRBUFHDR_APPEND)) + if (ira->o == IR_BUFHDR && ira->op2 == IRBUFHDR_RESET) return ref; /* CSE succeeded. */ if (ira->o == IR_CALLL && ira->op2 == IRCALL_lj_buf_puttab) break; @@ -1297,6 +1325,10 @@ LJFOLD(CONV SUB IRCONV_U32_U64) LJFOLD(CONV MUL IRCONV_U32_U64) LJFOLDF(simplify_conv_narrow) { +#if LJ_64 + UNUSED(J); + return NEXTFOLD; +#else IROp op = (IROp)fleft->o; IRType t = irt_type(fins->t); IRRef op1 = fleft->op1, op2 = fleft->op2, mode = fins->op2; @@ -1307,6 +1339,7 @@ LJFOLDF(simplify_conv_narrow) fins->op1 = op1; fins->op2 = op2; return RETRYFOLD; +#endif } /* Special CSE rule for CONV. */ @@ -2275,6 +2308,27 @@ LJFOLDF(fload_str_len_tostr) return NEXTFOLD; } +LJFOLD(FLOAD any IRFL_SBUF_W) +LJFOLD(FLOAD any IRFL_SBUF_E) +LJFOLD(FLOAD any IRFL_SBUF_B) +LJFOLD(FLOAD any IRFL_SBUF_L) +LJFOLD(FLOAD any IRFL_SBUF_REF) +LJFOLD(FLOAD any IRFL_SBUF_R) +LJFOLDF(fload_sbuf) +{ + TRef tr = lj_opt_fwd_fload(J); + return lj_opt_fwd_sbuf(J, tref_ref(tr)) ? tr : EMITFOLD; +} + +/* The fast function ID of function objects is immutable. */ +LJFOLD(FLOAD KGC IRFL_FUNC_FFID) +LJFOLDF(fload_func_ffid_kgc) +{ + if (LJ_LIKELY(J->flags & JIT_F_OPT_FOLD)) + return INTFOLD((int32_t)ir_kfunc(fleft)->c.ffid); + return NEXTFOLD; +} + /* The C type ID of cdata objects is immutable. */ LJFOLD(FLOAD KGC IRFL_CDATA_CTYPEID) LJFOLDF(fload_cdata_typeid_kgc) @@ -2421,6 +2475,7 @@ LJFOLD(XSTORE any any) LJFOLDX(lj_opt_dse_xstore) LJFOLD(NEWREF any any) /* Treated like a store. */ +LJFOLD(TMPREF any any) LJFOLD(CALLA any any) LJFOLD(CALLL any any) /* Safeguard fallback. */ LJFOLD(CALLS any any) @@ -2431,7 +2486,6 @@ LJFOLD(TNEW any any) LJFOLD(TDUP any) LJFOLD(CNEW any any) LJFOLD(XSNEW any any) -LJFOLD(BUFHDR any any) LJFOLDX(lj_ir_emit) /* ------------------------------------------------------------------------ */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_opt_loop.c b/love/src/jni/LuaJIT-2.1/src/lj_opt_loop.c index 4c709c63..df5811a9 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_opt_loop.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_opt_loop.c @@ -225,6 +225,7 @@ static void loop_subst_snap(jit_State *J, SnapShot *osnap, /* Setup new snapshot. */ snap->mapofs = (uint32_t)nmapofs; snap->ref = (IRRef1)J->cur.nins; + snap->mcofs = 0; snap->nslots = nslots; snap->topslot = osnap->topslot; snap->count = 0; diff --git a/love/src/jni/LuaJIT-2.1/src/lj_opt_mem.c b/love/src/jni/LuaJIT-2.1/src/lj_opt_mem.c index f430fc37..d6a419e4 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_opt_mem.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_opt_mem.c @@ -364,7 +364,10 @@ TRef LJ_FASTCALL lj_opt_dse_ahstore(jit_State *J) /* Different value: try to eliminate the redundant store. */ if (ref > J->chain[IR_LOOP]) { /* Quick check to avoid crossing LOOP. */ IRIns *ir; - /* Check for any intervening guards (includes conflicting loads). */ + /* Check for any intervening guards (includes conflicting loads). + ** Note that lj_tab_keyindex and lj_vm_next don't need guards, + ** since they are followed by at least one guarded VLOAD. + */ for (ir = IR(J->cur.nins-1); ir > store; ir--) if (irt_isguard(ir->t) || ir->o == IR_ALEN) goto doemit; /* No elimination possible. */ @@ -620,8 +623,9 @@ TRef LJ_FASTCALL lj_opt_dse_fstore(jit_State *J) goto doemit; break; /* Otherwise continue searching. */ case ALIAS_MUST: - if (store->op2 == val) /* Same value: drop the new store. */ - return DROPFOLD; + if (store->op2 == val && + !(xr->op2 >= IRFL_SBUF_W && xr->op2 <= IRFL_SBUF_R)) + return DROPFOLD; /* Same value: drop the new store. */ /* Different value: try to eliminate the redundant store. */ if (ref > J->chain[IR_LOOP]) { /* Quick check to avoid crossing LOOP. */ IRIns *ir; @@ -642,6 +646,29 @@ doemit: return EMITFOLD; /* Otherwise we have a conflict or simply no match. */ } +/* Check whether there's no aliasing buffer op between IRFL_SBUF_*. */ +int LJ_FASTCALL lj_opt_fwd_sbuf(jit_State *J, IRRef lim) +{ + IRRef ref; + if (J->chain[IR_BUFPUT] > lim) + return 0; /* Conflict. */ + ref = J->chain[IR_CALLS]; + while (ref > lim) { + IRIns *ir = IR(ref); + if (ir->op2 >= IRCALL_lj_strfmt_putint && ir->op2 < IRCALL_lj_buf_tostr) + return 0; /* Conflict. */ + ref = ir->prev; + } + ref = J->chain[IR_CALLL]; + while (ref > lim) { + IRIns *ir = IR(ref); + if (ir->op2 >= IRCALL_lj_strfmt_putint && ir->op2 < IRCALL_lj_buf_tostr) + return 0; /* Conflict. */ + ref = ir->prev; + } + return 1; /* No conflict. Can safely FOLD/CSE. */ +} + /* -- XLOAD forwarding and XSTORE elimination ----------------------------- */ /* Find cdata allocation for a reference (if any). */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_opt_split.c b/love/src/jni/LuaJIT-2.1/src/lj_opt_split.c index 3a02abd4..25c1c234 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_opt_split.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_opt_split.c @@ -645,7 +645,7 @@ static void split_ir(jit_State *J) tmp = split_emit(J, IRT(IR_CARG, IRT_NIL), hisubst[op1], oir[op1].prev); #endif ir->prev = split_emit(J, IRTI(IR_CALLN), tmp, IRCALL_lj_vm_tobit); - } else if (ir->o == IR_TOSTR) { + } else if (ir->o == IR_TOSTR || ir->o == IR_TMPREF) { if (hisubst[ir->op1]) { if (irref_isk(ir->op1)) nir->op1 = ir->op1; diff --git a/love/src/jni/LuaJIT-2.1/src/lj_parse.c b/love/src/jni/LuaJIT-2.1/src/lj_parse.c index 27fcf53c..ea64677f 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_parse.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_parse.c @@ -1465,7 +1465,7 @@ static size_t fs_prep_var(LexState *ls, FuncState *fs, size_t *ofsvar) MSize len = s->len+1; char *p = lj_buf_more(&ls->sb, len); p = lj_buf_wmem(p, strdata(s), len); - setsbufP(&ls->sb, p); + ls->sb.w = p; } *ofsvar = sbuflen(&ls->sb); lastpc = 0; @@ -1486,7 +1486,7 @@ static size_t fs_prep_var(LexState *ls, FuncState *fs, size_t *ofsvar) startpc = vs->startpc; p = lj_strfmt_wuleb128(p, startpc-lastpc); p = lj_strfmt_wuleb128(p, vs->endpc-startpc); - setsbufP(&ls->sb, p); + ls->sb.w = p; lastpc = startpc; } } @@ -1499,7 +1499,7 @@ static void fs_fixup_var(LexState *ls, GCproto *pt, uint8_t *p, size_t ofsvar) { setmref(pt->uvinfo, p); setmref(pt->varinfo, (char *)p + ofsvar); - memcpy(p, sbufB(&ls->sb), sbuflen(&ls->sb)); /* Copy from temp. buffer. */ + memcpy(p, ls->sb.b, sbuflen(&ls->sb)); /* Copy from temp. buffer. */ } #else diff --git a/love/src/jni/LuaJIT-2.1/src/lj_prng.c b/love/src/jni/LuaJIT-2.1/src/lj_prng.c index f09aaffa..bb32da8b 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_prng.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_prng.c @@ -109,13 +109,19 @@ static PRGR libfunc_rgr; #include #else -#if LJ_TARGET_OSX +#if LJ_TARGET_OSX && !LJ_TARGET_IOS +/* +** In their infinite wisdom Apple decided to disallow getentropy() in the +** iOS App Store. Even though the call is common to all BSD-ish OS, it's +** recommended by Apple in their own security-related docs, and, to top +** off the foolery, /dev/urandom is handled by the same kernel code, +** yet accessing it is actually permitted (but less efficient). +*/ #include -#if __MAC_OS_X_VERSION_MIN_REQUIRED >= 101200 || \ - __IPHONE_OS_VERSION_MIN_REQUIRED >= 100000 +#if __MAC_OS_X_VERSION_MIN_REQUIRED >= 101200 #define LJ_TARGET_HAS_GETENTROPY 1 #endif -#elif LJ_TARGET_BSD || LJ_TARGET_SOLARIS || LJ_TARGET_CYGWIN +#elif (LJ_TARGET_BSD && !defined(__NetBSD__)) || LJ_TARGET_SOLARIS || LJ_TARGET_CYGWIN #define LJ_TARGET_HAS_GETENTROPY 1 #endif @@ -167,7 +173,7 @@ int LJ_FASTCALL lj_prng_seed_secure(PRNGState *rs) #elif LJ_TARGET_PS4 || LJ_TARGET_PSVITA - if (sceRandomGetRandomNumber(rs->u, sizeof(rs->u) == 0) + if (sceRandomGetRandomNumber(rs->u, sizeof(rs->u)) == 0) goto ok; #elif LJ_TARGET_UWP || LJ_TARGET_XBOXONE diff --git a/love/src/jni/LuaJIT-2.1/src/lj_profile.c b/love/src/jni/LuaJIT-2.1/src/lj_profile.c index 00fdd6da..fbcb9878 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_profile.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_profile.c @@ -346,8 +346,7 @@ LUA_API void luaJIT_profile_stop(lua_State *L) lj_trace_flushall(L); #endif lj_buf_free(g, &ps->sb); - setmref(ps->sb.b, NULL); - setmref(ps->sb.e, NULL); + ps->sb.w = ps->sb.e = NULL; ps->g = NULL; } } @@ -362,7 +361,7 @@ LUA_API const char *luaJIT_profile_dumpstack(lua_State *L, const char *fmt, lj_buf_reset(sb); lj_debug_dumpstack(L, sb, fmt, depth); *len = (size_t)sbuflen(sb); - return sbufB(sb); + return sb->b; } #endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_record.c b/love/src/jni/LuaJIT-2.1/src/lj_record.c index c1adaa74..30722814 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_record.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_record.c @@ -156,6 +156,9 @@ static void rec_check_slots(jit_State *J) lj_assertJ((J->slot[s+1+LJ_FR2] & TREF_FRAME), "cont slot %d not followed by frame", s); depth++; + } else if ((tr & TREF_KEYINDEX)) { + lj_assertJ(tref_isint(tr), "keyindex slot %d bad type %d", + s, tref_type(tr)); } else { /* Number repr. may differ, but other types must be the same. */ lj_assertJ(tvisnumber(tv) ? tref_isnumber(tr) : @@ -259,6 +262,14 @@ TRef lj_record_constify(jit_State *J, cTValue *o) return 0; /* Can't represent lightuserdata (pointless). */ } +/* Emit a VLOAD with the correct type. */ +TRef lj_record_vload(jit_State *J, TRef ref, MSize idx, IRType t) +{ + TRef tr = emitir(IRTG(IR_VLOAD, t), ref, idx); + if (irtype_ispri(t)) tr = TREF_PRI(t); /* Canonicalize primitives. */ + return tr; +} + /* -- Record loop ops ----------------------------------------------------- */ /* Loop event. */ @@ -275,9 +286,9 @@ static void canonicalize_slots(jit_State *J) if (LJ_DUALNUM) return; for (s = J->baseslot+J->maxslot-1; s >= 1; s--) { TRef tr = J->slot[s]; - if (tref_isinteger(tr)) { + if (tref_isinteger(tr) && !(tr & TREF_KEYINDEX)) { IRIns *ir = IR(tref_ref(tr)); - if (!(ir->o == IR_SLOAD && (ir->op2 & IRSLOAD_READONLY))) + if (!(ir->o == IR_SLOAD && (ir->op2 & (IRSLOAD_READONLY)))) J->slot[s] = emitir(IRTN(IR_CONV), tr, IRCONV_NUM_INT); } } @@ -598,6 +609,7 @@ static void rec_loop_interp(jit_State *J, const BCIns *pc, LoopEvent ev) { if (J->parent == 0 && J->exitno == 0) { if (pc == J->startpc && J->framedepth + J->retdepth == 0) { + if (bc_op(J->cur.startins) == BC_ITERN) return; /* See rec_itern(). */ /* Same loop? */ if (ev == LOOPEV_LEAVE) /* Must loop back to form a root trace. */ lj_trace_err(J, LJ_TRERR_LLEAVE); @@ -638,6 +650,70 @@ static void rec_loop_jit(jit_State *J, TraceNo lnk, LoopEvent ev) } /* Side trace continues across a loop that's left or not entered. */ } +/* Record ITERN. */ +static LoopEvent rec_itern(jit_State *J, BCReg ra, BCReg rb) +{ +#if LJ_BE + /* YAGNI: Disabled on big-endian due to issues with lj_vm_next, + ** IR_HIOP, RID_RETLO/RID_RETHI and ra_destpair. + */ + UNUSED(ra); UNUSED(rb); + setintV(&J->errinfo, (int32_t)BC_ITERN); + lj_trace_err_info(J, LJ_TRERR_NYIBC); +#else + RecordIndex ix; + /* Since ITERN is recorded at the start, we need our own loop detection. */ + if (J->pc == J->startpc && + (J->cur.nins > REF_FIRST+1 || + (J->cur.nins == REF_FIRST+1 && J->cur.ir[REF_FIRST].o != IR_PROF)) && + J->framedepth + J->retdepth == 0 && J->parent == 0 && J->exitno == 0) { + lj_record_stop(J, LJ_TRLINK_LOOP, J->cur.traceno); /* Looping trace. */ + return LOOPEV_ENTER; + } + J->maxslot = ra; + lj_snap_add(J); /* Required to make JLOOP the first ins in a side-trace. */ + ix.tab = getslot(J, ra-2); + ix.key = J->base[ra-1] ? J->base[ra-1] : + sloadt(J, (int32_t)(ra-1), IRT_INT, IRSLOAD_KEYINDEX); + copyTV(J->L, &ix.tabv, &J->L->base[ra-2]); + copyTV(J->L, &ix.keyv, &J->L->base[ra-1]); + ix.idxchain = (rb < 3); /* Omit value type check, if unused. */ + ix.mobj = 1; /* We need the next index, too. */ + J->maxslot = ra + lj_record_next(J, &ix); + J->needsnap = 1; + if (!tref_isnil(ix.key)) { /* Looping back? */ + J->base[ra-1] = ix.mobj | TREF_KEYINDEX; /* Control var has next index. */ + J->base[ra] = ix.key; + J->base[ra+1] = ix.val; + J->pc += bc_j(J->pc[1])+2; + return LOOPEV_ENTER; + } else { + J->maxslot = ra-3; + J->pc += 2; + return LOOPEV_LEAVE; + } +#endif +} + +/* Record ISNEXT. */ +static void rec_isnext(jit_State *J, BCReg ra) +{ + cTValue *b = &J->L->base[ra-3]; + if (tvisfunc(b) && funcV(b)->c.ffid == FF_next && + tvistab(b+1) && tvisnil(b+2)) { + /* These checks are folded away for a compiled pairs(). */ + TRef func = getslot(J, ra-3); + TRef trid = emitir(IRT(IR_FLOAD, IRT_U8), func, IRFL_FUNC_FFID); + emitir(IRTGI(IR_EQ), trid, lj_ir_kint(J, FF_next)); + (void)getslot(J, ra-2); /* Type check for table. */ + (void)getslot(J, ra-1); /* Type check for nil key. */ + J->base[ra-1] = lj_ir_kint(J, 0) | TREF_KEYINDEX; + J->maxslot = ra; + } else { /* Abort trace. Interpreter will despecialize bytecode. */ + lj_trace_err(J, LJ_TRERR_RECERR); + } +} + /* -- Record profiler hook checks ----------------------------------------- */ #if LJ_HASPROFILE @@ -708,7 +784,7 @@ static TRef rec_call_specialize(jit_State *J, GCfunc *fn, TRef tr) /* NYI: io_file_iter doesn't have an ffid, yet. */ { /* Specialize to the ffid. */ TRef trid = emitir(IRT(IR_FLOAD, IRT_U8), tr, IRFL_FUNC_FFID); - emitir(IRTG(IR_EQ, IRT_INT), trid, lj_ir_kint(J, fn->c.ffid)); + emitir(IRTGI(IR_EQ), trid, lj_ir_kint(J, fn->c.ffid)); } return tr; default: @@ -832,6 +908,7 @@ void lj_record_ret(jit_State *J, BCReg rbase, ptrdiff_t gotresults) J->base -= cbase; J->base[--rbase] = TREF_TRUE; /* Prepend true to results. */ frame = frame_prevd(frame); + J->needsnap = 1; /* Stop catching on-trace errors. */ } /* Return to lower frame via interpreter for unhandled cases. */ if (J->framedepth == 0 && J->pt && bc_isret(bc_op(*J->pc)) && @@ -918,6 +995,9 @@ void lj_record_ret(jit_State *J, BCReg rbase, ptrdiff_t gotresults) TRef tr = gotresults ? J->base[cbase+rbase] : TREF_NIL; if (bslot != J->maxslot) { /* Concatenate the remainder. */ TValue *b = J->L->base, save; /* Simulate lower frame and result. */ + /* Can't handle MM_concat + CALLT + fast func side-effects. */ + if (J->postproc != LJ_POST_NONE) + lj_trace_err(J, LJ_TRERR_NYIRETL); J->base[J->maxslot] = tr; copyTV(J->L, &save, b-(2<mt == TREF_NIL && !ix->val && + tref_isudata(ix->tab) && udataV(&ix->tabv)->udtype == UDTYPE_BUFFER && + tref_istab(ix->mobj) && tref_isstr(ix->key) && tref_isk(ix->key)) { + cTValue *val = lj_tab_getstr(tabV(&ix->mobjv), strV(&ix->keyv)); + TRef tr = lj_record_constify(J, val); + if (tr) return tr; /* Specialize to the value, i.e. a method. */ + } +#endif /* Otherwise retry lookup with metaobject. */ ix->tab = ix->mobj; copyTV(J->L, &ix->tabv, &ix->mobjv); @@ -1543,6 +1633,47 @@ TRef lj_record_idx(jit_State *J, RecordIndex *ix) } } +/* Determine result type of table traversal. */ +static IRType rec_next_types(GCtab *t, uint32_t idx) +{ + for (; idx < t->asize; idx++) { + cTValue *a = arrayslot(t, idx); + if (LJ_LIKELY(!tvisnil(a))) + return (LJ_DUALNUM ? IRT_INT : IRT_NUM) + (itype2irt(a) << 8); + } + idx -= t->asize; + for (; idx <= t->hmask; idx++) { + Node *n = &noderef(t->node)[idx]; + if (!tvisnil(&n->val)) + return itype2irt(&n->key) + (itype2irt(&n->val) << 8); + } + return IRT_NIL + (IRT_NIL << 8); +} + +/* Record a table traversal step aka next(). */ +int lj_record_next(jit_State *J, RecordIndex *ix) +{ + IRType t, tkey, tval; + TRef trvk; + t = rec_next_types(tabV(&ix->tabv), ix->keyv.u32.lo); + tkey = (t & 0xff); tval = (t >> 8); + trvk = lj_ir_call(J, IRCALL_lj_vm_next, ix->tab, ix->key); + if (ix->mobj || tkey == IRT_NIL) { + TRef idx = emitir(IRTI(IR_HIOP), trvk, trvk); + /* Always check for invalid key from next() for nil result. */ + if (!ix->mobj) emitir(IRTGI(IR_NE), idx, lj_ir_kint(J, -1)); + ix->mobj = idx; + } + ix->key = lj_record_vload(J, trvk, 1, tkey); + if (tkey == IRT_NIL || ix->idxchain) { /* Omit value type check. */ + ix->val = TREF_NIL; + return 1; + } else { /* Need value. */ + ix->val = lj_record_vload(J, trvk, 0, tval); + return 2; + } +} + static void rec_tsetm(jit_State *J, BCReg ra, BCReg rn, int32_t i) { RecordIndex ix; @@ -1826,11 +1957,7 @@ static void rec_varg(jit_State *J, BCReg dst, ptrdiff_t nresults) vbase = emitir(IRT(IR_ADD, IRT_PGC), vbase, lj_ir_kint(J, frofs-8)); for (i = 0; i < nload; i++) { IRType t = itype2irt(&J->L->base[i-1-LJ_FR2-nvararg]); - TRef aref = emitir(IRT(IR_AREF, IRT_PGC), - vbase, lj_ir_kint(J, (int32_t)i)); - TRef tr = emitir(IRTG(IR_VLOAD, t), aref, 0); - if (irtype_ispri(t)) tr = TREF_PRI(t); /* Canonicalize primitives. */ - J->base[dst+i] = tr; + J->base[dst+i] = lj_record_vload(J, vbase, i, t); } } else { emitir(IRTGI(IR_LE), fr, lj_ir_kint(J, frofs)); @@ -1877,8 +2004,7 @@ static void rec_varg(jit_State *J, BCReg dst, ptrdiff_t nresults) lj_ir_kint(J, frofs-(8<L->base[idx-2-LJ_FR2-nvararg]); aref = emitir(IRT(IR_AREF, IRT_PGC), vbase, tridx); - tr = emitir(IRTG(IR_VLOAD, t), aref, 0); - if (irtype_ispri(t)) tr = TREF_PRI(t); /* Canonicalize primitives. */ + tr = lj_record_vload(J, aref, 0, t); } J->base[dst-2-LJ_FR2] = tr; J->maxslot = dst-1-LJ_FR2; @@ -1935,9 +2061,9 @@ static TRef rec_cat(jit_State *J, BCReg baseslot, BCReg topslot) tr = hdr = emitir(IRT(IR_BUFHDR, IRT_PGC), lj_ir_kptr(J, &J2G(J)->tmpbuf), IRBUFHDR_RESET); do { - tr = emitir(IRT(IR_BUFPUT, IRT_PGC), tr, *trp++); + tr = emitir(IRTG(IR_BUFPUT, IRT_PGC), tr, *trp++); } while (trp <= top); - tr = emitir(IRT(IR_BUFSTR, IRT_STR), tr, hdr); + tr = emitir(IRTG(IR_BUFSTR, IRT_STR), tr, hdr); J->maxslot = (BCReg)(xbase - J->base); if (xbase == base) return tr; /* Return simple concatenation result. */ /* Pass partial result. */ @@ -2050,7 +2176,7 @@ void lj_record_ins(jit_State *J) /* Need snapshot before recording next bytecode (e.g. after a store). */ if (J->needsnap) { J->needsnap = 0; - lj_snap_purge(J); + if (J->pt) lj_snap_purge(J); lj_snap_add(J); J->mergesnap = 1; } @@ -2423,6 +2549,9 @@ void lj_record_ins(jit_State *J) case BC_ITERL: rec_loop_interp(J, pc, rec_iterl(J, *pc)); break; + case BC_ITERN: + rec_loop_interp(J, pc, rec_itern(J, ra, rb)); + break; case BC_LOOP: rec_loop_interp(J, pc, rec_loop(J, ra, 1)); break; @@ -2451,6 +2580,10 @@ void lj_record_ins(jit_State *J) J->maxslot = ra; /* Shrink used slots. */ break; + case BC_ISNEXT: + rec_isnext(J, ra); + break; + /* -- Function headers -------------------------------------------------- */ case BC_FUNCF: @@ -2480,8 +2613,6 @@ void lj_record_ins(jit_State *J) break; } /* fallthrough */ - case BC_ITERN: - case BC_ISNEXT: case BC_UCLO: case BC_FNEW: setintV(&J->errinfo, (int32_t)op); @@ -2533,6 +2664,13 @@ static const BCIns *rec_setup_root(jit_State *J) lj_assertJ(bc_op(pc[-1]) == BC_JMP, "ITERL does not point to JMP+1"); J->bc_min = pc; break; + case BC_ITERN: + lj_assertJ(bc_op(pc[1]) == BC_ITERL, "no ITERL after ITERN"); + J->maxslot = ra; + J->bc_extent = (MSize)(-bc_j(pc[1]))*sizeof(BCIns); + J->bc_min = pc+2 + bc_j(pc[1]); + J->state = LJ_TRACE_RECORD_1ST; /* Record the first ITERN, too. */ + break; case BC_LOOP: /* Only check BC range for real loops, but not for "repeat until true". */ pcj = pc + bc_j(ins); @@ -2629,9 +2767,14 @@ void lj_record_setup(jit_State *J) } lj_snap_replay(J, T); sidecheck: - if (traceref(J, J->cur.root)->nchild >= J->param[JIT_P_maxside] || - T->snap[J->exitno].count >= J->param[JIT_P_hotexit] + - J->param[JIT_P_tryside]) { + if ((traceref(J, J->cur.root)->nchild >= J->param[JIT_P_maxside] || + T->snap[J->exitno].count >= J->param[JIT_P_hotexit] + + J->param[JIT_P_tryside])) { + if (bc_op(*J->pc) == BC_JLOOP) { + BCIns startins = traceref(J, bc_d(*J->pc))->startins; + if (bc_op(startins) == BC_ITERN) + rec_itern(J, bc_a(startins), bc_b(startins)); + } lj_record_stop(J, LJ_TRLINK_INTERP, 0); } } else { /* Root trace. */ @@ -2640,6 +2783,7 @@ void lj_record_setup(jit_State *J) J->pc = rec_setup_root(J); /* Note: the loop instruction itself is recorded at the end and not ** at the start! So snapshot #0 needs to point to the *next* instruction. + ** The one exception is BC_ITERN, which sets LJ_TRACE_RECORD_1ST. */ lj_snap_add(J); if (bc_op(J->cur.startins) == BC_FORL) diff --git a/love/src/jni/LuaJIT-2.1/src/lj_record.h b/love/src/jni/LuaJIT-2.1/src/lj_record.h index de74277e..01cc6041 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_record.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_record.h @@ -30,6 +30,7 @@ LJ_FUNC int lj_record_objcmp(jit_State *J, TRef a, TRef b, cTValue *av, cTValue *bv); LJ_FUNC void lj_record_stop(jit_State *J, TraceLink linktype, TraceNo lnk); LJ_FUNC TRef lj_record_constify(jit_State *J, cTValue *o); +LJ_FUNC TRef lj_record_vload(jit_State *J, TRef ref, MSize idx, IRType t); LJ_FUNC void lj_record_call(jit_State *J, BCReg func, ptrdiff_t nargs); LJ_FUNC void lj_record_tailcall(jit_State *J, BCReg func, ptrdiff_t nargs); @@ -37,6 +38,7 @@ LJ_FUNC void lj_record_ret(jit_State *J, BCReg rbase, ptrdiff_t gotresults); LJ_FUNC int lj_record_mm_lookup(jit_State *J, RecordIndex *ix, MMS mm); LJ_FUNC TRef lj_record_idx(jit_State *J, RecordIndex *ix); +LJ_FUNC int lj_record_next(jit_State *J, RecordIndex *ix); LJ_FUNC void lj_record_ins(jit_State *J); LJ_FUNC void lj_record_setup(jit_State *J); diff --git a/love/src/jni/LuaJIT-2.1/src/lj_serialize.c b/love/src/jni/LuaJIT-2.1/src/lj_serialize.c new file mode 100644 index 00000000..d6551b11 --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/src/lj_serialize.c @@ -0,0 +1,538 @@ +/* +** Object de/serialization. +** Copyright (C) 2005-2021 Mike Pall. See Copyright Notice in luajit.h +*/ + +#define lj_serialize_c +#define LUA_CORE + +#include "lj_obj.h" + +#if LJ_HASBUFFER +#include "lj_err.h" +#include "lj_buf.h" +#include "lj_str.h" +#include "lj_tab.h" +#include "lj_udata.h" +#if LJ_HASFFI +#include "lj_ctype.h" +#include "lj_cdata.h" +#endif +#if LJ_HASJIT +#include "lj_ir.h" +#endif +#include "lj_serialize.h" + +/* Tags for internal serialization format. */ +enum { + SER_TAG_NIL, /* 0x00 */ + SER_TAG_FALSE, + SER_TAG_TRUE, + SER_TAG_NULL, + SER_TAG_LIGHTUD32, + SER_TAG_LIGHTUD64, + SER_TAG_INT, + SER_TAG_NUM, + SER_TAG_TAB, /* 0x08 */ + SER_TAG_DICT_MT = SER_TAG_TAB+6, + SER_TAG_DICT_STR, + SER_TAG_INT64, /* 0x10 */ + SER_TAG_UINT64, + SER_TAG_COMPLEX, + SER_TAG_0x13, + SER_TAG_0x14, + SER_TAG_0x15, + SER_TAG_0x16, + SER_TAG_0x17, + SER_TAG_0x18, /* 0x18 */ + SER_TAG_0x19, + SER_TAG_0x1a, + SER_TAG_0x1b, + SER_TAG_0x1c, + SER_TAG_0x1d, + SER_TAG_0x1e, + SER_TAG_0x1f, + SER_TAG_STR, /* 0x20 + str->len */ +}; +LJ_STATIC_ASSERT((SER_TAG_TAB & 7) == 0); + +/* -- Helper functions ---------------------------------------------------- */ + +static LJ_AINLINE char *serialize_more(char *w, SBufExt *sbx, MSize sz) +{ + if (LJ_UNLIKELY(sz > (MSize)(sbx->e - w))) { + sbx->w = w; + w = lj_buf_more2((SBuf *)sbx, sz); + } + return w; +} + +/* Write U124 to buffer. */ +static LJ_NOINLINE char *serialize_wu124_(char *w, uint32_t v) +{ + if (v < 0x1fe0) { + v -= 0xe0; + *w++ = (char)(0xe0 | (v >> 8)); *w++ = (char)v; + } else { + *w++ = (char)0xff; +#if LJ_BE + v = lj_bswap(v); +#endif + memcpy(w, &v, 4); w += 4; + } + return w; +} + +static LJ_AINLINE char *serialize_wu124(char *w, uint32_t v) +{ + if (LJ_LIKELY(v < 0xe0)) { + *w++ = (char)v; + return w; + } else { + return serialize_wu124_(w, v); + } +} + +static LJ_NOINLINE char *serialize_ru124_(char *r, char *w, uint32_t *pv) +{ + uint32_t v = *pv; + if (v != 0xff) { + if (r >= w) return NULL; + v = ((v & 0x1f) << 8) + *(uint8_t *)r + 0xe0; r++; + } else { + if (r + 4 > w) return NULL; + v = lj_getu32(r); r += 4; +#if LJ_BE + v = lj_bswap(v); +#endif + } + *pv = v; + return r; +} + +static LJ_AINLINE char *serialize_ru124(char *r, char *w, uint32_t *pv) +{ + if (LJ_LIKELY(r < w)) { + uint32_t v = *(uint8_t *)r; r++; + *pv = v; + if (LJ_UNLIKELY(v >= 0xe0)) { + r = serialize_ru124_(r, w, pv); + } + return r; + } + return NULL; +} + +/* Prepare string dictionary for use (once). */ +void LJ_FASTCALL lj_serialize_dict_prep_str(lua_State *L, GCtab *dict) +{ + if (!dict->hmask) { /* No hash part means not prepared, yet. */ + MSize i, len = lj_tab_len(dict); + if (!len) return; + lj_tab_resize(L, dict, dict->asize, hsize2hbits(len)); + for (i = 1; i <= len && i < dict->asize; i++) { + cTValue *o = arrayslot(dict, i); + if (tvisstr(o)) { + if (!lj_tab_getstr(dict, strV(o))) { /* Ignore dups. */ + lj_tab_newkey(L, dict, o)->u64 = (uint64_t)(i-1); + } + } else if (!tvisfalse(o)) { + lj_err_caller(L, LJ_ERR_BUFFER_BADOPT); + } + } + } +} + +/* Prepare metatable dictionary for use (once). */ +void LJ_FASTCALL lj_serialize_dict_prep_mt(lua_State *L, GCtab *dict) +{ + if (!dict->hmask) { /* No hash part means not prepared, yet. */ + MSize i, len = lj_tab_len(dict); + if (!len) return; + lj_tab_resize(L, dict, dict->asize, hsize2hbits(len)); + for (i = 1; i <= len && i < dict->asize; i++) { + cTValue *o = arrayslot(dict, i); + if (tvistab(o)) { + if (tvisnil(lj_tab_get(L, dict, o))) { /* Ignore dups. */ + lj_tab_newkey(L, dict, o)->u64 = (uint64_t)(i-1); + } + } else if (!tvisfalse(o)) { + lj_err_caller(L, LJ_ERR_BUFFER_BADOPT); + } + } + } +} + +/* -- Internal serializer ------------------------------------------------- */ + +/* Put serialized object into buffer. */ +static char *serialize_put(char *w, SBufExt *sbx, cTValue *o) +{ + if (LJ_LIKELY(tvisstr(o))) { + const GCstr *str = strV(o); + MSize len = str->len; + w = serialize_more(w, sbx, 5+len); + w = serialize_wu124(w, SER_TAG_STR + len); + w = lj_buf_wmem(w, strdata(str), len); + } else if (tvisint(o)) { + uint32_t x = LJ_BE ? lj_bswap((uint32_t)intV(o)) : (uint32_t)intV(o); + w = serialize_more(w, sbx, 1+4); + *w++ = SER_TAG_INT; memcpy(w, &x, 4); w += 4; + } else if (tvisnum(o)) { + uint64_t x = LJ_BE ? lj_bswap64(o->u64) : o->u64; + w = serialize_more(w, sbx, 1+sizeof(lua_Number)); + *w++ = SER_TAG_NUM; memcpy(w, &x, 8); w += 8; + } else if (tvispri(o)) { + w = serialize_more(w, sbx, 1); + *w++ = (char)(SER_TAG_NIL + ~itype(o)); + } else if (tvistab(o)) { + const GCtab *t = tabV(o); + uint32_t narray = 0, nhash = 0, one = 2; + if (sbx->depth <= 0) lj_err_caller(sbufL(sbx), LJ_ERR_BUFFER_DEPTH); + sbx->depth--; + if (t->asize > 0) { /* Determine max. length of array part. */ + ptrdiff_t i; + TValue *array = tvref(t->array); + for (i = (ptrdiff_t)t->asize-1; i >= 0; i--) + if (!tvisnil(&array[i])) + break; + narray = (uint32_t)(i+1); + if (narray && tvisnil(&array[0])) one = 4; + } + if (t->hmask > 0) { /* Count number of used hash slots. */ + uint32_t i, hmask = t->hmask; + Node *node = noderef(t->node); + for (i = 0; i <= hmask; i++) + nhash += !tvisnil(&node[i].val); + } + /* Write metatable index. */ + if (LJ_UNLIKELY(tabref(sbx->dict_mt)) && tabref(t->metatable)) { + TValue mto; + Node *n; + settabV(sbufL(sbx), &mto, tabref(t->metatable)); + n = hashgcref(tabref(sbx->dict_mt), mto.gcr); + do { + if (n->key.u64 == mto.u64) { + uint32_t idx = n->val.u32.lo; + w = serialize_more(w, sbx, 1+5); + *w++ = SER_TAG_DICT_MT; + w = serialize_wu124(w, idx); + break; + } + } while ((n = nextnode(n))); + } + /* Write number of array slots and hash slots. */ + w = serialize_more(w, sbx, 1+2*5); + *w++ = (char)(SER_TAG_TAB + (nhash ? 1 : 0) + (narray ? one : 0)); + if (narray) w = serialize_wu124(w, narray); + if (nhash) w = serialize_wu124(w, nhash); + if (narray) { /* Write array entries. */ + cTValue *oa = tvref(t->array) + (one >> 2); + cTValue *oe = tvref(t->array) + narray; + while (oa < oe) w = serialize_put(w, sbx, oa++); + } + if (nhash) { /* Write hash entries. */ + const Node *node = noderef(t->node) + t->hmask; + GCtab *dict_str = tabref(sbx->dict_str); + if (LJ_UNLIKELY(dict_str)) { + for (;; node--) + if (!tvisnil(&node->val)) { + if (LJ_LIKELY(tvisstr(&node->key))) { + /* Inlined lj_tab_getstr is 30% faster. */ + const GCstr *str = strV(&node->key); + Node *n = hashstr(dict_str, str); + do { + if (tvisstr(&n->key) && strV(&n->key) == str) { + uint32_t idx = n->val.u32.lo; + w = serialize_more(w, sbx, 1+5); + *w++ = SER_TAG_DICT_STR; + w = serialize_wu124(w, idx); + break; + } + n = nextnode(n); + if (!n) { + MSize len = str->len; + w = serialize_more(w, sbx, 5+len); + w = serialize_wu124(w, SER_TAG_STR + len); + w = lj_buf_wmem(w, strdata(str), len); + break; + } + } while (1); + } else { + w = serialize_put(w, sbx, &node->key); + } + w = serialize_put(w, sbx, &node->val); + if (--nhash == 0) break; + } + } else { + for (;; node--) + if (!tvisnil(&node->val)) { + w = serialize_put(w, sbx, &node->key); + w = serialize_put(w, sbx, &node->val); + if (--nhash == 0) break; + } + } + } + sbx->depth++; +#if LJ_HASFFI + } else if (tviscdata(o)) { + CTState *cts = ctype_cts(sbufL(sbx)); + CType *s = ctype_raw(cts, cdataV(o)->ctypeid); + uint8_t *sp = cdataptr(cdataV(o)); + if (ctype_isinteger(s->info) && s->size == 8) { + w = serialize_more(w, sbx, 1+8); + *w++ = (s->info & CTF_UNSIGNED) ? SER_TAG_UINT64 : SER_TAG_INT64; +#if LJ_BE + { uint64_t u = lj_bswap64(*(uint64_t *)sp); memcpy(w, &u, 8); } +#else + memcpy(w, sp, 8); +#endif + w += 8; + } else if (ctype_iscomplex(s->info) && s->size == 16) { + w = serialize_more(w, sbx, 1+16); + *w++ = SER_TAG_COMPLEX; +#if LJ_BE + { /* Only swap the doubles. The re/im order stays the same. */ + uint64_t u = lj_bswap64(((uint64_t *)sp)[0]); memcpy(w, &u, 8); + u = lj_bswap64(((uint64_t *)sp)[1]); memcpy(w+8, &u, 8); + } +#else + memcpy(w, sp, 16); +#endif + w += 16; + } else { + goto badenc; /* NYI other cdata */ + } +#endif + } else if (tvislightud(o)) { + uintptr_t ud = (uintptr_t)lightudV(G(sbufL(sbx)), o); + w = serialize_more(w, sbx, 1+sizeof(ud)); + if (ud == 0) { + *w++ = SER_TAG_NULL; + } else if (LJ_32 || checku32(ud)) { +#if LJ_BE && LJ_64 + ud = lj_bswap64(ud); +#elif LJ_BE + ud = lj_bswap(ud); +#endif + *w++ = SER_TAG_LIGHTUD32; memcpy(w, &ud, 4); w += 4; +#if LJ_64 + } else { +#if LJ_BE + ud = lj_bswap64(ud); +#endif + *w++ = SER_TAG_LIGHTUD64; memcpy(w, &ud, 8); w += 8; +#endif + } + } else { + /* NYI userdata */ +#if LJ_HASFFI + badenc: +#endif + lj_err_callerv(sbufL(sbx), LJ_ERR_BUFFER_BADENC, lj_typename(o)); + } + return w; +} + +/* Get serialized object from buffer. */ +static char *serialize_get(char *r, SBufExt *sbx, TValue *o) +{ + char *w = sbx->w; + uint32_t tp; + r = serialize_ru124(r, w, &tp); if (LJ_UNLIKELY(!r)) goto eob; + if (LJ_LIKELY(tp >= SER_TAG_STR)) { + uint32_t len = tp - SER_TAG_STR; + if (LJ_UNLIKELY(len > (uint32_t)(w - r))) goto eob; + setstrV(sbufL(sbx), o, lj_str_new(sbufL(sbx), r, len)); + r += len; + } else if (tp == SER_TAG_INT) { + if (LJ_UNLIKELY(r + 4 > w)) goto eob; + setintV(o, (int32_t)(LJ_BE ? lj_bswap(lj_getu32(r)) : lj_getu32(r))); + r += 4; + } else if (tp == SER_TAG_NUM) { + if (LJ_UNLIKELY(r + 8 > w)) goto eob; + memcpy(o, r, 8); r += 8; +#if LJ_BE + o->u64 = lj_bswap64(o->u64); +#endif + if (!tvisnum(o)) setnanV(o); /* Fix non-canonical NaNs. */ + } else if (tp <= SER_TAG_TRUE) { + setpriV(o, ~tp); + } else if (tp == SER_TAG_DICT_STR) { + GCtab *dict_str; + uint32_t idx; + r = serialize_ru124(r, w, &idx); if (LJ_UNLIKELY(!r)) goto eob; + idx++; + dict_str = tabref(sbx->dict_str); + if (dict_str && idx < dict_str->asize && tvisstr(arrayslot(dict_str, idx))) + copyTV(sbufL(sbx), o, arrayslot(dict_str, idx)); + else + lj_err_callerv(sbufL(sbx), LJ_ERR_BUFFER_BADDICTX, idx); + } else if (tp >= SER_TAG_TAB && tp <= SER_TAG_DICT_MT) { + uint32_t narray = 0, nhash = 0; + GCtab *t, *mt = NULL; + if (sbx->depth <= 0) lj_err_caller(sbufL(sbx), LJ_ERR_BUFFER_DEPTH); + sbx->depth--; + if (tp == SER_TAG_DICT_MT) { + GCtab *dict_mt; + uint32_t idx; + r = serialize_ru124(r, w, &idx); if (LJ_UNLIKELY(!r)) goto eob; + idx++; + dict_mt = tabref(sbx->dict_mt); + if (dict_mt && idx < dict_mt->asize && tvistab(arrayslot(dict_mt, idx))) + mt = tabV(arrayslot(dict_mt, idx)); + else + lj_err_callerv(sbufL(sbx), LJ_ERR_BUFFER_BADDICTX, idx); + r = serialize_ru124(r, w, &tp); if (LJ_UNLIKELY(!r)) goto eob; + if (!(tp >= SER_TAG_TAB && tp < SER_TAG_DICT_MT)) goto badtag; + } + if (tp >= SER_TAG_TAB+2) { + r = serialize_ru124(r, w, &narray); if (LJ_UNLIKELY(!r)) goto eob; + } + if ((tp & 1)) { + r = serialize_ru124(r, w, &nhash); if (LJ_UNLIKELY(!r)) goto eob; + } + t = lj_tab_new(sbufL(sbx), narray, hsize2hbits(nhash)); + /* NOBARRIER: The table is new (marked white). */ + setgcref(t->metatable, obj2gco(mt)); + settabV(sbufL(sbx), o, t); + if (narray) { + TValue *oa = tvref(t->array) + (tp >= SER_TAG_TAB+4); + TValue *oe = tvref(t->array) + narray; + while (oa < oe) r = serialize_get(r, sbx, oa++); + } + if (nhash) { + do { + TValue k, *v; + r = serialize_get(r, sbx, &k); + v = lj_tab_set(sbufL(sbx), t, &k); + if (LJ_UNLIKELY(!tvisnil(v))) + lj_err_caller(sbufL(sbx), LJ_ERR_BUFFER_DUPKEY); + r = serialize_get(r, sbx, v); + } while (--nhash); + } + sbx->depth++; +#if LJ_HASFFI + } else if (tp >= SER_TAG_INT64 && tp <= SER_TAG_COMPLEX) { + uint32_t sz = tp == SER_TAG_COMPLEX ? 16 : 8; + GCcdata *cd; + if (LJ_UNLIKELY(r + sz > w)) goto eob; + cd = lj_cdata_new_(sbufL(sbx), + tp == SER_TAG_INT64 ? CTID_INT64 : + tp == SER_TAG_UINT64 ? CTID_UINT64 : CTID_COMPLEX_DOUBLE, + sz); + memcpy(cdataptr(cd), r, sz); r += sz; +#if LJ_BE + *(uint64_t *)cdataptr(cd) = lj_bswap64(*(uint64_t *)cdataptr(cd)); + if (sz == 16) + ((uint64_t *)cdataptr(cd))[1] = lj_bswap64(((uint64_t *)cdataptr(cd))[1]); +#endif + if (sz == 16) { /* Fix non-canonical NaNs. */ + TValue *cdo = (TValue *)cdataptr(cd); + if (!tvisnum(&cdo[0])) setnanV(&cdo[0]); + if (!tvisnum(&cdo[1])) setnanV(&cdo[1]); + } + setcdataV(sbufL(sbx), o, cd); +#endif + } else if (tp <= (LJ_64 ? SER_TAG_LIGHTUD64 : SER_TAG_LIGHTUD32)) { + uintptr_t ud = 0; + if (tp == SER_TAG_LIGHTUD32) { + if (LJ_UNLIKELY(r + 4 > w)) goto eob; + ud = (uintptr_t)(LJ_BE ? lj_bswap(lj_getu32(r)) : lj_getu32(r)); + r += 4; + } +#if LJ_64 + else if (tp == SER_TAG_LIGHTUD64) { + if (LJ_UNLIKELY(r + 8 > w)) goto eob; + memcpy(&ud, r, 8); r += 8; +#if LJ_BE + ud = lj_bswap64(ud); +#endif + } + setrawlightudV(o, lj_lightud_intern(sbufL(sbx), (void *)ud)); +#else + setrawlightudV(o, (void *)ud); +#endif + } else { +badtag: + lj_err_callerv(sbufL(sbx), LJ_ERR_BUFFER_BADDEC, tp); + } + return r; +eob: + lj_err_caller(sbufL(sbx), LJ_ERR_BUFFER_EOB); + return NULL; +} + +/* -- External serialization API ------------------------------------------ */ + +/* Encode to buffer. */ +SBufExt * LJ_FASTCALL lj_serialize_put(SBufExt *sbx, cTValue *o) +{ + sbx->depth = LJ_SERIALIZE_DEPTH; + sbx->w = serialize_put(sbx->w, sbx, o); + return sbx; +} + +/* Decode from buffer. */ +char * LJ_FASTCALL lj_serialize_get(SBufExt *sbx, TValue *o) +{ + sbx->depth = LJ_SERIALIZE_DEPTH; + return serialize_get(sbx->r, sbx, o); +} + +/* Stand-alone encoding, borrowing from global temporary buffer. */ +GCstr * LJ_FASTCALL lj_serialize_encode(lua_State *L, cTValue *o) +{ + SBufExt sbx; + char *w; + memset(&sbx, 0, sizeof(SBufExt)); + lj_bufx_set_borrow(L, &sbx, &G(L)->tmpbuf); + sbx.depth = LJ_SERIALIZE_DEPTH; + w = serialize_put(sbx.w, &sbx, o); + return lj_str_new(L, sbx.b, (size_t)(w - sbx.b)); +} + +/* Stand-alone decoding, copy-on-write from string. */ +void lj_serialize_decode(lua_State *L, TValue *o, GCstr *str) +{ + SBufExt sbx; + char *r; + memset(&sbx, 0, sizeof(SBufExt)); + lj_bufx_set_cow(L, &sbx, strdata(str), str->len); + /* No need to set sbx.cowref here. */ + sbx.depth = LJ_SERIALIZE_DEPTH; + r = serialize_get(sbx.r, &sbx, o); + if (r != sbx.w) lj_err_caller(L, LJ_ERR_BUFFER_LEFTOV); +} + +#if LJ_HASJIT +/* Peek into buffer to find the result IRType for specialization purposes. */ +LJ_FUNC MSize LJ_FASTCALL lj_serialize_peektype(SBufExt *sbx) +{ + uint32_t tp; + if (serialize_ru124(sbx->r, sbx->w, &tp)) { + /* This must match the handling of all tags in the decoder above. */ + switch (tp) { + case SER_TAG_NIL: return IRT_NIL; + case SER_TAG_FALSE: return IRT_FALSE; + case SER_TAG_TRUE: return IRT_TRUE; + case SER_TAG_NULL: case SER_TAG_LIGHTUD32: case SER_TAG_LIGHTUD64: + return IRT_LIGHTUD; + case SER_TAG_INT: return LJ_DUALNUM ? IRT_INT : IRT_NUM; + case SER_TAG_NUM: return IRT_NUM; + case SER_TAG_TAB: case SER_TAG_TAB+1: case SER_TAG_TAB+2: + case SER_TAG_TAB+3: case SER_TAG_TAB+4: case SER_TAG_TAB+5: + case SER_TAG_DICT_MT: + return IRT_TAB; + case SER_TAG_INT64: case SER_TAG_UINT64: case SER_TAG_COMPLEX: + return IRT_CDATA; + case SER_TAG_DICT_STR: + default: + return IRT_STR; + } + } + return IRT_NIL; /* Will fail on actual decode. */ +} +#endif + +#endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_serialize.h b/love/src/jni/LuaJIT-2.1/src/lj_serialize.h new file mode 100644 index 00000000..1fda23eb --- /dev/null +++ b/love/src/jni/LuaJIT-2.1/src/lj_serialize.h @@ -0,0 +1,28 @@ +/* +** Object de/serialization. +** Copyright (C) 2005-2021 Mike Pall. See Copyright Notice in luajit.h +*/ + +#ifndef _LJ_SERIALIZE_H +#define _LJ_SERIALIZE_H + +#include "lj_obj.h" +#include "lj_buf.h" + +#if LJ_HASBUFFER + +#define LJ_SERIALIZE_DEPTH 100 /* Default depth. */ + +LJ_FUNC void LJ_FASTCALL lj_serialize_dict_prep_str(lua_State *L, GCtab *dict); +LJ_FUNC void LJ_FASTCALL lj_serialize_dict_prep_mt(lua_State *L, GCtab *dict); +LJ_FUNC SBufExt * LJ_FASTCALL lj_serialize_put(SBufExt *sbx, cTValue *o); +LJ_FUNC char * LJ_FASTCALL lj_serialize_get(SBufExt *sbx, TValue *o); +LJ_FUNC GCstr * LJ_FASTCALL lj_serialize_encode(lua_State *L, cTValue *o); +LJ_FUNC void lj_serialize_decode(lua_State *L, TValue *o, GCstr *str); +#if LJ_HASJIT +LJ_FUNC MSize LJ_FASTCALL lj_serialize_peektype(SBufExt *sbx); +#endif + +#endif + +#endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_snap.c b/love/src/jni/LuaJIT-2.1/src/lj_snap.c index 21f27e1f..97097a5b 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_snap.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_snap.c @@ -171,6 +171,7 @@ static void snapshot_stack(jit_State *J, SnapShot *snap, MSize nsnapmap) nent += snapshot_framelinks(J, p + nent, &snap->topslot); snap->mapofs = (uint32_t)nsnapmap; snap->ref = (IRRef1)J->cur.nins; + snap->mcofs = 0; snap->nslots = (uint8_t)nslots; snap->count = 0; J->cur.nsnapmap = (uint32_t)(nsnapmap + nent); @@ -251,7 +252,12 @@ static BCReg snap_usedef(jit_State *J, uint8_t *udf, BCReg minslot = bc_a(ins); if (op >= BC_FORI && op <= BC_JFORL) minslot += FORL_EXT; else if (op >= BC_ITERL && op <= BC_JITERL) minslot += bc_b(pc[-2])-1; - else if (op == BC_UCLO) { pc += bc_j(ins); break; } + else if (op == BC_UCLO) { + ptrdiff_t delta = bc_j(ins); + if (delta < 0) return maxslot; /* Prevent loop. */ + pc += delta; + break; + } for (s = minslot; s < maxslot; s++) DEF_SLOT(s); return minslot < maxslot ? minslot : maxslot; } @@ -275,7 +281,7 @@ static BCReg snap_usedef(jit_State *J, uint8_t *udf, if (!(op == BC_ISTC || op == BC_ISFC)) DEF_SLOT(bc_a(ins)); break; case BCMbase: - if (op >= BC_CALLM && op <= BC_VARG) { + if (op >= BC_CALLM && op <= BC_ITERN) { BCReg top = (op == BC_CALLM || op == BC_CALLMT || bc_c(ins) == 0) ? maxslot : (bc_a(ins) + bc_c(ins)+LJ_FR2); if (LJ_FR2) DEF_SLOT(bc_a(ins)+1); @@ -286,6 +292,8 @@ static BCReg snap_usedef(jit_State *J, uint8_t *udf, for (s = 0; s < bc_a(ins); s++) DEF_SLOT(s); return 0; } + } else if (op == BC_VARG) { + return maxslot; /* NYI: punt. */ } else if (op == BC_KNIL) { for (s = bc_a(ins); s <= bc_d(ins); s++) DEF_SLOT(s); } else if (op == BC_TSETM) { @@ -304,15 +312,45 @@ static BCReg snap_usedef(jit_State *J, uint8_t *udf, return 0; /* unreachable */ } +/* Mark slots used by upvalues of child prototypes as used. */ +void snap_useuv(GCproto *pt, uint8_t *udf) +{ + /* This is a coarse check, because it's difficult to correlate the lifetime + ** of slots and closures. But the number of false positives is quite low. + ** A false positive may cause a slot not to be purged, which is just + ** a missed optimization. + */ + if ((pt->flags & PROTO_CHILD)) { + ptrdiff_t i, j, n = pt->sizekgc; + GCRef *kr = mref(pt->k, GCRef) - 1; + for (i = 0; i < n; i++, kr--) { + GCobj *o = gcref(*kr); + if (o->gch.gct == ~LJ_TPROTO) { + for (j = 0; j < gco2pt(o)->sizeuv; j++) { + uint32_t v = proto_uv(gco2pt(o))[j]; + if ((v & PROTO_UV_LOCAL)) { + udf[(v & 0xff)] = 0; + } + } + } + } + } +} + /* Purge dead slots before the next snapshot. */ void lj_snap_purge(jit_State *J) { uint8_t udf[SNAP_USEDEF_SLOTS]; - BCReg maxslot = J->maxslot; - BCReg s = snap_usedef(J, udf, J->pc, maxslot); - for (; s < maxslot; s++) - if (udf[s] != 0) - J->base[s] = 0; /* Purge dead slots. */ + BCReg s, maxslot = J->maxslot; + if (bc_op(*J->pc) == BC_FUNCV && maxslot > J->pt->numparams) + maxslot = J->pt->numparams; + s = snap_usedef(J, udf, J->pc, maxslot); + if (s < maxslot) { + snap_useuv(J->pt, udf); + for (; s < maxslot; s++) + if (udf[s] != 0) + J->base[s] = 0; /* Purge dead slots. */ + } } /* Shrink last snapshot. */ @@ -325,6 +363,7 @@ void lj_snap_shrink(jit_State *J) BCReg maxslot = J->maxslot; BCReg baseslot = J->baseslot; BCReg minslot = snap_usedef(J, udf, snap_pc(&map[nent]), maxslot); + if (minslot < maxslot) snap_useuv(J->pt, udf); maxslot += baseslot; minslot += baseslot; snap->nslots = (uint8_t)maxslot; @@ -424,7 +463,7 @@ static TRef snap_dedup(jit_State *J, SnapEntry *map, MSize nmax, IRRef ref) MSize j; for (j = 0; j < nmax; j++) if (snap_ref(map[j]) == ref) - return J->slot[snap_slot(map[j])] & ~(SNAP_CONT|SNAP_FRAME); + return J->slot[snap_slot(map[j])] & ~(SNAP_KEYINDEX|SNAP_CONT|SNAP_FRAME); return 0; } @@ -499,10 +538,12 @@ void lj_snap_replay(jit_State *J, GCtrace *T) uint32_t mode = IRSLOAD_INHERIT|IRSLOAD_PARENT; if (LJ_SOFTFP32 && (sn & SNAP_SOFTFPNUM)) t = IRT_NUM; if (ir->o == IR_SLOAD) mode |= (ir->op2 & IRSLOAD_READONLY); + if ((sn & SNAP_KEYINDEX)) mode |= IRSLOAD_KEYINDEX; tr = emitir_raw(IRT(IR_SLOAD, t), s, mode); } setslot: - J->slot[s] = tr | (sn&(SNAP_CONT|SNAP_FRAME)); /* Same as TREF_* flags. */ + /* Same as TREF_* flags. */ + J->slot[s] = tr | (sn&(SNAP_KEYINDEX|SNAP_CONT|SNAP_FRAME)); J->framedepth += ((sn & (SNAP_CONT|SNAP_FRAME)) && (s != LJ_FR2)); if ((sn & SNAP_FRAME)) J->baseslot = s+1; @@ -922,6 +963,10 @@ const BCIns *lj_snap_restore(jit_State *J, void *exptr) setframe_ftsz(o, snap_slot(sn) != 0 ? (int32_t)*flinks-- : ftsz0); L->base = o+1; #endif + } else if ((sn & SNAP_KEYINDEX)) { + /* A IRT_INT key index slot is restored as a number. Undo this. */ + o->u32.lo = (uint32_t)(LJ_DUALNUM ? intV(o) : lj_num2int(numV(o))); + o->u32.hi = LJ_KEYINDEX; } } } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_state.c b/love/src/jni/LuaJIT-2.1/src/lj_state.c index 9585c18a..e87b945a 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_state.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_state.c @@ -156,6 +156,7 @@ static TValue *cpluaopen(lua_State *L, lua_CFunction dummy, void *ud) fixstring(lj_err_str(L, LJ_ERR_ERRMEM)); /* Preallocate memory error msg. */ g->gc.threshold = 4*g->gc.total; lj_trace_initstate(g); + lj_err_verify(); return NULL; } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_strfmt.c b/love/src/jni/LuaJIT-2.1/src/lj_strfmt.c index b1c74f88..945954aa 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_strfmt.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_strfmt.c @@ -9,11 +9,17 @@ #define LUA_CORE #include "lj_obj.h" +#include "lj_err.h" #include "lj_buf.h" #include "lj_str.h" +#include "lj_meta.h" #include "lj_state.h" #include "lj_char.h" #include "lj_strfmt.h" +#if LJ_HASFFI +#include "lj_ctype.h" +#endif +#include "lj_lib.h" /* -- Format parser ------------------------------------------------------- */ @@ -161,6 +167,10 @@ const char *lj_strfmt_wstrnum(lua_State *L, cTValue *o, MSize *lenp) if (tvisstr(o)) { *lenp = strV(o)->len; return strVdata(o); + } else if (tvisbuf(o)) { + SBufExt *sbx = bufV(o); + *lenp = sbufxlen(sbx); + return sbx->r; } else if (tvisint(o)) { sb = lj_strfmt_putint(lj_buf_tmp_(L), intV(o)); } else if (tvisnum(o)) { @@ -169,7 +179,7 @@ const char *lj_strfmt_wstrnum(lua_State *L, cTValue *o, MSize *lenp) return NULL; } *lenp = sbuflen(sb); - return sbufB(sb); + return sb->b; } /* -- Unformatted conversions to buffer ----------------------------------- */ @@ -177,7 +187,7 @@ const char *lj_strfmt_wstrnum(lua_State *L, cTValue *o, MSize *lenp) /* Add integer to buffer. */ SBuf * LJ_FASTCALL lj_strfmt_putint(SBuf *sb, int32_t k) { - setsbufP(sb, lj_strfmt_wint(lj_buf_more(sb, STRFMT_MAXBUF_INT), k)); + sb->w = lj_strfmt_wint(lj_buf_more(sb, STRFMT_MAXBUF_INT), k); return sb; } @@ -191,73 +201,86 @@ SBuf * LJ_FASTCALL lj_strfmt_putnum(SBuf *sb, cTValue *o) SBuf * LJ_FASTCALL lj_strfmt_putptr(SBuf *sb, const void *v) { - setsbufP(sb, lj_strfmt_wptr(lj_buf_more(sb, STRFMT_MAXBUF_PTR), v)); + sb->w = lj_strfmt_wptr(lj_buf_more(sb, STRFMT_MAXBUF_PTR), v); return sb; } /* Add quoted string to buffer. */ -SBuf * LJ_FASTCALL lj_strfmt_putquoted(SBuf *sb, GCstr *str) +static SBuf *strfmt_putquotedlen(SBuf *sb, const char *s, MSize len) { - const char *s = strdata(str); - MSize len = str->len; lj_buf_putb(sb, '"'); while (len--) { uint32_t c = (uint32_t)(uint8_t)*s++; - char *p = lj_buf_more(sb, 4); + char *w = lj_buf_more(sb, 4); if (c == '"' || c == '\\' || c == '\n') { - *p++ = '\\'; + *w++ = '\\'; } else if (lj_char_iscntrl(c)) { /* This can only be 0-31 or 127. */ uint32_t d; - *p++ = '\\'; + *w++ = '\\'; if (c >= 100 || lj_char_isdigit((uint8_t)*s)) { - *p++ = (char)('0'+(c >= 100)); if (c >= 100) c -= 100; + *w++ = (char)('0'+(c >= 100)); if (c >= 100) c -= 100; goto tens; } else if (c >= 10) { tens: - d = (c * 205) >> 11; c -= d * 10; *p++ = (char)('0'+d); + d = (c * 205) >> 11; c -= d * 10; *w++ = (char)('0'+d); } c += '0'; } - *p++ = (char)c; - setsbufP(sb, p); + *w++ = (char)c; + sb->w = w; } lj_buf_putb(sb, '"'); return sb; } +#if LJ_HASJIT +SBuf * LJ_FASTCALL lj_strfmt_putquoted(SBuf *sb, GCstr *str) +{ + return strfmt_putquotedlen(sb, strdata(str), str->len); +} +#endif + /* -- Formatted conversions to buffer ------------------------------------- */ /* Add formatted char to buffer. */ SBuf *lj_strfmt_putfchar(SBuf *sb, SFormat sf, int32_t c) { MSize width = STRFMT_WIDTH(sf); - char *p = lj_buf_more(sb, width > 1 ? width : 1); - if ((sf & STRFMT_F_LEFT)) *p++ = (char)c; - while (width-- > 1) *p++ = ' '; - if (!(sf & STRFMT_F_LEFT)) *p++ = (char)c; - setsbufP(sb, p); + char *w = lj_buf_more(sb, width > 1 ? width : 1); + if ((sf & STRFMT_F_LEFT)) *w++ = (char)c; + while (width-- > 1) *w++ = ' '; + if (!(sf & STRFMT_F_LEFT)) *w++ = (char)c; + sb->w = w; return sb; } /* Add formatted string to buffer. */ -SBuf *lj_strfmt_putfstr(SBuf *sb, SFormat sf, GCstr *str) +static SBuf *strfmt_putfstrlen(SBuf *sb, SFormat sf, const char *s, MSize len) { - MSize len = str->len <= STRFMT_PREC(sf) ? str->len : STRFMT_PREC(sf); MSize width = STRFMT_WIDTH(sf); - char *p = lj_buf_more(sb, width > len ? width : len); - if ((sf & STRFMT_F_LEFT)) p = lj_buf_wmem(p, strdata(str), len); - while (width-- > len) *p++ = ' '; - if (!(sf & STRFMT_F_LEFT)) p = lj_buf_wmem(p, strdata(str), len); - setsbufP(sb, p); + char *w; + if (len > STRFMT_PREC(sf)) len = STRFMT_PREC(sf); + w = lj_buf_more(sb, width > len ? width : len); + if ((sf & STRFMT_F_LEFT)) w = lj_buf_wmem(w, s, len); + while (width-- > len) *w++ = ' '; + if (!(sf & STRFMT_F_LEFT)) w = lj_buf_wmem(w, s, len); + sb->w = w; return sb; } +#if LJ_HASJIT +SBuf *lj_strfmt_putfstr(SBuf *sb, SFormat sf, GCstr *str) +{ + return strfmt_putfstrlen(sb, sf, strdata(str), str->len); +} +#endif + /* Add formatted signed/unsigned integer to buffer. */ SBuf *lj_strfmt_putfxint(SBuf *sb, SFormat sf, uint64_t k) { - char buf[STRFMT_MAXBUF_XINT], *q = buf + sizeof(buf), *p; + char buf[STRFMT_MAXBUF_XINT], *q = buf + sizeof(buf), *w; #ifdef LUA_USE_ASSERT - char *ps; + char *ws; #endif MSize prefix = 0, len, prec, pprec, width, need; @@ -301,27 +324,27 @@ SBuf *lj_strfmt_putfxint(SBuf *sb, SFormat sf, uint64_t k) width = STRFMT_WIDTH(sf); pprec = prec + (prefix >> 8); need = width > pprec ? width : pprec; - p = lj_buf_more(sb, need); + w = lj_buf_more(sb, need); #ifdef LUA_USE_ASSERT - ps = p; + ws = w; #endif /* Format number with leading/trailing whitespace and zeros. */ if ((sf & (STRFMT_F_LEFT|STRFMT_F_ZERO)) == 0) - while (width-- > pprec) *p++ = ' '; + while (width-- > pprec) *w++ = ' '; if (prefix) { - if ((char)prefix >= 'X') *p++ = '0'; - *p++ = (char)prefix; + if ((char)prefix >= 'X') *w++ = '0'; + *w++ = (char)prefix; } if ((sf & (STRFMT_F_LEFT|STRFMT_F_ZERO)) == STRFMT_F_ZERO) - while (width-- > pprec) *p++ = '0'; - while (prec-- > len) *p++ = '0'; - while (q < buf + sizeof(buf)) *p++ = *q++; /* Add number itself. */ + while (width-- > pprec) *w++ = '0'; + while (prec-- > len) *w++ = '0'; + while (q < buf + sizeof(buf)) *w++ = *q++; /* Add number itself. */ if ((sf & STRFMT_F_LEFT)) - while (width-- > pprec) *p++ = ' '; + while (width-- > pprec) *w++ = ' '; - lj_assertX(need == (MSize)(p - ps), "miscalculated format size"); - setsbufP(sb, p); + lj_assertX(need == (MSize)(w - ws), "miscalculated format size"); + sb->w = w; return sb; } @@ -346,6 +369,117 @@ SBuf *lj_strfmt_putfnum_uint(SBuf *sb, SFormat sf, lua_Number n) return lj_strfmt_putfxint(sb, sf, (uint64_t)k); } +/* Format stack arguments to buffer. */ +int lj_strfmt_putarg(lua_State *L, SBuf *sb, int arg, int retry) +{ + int narg = (int)(L->top - L->base); + GCstr *fmt = lj_lib_checkstr(L, arg); + FormatState fs; + SFormat sf; + lj_strfmt_init(&fs, strdata(fmt), fmt->len); + while ((sf = lj_strfmt_parse(&fs)) != STRFMT_EOF) { + if (sf == STRFMT_LIT) { + lj_buf_putmem(sb, fs.str, fs.len); + } else if (sf == STRFMT_ERR) { + lj_err_callerv(L, LJ_ERR_STRFMT, + strdata(lj_str_new(L, fs.str, fs.len))); + } else { + TValue *o = &L->base[arg++]; + if (arg > narg) + lj_err_arg(L, arg, LJ_ERR_NOVAL); + switch (STRFMT_TYPE(sf)) { + case STRFMT_INT: + if (tvisint(o)) { + int32_t k = intV(o); + if (sf == STRFMT_INT) + lj_strfmt_putint(sb, k); /* Shortcut for plain %d. */ + else + lj_strfmt_putfxint(sb, sf, k); + break; + } +#if LJ_HASFFI + if (tviscdata(o)) { + GCcdata *cd = cdataV(o); + if (cd->ctypeid == CTID_INT64 || cd->ctypeid == CTID_UINT64) { + lj_strfmt_putfxint(sb, sf, *(uint64_t *)cdataptr(cd)); + break; + } + } +#endif + lj_strfmt_putfnum_int(sb, sf, lj_lib_checknum(L, arg)); + break; + case STRFMT_UINT: + if (tvisint(o)) { + lj_strfmt_putfxint(sb, sf, intV(o)); + break; + } +#if LJ_HASFFI + if (tviscdata(o)) { + GCcdata *cd = cdataV(o); + if (cd->ctypeid == CTID_INT64 || cd->ctypeid == CTID_UINT64) { + lj_strfmt_putfxint(sb, sf, *(uint64_t *)cdataptr(cd)); + break; + } + } +#endif + lj_strfmt_putfnum_uint(sb, sf, lj_lib_checknum(L, arg)); + break; + case STRFMT_NUM: + lj_strfmt_putfnum(sb, sf, lj_lib_checknum(L, arg)); + break; + case STRFMT_STR: { + MSize len; + const char *s; + cTValue *mo; + if (LJ_UNLIKELY(!tvisstr(o) && !tvisbuf(o)) && retry >= 0 && + !tvisnil(mo = lj_meta_lookup(L, o, MM_tostring))) { + /* Call __tostring metamethod once. */ + copyTV(L, L->top++, mo); + copyTV(L, L->top++, o); + lua_call(L, 1, 1); + o = &L->base[arg-1]; /* Stack may have been reallocated. */ + copyTV(L, o, --L->top); /* Replace inline for retry. */ + if (retry < 2) { /* Global buffer may have been overwritten. */ + retry = 1; + break; + } + } + if (LJ_LIKELY(tvisstr(o))) { + len = strV(o)->len; + s = strVdata(o); +#if LJ_HASBUFFER + } else if (tvisbuf(o)) { + SBufExt *sbx = bufV(o); + if (sbx == (SBufExt *)sb) lj_err_arg(L, arg+1, LJ_ERR_BUFFER_SELF); + len = sbufxlen(sbx); + s = sbx->r; +#endif + } else { + GCstr *str = lj_strfmt_obj(L, o); + len = str->len; + s = strdata(str); + } + if ((sf & STRFMT_T_QUOTED)) + strfmt_putquotedlen(sb, s, len); /* No formatting. */ + else + strfmt_putfstrlen(sb, sf, s, len); + break; + } + case STRFMT_CHAR: + lj_strfmt_putfchar(sb, sf, lj_lib_checkint(L, arg)); + break; + case STRFMT_PTR: /* No formatting. */ + lj_strfmt_putptr(sb, lj_obj_ptr(G(L), o)); + break; + default: + lj_assertL(0, "bad string format type"); + break; + } + } + } + return retry; +} + /* -- Conversions to strings ---------------------------------------------- */ /* Convert integer to string. */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_strfmt.h b/love/src/jni/LuaJIT-2.1/src/lj_strfmt.h index fa25dcd0..cb2c7360 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_strfmt.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_strfmt.h @@ -95,7 +95,9 @@ LJ_FUNC SBuf * LJ_FASTCALL lj_strfmt_putint(SBuf *sb, int32_t k); LJ_FUNC SBuf * LJ_FASTCALL lj_strfmt_putnum(SBuf *sb, cTValue *o); #endif LJ_FUNC SBuf * LJ_FASTCALL lj_strfmt_putptr(SBuf *sb, const void *v); +#if LJ_HASJIT LJ_FUNC SBuf * LJ_FASTCALL lj_strfmt_putquoted(SBuf *sb, GCstr *str); +#endif /* Formatted conversions to buffer. */ LJ_FUNC SBuf *lj_strfmt_putfxint(SBuf *sb, SFormat sf, uint64_t k); @@ -103,7 +105,10 @@ LJ_FUNC SBuf *lj_strfmt_putfnum_int(SBuf *sb, SFormat sf, lua_Number n); LJ_FUNC SBuf *lj_strfmt_putfnum_uint(SBuf *sb, SFormat sf, lua_Number n); LJ_FUNC SBuf *lj_strfmt_putfnum(SBuf *sb, SFormat, lua_Number n); LJ_FUNC SBuf *lj_strfmt_putfchar(SBuf *sb, SFormat, int32_t c); +#if LJ_HASJIT LJ_FUNC SBuf *lj_strfmt_putfstr(SBuf *sb, SFormat, GCstr *str); +#endif +LJ_FUNC int lj_strfmt_putarg(lua_State *L, SBuf *sb, int arg, int retry); /* Conversions to strings. */ LJ_FUNC GCstr * LJ_FASTCALL lj_strfmt_int(lua_State *L, int32_t k); diff --git a/love/src/jni/LuaJIT-2.1/src/lj_strfmt_num.c b/love/src/jni/LuaJIT-2.1/src/lj_strfmt_num.c index 66f83af3..dfd56bd4 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_strfmt_num.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_strfmt_num.c @@ -576,7 +576,7 @@ static char *lj_strfmt_wfnum(SBuf *sb, SFormat sf, lua_Number n, char *p) /* Add formatted floating-point number to buffer. */ SBuf *lj_strfmt_putfnum(SBuf *sb, SFormat sf, lua_Number n) { - setsbufP(sb, lj_strfmt_wfnum(sb, sf, n, NULL)); + sb->w = lj_strfmt_wfnum(sb, sf, n, NULL); return sb; } diff --git a/love/src/jni/LuaJIT-2.1/src/lj_tab.c b/love/src/jni/LuaJIT-2.1/src/lj_tab.c index 27e58f0a..4113839f 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_tab.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_tab.c @@ -16,25 +16,6 @@ /* -- Object hashing ------------------------------------------------------ */ -/* Hash values are masked with the table hash mask and used as an index. */ -static LJ_AINLINE Node *hashmask(const GCtab *t, uint32_t hash) -{ - Node *n = noderef(t->node); - return &n[hash & t->hmask]; -} - -/* String IDs are generated when a string is interned. */ -#define hashstr(t, s) hashmask(t, (s)->sid) - -#define hashlohi(t, lo, hi) hashmask((t), hashrot((lo), (hi))) -#define hashnum(t, o) hashlohi((t), (o)->u32.lo, ((o)->u32.hi << 1)) -#if LJ_GC64 -#define hashgcref(t, r) \ - hashlohi((t), (uint32_t)gcrefu(r), (uint32_t)(gcrefu(r) >> 32)) -#else -#define hashgcref(t, r) hashlohi((t), gcrefu(r), gcrefu(r) + HASH_BIAS) -#endif - /* Hash an arbitrary key and return its anchor position in the hash table. */ static Node *hashkey(const GCtab *t, cTValue *key) { @@ -413,7 +394,7 @@ cTValue * LJ_FASTCALL lj_tab_getinth(GCtab *t, int32_t key) return NULL; } -cTValue *lj_tab_getstr(GCtab *t, GCstr *key) +cTValue *lj_tab_getstr(GCtab *t, const GCstr *key) { Node *n = hashstr(t, key); do { @@ -546,7 +527,7 @@ TValue *lj_tab_setinth(lua_State *L, GCtab *t, int32_t key) return lj_tab_newkey(L, t, &k); } -TValue *lj_tab_setstr(lua_State *L, GCtab *t, GCstr *key) +TValue *lj_tab_setstr(lua_State *L, GCtab *t, const GCstr *key) { TValue k; Node *n = hashstr(t, key); @@ -587,56 +568,66 @@ TValue *lj_tab_set(lua_State *L, GCtab *t, cTValue *key) /* -- Table traversal ----------------------------------------------------- */ -/* Get the traversal index of a key. */ -static uint32_t keyindex(lua_State *L, GCtab *t, cTValue *key) +/* Table traversal indexes: +** +** Array key index: [0 .. t->asize-1] +** Hash key index: [t->asize .. t->asize+t->hmask] +** Invalid key: ~0 +*/ + +/* Get the successor traversal index of a key. */ +uint32_t LJ_FASTCALL lj_tab_keyindex(GCtab *t, cTValue *key) { TValue tmp; if (tvisint(key)) { int32_t k = intV(key); if ((uint32_t)k < t->asize) - return (uint32_t)k; /* Array key indexes: [0..t->asize-1] */ + return (uint32_t)k + 1; setnumV(&tmp, (lua_Number)k); key = &tmp; } else if (tvisnum(key)) { lua_Number nk = numV(key); int32_t k = lj_num2int(nk); if ((uint32_t)k < t->asize && nk == (lua_Number)k) - return (uint32_t)k; /* Array key indexes: [0..t->asize-1] */ + return (uint32_t)k + 1; } if (!tvisnil(key)) { Node *n = hashkey(t, key); do { if (lj_obj_equal(&n->key, key)) - return t->asize + (uint32_t)(n - noderef(t->node)); - /* Hash key indexes: [t->asize..t->asize+t->nmask] */ + return t->asize + (uint32_t)((n+1) - noderef(t->node)); } while ((n = nextnode(n))); - if (key->u32.hi == 0xfffe7fff) /* ITERN was despecialized while running. */ - return key->u32.lo - 1; - lj_err_msg(L, LJ_ERR_NEXTIDX); - return 0; /* unreachable */ + if (key->u32.hi == LJ_KEYINDEX) /* Despecialized ITERN while running. */ + return key->u32.lo; + return ~0u; /* Invalid key to next. */ } - return ~0u; /* A nil key starts the traversal. */ + return 0; /* A nil key starts the traversal. */ } -/* Advance to the next step in a table traversal. */ -int lj_tab_next(lua_State *L, GCtab *t, TValue *key) +/* Get the next key/value pair of a table traversal. */ +int lj_tab_next(GCtab *t, cTValue *key, TValue *o) { - uint32_t i = keyindex(L, t, key); /* Find predecessor key index. */ - for (i++; i < t->asize; i++) /* First traverse the array keys. */ - if (!tvisnil(arrayslot(t, i))) { - setintV(key, i); - copyTV(L, key+1, arrayslot(t, i)); - return 1; - } - for (i -= t->asize; i <= t->hmask; i++) { /* Then traverse the hash keys. */ - Node *n = &noderef(t->node)[i]; - if (!tvisnil(&n->val)) { - copyTV(L, key, &n->key); - copyTV(L, key+1, &n->val); + uint32_t idx = lj_tab_keyindex(t, key); /* Find successor index of key. */ + /* First traverse the array part. */ + for (; idx < t->asize; idx++) { + cTValue *a = arrayslot(t, idx); + if (LJ_LIKELY(!tvisnil(a))) { + setintV(o, idx); + o[1] = *a; return 1; } } - return 0; /* End of traversal. */ + idx -= t->asize; + /* Then traverse the hash part. */ + for (; idx <= t->hmask; idx++) { + Node *n = &noderef(t->node)[idx]; + if (!tvisnil(&n->val)) { + o[0] = n->key; + o[1] = n->val; + return 1; + } + } + return (int32_t)idx < 0 ? -1 : 0; /* Invalid key or end of traversal. */ } /* -- Table length calculation -------------------------------------------- */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_tab.h b/love/src/jni/LuaJIT-2.1/src/lj_tab.h index 97436cc0..e0e81ff7 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_tab.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_tab.h @@ -31,6 +31,25 @@ static LJ_AINLINE uint32_t hashrot(uint32_t lo, uint32_t hi) return hi; } +/* Hash values are masked with the table hash mask and used as an index. */ +static LJ_AINLINE Node *hashmask(const GCtab *t, uint32_t hash) +{ + Node *n = noderef(t->node); + return &n[hash & t->hmask]; +} + +/* String IDs are generated when a string is interned. */ +#define hashstr(t, s) hashmask(t, (s)->sid) + +#define hashlohi(t, lo, hi) hashmask((t), hashrot((lo), (hi))) +#define hashnum(t, o) hashlohi((t), (o)->u32.lo, ((o)->u32.hi << 1)) +#if LJ_GC64 +#define hashgcref(t, r) \ + hashlohi((t), (uint32_t)gcrefu(r), (uint32_t)(gcrefu(r) >> 32)) +#else +#define hashgcref(t, r) hashlohi((t), gcrefu(r), gcrefu(r) + HASH_BIAS) +#endif + #define hsize2hbits(s) ((s) ? ((s)==1 ? 1 : 1+lj_fls((uint32_t)((s)-1))) : 0) LJ_FUNCA GCtab *lj_tab_new(lua_State *L, uint32_t asize, uint32_t hbits); @@ -50,14 +69,14 @@ LJ_FUNCA void lj_tab_reasize(lua_State *L, GCtab *t, uint32_t nasize); /* Caveat: all getters except lj_tab_get() can return NULL! */ LJ_FUNCA cTValue * LJ_FASTCALL lj_tab_getinth(GCtab *t, int32_t key); -LJ_FUNC cTValue *lj_tab_getstr(GCtab *t, GCstr *key); +LJ_FUNC cTValue *lj_tab_getstr(GCtab *t, const GCstr *key); LJ_FUNCA cTValue *lj_tab_get(lua_State *L, GCtab *t, cTValue *key); /* Caveat: all setters require a write barrier for the stored value. */ LJ_FUNCA TValue *lj_tab_newkey(lua_State *L, GCtab *t, cTValue *key); LJ_FUNCA TValue *lj_tab_setinth(lua_State *L, GCtab *t, int32_t key); -LJ_FUNC TValue *lj_tab_setstr(lua_State *L, GCtab *t, GCstr *key); +LJ_FUNC TValue *lj_tab_setstr(lua_State *L, GCtab *t, const GCstr *key); LJ_FUNC TValue *lj_tab_set(lua_State *L, GCtab *t, cTValue *key); #define inarray(t, key) ((MSize)(key) < (MSize)(t)->asize) @@ -67,7 +86,8 @@ LJ_FUNC TValue *lj_tab_set(lua_State *L, GCtab *t, cTValue *key); #define lj_tab_setint(L, t, key) \ (inarray((t), (key)) ? arrayslot((t), (key)) : lj_tab_setinth(L, (t), (key))) -LJ_FUNCA int lj_tab_next(lua_State *L, GCtab *t, TValue *key); +LJ_FUNC uint32_t LJ_FASTCALL lj_tab_keyindex(GCtab *t, cTValue *key); +LJ_FUNCA int lj_tab_next(GCtab *t, cTValue *key, TValue *o); LJ_FUNCA MSize LJ_FASTCALL lj_tab_len(GCtab *t); #if LJ_HASJIT LJ_FUNC MSize LJ_FASTCALL lj_tab_len_hint(GCtab *t, size_t hint); diff --git a/love/src/jni/LuaJIT-2.1/src/lj_target_arm.h b/love/src/jni/LuaJIT-2.1/src/lj_target_arm.h index f41f53e9..72516bc2 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_target_arm.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_target_arm.h @@ -211,6 +211,7 @@ typedef enum ARMIns { /* ARMv6T2 */ ARMI_MOVW = 0xe3000000, ARMI_MOVT = 0xe3400000, + ARMI_BFI = 0xe7c00010, /* VFP */ ARMI_VMOV_D = 0xeeb00b40, diff --git a/love/src/jni/LuaJIT-2.1/src/lj_target_arm64.h b/love/src/jni/LuaJIT-2.1/src/lj_target_arm64.h index c2f71153..6d39ffb8 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_target_arm64.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_target_arm64.h @@ -31,6 +31,8 @@ enum { /* Calling conventions. */ RID_RET = RID_X0, + RID_RETLO = RID_X0, + RID_RETHI = RID_X1, RID_FPRET = RID_D0, /* These definitions must match with the *.dasc file(s): */ @@ -210,6 +212,8 @@ typedef enum A64Ins { A64I_EXTRw = 0x13800000, A64I_EXTRx = 0x93c00000, + A64I_BFMw = 0x33000000, + A64I_BFMx = 0xb3400000, A64I_SBFMw = 0x13000000, A64I_SBFMx = 0x93400000, A64I_SXTBw = 0x13001c00, diff --git a/love/src/jni/LuaJIT-2.1/src/lj_target_mips.h b/love/src/jni/LuaJIT-2.1/src/lj_target_mips.h index 8d8f68ec..5da94605 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_target_mips.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_target_mips.h @@ -256,6 +256,8 @@ typedef enum MIPSIns { MIPSI_ROTRV = 0x00000046, /* MIPSXXR2 */ MIPSI_DROTRV = 0x00000056, + MIPSI_INS = 0x7c000004, /* MIPSXXR2 */ + MIPSI_SEB = 0x7c000420, /* MIPSXXR2 */ MIPSI_SEH = 0x7c000620, /* MIPSXXR2 */ MIPSI_WSBH = 0x7c0000a0, /* MIPSXXR2 */ diff --git a/love/src/jni/LuaJIT-2.1/src/lj_target_x86.h b/love/src/jni/LuaJIT-2.1/src/lj_target_x86.h index 4be5f04c..d0ce196c 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_target_x86.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_target_x86.h @@ -38,10 +38,9 @@ enum { RID_RET = RID_EAX, #if LJ_64 RID_FPRET = RID_XMM0, -#else +#endif RID_RETLO = RID_EAX, RID_RETHI = RID_EDX, -#endif /* These definitions must match with the *.dasc file(s): */ RID_BASE = RID_EDX, /* Interpreter BASE. */ @@ -165,6 +164,8 @@ typedef struct { #define EXITSTUB_SPACING (2+2) #define EXITSTUBS_PER_GROUP 32 +#define EXITTRACE_VMSTATE 1 /* g->vmstate has traceno on exit. */ + /* -- x86 ModRM operand encoding ------------------------------------------ */ typedef enum { diff --git a/love/src/jni/LuaJIT-2.1/src/lj_trace.c b/love/src/jni/LuaJIT-2.1/src/lj_trace.c index 4608f81f..43b86e4f 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_trace.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_trace.c @@ -215,8 +215,8 @@ static void trace_unpatch(jit_State *J, GCtrace *T) break; case BC_JITERL: case BC_JLOOP: - lj_assertJ(op == BC_ITERL || op == BC_LOOP || bc_isret(op), - "bad original bytecode %d", op); + lj_assertJ(op == BC_ITERL || op == BC_ITERN || op == BC_LOOP || + bc_isret(op), "bad original bytecode %d", op); *pc = T->startins; break; case BC_JMP: @@ -373,8 +373,13 @@ void lj_trace_freestate(global_State *g) /* Blacklist a bytecode instruction. */ static void blacklist_pc(GCproto *pt, BCIns *pc) { - setbc_op(pc, (int)bc_op(*pc)+(int)BC_ILOOP-(int)BC_LOOP); - pt->flags |= PROTO_ILOOP; + if (bc_op(*pc) == BC_ITERN) { + setbc_op(pc, BC_ITERC); + setbc_op(pc+1+bc_j(pc[1]), BC_JMP); + } else { + setbc_op(pc, (int)bc_op(*pc)+(int)BC_ILOOP-(int)BC_LOOP); + pt->flags |= PROTO_ILOOP; + } } /* Penalize a bytecode instruction. */ @@ -411,7 +416,7 @@ static void trace_start(jit_State *J) TraceNo traceno; if ((J->pt->flags & PROTO_NOJIT)) { /* JIT disabled for this proto? */ - if (J->parent == 0 && J->exitno == 0) { + if (J->parent == 0 && J->exitno == 0 && bc_op(*J->pc) != BC_ITERN) { /* Lazy bytecode patching to disable hotcount events. */ lj_assertJ(bc_op(*J->pc) == BC_FORL || bc_op(*J->pc) == BC_ITERL || bc_op(*J->pc) == BC_LOOP || bc_op(*J->pc) == BC_FUNCF, @@ -496,6 +501,7 @@ static void trace_stop(jit_State *J) J->cur.nextroot = pt->trace; pt->trace = (TraceNo1)traceno; break; + case BC_ITERN: case BC_RET: case BC_RET0: case BC_RET1: @@ -506,7 +512,11 @@ static void trace_stop(jit_State *J) lj_assertJ(J->parent != 0 && J->cur.root != 0, "not a side trace"); lj_asm_patchexit(J, traceref(J, J->parent), J->exitno, J->cur.mcode); /* Avoid compiling a side trace twice (stack resizing uses parent exit). */ - traceref(J, J->parent)->snap[J->exitno].count = SNAPCOUNT_DONE; + { + SnapShot *snap = &traceref(J, J->parent)->snap[J->exitno]; + snap->count = SNAPCOUNT_DONE; + if (J->cur.topslot > snap->topslot) snap->topslot = J->cur.topslot; + } /* Add to side trace chain in root trace. */ { GCtrace *root = traceref(J, J->cur.root); @@ -651,8 +661,13 @@ static TValue *trace_state(lua_State *L, lua_CFunction dummy, void *ud) J->state = LJ_TRACE_RECORD; /* trace_start() may change state. */ trace_start(J); lj_dispatch_update(J2G(J)); - break; + if (J->state != LJ_TRACE_RECORD_1ST) + break; + /* fallthrough */ + case LJ_TRACE_RECORD_1ST: + J->state = LJ_TRACE_RECORD; + /* fallthrough */ case LJ_TRACE_RECORD: trace_pendpatch(J, 0); setvmstate(J2G(J), RECORD); @@ -821,7 +836,7 @@ static void trace_exit_regs(lua_State *L, ExitState *ex) } #endif -#ifdef EXITSTATE_PCREG +#if defined(EXITSTATE_PCREG) || (LJ_UNWIND_JIT && !EXITTRACE_VMSTATE) /* Determine trace number from pc of exit instruction. */ static TraceNo trace_exit_find(jit_State *J, MCode *pc) { @@ -843,10 +858,18 @@ int LJ_FASTCALL lj_trace_exit(jit_State *J, void *exptr) lua_State *L = J->L; ExitState *ex = (ExitState *)exptr; ExitDataCP exd; - int errcode; + int errcode, exitcode = J->exitcode; + TValue exiterr; const BCIns *pc; void *cf; GCtrace *T; + + setnilV(&exiterr); + if (exitcode) { /* Trace unwound with error code. */ + J->exitcode = 0; + copyTV(L, &exiterr, L->top-1); + } + #ifdef EXITSTATE_PCREG J->parent = trace_exit_find(J, (MCode *)(intptr_t)ex->gpr[EXITSTATE_PCREG]); #endif @@ -866,6 +889,8 @@ int LJ_FASTCALL lj_trace_exit(jit_State *J, void *exptr) if (errcode) return -errcode; /* Return negated error code. */ + if (exitcode) copyTV(L, L->top++, &exiterr); /* Anchor the error object. */ + if (!(LJ_HASPROFILE && (G(L)->hookmask & HOOK_PROFILE))) lj_vmevent_send(L, TEXIT, lj_state_checkstack(L, 4+RID_NUM_GPR+RID_NUM_FPR+LUA_MINSTACK); @@ -877,7 +902,9 @@ int LJ_FASTCALL lj_trace_exit(jit_State *J, void *exptr) pc = exd.pc; cf = cframe_raw(L->cframe); setcframe_pc(cf, pc); - if (LJ_HASPROFILE && (G(L)->hookmask & HOOK_PROFILE)) { + if (exitcode) { + return -exitcode; + } else if (LJ_HASPROFILE && (G(L)->hookmask & HOOK_PROFILE)) { /* Just exit to interpreter. */ } else if (G(L)->gc.state == GCSatomic || G(L)->gc.state == GCSfinalize) { if (!(G(L)->hookmask & HOOK_GC)) @@ -887,13 +914,14 @@ int LJ_FASTCALL lj_trace_exit(jit_State *J, void *exptr) } if (bc_op(*pc) == BC_JLOOP) { BCIns *retpc = &traceref(J, bc_d(*pc))->startins; - if (bc_isret(bc_op(*retpc))) { + int isret = bc_isret(bc_op(*retpc)); + if (isret || bc_op(*retpc) == BC_ITERN) { if (J->state == LJ_TRACE_RECORD) { J->patchins = *pc; J->patchpc = (BCIns *)pc; *J->patchpc = *retpc; J->bcskip = 1; - } else { + } else if (isret) { pc = retpc; setcframe_pc(cf, pc); } @@ -915,4 +943,41 @@ int LJ_FASTCALL lj_trace_exit(jit_State *J, void *exptr) } } +#if LJ_UNWIND_JIT +/* Given an mcode address determine trace exit address for unwinding. */ +uintptr_t LJ_FASTCALL lj_trace_unwind(jit_State *J, uintptr_t addr, ExitNo *ep) +{ +#if EXITTRACE_VMSTATE + TraceNo traceno = J2G(J)->vmstate; +#else + TraceNo traceno = trace_exit_find(J, (MCode *)addr); +#endif + GCtrace *T = traceref(J, traceno); + if (T +#if EXITTRACE_VMSTATE + && addr >= (uintptr_t)T->mcode && addr < (uintptr_t)T->mcode + T->szmcode +#endif + ) { + SnapShot *snap = T->snap; + SnapNo lo = 0, exitno = T->nsnap; + uintptr_t ofs = (uintptr_t)((MCode *)addr - T->mcode); /* MCode units! */ + /* Rightmost binary search for mcode offset to determine exit number. */ + do { + SnapNo mid = (lo+exitno) >> 1; + if (ofs < snap[mid].mcofs) exitno = mid; else lo = mid + 1; + } while (lo < exitno); + exitno--; + *ep = exitno; +#ifdef EXITSTUBS_PER_GROUP + return (uintptr_t)exitstub_addr(J, exitno); +#else + return (uintptr_t)exitstub_trace_addr(T, exitno); +#endif + } + /* Cannot correlate addr with trace/exit. This will be fatal. */ + lj_assertJ(0, "bad exit pc"); + return 0; +} +#endif + #endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_trace.h b/love/src/jni/LuaJIT-2.1/src/lj_trace.h index 7e86d963..e4cf2dc4 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_trace.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_trace.h @@ -37,6 +37,9 @@ LJ_FUNC void lj_trace_ins(jit_State *J, const BCIns *pc); LJ_FUNCA void LJ_FASTCALL lj_trace_hot(jit_State *J, const BCIns *pc); LJ_FUNCA void LJ_FASTCALL lj_trace_stitch(jit_State *J, const BCIns *pc); LJ_FUNCA int LJ_FASTCALL lj_trace_exit(jit_State *J, void *exptr); +#if LJ_UNWIND_EXT +LJ_FUNC uintptr_t LJ_FASTCALL lj_trace_unwind(jit_State *J, uintptr_t addr, ExitNo *ep); +#endif /* Signal asynchronous abort of trace or end of trace. */ #define lj_trace_abort(g) (G2J(g)->state &= ~LJ_TRACE_ACTIVE) diff --git a/love/src/jni/LuaJIT-2.1/src/lj_udata.c b/love/src/jni/LuaJIT-2.1/src/lj_udata.c index ec3478c8..a0edd0df 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_udata.c +++ b/love/src/jni/LuaJIT-2.1/src/lj_udata.c @@ -8,6 +8,7 @@ #include "lj_obj.h" #include "lj_gc.h" +#include "lj_err.h" #include "lj_udata.h" GCudata *lj_udata_new(lua_State *L, MSize sz, GCtab *env) @@ -32,3 +33,30 @@ void LJ_FASTCALL lj_udata_free(global_State *g, GCudata *ud) lj_mem_free(g, ud, sizeudata(ud)); } +#if LJ_64 +void *lj_lightud_intern(lua_State *L, void *p) +{ + global_State *g = G(L); + uint64_t u = (uint64_t)p; + uint32_t up = lightudup(u); + uint32_t *segmap = mref(g->gc.lightudseg, uint32_t); + MSize segnum = g->gc.lightudnum; + if (segmap) { + MSize seg; + for (seg = 0; seg <= segnum; seg++) + if (segmap[seg] == up) /* Fast path. */ + return (void *)(((uint64_t)seg << LJ_LIGHTUD_BITS_LO) | lightudlo(u)); + segnum++; + /* Leave last segment unused to avoid clash with ITERN key. */ + if (segnum >= (1 << LJ_LIGHTUD_BITS_SEG)-1) lj_err_msg(L, LJ_ERR_BADLU); + } + if (!((segnum-1) & segnum) && segnum != 1) { + lj_mem_reallocvec(L, segmap, segnum, segnum ? 2*segnum : 2u, uint32_t); + setmref(g->gc.lightudseg, segmap); + } + g->gc.lightudnum = segnum; + segmap[segnum] = up; + return (void *)(((uint64_t)segnum << LJ_LIGHTUD_BITS_LO) | lightudlo(u)); +} +#endif + diff --git a/love/src/jni/LuaJIT-2.1/src/lj_udata.h b/love/src/jni/LuaJIT-2.1/src/lj_udata.h index 1873b694..78522ecc 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_udata.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_udata.h @@ -10,5 +10,8 @@ LJ_FUNC GCudata *lj_udata_new(lua_State *L, MSize sz, GCtab *env); LJ_FUNC void LJ_FASTCALL lj_udata_free(global_State *g, GCudata *ud); +#if LJ_64 +LJ_FUNC void * LJ_FASTCALL lj_lightud_intern(lua_State *L, void *p); +#endif #endif diff --git a/love/src/jni/LuaJIT-2.1/src/lj_vm.h b/love/src/jni/LuaJIT-2.1/src/lj_vm.h index 6358dc58..81ee8e28 100644 --- a/love/src/jni/LuaJIT-2.1/src/lj_vm.h +++ b/love/src/jni/LuaJIT-2.1/src/lj_vm.h @@ -26,6 +26,9 @@ LJ_ASMF void lj_vm_unwind_ff_eh(void); #if LJ_TARGET_X86ORX64 LJ_ASMF void lj_vm_unwind_rethrow(void); #endif +#if LJ_TARGET_MIPS +LJ_ASMF void lj_vm_unwind_stub(void); +#endif /* Miscellaneous functions. */ #if LJ_TARGET_X86ORX64 @@ -48,6 +51,7 @@ LJ_ASMF void lj_vm_inshook(void); LJ_ASMF void lj_vm_rethook(void); LJ_ASMF void lj_vm_callhook(void); LJ_ASMF void lj_vm_profhook(void); +LJ_ASMF void lj_vm_IITERN(void); /* Trace exit handling. */ LJ_ASMF void lj_vm_exit_handler(void); @@ -95,6 +99,7 @@ LJ_ASMF double lj_vm_trunc_sf(double); #if LJ_HASFFI LJ_ASMF int lj_vm_errno(void); #endif +LJ_ASMF TValue *lj_vm_next(GCtab *t, uint32_t idx); #endif /* Continuations for metamethods. */ diff --git a/love/src/jni/LuaJIT-2.1/src/ljamalg.c b/love/src/jni/LuaJIT-2.1/src/ljamalg.c index 8e2d4937..384b3cc1 100644 --- a/love/src/jni/LuaJIT-2.1/src/ljamalg.c +++ b/love/src/jni/LuaJIT-2.1/src/ljamalg.c @@ -39,6 +39,7 @@ #include "lj_strscan.c" #include "lj_strfmt.c" #include "lj_strfmt_num.c" +#include "lj_serialize.c" #include "lj_api.c" #include "lj_profile.c" #include "lj_lex.c" @@ -85,5 +86,6 @@ #include "lib_bit.c" #include "lib_jit.c" #include "lib_ffi.c" +#include "lib_buffer.c" #include "lib_init.c" diff --git a/love/src/jni/LuaJIT-2.1/src/lualib.h b/love/src/jni/LuaJIT-2.1/src/lualib.h index 9cd39880..5c18e9ec 100644 --- a/love/src/jni/LuaJIT-2.1/src/lualib.h +++ b/love/src/jni/LuaJIT-2.1/src/lualib.h @@ -33,6 +33,7 @@ LUALIB_API int luaopen_debug(lua_State *L); LUALIB_API int luaopen_bit(lua_State *L); LUALIB_API int luaopen_jit(lua_State *L); LUALIB_API int luaopen_ffi(lua_State *L); +LUALIB_API int luaopen_string_buffer(lua_State *L); LUALIB_API void luaL_openlibs(lua_State *L); diff --git a/love/src/jni/LuaJIT-2.1/src/msvcbuild.bat b/love/src/jni/LuaJIT-2.1/src/msvcbuild.bat index d0768948..7e1a6e04 100644 --- a/love/src/jni/LuaJIT-2.1/src/msvcbuild.bat +++ b/love/src/jni/LuaJIT-2.1/src/msvcbuild.bat @@ -25,7 +25,7 @@ @set LJDLLNAME=lua51.dll @set LJLIBNAME=lua51.lib @set BUILDTYPE=release -@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c +@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c lib_buffer.c %LJCOMPILE% host\minilua.c @if errorlevel 1 goto :BAD diff --git a/love/src/jni/LuaJIT-2.1/src/ps4build.bat b/love/src/jni/LuaJIT-2.1/src/ps4build.bat index e83c674a..fdd09d81 100644 --- a/love/src/jni/LuaJIT-2.1/src/ps4build.bat +++ b/love/src/jni/LuaJIT-2.1/src/ps4build.bat @@ -26,7 +26,7 @@ @set LJMT=mt /nologo @set DASMDIR=..\dynasm @set DASM=%DASMDIR%\dynasm.lua -@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c +@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c lib_buffer.c @set GC64= @set DASC=vm_x64.dasc diff --git a/love/src/jni/LuaJIT-2.1/src/psvitabuild.bat b/love/src/jni/LuaJIT-2.1/src/psvitabuild.bat index 3991dc65..2980e157 100644 --- a/love/src/jni/LuaJIT-2.1/src/psvitabuild.bat +++ b/love/src/jni/LuaJIT-2.1/src/psvitabuild.bat @@ -14,7 +14,7 @@ @set LJMT=mt /nologo @set DASMDIR=..\dynasm @set DASM=%DASMDIR%\dynasm.lua -@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c +@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c lib_buffer.c %LJCOMPILE% host\minilua.c @if errorlevel 1 goto :BAD diff --git a/love/src/jni/LuaJIT-2.1/src/vm_arm.dasc b/love/src/jni/LuaJIT-2.1/src/vm_arm.dasc index 5efda227..405be30a 100644 --- a/love/src/jni/LuaJIT-2.1/src/vm_arm.dasc +++ b/love/src/jni/LuaJIT-2.1/src/vm_arm.dasc @@ -539,13 +539,13 @@ static void build_subroutines(BuildCtx *ctx) | cmp CARG1, #1 |.endif | ldr PC, [CARG4, #-12] // Restore PC from [cont|PC]. - | ldr CARG3, LFUNC:CARG3->field_pc | mvn INS, #~LJ_TNIL | add CARG2, RA, RC | str INS, [CARG2, #-4] // Ensure one valid arg. |.if FFI | bls >1 |.endif + | ldr CARG3, LFUNC:CARG3->field_pc | ldr KBASE, [CARG3, #PC2PROTO(k)] | // BASE = base, RA = resultptr, CARG4 = meta base | bx CARG1 @@ -1111,24 +1111,18 @@ static void build_subroutines(BuildCtx *ctx) | checktab CARG2, ->fff_fallback | strd CARG34, [BASE, NARGS8:RC] // Set missing 2nd arg to nil. | ldr PC, [BASE, FRAME_PC] - | mov CARG2, CARG1 - | str BASE, L->base // Add frame since C call can throw. - | mov CARG1, L - | str BASE, L->top // Dummy frame length is ok. - | add CARG3, BASE, #8 - | str PC, SAVE_PC - | bl extern lj_tab_next // (lua_State *L, GCtab *t, TValue *key) - | // Returns 0 at end of traversal. + | add CARG2, BASE, #8 + | sub CARG3, BASE, #8 + | bl extern lj_tab_next // (GCtab *t, cTValue *key, TValue *o) + | // Returns 1=found, 0=end, -1=error. | .IOS ldr BASE, L->base | cmp CRET1, #0 - | mvneq CRET2, #~LJ_TNIL - | beq ->fff_restv // End of traversal: return nil. - | ldrd CARG12, [BASE, #8] // Copy key and value to results. - | ldrd CARG34, [BASE, #16] - | mov RC, #(2+1)*8 - | strd CARG12, [BASE, #-8] - | strd CARG34, [BASE] - | b ->fff_res + | mov RC, #(2+1)*8 + | bgt ->fff_res // Found key/value. + | bmi ->fff_fallback // Invalid key. + | // End of traversal: return nil. + | mvn CRET2, #~LJ_TNIL + | b ->fff_restv | |.ffunc_1 pairs | checktab CARG2, ->fff_fallback @@ -1810,7 +1804,7 @@ static void build_subroutines(BuildCtx *ctx) | str BASE, L->base | str PC, SAVE_PC | str L, SBUF:CARG1->L - | str CARG4, SBUF:CARG1->p + | str CARG4, SBUF:CARG1->w | bl extern lj_buf_putstr_ .. name | bl extern lj_buf_tostr | b ->fff_resstr @@ -2245,8 +2239,9 @@ static void build_subroutines(BuildCtx *ctx) | b <2 | |9: // Rethrow error from the right C frame. + | rsb CARG2, CARG1, #0 | mov CARG1, L - | bl extern lj_err_run // (lua_State *L) + | bl extern lj_err_trace // (lua_State *L, int errcode) |.endif | |//----------------------------------------------------------------------- @@ -2429,6 +2424,64 @@ static void build_subroutines(BuildCtx *ctx) |//-- Miscellaneous functions -------------------------------------------- |//----------------------------------------------------------------------- | + |.define NEXT_TAB, TAB:CARG1 + |.define NEXT_RES, CARG1 + |.define NEXT_IDX, CARG2 + |.define NEXT_TMP0, CARG3 + |.define NEXT_TMP1, CARG4 + |.define NEXT_LIM, r12 + |.define NEXT_RES_PTR, sp + |.define NEXT_RES_VAL, [sp] + |.define NEXT_RES_KEY_I, [sp, #8] + |.define NEXT_RES_KEY_IT, [sp, #12] + | + |// TValue *lj_vm_next(GCtab *t, uint32_t idx) + |// Next idx returned in CRET2. + |->vm_next: + |.if JIT + | ldr NEXT_TMP0, NEXT_TAB->array + | ldr NEXT_LIM, NEXT_TAB->asize + | add NEXT_TMP0, NEXT_TMP0, NEXT_IDX, lsl #3 + |1: // Traverse array part. + | subs NEXT_TMP1, NEXT_IDX, NEXT_LIM + | bhs >5 + | ldr NEXT_TMP1, [NEXT_TMP0, #4] + | str NEXT_IDX, NEXT_RES_KEY_I + | add NEXT_TMP0, NEXT_TMP0, #8 + | add NEXT_IDX, NEXT_IDX, #1 + | checktp NEXT_TMP1, LJ_TNIL + | beq <1 // Skip holes in array part. + | ldr NEXT_TMP0, [NEXT_TMP0, #-8] + | mov NEXT_RES, NEXT_RES_PTR + | strd NEXT_TMP0, NEXT_RES_VAL // Stores NEXT_TMP1, too. + | mvn NEXT_TMP0, #~LJ_TISNUM + | str NEXT_TMP0, NEXT_RES_KEY_IT + | bx lr + | + |5: // Traverse hash part. + | ldr NEXT_TMP0, NEXT_TAB->hmask + | ldr NODE:NEXT_RES, NEXT_TAB->node + | add NEXT_TMP1, NEXT_TMP1, NEXT_TMP1, lsl #1 + | add NEXT_LIM, NEXT_LIM, NEXT_TMP0 + | add NODE:NEXT_RES, NODE:NEXT_RES, NEXT_TMP1, lsl #3 + |6: + | cmp NEXT_IDX, NEXT_LIM + | bhi >9 + | ldr NEXT_TMP1, NODE:NEXT_RES->val.it + | checktp NEXT_TMP1, LJ_TNIL + | add NEXT_IDX, NEXT_IDX, #1 + | bxne lr + | // Skip holes in hash part. + | add NEXT_RES, NEXT_RES, #sizeof(Node) + | b <6 + | + |9: // End of iteration. Set the key to nil (not the value). + | mvn NEXT_TMP0, #0 + | mov NEXT_RES, NEXT_RES_PTR + | str NEXT_TMP0, NEXT_RES_KEY_IT + | bx lr + |.endif + | |//----------------------------------------------------------------------- |//-- FFI helper functions ----------------------------------------------- |//----------------------------------------------------------------------- @@ -3919,10 +3972,11 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) break; case BC_ITERN: - | // RA = base*8, (RB = nresults+1, RC = nargs+1 (2+1)) |.if JIT - | // NYI: add hotloop, record BC_ITERN. + | hotloop |.endif + |->vm_IITERN: + | // RA = base*8, (RB = nresults+1, RC = nargs+1 (2+1)) | add RA, BASE, RA | ldr TAB:RB, [RA, #-16] | ldr CARG1, [RA, #-8] // Get index from control var. @@ -3988,7 +4042,7 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | ins_next1 | ins_next2 | mov CARG1, #0 - | mvn CARG2, #0x00018000 + | mvn CARG2, #~LJ_KEYINDEX | strd CARG1, [RA, #-8] // Initialize control var. |1: | ins_next3 @@ -3997,9 +4051,25 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | mov OP, #BC_ITERC | strb CARG1, [PC, #-4] | sub PC, RC, #0x20000 + |.if JIT + | ldrb CARG1, [PC] + | cmp CARG1, #BC_ITERN + | bne >6 + |.endif | strb OP, [PC] // Subsumes ins_next1. | ins_next2 | b <1 + |.if JIT + |6: // Unpatch JLOOP. + | ldr CARG1, [DISPATCH, #DISPATCH_J(trace)] + | ldrh CARG2, [PC, #2] + | ldr TRACE:CARG1, [CARG1, CARG2, lsl #2] + | // Subsumes ins_next1 and ins_next2. + | ldr INS, TRACE:CARG1->startins + | bfi INS, OP, #0, #8 + | str INS, [PC], #4 + | b <1 + |.endif break; case BC_VARG: diff --git a/love/src/jni/LuaJIT-2.1/src/vm_arm64.dasc b/love/src/jni/LuaJIT-2.1/src/vm_arm64.dasc index 755d9a64..eb87857f 100644 --- a/love/src/jni/LuaJIT-2.1/src/vm_arm64.dasc +++ b/love/src/jni/LuaJIT-2.1/src/vm_arm64.dasc @@ -81,47 +81,48 @@ | |.define CFRAME_SPACE, 208 |//----- 16 byte aligned, <-- sp entering interpreter -|// Unused [sp, #204] // 32 bit values -|.define SAVE_NRES, [sp, #200] -|.define SAVE_ERRF, [sp, #196] -|.define SAVE_MULTRES, [sp, #192] -|.define TMPD, [sp, #184] // 64 bit values -|.define SAVE_L, [sp, #176] -|.define SAVE_PC, [sp, #168] -|.define SAVE_CFRAME, [sp, #160] -|.define SAVE_FPR_, 96 // 96+8*8: 64 bit FPR saves -|.define SAVE_GPR_, 16 // 16+10*8: 64 bit GPR saves -|.define SAVE_LR, [sp, #8] -|.define SAVE_FP, [sp] +|.define SAVE_FP_LR_, 192 +|.define SAVE_GPR_, 112 // 112+10*8: 64 bit GPR saves +|.define SAVE_FPR_, 48 // 48+8*8: 64 bit FPR saves +|// Unused [sp, #44] // 32 bit values +|.define SAVE_NRES, [sp, #40] +|.define SAVE_ERRF, [sp, #36] +|.define SAVE_MULTRES, [sp, #32] +|.define TMPD, [sp, #24] // 64 bit values +|.define SAVE_L, [sp, #16] +|.define SAVE_PC, [sp, #8] +|.define SAVE_CFRAME, [sp, #0] |//----- 16 byte aligned, <-- sp while in interpreter. | -|.define TMPDofs, #184 +|.define TMPDofs, #24 | |.macro save_, gpr1, gpr2, fpr1, fpr2 -| stp d..fpr1, d..fpr2, [sp, # SAVE_FPR_+(fpr1-8)*8] -| stp x..gpr1, x..gpr2, [sp, # SAVE_GPR_+(gpr1-19)*8] +| stp d..fpr2, d..fpr1, [sp, # SAVE_FPR_+(14-fpr1)*8] +| stp x..gpr2, x..gpr1, [sp, # SAVE_GPR_+(27-gpr1)*8] |.endmacro |.macro rest_, gpr1, gpr2, fpr1, fpr2 -| ldp d..fpr1, d..fpr2, [sp, # SAVE_FPR_+(fpr1-8)*8] -| ldp x..gpr1, x..gpr2, [sp, # SAVE_GPR_+(gpr1-19)*8] +| ldp d..fpr2, d..fpr1, [sp, # SAVE_FPR_+(14-fpr1)*8] +| ldp x..gpr2, x..gpr1, [sp, # SAVE_GPR_+(27-gpr1)*8] |.endmacro | |.macro saveregs -| stp fp, lr, [sp, #-CFRAME_SPACE]! -| add fp, sp, #0 -| stp x19, x20, [sp, # SAVE_GPR_] +| sub sp, sp, # CFRAME_SPACE +| stp fp, lr, [sp, # SAVE_FP_LR_] +| add fp, sp, # SAVE_FP_LR_ +| stp x20, x19, [sp, # SAVE_GPR_+(27-19)*8] | save_ 21, 22, 8, 9 | save_ 23, 24, 10, 11 | save_ 25, 26, 12, 13 | save_ 27, 28, 14, 15 |.endmacro |.macro restoreregs -| ldp x19, x20, [sp, # SAVE_GPR_] +| ldp x20, x19, [sp, # SAVE_GPR_+(27-19)*8] | rest_ 21, 22, 8, 9 | rest_ 23, 24, 10, 11 | rest_ 25, 26, 12, 13 | rest_ 27, 28, 14, 15 -| ldp fp, lr, [sp], # CFRAME_SPACE +| ldp fp, lr, [sp, # SAVE_FP_LR_] +| add sp, sp, # CFRAME_SPACE |.endmacro | |// Type definitions. Some of these are only used for documentation. @@ -500,8 +501,9 @@ static void build_subroutines(BuildCtx *ctx) | ldr GL, L->glref // Setup pointer to global state. | mov BASE, CARG2 | str CARG1, SAVE_PC // Any value outside of bytecode is ok. - | str RC, SAVE_CFRAME - | str fp, L->cframe // Add our C frame to cframe chain. + | add TMP0, sp, #0 + | str RC, SAVE_CFRAME + | str TMP0, L->cframe // Add our C frame to cframe chain. | |3: // Entry point for vm_cpcall/vm_resume (BASE = base, PC = ftype). | str L, GL->cur_L @@ -536,8 +538,9 @@ static void build_subroutines(BuildCtx *ctx) | sub RA, RA, RB // Compute -savestack(L, L->top). | str RAw, SAVE_NRES // Neg. delta means cframe w/o frame. | str wzr, SAVE_ERRF // No error function. - | str RC, SAVE_CFRAME - | str fp, L->cframe // Add our C frame to cframe chain. + | add TMP0, sp, #0 + | str RC, SAVE_CFRAME + | str TMP0, L->cframe // Add our C frame to cframe chain. | str L, GL->cur_L | blr CARG4 // (lua_State *L, lua_CFunction func, void *ud) | mov BASE, CRET1 @@ -562,12 +565,12 @@ static void build_subroutines(BuildCtx *ctx) | cmp CARG1, #1 |.endif | ldr PC, [CARG4, #-24] // Restore PC from [cont|PC]. - | ldr CARG3, LFUNC:CARG3->pc | add TMP0, RA, RC | str TISNIL, [TMP0, #-8] // Ensure one valid arg. |.if FFI | bls >1 |.endif + | ldr CARG3, LFUNC:CARG3->pc | ldr KBASE, [CARG3, #PC2PROTO(k)] | // BASE = base, RA = resultptr, CARG4 = meta base | br CARG1 @@ -1086,21 +1089,19 @@ static void build_subroutines(BuildCtx *ctx) |//-- Base library: iterators ------------------------------------------- | |.ffunc_1 next - | checktp CARG2, CARG1, LJ_TTAB, ->fff_fallback + | checktp CARG1, LJ_TTAB, ->fff_fallback | str TISNIL, [BASE, NARGS8:RC] // Set missing 2nd arg to nil. | ldr PC, [BASE, FRAME_PC] - | stp BASE, BASE, L->base // Add frame since C call can throw. - | mov CARG1, L - | add CARG3, BASE, #8 - | str PC, SAVE_PC - | bl extern lj_tab_next // (lua_State *L, GCtab *t, TValue *key) - | // Returns 0 at end of traversal. + | add CARG2, BASE, #8 + | sub CARG3, BASE, #16 + | bl extern lj_tab_next // (GCtab *t, cTValue *key, TValue *o) + | // Returns 1=found, 0=end, -1=error. + | mov RC, #(2+1)*8 + | tbnz CRET1w, #31, ->fff_fallback // Invalid key. + | cbnz CRET1, ->fff_res // Found key/value. + | // End of traversal: return nil. | str TISNIL, [BASE, #-16] - | cbz CRET1, ->fff_res1 // End of traversal: return nil. - | ldp CARG1, CARG2, [BASE, #8] // Copy key and value to results. - | mov RC, #(2+1)*8 - | stp CARG1, CARG2, [BASE, #-16] - | b ->fff_res + | b ->fff_res1 | |.ffunc_1 pairs | checktp TMP1, CARG1, LJ_TTAB, ->fff_fallback @@ -1163,9 +1164,10 @@ static void build_subroutines(BuildCtx *ctx) |//-- Base library: catch errors ---------------------------------------- | |.ffunc pcall + | cmp NARGS8:RC, #8 | ldrb TMP0w, GL->hookmask - | subs NARGS8:RC, NARGS8:RC, #8 | blo ->fff_fallback + | sub NARGS8:RC, NARGS8:RC, #8 | mov RB, BASE | add BASE, BASE, #16 | ubfx TMP0w, TMP0w, #HOOK_ACTIVE_SHIFT, #1 @@ -1589,7 +1591,7 @@ static void build_subroutines(BuildCtx *ctx) | str BASE, L->base | str PC, SAVE_PC | str L, GL->tmpbuf.L - | str TMP0, GL->tmpbuf.p + | str TMP0, GL->tmpbuf.w | bl extern lj_buf_putstr_ .. name | bl extern lj_buf_tostr | b ->fff_resstr @@ -2035,8 +2037,9 @@ static void build_subroutines(BuildCtx *ctx) | b <2 | |9: // Rethrow error from the right C frame. + | neg CARG2w, CARG1w | mov CARG1, L - | bl extern lj_err_run // (lua_State *L) + | bl extern lj_err_trace // (lua_State *L, int errcode) |.endif | |//----------------------------------------------------------------------- @@ -2065,6 +2068,63 @@ static void build_subroutines(BuildCtx *ctx) |//-- Miscellaneous functions -------------------------------------------- |//----------------------------------------------------------------------- | + |.define NEXT_TAB, TAB:CARG1 + |.define NEXT_RES, CARG1 + |.define NEXT_IDX, CARG2w + |.define NEXT_LIM, CARG3w + |.define NEXT_TMP0, TMP0 + |.define NEXT_TMP0w, TMP0w + |.define NEXT_TMP1, TMP1 + |.define NEXT_TMP1w, TMP1w + |.define NEXT_RES_PTR, sp + |.define NEXT_RES_VAL, [sp] + |.define NEXT_RES_KEY, [sp, #8] + | + |// TValue *lj_vm_next(GCtab *t, uint32_t idx) + |// Next idx returned in CRET2w. + |->vm_next: + |.if JIT + | ldr NEXT_LIM, NEXT_TAB->asize + | ldr NEXT_TMP1, NEXT_TAB->array + |1: // Traverse array part. + | subs NEXT_TMP0w, NEXT_IDX, NEXT_LIM + | bhs >5 // Index points after array part? + | ldr NEXT_TMP0, [NEXT_TMP1, NEXT_IDX, uxtw #3] + | cmn NEXT_TMP0, #-LJ_TNIL + | cinc NEXT_IDX, NEXT_IDX, eq + | beq <1 // Skip holes in array part. + | str NEXT_TMP0, NEXT_RES_VAL + | movz NEXT_TMP0w, #(LJ_TISNUM>>1)&0xffff, lsl #16 + | stp NEXT_IDX, NEXT_TMP0w, NEXT_RES_KEY + | add NEXT_IDX, NEXT_IDX, #1 + | mov NEXT_RES, NEXT_RES_PTR + |4: + | ret + | + |5: // Traverse hash part. + | ldr NEXT_TMP1w, NEXT_TAB->hmask + | ldr NODE:NEXT_RES, NEXT_TAB->node + | add NEXT_TMP0w, NEXT_TMP0w, NEXT_TMP0w, lsl #1 + | add NEXT_LIM, NEXT_LIM, NEXT_TMP1w + | add NODE:NEXT_RES, NODE:NEXT_RES, NEXT_TMP0w, uxtw #3 + |6: + | cmp NEXT_IDX, NEXT_LIM + | bhi >9 + | ldr NEXT_TMP0, NODE:NEXT_RES->val + | cmn NEXT_TMP0, #-LJ_TNIL + | add NEXT_IDX, NEXT_IDX, #1 + | bne <4 + | // Skip holes in hash part. + | add NODE:NEXT_RES, NODE:NEXT_RES, #sizeof(Node) + | b <6 + | + |9: // End of iteration. Set the key to nil (not the value). + | movn NEXT_TMP0, #0 + | str NEXT_TMP0, NEXT_RES_KEY + | mov NEXT_RES, NEXT_RES_PTR + | ret + |.endif + | |//----------------------------------------------------------------------- |//-- FFI helper functions ----------------------------------------------- |//----------------------------------------------------------------------- @@ -2122,16 +2182,16 @@ static void build_subroutines(BuildCtx *ctx) | // Caveat: needs special frame unwinding, see below. |.if FFI | .type CCSTATE, CCallState, x19 - | stp fp, lr, [sp, #-32]! - | add fp, sp, #0 - | str CCSTATE, [sp, #16] + | stp x20, CCSTATE, [sp, #-32]! + | stp fp, lr, [sp, #16] + | add fp, sp, #16 | mov CCSTATE, x0 | ldr TMP0w, CCSTATE:x0->spadj | ldrb TMP1w, CCSTATE->nsp | add TMP2, CCSTATE, #offsetof(CCallState, stack) | subs TMP1, TMP1, #1 | ldr TMP3, CCSTATE->func - | sub sp, fp, TMP0 + | sub sp, sp, TMP0 | bmi >2 |1: // Copy stack slots | ldr TMP0, [TMP2, TMP1, lsl #3] @@ -2149,12 +2209,12 @@ static void build_subroutines(BuildCtx *ctx) | ldp d6, d7, CCSTATE->fpr[6] | ldr x8, CCSTATE->retp | blr TMP3 - | mov sp, fp + | sub sp, fp, #16 | stp x0, x1, CCSTATE->gpr[0] | stp d0, d1, CCSTATE->fpr[0] | stp d2, d3, CCSTATE->fpr[2] - | ldr CCSTATE, [sp, #16] - | ldp fp, lr, [sp], #32 + | ldp fp, lr, [sp, #16] + | ldp x20, CCSTATE, [sp], #32 | ret |.endif |// Note: vm_ffi_call must be the last function in this object file! @@ -2884,7 +2944,7 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) case BC_GGET: | // RA = dst, RC = str_const (~) case BC_GSET: - | // RA = dst, RC = str_const (~) + | // RA = src, RC = str_const (~) | ldr LFUNC:CARG1, [BASE, FRAME_FUNC] | mvn RC, RC | and LFUNC:CARG1, CARG1, #LJ_GCVMASK @@ -3321,10 +3381,11 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) break; case BC_ITERN: - | // RA = base, (RB = nresults+1, RC = nargs+1 (2+1)) |.if JIT - | // NYI: add hotloop, record BC_ITERN. + | hotloop |.endif + |->vm_IITERN: + | // RA = base, (RB = nresults+1, RC = nargs+1 (2+1)) | add RA, BASE, RA, lsl #3 | ldr TAB:RB, [RA, #-16] | ldrh TMP3w, [PC, # OFS_RD] @@ -3383,7 +3444,7 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | ccmp CARG4, TISNIL, #0, eq | ccmp TMP1w, #FF_next_N, #0, eq | bne >5 - | mov TMP0w, #0xfffe7fff + | mov TMP0w, #0xfffe7fff // LJ_KEYINDEX | lsl TMP0, TMP0, #32 | str TMP0, [RA, #-8] // Initialize control var. |1: @@ -3391,11 +3452,28 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | ins_next | |5: // Despecialize bytecode if any of the checks fail. + |.if JIT + | ldrb TMP2w, [RC, # OFS_OP] + |.endif | mov TMP0, #BC_JMP | mov TMP1, #BC_ITERC | strb TMP0w, [PC, #-4+OFS_OP] + |.if JIT + | cmp TMP2w, #BC_ITERN + | bne >6 + |.endif | strb TMP1w, [RC, # OFS_OP] | b <1 + |.if JIT + |6: // Unpatch JLOOP. + | ldr RA, [GL, #GL_J(trace)] + | ldrh TMP2w, [RC, # OFS_RD] + | ldr TRACE:RA, [RA, TMP2, lsl #3] + | ldr TMP2w, TRACE:RA->startins + | bfxil TMP2w, TMP1w, #0, #8 + | str TMP2w, [RC] + | b <1 + |.endif break; case BC_VARG: @@ -3860,7 +3938,7 @@ static int build_backend(BuildCtx *ctx) static void emit_asm_debug(BuildCtx *ctx) { int fcofs = (int)((uint8_t *)ctx->glob[GLOB_vm_ffi_call] - ctx->code); - int i, cf = CFRAME_SIZE >> 3; + int i; switch (ctx->mode) { case BUILD_elfasm: fprintf(ctx->fp, "\t.section .debug_frame,\"\",%%progbits\n"); @@ -3874,7 +3952,7 @@ static void emit_asm_debug(BuildCtx *ctx) "\t.uleb128 0x1\n" "\t.sleb128 -8\n" "\t.byte 30\n" /* Return address is in lr. */ - "\t.byte 0xc\n\t.uleb128 31\n\t.uleb128 0\n" /* def_cfa sp */ + "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 16\n" /* def_cfa fp 16 */ "\t.align 3\n" ".LECIE0:\n\n"); fprintf(ctx->fp, @@ -3884,15 +3962,14 @@ static void emit_asm_debug(BuildCtx *ctx) "\t.long .Lframe0\n" "\t.quad .Lbegin\n" "\t.quad %d\n" - "\t.byte 0xe\n\t.uleb128 %d\n" /* def_cfa_offset */ - "\t.byte 0x9d\n\t.uleb128 %d\n" /* offset fp */ - "\t.byte 0x9e\n\t.uleb128 %d\n", /* offset lr */ - fcofs, CFRAME_SIZE, cf, cf-1); + "\t.byte 0x9e\n\t.uleb128 1\n" /* offset lr */ + "\t.byte 0x9d\n\t.uleb128 2\n", /* offset fp */ + fcofs); for (i = 19; i <= 28; i++) /* offset x19-x28 */ - fprintf(ctx->fp, "\t.byte 0x%x\n\t.uleb128 %d\n", 0x80+i, cf-i+17); + fprintf(ctx->fp, "\t.byte 0x%x\n\t.uleb128 %d\n", 0x80+i, i+(3-19)); for (i = 8; i <= 15; i++) /* offset d8-d15 */ fprintf(ctx->fp, "\t.byte 5\n\t.uleb128 0x%x\n\t.uleb128 %d\n", - 64+i, cf-i-4); + 64+i, i+(3+(28-19+1)-8)); fprintf(ctx->fp, "\t.align 3\n" ".LEFDE0:\n\n"); @@ -3904,10 +3981,10 @@ static void emit_asm_debug(BuildCtx *ctx) "\t.long .Lframe0\n" "\t.quad lj_vm_ffi_call\n" "\t.quad %d\n" - "\t.byte 0xe\n\t.uleb128 32\n" /* def_cfa_offset */ - "\t.byte 0x9d\n\t.uleb128 4\n" /* offset fp */ - "\t.byte 0x9e\n\t.uleb128 3\n" /* offset lr */ - "\t.byte 0x93\n\t.uleb128 2\n" /* offset x19 */ + "\t.byte 0x9e\n\t.uleb128 1\n" /* offset lr */ + "\t.byte 0x9d\n\t.uleb128 2\n" /* offset fp */ + "\t.byte 0x93\n\t.uleb128 3\n" /* offset x19 */ + "\t.byte 0x94\n\t.uleb128 4\n" /* offset x20 */ "\t.align 3\n" ".LEFDE1:\n\n", (int)ctx->codesz - fcofs); #endif @@ -3926,7 +4003,7 @@ static void emit_asm_debug(BuildCtx *ctx) "\t.byte 0x1b\n" /* pcrel|sdata4 */ "\t.long lj_err_unwind_dwarf-.\n" "\t.byte 0x1b\n" /* pcrel|sdata4 */ - "\t.byte 0xc\n\t.uleb128 31\n\t.uleb128 0\n" /* def_cfa sp */ + "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 16\n" /* def_cfa fp 16 */ "\t.align 3\n" ".LECIE1:\n\n"); fprintf(ctx->fp, @@ -3937,15 +4014,14 @@ static void emit_asm_debug(BuildCtx *ctx) "\t.long .Lbegin-.\n" "\t.long %d\n" "\t.uleb128 0\n" /* augmentation length */ - "\t.byte 0xe\n\t.uleb128 %d\n" /* def_cfa_offset */ - "\t.byte 0x9d\n\t.uleb128 %d\n" /* offset fp */ - "\t.byte 0x9e\n\t.uleb128 %d\n", /* offset lr */ - fcofs, CFRAME_SIZE, cf, cf-1); + "\t.byte 0x9e\n\t.uleb128 1\n" /* offset lr */ + "\t.byte 0x9d\n\t.uleb128 2\n", /* offset fp */ + fcofs); for (i = 19; i <= 28; i++) /* offset x19-x28 */ - fprintf(ctx->fp, "\t.byte 0x%x\n\t.uleb128 %d\n", 0x80+i, cf-i+17); + fprintf(ctx->fp, "\t.byte 0x%x\n\t.uleb128 %d\n", 0x80+i, i+(3-19)); for (i = 8; i <= 15; i++) /* offset d8-d15 */ fprintf(ctx->fp, "\t.byte 5\n\t.uleb128 0x%x\n\t.uleb128 %d\n", - 64+i, cf-i-4); + 64+i, i+(3+(28-19+1)-8)); fprintf(ctx->fp, "\t.align 3\n" ".LEFDE2:\n\n"); @@ -3962,7 +4038,7 @@ static void emit_asm_debug(BuildCtx *ctx) "\t.byte 30\n" /* Return address is in lr. */ "\t.uleb128 1\n" /* augmentation length */ "\t.byte 0x1b\n" /* pcrel|sdata4 */ - "\t.byte 0xc\n\t.uleb128 31\n\t.uleb128 0\n" /* def_cfa sp */ + "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 16\n" /* def_cfa fp 16 */ "\t.align 3\n" ".LECIE2:\n\n"); fprintf(ctx->fp, @@ -3973,14 +4049,106 @@ static void emit_asm_debug(BuildCtx *ctx) "\t.long lj_vm_ffi_call-.\n" "\t.long %d\n" "\t.uleb128 0\n" /* augmentation length */ - "\t.byte 0xe\n\t.uleb128 32\n" /* def_cfa_offset */ - "\t.byte 0x9d\n\t.uleb128 4\n" /* offset fp */ - "\t.byte 0x9e\n\t.uleb128 3\n" /* offset lr */ - "\t.byte 0x93\n\t.uleb128 2\n" /* offset x19 */ + "\t.byte 0x9e\n\t.uleb128 1\n" /* offset lr */ + "\t.byte 0x9d\n\t.uleb128 2\n" /* offset fp */ + "\t.byte 0x93\n\t.uleb128 3\n" /* offset x19 */ + "\t.byte 0x94\n\t.uleb128 4\n" /* offset x20 */ "\t.align 3\n" ".LEFDE3:\n\n", (int)ctx->codesz - fcofs); #endif break; +#if !LJ_NO_UNWIND + case BUILD_machasm: { +#if LJ_HASFFI + int fcsize = 0; +#endif + int j; + fprintf(ctx->fp, "\t.section __TEXT,__eh_frame,coalesced,no_toc+strip_static_syms+live_support\n"); + fprintf(ctx->fp, + "EH_frame1:\n" + "\t.set L$set$x,LECIEX-LSCIEX\n" + "\t.long L$set$x\n" + "LSCIEX:\n" + "\t.long 0\n" + "\t.byte 0x1\n" + "\t.ascii \"zPR\\0\"\n" + "\t.uleb128 0x1\n" + "\t.sleb128 -8\n" + "\t.byte 30\n" /* Return address is in lr. */ + "\t.uleb128 6\n" /* augmentation length */ + "\t.byte 0x9b\n" /* indirect|pcrel|sdata4 */ + "\t.long _lj_err_unwind_dwarf@GOT-.\n" + "\t.byte 0x1b\n" /* pcrel|sdata4 */ + "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 16\n" /* def_cfa fp 16 */ + "\t.align 3\n" + "LECIEX:\n\n"); + for (j = 0; j < ctx->nsym; j++) { + const char *name = ctx->sym[j].name; + int32_t size = ctx->sym[j+1].ofs - ctx->sym[j].ofs; + if (size == 0) continue; +#if LJ_HASFFI + if (!strcmp(name, "_lj_vm_ffi_call")) { fcsize = size; continue; } +#endif + fprintf(ctx->fp, + "LSFDE%d:\n" + "\t.set L$set$%d,LEFDE%d-LASFDE%d\n" + "\t.long L$set$%d\n" + "LASFDE%d:\n" + "\t.long LASFDE%d-EH_frame1\n" + "\t.long %s-.\n" + "\t.long %d\n" + "\t.uleb128 0\n" /* augmentation length */ + "\t.byte 0x9e\n\t.uleb128 1\n" /* offset lr */ + "\t.byte 0x9d\n\t.uleb128 2\n", /* offset fp */ + j, j, j, j, j, j, j, name, size); + for (i = 19; i <= 28; i++) /* offset x19-x28 */ + fprintf(ctx->fp, "\t.byte 0x%x\n\t.uleb128 %d\n", 0x80+i, i+(3-19)); + for (i = 8; i <= 15; i++) /* offset d8-d15 */ + fprintf(ctx->fp, "\t.byte 5\n\t.uleb128 0x%x\n\t.uleb128 %d\n", + 64+i, i+(3+(28-19+1)-8)); + fprintf(ctx->fp, + "\t.align 3\n" + "LEFDE%d:\n\n", j); + } +#if LJ_HASFFI + if (fcsize) { + fprintf(ctx->fp, + "EH_frame2:\n" + "\t.set L$set$y,LECIEY-LSCIEY\n" + "\t.long L$set$y\n" + "LSCIEY:\n" + "\t.long 0\n" + "\t.byte 0x1\n" + "\t.ascii \"zR\\0\"\n" + "\t.uleb128 0x1\n" + "\t.sleb128 -8\n" + "\t.byte 30\n" /* Return address is in lr. */ + "\t.uleb128 1\n" /* augmentation length */ + "\t.byte 0x1b\n" /* pcrel|sdata4 */ + "\t.byte 0xc\n\t.uleb128 29\n\t.uleb128 16\n" /* def_cfa fp 16 */ + "\t.align 3\n" + "LECIEY:\n\n"); + fprintf(ctx->fp, + "LSFDEY:\n" + "\t.set L$set$yy,LEFDEY-LASFDEY\n" + "\t.long L$set$yy\n" + "LASFDEY:\n" + "\t.long LASFDEY-EH_frame2\n" + "\t.long _lj_vm_ffi_call-.\n" + "\t.long %d\n" + "\t.uleb128 0\n" /* augmentation length */ + "\t.byte 0x9e\n\t.uleb128 1\n" /* offset lr */ + "\t.byte 0x9d\n\t.uleb128 2\n" /* offset fp */ + "\t.byte 0x93\n\t.uleb128 3\n" /* offset x19 */ + "\t.byte 0x94\n\t.uleb128 4\n" /* offset x20 */ + "\t.align 3\n" + "LEFDEY:\n\n", fcsize); + } +#endif + fprintf(ctx->fp, ".subsections_via_symbols\n"); + } + break; +#endif default: break; } diff --git a/love/src/jni/LuaJIT-2.1/src/vm_mips.dasc b/love/src/jni/LuaJIT-2.1/src/vm_mips.dasc index 7a6a83be..e3cc42a5 100644 --- a/love/src/jni/LuaJIT-2.1/src/vm_mips.dasc +++ b/love/src/jni/LuaJIT-2.1/src/vm_mips.dasc @@ -190,7 +190,7 @@ |//----------------------------------------------------------------------- | |// Trap for not-yet-implemented parts. -|.macro NYI; .long 0xf0f0f0f0; .endmacro +|.macro NYI; .long 0xec1cf0f0; .endmacro | |// Macros to mark delay slots. |.macro ., a; a; .endmacro @@ -501,6 +501,10 @@ static void build_subroutines(BuildCtx *ctx) | b ->vm_returnc |. li RD, 16 // 2 results: false + error message. | + |->vm_unwind_stub: // Jump to exit stub from unwinder. + | jr CARG1 + |. move ra, CARG2 + | |//----------------------------------------------------------------------- |//-- Grow stack for calls ----------------------------------------------- |//----------------------------------------------------------------------- @@ -669,11 +673,11 @@ static void build_subroutines(BuildCtx *ctx) |.endif | lw PC, -16+HI(RB) // Restore PC from [cont|PC]. | addu TMP2, RA, RD - | lw TMP1, LFUNC:TMP1->pc |.if FFI | bnez AT, >1 |.endif |. sw TISNIL, -8+HI(TMP2) // Ensure one valid arg. + | lw TMP1, LFUNC:TMP1->pc | // BASE = base, RA = resultptr, RB = meta base | jr TMP0 // Jump to continuation. |. lw KBASE, PC2PROTO(k)(TMP1) @@ -1258,35 +1262,27 @@ static void build_subroutines(BuildCtx *ctx) |//-- Base library: iterators ------------------------------------------- | |.ffunc next - | lw CARG1, HI(BASE) - | lw TAB:CARG2, LO(BASE) + | lw CARG2, HI(BASE) + | lw TAB:CARG1, LO(BASE) | beqz NARGS8:RC, ->fff_fallback |. addu TMP2, BASE, NARGS8:RC | li AT, LJ_TTAB | sw TISNIL, HI(TMP2) // Set missing 2nd arg to nil. - | bne CARG1, AT, ->fff_fallback + | bne CARG2, AT, ->fff_fallback |. lw PC, FRAME_PC(BASE) | load_got lj_tab_next - | sw BASE, L->base // Add frame since C call can throw. - | sw BASE, L->top // Dummy frame length is ok. - | addiu CARG3, BASE, 8 - | sw PC, SAVE_PC - | call_intern lj_tab_next // (lua_State *L, GCtab *t, TValue *key) - |. move CARG1, L - | // Returns 0 at end of traversal. + | addiu CARG2, BASE, 8 + | call_intern lj_tab_next // (GCtab *t, cTValue *key, TValue *o) + |. addiu CARG3, BASE, -8 + | // Returns 1=found, 0=end, -1=error. + | addiu RA, BASE, -8 + | bgtz CRET1, ->fff_res // Found key/value. + |. li RD, (2+1)*8 | beqz CRET1, ->fff_restv // End of traversal: return nil. |. li SFARG1HI, LJ_TNIL - | lw TMP0, 8+HI(BASE) - | lw TMP1, 8+LO(BASE) - | addiu RA, BASE, -8 - | lw TMP2, 16+HI(BASE) - | lw TMP3, 16+LO(BASE) - | sw TMP0, HI(RA) - | sw TMP1, LO(RA) - | sw TMP2, 8+HI(RA) - | sw TMP3, 8+LO(RA) - | b ->fff_res - |. li RD, (2+1)*8 + | lw CFUNC:RB, FRAME_FUNC(BASE) + | b ->fff_fallback // Invalid key. + |. li RC, 2*8 | |.ffunc_1 pairs | li AT, LJ_TTAB @@ -1967,7 +1963,7 @@ static void build_subroutines(BuildCtx *ctx) | lw TMP0, SBUF:CARG1->b | sw L, SBUF:CARG1->L | sw BASE, L->base - | sw TMP0, SBUF:CARG1->p + | sw TMP0, SBUF:CARG1->w | call_intern extern lj_buf_putstr_ .. name |. sw PC, SAVE_PC | load_got lj_buf_tostr @@ -2520,8 +2516,9 @@ static void build_subroutines(BuildCtx *ctx) |. addu RA, RA, BASE | |9: // Rethrow error from the right C frame. - | load_got lj_err_run - | call_intern lj_err_run // (lua_State *L) + | load_got lj_err_trace + | sub CARG2, r0, CRET1 + | call_intern lj_err_trace // (lua_State *L, int errcode) |. move CARG1, L |.endif | @@ -2801,6 +2798,73 @@ static void build_subroutines(BuildCtx *ctx) |//-- Miscellaneous functions -------------------------------------------- |//----------------------------------------------------------------------- | + |.define NEXT_TAB, TAB:CARG1 + |.define NEXT_IDX, CARG2 + |.define NEXT_ASIZE, CARG3 + |.define NEXT_NIL, CARG4 + |.define NEXT_TMP0, r12 + |.define NEXT_TMP1, r13 + |.define NEXT_TMP2, r14 + |.define NEXT_RES_VK, CRET1 + |.define NEXT_RES_IDX, CRET2 + |.define NEXT_RES_PTR, sp + |.define NEXT_RES_VAL_I, 0(sp) + |.define NEXT_RES_VAL_IT, 4(sp) + |.define NEXT_RES_KEY_I, 8(sp) + |.define NEXT_RES_KEY_IT, 12(sp) + | + |// TValue *lj_vm_next(GCtab *t, uint32_t idx) + |// Next idx returned in CRET2. + |->vm_next: + |.if JIT and ENDIAN_LE + | lw NEXT_ASIZE, NEXT_TAB->asize + | lw NEXT_TMP0, NEXT_TAB->array + | li NEXT_NIL, LJ_TNIL + |1: // Traverse array part. + | sltu AT, NEXT_IDX, NEXT_ASIZE + | sll NEXT_TMP1, NEXT_IDX, 3 + | beqz AT, >5 + |. addu NEXT_TMP1, NEXT_TMP0, NEXT_TMP1 + | lw NEXT_TMP2, 4(NEXT_TMP1) + | sw NEXT_IDX, NEXT_RES_KEY_I + | beq NEXT_TMP2, NEXT_NIL, <1 + |. addiu NEXT_IDX, NEXT_IDX, 1 + | lw NEXT_TMP0, 0(NEXT_TMP1) + | li AT, LJ_TISNUM + | sw NEXT_TMP2, NEXT_RES_VAL_IT + | sw AT, NEXT_RES_KEY_IT + | sw NEXT_TMP0, NEXT_RES_VAL_I + | move NEXT_RES_VK, NEXT_RES_PTR + | jr ra + |. move NEXT_RES_IDX, NEXT_IDX + | + |5: // Traverse hash part. + | subu NEXT_RES_IDX, NEXT_IDX, NEXT_ASIZE + | lw NODE:NEXT_RES_VK, NEXT_TAB->node + | sll NEXT_TMP2, NEXT_RES_IDX, 5 + | lw NEXT_TMP0, NEXT_TAB->hmask + | sll AT, NEXT_RES_IDX, 3 + | subu AT, NEXT_TMP2, AT + | addu NODE:NEXT_RES_VK, NODE:NEXT_RES_VK, AT + |6: + | sltu AT, NEXT_TMP0, NEXT_RES_IDX + | bnez AT, >8 + |. nop + | lw NEXT_TMP2, NODE:NEXT_RES_VK->val.it + | bne NEXT_TMP2, NEXT_NIL, >9 + |. addiu NEXT_RES_IDX, NEXT_RES_IDX, 1 + | // Skip holes in hash part. + | b <6 + |. addiu NODE:NEXT_RES_VK, NODE:NEXT_RES_VK, sizeof(Node) + | + |8: // End of iteration. Set the key to nil (not the value). + | sw NEXT_NIL, NEXT_RES_KEY_IT + | move NEXT_RES_VK, NEXT_RES_PTR + |9: + | jr ra + |. addu NEXT_RES_IDX, NEXT_RES_IDX, NEXT_ASIZE + |.endif + | |//----------------------------------------------------------------------- |//-- FFI helper functions ----------------------------------------------- |//----------------------------------------------------------------------- @@ -4524,10 +4588,11 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) break; case BC_ITERN: - | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8) - |.if JIT - | // NYI: add hotloop, record BC_ITERN. + |.if JIT and ENDIAN_LE + | hotloop |.endif + |->vm_IITERN: + | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8) | addu RA, BASE, RA | lw TAB:RB, -16+LO(RA) | lw RC, -8+LO(RA) // Get index from control var. @@ -4606,9 +4671,9 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | addiu CARG2, CARG2, -FF_next_N | or CARG2, CARG2, CARG3 | bnez CARG2, >5 - |. lui TMP1, 0xfffe + |. lui TMP1, (LJ_KEYINDEX >> 16) | addu PC, TMP0, TMP2 - | ori TMP1, TMP1, 0x7fff + | ori TMP1, TMP1, (LJ_KEYINDEX & 0xffff) | sw r0, -8+LO(RA) // Initialize control var. | sw TMP1, -8+HI(RA) |1: @@ -4617,9 +4682,28 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | li TMP3, BC_JMP | li TMP1, BC_ITERC | sb TMP3, -4+OFS_OP(PC) - | addu PC, TMP0, TMP2 + | addu PC, TMP0, TMP2 + |.if JIT + | lb TMP0, OFS_OP(PC) + | li AT, BC_ITERN + | bne TMP0, AT, >6 + |. lhu TMP2, OFS_RD(PC) + |.endif | b <1 |. sb TMP1, OFS_OP(PC) + |.if JIT + |6: // Unpatch JLOOP. + | lw TMP0, DISPATCH_J(trace)(DISPATCH) + | sll TMP2, TMP2, 2 + | addu TMP0, TMP0, TMP2 + | lw TRACE:TMP2, 0(TMP0) + | lw TMP0, TRACE:TMP2->startins + | li AT, -256 + | and TMP0, TMP0, AT + | or TMP0, TMP0, TMP1 + | b <1 + |. sw TMP0, 0(PC) + |.endif break; case BC_VARG: diff --git a/love/src/jni/LuaJIT-2.1/src/vm_mips64.dasc b/love/src/jni/LuaJIT-2.1/src/vm_mips64.dasc index 0b642095..4ddb2f9c 100644 --- a/love/src/jni/LuaJIT-2.1/src/vm_mips64.dasc +++ b/love/src/jni/LuaJIT-2.1/src/vm_mips64.dasc @@ -193,7 +193,7 @@ |//----------------------------------------------------------------------- | |// Trap for not-yet-implemented parts. -|.macro NYI; .long 0xf0f0f0f0; .endmacro +|.macro NYI; .long 0xec1cf0f0; .endmacro | |// Macros to mark delay slots. |.macro ., a; a; .endmacro @@ -556,6 +556,10 @@ static void build_subroutines(BuildCtx *ctx) | b ->vm_returnc |. li RD, 16 // 2 results: false + error message. | + |->vm_unwind_stub: // Jump to exit stub from unwinder. + | jr CARG1 + |. move ra, CARG2 + | |//----------------------------------------------------------------------- |//-- Grow stack for calls ----------------------------------------------- |//----------------------------------------------------------------------- @@ -724,11 +728,11 @@ static void build_subroutines(BuildCtx *ctx) | ld PC, -24(RB) // Restore PC from [cont|PC]. | cleartp LFUNC:TMP1 | daddu TMP2, RA, RD - | ld TMP1, LFUNC:TMP1->pc |.if FFI | bnez AT, >1 |.endif |. sd TISNIL, -8(TMP2) // Ensure one valid arg. + | ld TMP1, LFUNC:TMP1->pc | // BASE = base, RA = resultptr, RB = meta base | jr TMP0 // Jump to continuation. |. ld KBASE, PC2PROTO(k)(TMP1) @@ -1318,27 +1322,24 @@ static void build_subroutines(BuildCtx *ctx) |//-- Base library: iterators ------------------------------------------- | |.ffunc_1 next - | checktp CARG2, CARG1, -LJ_TTAB, ->fff_fallback + | checktp CARG1, -LJ_TTAB, ->fff_fallback | daddu TMP2, BASE, NARGS8:RC | sd TISNIL, 0(TMP2) // Set missing 2nd arg to nil. - | ld PC, FRAME_PC(BASE) | load_got lj_tab_next - | sd BASE, L->base // Add frame since C call can throw. - | sd BASE, L->top // Dummy frame length is ok. - | daddiu CARG3, BASE, 8 - | sd PC, SAVE_PC - | call_intern lj_tab_next // (lua_State *L, GCtab *t, TValue *key) - |. move CARG1, L - | // Returns 0 at end of traversal. + | ld PC, FRAME_PC(BASE) + | daddiu CARG2, BASE, 8 + | call_intern lj_tab_next // (GCtab *t, cTValue *key, TValue *o) + |. daddiu CARG3, BASE, -16 + | // Returns 1=found, 0=end, -1=error. + | daddiu RA, BASE, -16 + | bgtz CRET1, ->fff_res // Found key/value. + |. li RD, (2+1)*8 | beqz CRET1, ->fff_restv // End of traversal: return nil. |. move CARG1, TISNIL - | ld TMP0, 8(BASE) - | daddiu RA, BASE, -16 - | ld TMP2, 16(BASE) - | sd TMP0, 0(RA) - | sd TMP2, 8(RA) - | b ->fff_res - |. li RD, (2+1)*8 + | ld CFUNC:RB, FRAME_FUNC(BASE) + | cleartp CFUNC:RB + | b ->fff_fallback // Invalid key. + |. li RC, 2*8 | |.ffunc_1 pairs | checktp TAB:TMP1, CARG1, -LJ_TTAB, ->fff_fallback @@ -2037,7 +2038,7 @@ static void build_subroutines(BuildCtx *ctx) | ld TMP0, SBUF:CARG1->b | sd L, SBUF:CARG1->L | sd BASE, L->base - | sd TMP0, SBUF:CARG1->p + | sd TMP0, SBUF:CARG1->w | call_intern extern lj_buf_putstr_ .. name |. sd PC, SAVE_PC | load_got lj_buf_tostr @@ -2622,8 +2623,9 @@ static void build_subroutines(BuildCtx *ctx) |. daddu RA, RA, BASE | |9: // Rethrow error from the right C frame. - | load_got lj_err_run - | call_intern lj_err_run // (lua_State *L) + | load_got lj_err_trace + | sub CARG2, r0, CRET1 + | call_intern lj_err_trace // (lua_State *L, int errcode) |. move CARG1, L |.endif | @@ -2902,6 +2904,70 @@ static void build_subroutines(BuildCtx *ctx) |//-- Miscellaneous functions -------------------------------------------- |//----------------------------------------------------------------------- | + |.define NEXT_TAB, TAB:CARG1 + |.define NEXT_IDX, CARG2 + |.define NEXT_ASIZE, CARG3 + |.define NEXT_NIL, CARG4 + |.define NEXT_TMP0, r12 + |.define NEXT_TMP1, r13 + |.define NEXT_TMP2, r14 + |.define NEXT_RES_VK, CRET1 + |.define NEXT_RES_IDX, CRET2 + |.define NEXT_RES_PTR, sp + |.define NEXT_RES_VAL, 0(sp) + |.define NEXT_RES_KEY, 8(sp) + | + |// TValue *lj_vm_next(GCtab *t, uint32_t idx) + |// Next idx returned in CRET2. + |->vm_next: + |.if JIT and ENDIAN_LE + | lw NEXT_ASIZE, NEXT_TAB->asize + | ld NEXT_TMP0, NEXT_TAB->array + | li NEXT_NIL, LJ_TNIL + |1: // Traverse array part. + | sltu AT, NEXT_IDX, NEXT_ASIZE + | sll NEXT_TMP1, NEXT_IDX, 3 + | beqz AT, >5 + |. daddu NEXT_TMP1, NEXT_TMP0, NEXT_TMP1 + | li AT, LJ_TISNUM + | ld NEXT_TMP2, 0(NEXT_TMP1) + | dsll AT, AT, 47 + | or NEXT_TMP1, NEXT_IDX, AT + | beq NEXT_TMP2, NEXT_NIL, <1 + |. addiu NEXT_IDX, NEXT_IDX, 1 + | sd NEXT_TMP2, NEXT_RES_VAL + | sd NEXT_TMP1, NEXT_RES_KEY + | move NEXT_RES_VK, NEXT_RES_PTR + | jr ra + |. move NEXT_RES_IDX, NEXT_IDX + | + |5: // Traverse hash part. + | subu NEXT_RES_IDX, NEXT_IDX, NEXT_ASIZE + | ld NODE:NEXT_RES_VK, NEXT_TAB->node + | sll NEXT_TMP2, NEXT_RES_IDX, 5 + | lw NEXT_TMP0, NEXT_TAB->hmask + | sll AT, NEXT_RES_IDX, 3 + | subu AT, NEXT_TMP2, AT + | daddu NODE:NEXT_RES_VK, NODE:NEXT_RES_VK, AT + |6: + | sltu AT, NEXT_TMP0, NEXT_RES_IDX + | bnez AT, >8 + |. nop + | ld NEXT_TMP2, NODE:NEXT_RES_VK->val + | bne NEXT_TMP2, NEXT_NIL, >9 + |. addiu NEXT_RES_IDX, NEXT_RES_IDX, 1 + | // Skip holes in hash part. + | b <6 + |. daddiu NODE:NEXT_RES_VK, NODE:NEXT_RES_VK, sizeof(Node) + | + |8: // End of iteration. Set the key to nil (not the value). + | sd NEXT_NIL, NEXT_RES_KEY + | move NEXT_RES_VK, NEXT_RES_PTR + |9: + | jr ra + |. addu NEXT_RES_IDX, NEXT_RES_IDX, NEXT_ASIZE + |.endif + | |//----------------------------------------------------------------------- |//-- FFI helper functions ----------------------------------------------- |//----------------------------------------------------------------------- @@ -4698,10 +4764,11 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) break; case BC_ITERN: - | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8) - |.if JIT - | // NYI: add hotloop, record BC_ITERN. + |.if JIT and ENDIAN_LE + | hotloop |.endif + |->vm_IITERN: + | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8) | daddu RA, BASE, RA | ld TAB:RB, -16(RA) | lw RC, -8+LO(RA) // Get index from control var. @@ -4722,11 +4789,10 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) |. addiu RC, RC, 1 | sd TMP2, 0(RA) | sd CARG1, 8(RA) - | or TMP0, RC, CARG3 | lui TMP3, (-(BCBIAS_J*4 >> 16) & 65535) | decode_RD4b RD | daddu RD, RD, TMP3 - | sw TMP0, -8+LO(RA) // Update control var. + | sw RC, -8+LO(RA) // Update control var. | daddu PC, PC, RD |3: | ins_next @@ -4776,9 +4842,9 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | daddiu TMP1, TMP1, -FF_next_N | or AT, AT, TMP1 | bnez AT, >5 - |. lui TMP1, 0xfffe + |. lui TMP1, (LJ_KEYINDEX >> 16) | daddu PC, TMP0, TMP2 - | ori TMP1, TMP1, 0x7fff + | ori TMP1, TMP1, (LJ_KEYINDEX & 0xffff) | dsll TMP1, TMP1, 32 | sd TMP1, -8(RA) |1: @@ -4788,8 +4854,27 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | li TMP1, BC_ITERC | sb TMP3, -4+OFS_OP(PC) | daddu PC, TMP0, TMP2 + |.if JIT + | lb TMP0, OFS_OP(PC) + | li AT, BC_ITERN + | bne TMP0, AT, >6 + |. lhu TMP2, OFS_RD(PC) + |.endif | b <1 |. sb TMP1, OFS_OP(PC) + |.if JIT + |6: // Unpatch JLOOP. + | ld TMP0, DISPATCH_J(trace)(DISPATCH) + | sll TMP2, TMP2, 3 + | daddu TMP0, TMP0, TMP2 + | ld TRACE:TMP2, 0(TMP0) + | lw TMP0, TRACE:TMP2->startins + | li AT, -256 + | and TMP0, TMP0, AT + | or TMP0, TMP0, TMP1 + | b <1 + |. sw TMP0, 0(PC) + |.endif break; case BC_VARG: diff --git a/love/src/jni/LuaJIT-2.1/src/vm_ppc.dasc b/love/src/jni/LuaJIT-2.1/src/vm_ppc.dasc index f4ebb9db..a376c849 100644 --- a/love/src/jni/LuaJIT-2.1/src/vm_ppc.dasc +++ b/love/src/jni/LuaJIT-2.1/src/vm_ppc.dasc @@ -859,11 +859,11 @@ static void build_subroutines(BuildCtx *ctx) |.endif | lwz PC, -16(RB) // Restore PC from [cont|PC]. | subi TMP2, RD, 8 - | lwz TMP1, LFUNC:TMP1->pc | stwx TISNIL, RA, TMP2 // Ensure one valid arg. |.if FFI | ble >1 |.endif + | lwz TMP1, LFUNC:TMP1->pc | lwz KBASE, PC2PROTO(k)(TMP1) | // BASE = base, RA = resultptr, RB = meta base | mtctr TMP0 @@ -1559,43 +1559,24 @@ static void build_subroutines(BuildCtx *ctx) | |//-- Base library: iterators ------------------------------------------- | - |.ffunc next - | cmplwi NARGS8:RC, 8 - | lwz CARG1, 0(BASE) - | lwz TAB:CARG2, 4(BASE) - | blt ->fff_fallback + |.ffunc_1 next | stwx TISNIL, BASE, NARGS8:RC // Set missing 2nd arg to nil. - | checktab CARG1 + | checktab CARG3 | lwz PC, FRAME_PC(BASE) | bne ->fff_fallback - | stp BASE, L->base // Add frame since C call can throw. - | mr CARG1, L - | stp BASE, L->top // Dummy frame length is ok. - | la CARG3, 8(BASE) - | stw PC, SAVE_PC - | bl extern lj_tab_next // (lua_State *L, GCtab *t, TValue *key) - | // Returns 0 at end of traversal. - | cmplwi CRET1, 0 + | la CARG2, 8(BASE) + | la CARG3, -8(BASE) + | bl extern lj_tab_next // (GCtab *t, cTValue *key, TValue *o) + | // Returns 1=found, 0=end, -1=error. + | cmpwi CRET1, 0 + | la RA, -8(BASE) + | li RD, (2+1)*8 + | bgt ->fff_res // Found key/value. | li CARG3, LJ_TNIL | beq ->fff_restv // End of traversal: return nil. - | la RA, -8(BASE) - |.if FPU - | lfd f0, 8(BASE) // Copy key and value to results. - | lfd f1, 16(BASE) - | stfd f0, 0(RA) - | stfd f1, 8(RA) - |.else - | lwz CARG1, 8(BASE) - | lwz CARG2, 12(BASE) - | lwz CARG3, 16(BASE) - | lwz CARG4, 20(BASE) - | stw CARG1, 0(RA) - | stw CARG2, 4(RA) - | stw CARG3, 8(RA) - | stw CARG4, 12(RA) - |.endif - | li RD, (2+1)*8 - | b ->fff_res + | lwz CFUNC:RB, FRAME_FUNC(BASE) + | li NARGS8:RC, 2*8 + | b ->fff_fallback // Invalid key. | |.ffunc_1 pairs | checktab CARG3 @@ -2516,7 +2497,7 @@ static void build_subroutines(BuildCtx *ctx) | stw L, SBUF:CARG1->L | stp BASE, L->base | stw PC, SAVE_PC - | stw TMP0, SBUF:CARG1->p + | stw TMP0, SBUF:CARG1->w | bl extern lj_buf_putstr_ .. name | bl extern lj_buf_tostr | b ->fff_resstr @@ -3090,8 +3071,9 @@ static void build_subroutines(BuildCtx *ctx) | bctr | |9: // Rethrow error from the right C frame. + | neg CARG2, CARG1 | mr CARG1, L - | bl extern lj_err_run // (lua_State *L) + | bl extern lj_err_trace // (lua_State *L, int errcode) |.endif | |//----------------------------------------------------------------------- @@ -3181,6 +3163,11 @@ static void build_subroutines(BuildCtx *ctx) | blr |.endif | + |->vm_next: + |.if JIT + | NYI // On big-endian. + |.endif + | |//----------------------------------------------------------------------- |//-- FFI helper functions ----------------------------------------------- |//----------------------------------------------------------------------- @@ -4100,8 +4087,8 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) |.macro addo32., y, a, b | // Need to check overflow for (a<<32) + (b<<32). | rldicr TMP0, a, 32, 31 - | rldicr TMP3, b, 32, 31 - | addo. TMP0, TMP0, TMP3 + | rldicr TMP1, b, 32, 31 + | addo. TMP0, TMP0, TMP1 | add y, a, b |.endmacro | ins_arith addo32., fadd, __adddf3 @@ -4114,8 +4101,8 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) |.macro subo32., y, a, b | // Need to check overflow for (a<<32) - (b<<32). | rldicr TMP0, a, 32, 31 - | rldicr TMP3, b, 32, 31 - | subo. TMP0, TMP0, TMP3 + | rldicr TMP1, b, 32, 31 + | subo. TMP0, TMP0, TMP1 | sub y, a, b |.endmacro | ins_arith subo32., fsub, __subdf3 @@ -5130,8 +5117,9 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) case BC_ITERN: | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8) |.if JIT - | // NYI: add hotloop, record BC_ITERN. + | // NYI on big-endian |.endif + |->vm_IITERN: | add RA, BASE, RA | lwz TAB:RB, -12(RA) | lwz RC, -4(RA) // Get index from control var. @@ -5250,8 +5238,8 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | crand 4*cr0+eq, 4*cr0+eq, 4*cr7+eq | add TMP3, PC, TMP0 | bne cr0, >5 - | lus TMP1, 0xfffe - | ori TMP1, TMP1, 0x7fff + | lus TMP1, (LJ_KEYINDEX >> 16) + | ori TMP1, TMP1, (LJ_KEYINDEX & 0xffff) | stw ZERO, -4(RA) // Initialize control var. | stw TMP1, -8(RA) | addis PC, TMP3, -(BCBIAS_J*4 >> 16) @@ -5262,6 +5250,7 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | li TMP1, BC_ITERC | stb TMP0, -1(PC) | addis PC, TMP3, -(BCBIAS_J*4 >> 16) + | // NYI on big-endian: unpatch JLOOP. | stb TMP1, 3(PC) | b <1 break; diff --git a/love/src/jni/LuaJIT-2.1/src/vm_x64.dasc b/love/src/jni/LuaJIT-2.1/src/vm_x64.dasc index 8ade0a6a..fdffd4b6 100644 --- a/love/src/jni/LuaJIT-2.1/src/vm_x64.dasc +++ b/love/src/jni/LuaJIT-2.1/src/vm_x64.dasc @@ -1346,44 +1346,28 @@ static void build_subroutines(BuildCtx *ctx) |.ffunc_1 next | je >2 // Missing 2nd arg? |1: - |.if X64WIN - | mov RA, [BASE] - | checktab RA, ->fff_fallback - |.else - | mov CARG2, [BASE] - | checktab CARG2, ->fff_fallback - |.endif - | mov L:RB, SAVE_L - | mov L:RB->base, BASE // Add frame since C call can throw. - | mov L:RB->top, BASE // Dummy frame length is ok. + | mov CARG1, [BASE] | mov PC, [BASE-8] + | checktab CARG1, ->fff_fallback + | mov RB, BASE // Save BASE. |.if X64WIN - | lea CARG3, [BASE+8] - | mov CARG2, RA // Caveat: CARG2 == BASE. - | mov CARG1, L:RB + | lea CARG3, [BASE-16] + | lea CARG2, [BASE+8] // Caveat: CARG2 == BASE. |.else - | lea CARG3, [BASE+8] // Caveat: CARG3 == BASE. - | mov CARG1, L:RB + | lea CARG2, [BASE+8] + | lea CARG3, [BASE-16] // Caveat: CARG3 == BASE. |.endif - | mov SAVE_PC, PC // Needed for ITERN fallback. - | call extern lj_tab_next // (lua_State *L, GCtab *t, TValue *key) - | // Flag returned in eax (RD). - | mov BASE, L:RB->base - | test RDd, RDd; jz >3 // End of traversal? - | // Copy key and value to results. - | mov RB, [BASE+8] - | mov RD, [BASE+16] - | mov [BASE-16], RB - | mov [BASE-8], RD - |->fff_res2: - | mov RDd, 1+2 - | jmp ->fff_res + | call extern lj_tab_next // (GCtab *t, cTValue *key, TValue *o) + | // 1=found, 0=end, -1=error returned in eax (RD). + | mov BASE, RB // Restore BASE. + | test RDd, RDd; jg ->fff_res2 // Found key/value. + | js ->fff_fallback_2 // Invalid key. + | // End of traversal: return nil. + | mov aword [BASE-16], LJ_TNIL + | jmp ->fff_res1 |2: // Set missing 2nd arg to nil. | mov aword [BASE+8], LJ_TNIL | jmp <1 - |3: // End of traversal: return nil. - | mov aword [BASE-16], LJ_TNIL - | jmp ->fff_res1 | |.ffunc_1 pairs | mov TAB:RB, [BASE] @@ -1432,7 +1416,9 @@ static void build_subroutines(BuildCtx *ctx) | // Copy array slot. | mov RB, [RD] | mov [BASE-8], RB - | jmp ->fff_res2 + |->fff_res2: + | mov RDd, 1+2 + | jmp ->fff_res |2: // Check for empty hash part first. Otherwise call C function. | cmp dword TAB:RB->hmask, 0; je ->fff_res0 |.if X64WIN @@ -2011,7 +1997,7 @@ static void build_subroutines(BuildCtx *ctx) |.endif | mov RC, SBUF:CARG1->b | mov SBUF:CARG1->L, L:RB - | mov SBUF:CARG1->p, RC + | mov SBUF:CARG1->w, RC | mov SAVE_PC, PC | call extern lj_buf_putstr_ .. name | mov CARG1, rax @@ -2509,8 +2495,10 @@ static void build_subroutines(BuildCtx *ctx) | jmp <2 | |9: // Rethrow error from the right C frame. + | mov CARG2d, RDd | mov CARG1, L:RB - | call extern lj_err_run // (lua_State *L) + | neg CARG2d + | call extern lj_err_trace // (lua_State *L, int errcode) |.endif | |//----------------------------------------------------------------------- @@ -2645,6 +2633,67 @@ static void build_subroutines(BuildCtx *ctx) | .if X64WIN; pop rsi; .endif | ret | + |.define NEXT_TAB, TAB:CARG1 + |.define NEXT_IDX, CARG2d + |.define NEXT_IDXa, CARG2 + |.define NEXT_PTR, RC + |.define NEXT_PTRd, RCd + |.define NEXT_TMP, CARG3 + |.define NEXT_ASIZE, CARG4d + |.macro NEXT_RES_IDXL, op2; lea edx, [NEXT_IDX+op2]; .endmacro + |.if X64WIN + |.define NEXT_RES_PTR, [rsp+aword*5] + |.macro NEXT_RES_IDX, op2; add NEXT_IDX, op2; .endmacro + |.else + |.define NEXT_RES_PTR, [rsp+aword*1] + |.macro NEXT_RES_IDX, op2; lea edx, [NEXT_IDX+op2]; .endmacro + |.endif + | + |// TValue *lj_vm_next(GCtab *t, uint32_t idx) + |// Next idx returned in edx. + |->vm_next: + |.if JIT + | mov NEXT_ASIZE, NEXT_TAB->asize + |1: // Traverse array part. + | cmp NEXT_IDX, NEXT_ASIZE; jae >5 + | mov NEXT_TMP, NEXT_TAB->array + | mov NEXT_TMP, qword [NEXT_TMP+NEXT_IDX*8] + | cmp NEXT_TMP, LJ_TNIL; je >2 + | lea NEXT_PTR, NEXT_RES_PTR + | mov qword [NEXT_PTR], NEXT_TMP + |.if DUALNUM + | setint NEXT_TMP, NEXT_IDXa + | mov qword [NEXT_PTR+qword*1], NEXT_TMP + |.else + | cvtsi2sd xmm0, NEXT_IDX + | movsd qword [NEXT_PTR+qword*1], xmm0 + |.endif + | NEXT_RES_IDX 1 + | ret + |2: // Skip holes in array part. + | add NEXT_IDX, 1 + | jmp <1 + | + |5: // Traverse hash part. + | sub NEXT_IDX, NEXT_ASIZE + |6: + | cmp NEXT_IDX, NEXT_TAB->hmask; ja >9 + | imul NEXT_PTRd, NEXT_IDX, #NODE + | add NODE:NEXT_PTR, NEXT_TAB->node + | cmp qword NODE:NEXT_PTR->val, LJ_TNIL; je >7 + | NEXT_RES_IDXL NEXT_ASIZE+1 + | ret + |7: // Skip holes in hash part. + | add NEXT_IDX, 1 + | jmp <6 + | + |9: // End of iteration. Set the key to nil (not the value). + | NEXT_RES_IDX NEXT_ASIZE + | lea NEXT_PTR, NEXT_RES_PTR + | mov qword [NEXT_PTR+qword*1], LJ_TNIL + | ret + |.endif + | |//----------------------------------------------------------------------- |//-- Assertions --------------------------------------------------------- |//----------------------------------------------------------------------- @@ -4056,10 +4105,11 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) break; case BC_ITERN: - | ins_A // RA = base, (RB = nresults+1, RC = nargs+1 (2+1)) |.if JIT - | // NYI: add hotloop, record BC_ITERN. + | hotloop RBd |.endif + |->vm_IITERN: + | ins_A // RA = base, (RB = nresults+1, RC = nargs+1 (2+1)) | mov TAB:RB, [BASE+RA*8-16] | cleartp TAB:RB | mov RCd, [BASE+RA*8-8] // Get index from control var. @@ -4123,15 +4173,29 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | cmp aword [BASE+RA*8-8], LJ_TNIL; jne >5 | cmp byte CFUNC:RB->ffid, FF_next_N; jne >5 | branchPC RD - | mov64 TMPR, U64x(fffe7fff, 00000000) + | mov64 TMPR, ((uint64_t)LJ_KEYINDEX << 32) | mov [BASE+RA*8-8], TMPR // Initialize control var. |1: | ins_next |5: // Despecialize bytecode if any of the checks fail. | mov PC_OP, BC_JMP | branchPC RD + |.if JIT + | cmp byte [PC], BC_ITERN + | jne >6 + |.endif | mov byte [PC], BC_ITERC | jmp <1 + |.if JIT + |6: // Unpatch JLOOP. + | mov RA, [DISPATCH+DISPATCH_J(trace)] + | movzx RCd, word [PC+2] + | mov TRACE:RA, [RA+RC*8] + | mov eax, TRACE:RA->startins + | mov al, BC_ITERC + | mov dword [PC], eax + | jmp <1 + |.endif break; case BC_VARG: diff --git a/love/src/jni/LuaJIT-2.1/src/vm_x86.dasc b/love/src/jni/LuaJIT-2.1/src/vm_x86.dasc index 1fee4a2a..cbf0810c 100644 --- a/love/src/jni/LuaJIT-2.1/src/vm_x86.dasc +++ b/love/src/jni/LuaJIT-2.1/src/vm_x86.dasc @@ -1673,55 +1673,35 @@ static void build_subroutines(BuildCtx *ctx) | je >2 // Missing 2nd arg? |1: | cmp dword [BASE+4], LJ_TTAB; jne ->fff_fallback - | mov L:RB, SAVE_L - | mov L:RB->base, BASE // Add frame since C call can throw. - | mov L:RB->top, BASE // Dummy frame length is ok. | mov PC, [BASE-4] + | mov RB, BASE // Save BASE. |.if X64WIN - | lea CARG3d, [BASE+8] - | mov CARG2d, [BASE] // Caveat: CARG2d == BASE. - | mov CARG1d, L:RB + | mov CARG1d, [BASE] + | lea CARG3d, [BASE-8] + | lea CARG2d, [BASE+8] // Caveat: CARG2d == BASE. |.elif X64 - | mov CARG2d, [BASE] - | lea CARG3d, [BASE+8] // Caveat: CARG3d == BASE. - | mov CARG1d, L:RB + | mov CARG1d, [BASE] + | lea CARG2d, [BASE+8] + | lea CARG3d, [BASE-8] // Caveat: CARG3d == BASE. |.else | mov TAB:RD, [BASE] - | mov ARG2, TAB:RD - | mov ARG1, L:RB + | mov ARG1, TAB:RD | add BASE, 8 + | mov ARG2, BASE + | sub BASE, 8+8 | mov ARG3, BASE |.endif - | mov SAVE_PC, PC // Needed for ITERN fallback. - | call extern lj_tab_next // (lua_State *L, GCtab *t, TValue *key) - | // Flag returned in eax (RD). - | mov BASE, L:RB->base - | test RD, RD; jz >3 // End of traversal? - | // Copy key and value to results. - |.if X64 - | mov RBa, [BASE+8] - | mov RDa, [BASE+16] - | mov [BASE-8], RBa - | mov [BASE], RDa - |.else - | mov RB, [BASE+8] - | mov RD, [BASE+12] - | mov [BASE-8], RB - | mov [BASE-4], RD - | mov RB, [BASE+16] - | mov RD, [BASE+20] - | mov [BASE], RB - | mov [BASE+4], RD - |.endif - |->fff_res2: - | mov RD, 1+2 - | jmp ->fff_res + | call extern lj_tab_next // (GCtab *t, cTValue *key, TValue *o) + | // 1=found, 0=end, -1=error returned in eax (RD). + | mov BASE, RB // Restore BASE. + | test RD, RD; jg ->fff_res2 // Found key/value. + | js ->fff_fallback_2 // Invalid key. + | // End of traversal: return nil. + | mov dword [BASE-4], LJ_TNIL + | jmp ->fff_res1 |2: // Set missing 2nd arg to nil. | mov dword [BASE+12], LJ_TNIL | jmp <1 - |3: // End of traversal: return nil. - | mov dword [BASE-4], LJ_TNIL - | jmp ->fff_res1 | |.ffunc_1 pairs | mov TAB:RB, [BASE] @@ -1775,7 +1755,9 @@ static void build_subroutines(BuildCtx *ctx) | mov [BASE], RB | mov [BASE+4], RD |.endif - | jmp ->fff_res2 + |->fff_res2: + | mov RD, 1+2 + | jmp ->fff_res |2: // Check for empty hash part first. Otherwise call C function. | cmp dword TAB:RB->hmask, 0; je ->fff_res0 | mov FCARG1, TAB:RB @@ -2423,9 +2405,9 @@ static void build_subroutines(BuildCtx *ctx) | lea SBUF:FCARG1, [DISPATCH+DISPATCH_GL(tmpbuf)] | mov L:RB->base, BASE | mov STR:FCARG2, [BASE] // Caveat: FCARG2 == BASE - | mov RC, SBUF:FCARG1->b + | mov RCa, SBUF:FCARG1->b | mov SBUF:FCARG1->L, L:RB - | mov SBUF:FCARG1->p, RC + | mov SBUF:FCARG1->w, RCa | mov SAVE_PC, PC | call extern lj_buf_putstr_ .. name .. @8 | mov FCARG1, eax @@ -2964,8 +2946,10 @@ static void build_subroutines(BuildCtx *ctx) | jmp <2 | |9: // Rethrow error from the right C frame. + | mov FCARG2, RD | mov FCARG1, L:RB - | call extern lj_err_run@4 // (lua_State *L) + | neg FCARG2 + | call extern lj_err_trace@8 // (lua_State *L, int errcode) |.endif | |//----------------------------------------------------------------------- @@ -3136,6 +3120,86 @@ static void build_subroutines(BuildCtx *ctx) | ret |.endif | + |.define NEXT_TAB, TAB:FCARG1 + |.define NEXT_IDX, FCARG2 + |.define NEXT_PTR, RCa + |.define NEXT_PTRd, RC + |.macro NEXT_RES_IDXL, op2; lea edx, [NEXT_IDX+op2]; .endmacro + |.if X64 + |.define NEXT_TMP, CARG3d + |.define NEXT_TMPq, CARG3 + |.define NEXT_ASIZE, CARG4d + |.macro NEXT_ENTER; .endmacro + |.macro NEXT_LEAVE; ret; .endmacro + |.if X64WIN + |.define NEXT_RES_PTR, [rsp+aword*5] + |.macro NEXT_RES_IDX, op2; add NEXT_IDX, op2; .endmacro + |.else + |.define NEXT_RES_PTR, [rsp+aword*1] + |.macro NEXT_RES_IDX, op2; lea edx, [NEXT_IDX+op2]; .endmacro + |.endif + |.else + |.define NEXT_ASIZE, esi + |.define NEXT_TMP, edi + |.macro NEXT_ENTER; push esi; push edi; .endmacro + |.macro NEXT_LEAVE; pop edi; pop esi; ret; .endmacro + |.define NEXT_RES_PTR, [esp+dword*3] + |.macro NEXT_RES_IDX, op2; add NEXT_IDX, op2; .endmacro + |.endif + | + |// TValue *lj_vm_next(GCtab *t, uint32_t idx) + |// Next idx returned in edx. + |->vm_next: + |.if JIT + | NEXT_ENTER + | mov NEXT_ASIZE, NEXT_TAB->asize + |1: // Traverse array part. + | cmp NEXT_IDX, NEXT_ASIZE; jae >5 + | mov NEXT_TMP, NEXT_TAB->array + | cmp dword [NEXT_TMP+NEXT_IDX*8+4], LJ_TNIL; je >2 + | lea NEXT_PTR, NEXT_RES_PTR + |.if X64 + | mov NEXT_TMPq, qword [NEXT_TMP+NEXT_IDX*8] + | mov qword [NEXT_PTR], NEXT_TMPq + |.else + | mov NEXT_ASIZE, dword [NEXT_TMP+NEXT_IDX*8+4] + | mov NEXT_TMP, dword [NEXT_TMP+NEXT_IDX*8] + | mov dword [NEXT_PTR+4], NEXT_ASIZE + | mov dword [NEXT_PTR], NEXT_TMP + |.endif + |.if DUALNUM + | mov dword [NEXT_PTR+dword*3], LJ_TISNUM + | mov dword [NEXT_PTR+dword*2], NEXT_IDX + |.else + | cvtsi2sd xmm0, NEXT_IDX + | movsd qword [NEXT_PTR+dword*2], xmm0 + |.endif + | NEXT_RES_IDX 1 + | NEXT_LEAVE + |2: // Skip holes in array part. + | add NEXT_IDX, 1 + | jmp <1 + | + |5: // Traverse hash part. + | sub NEXT_IDX, NEXT_ASIZE + |6: + | cmp NEXT_IDX, NEXT_TAB->hmask; ja >9 + | imul NEXT_PTRd, NEXT_IDX, #NODE + | add NODE:NEXT_PTRd, dword NEXT_TAB->node + | cmp dword NODE:NEXT_PTR->val.it, LJ_TNIL; je >7 + | NEXT_RES_IDXL NEXT_ASIZE+1 + | NEXT_LEAVE + |7: // Skip holes in hash part. + | add NEXT_IDX, 1 + | jmp <6 + | + |9: // End of iteration. Set the key to nil (not the value). + | NEXT_RES_IDX NEXT_ASIZE + | lea NEXT_PTR, NEXT_RES_PTR + | mov dword [NEXT_PTR+dword*3], LJ_TNIL + | NEXT_LEAVE + |.endif + | |//----------------------------------------------------------------------- |//-- Assertions --------------------------------------------------------- |//----------------------------------------------------------------------- @@ -4787,10 +4851,11 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) break; case BC_ITERN: - | ins_A // RA = base, (RB = nresults+1, RC = nargs+1 (2+1)) |.if JIT - | // NYI: add hotloop, record BC_ITERN. + | hotloop RB |.endif + |->vm_IITERN: + | ins_A // RA = base, (RB = nresults+1, RC = nargs+1 (2+1)) | mov TMP1, KBASE // Need two more free registers. | mov TMP2, DISPATCH | mov TAB:RB, [BASE+RA*8-16] @@ -4878,14 +4943,28 @@ static void build_ins(BuildCtx *ctx, BCOp op, int defop) | cmp byte CFUNC:RB->ffid, FF_next_N; jne >5 | branchPC RD | mov dword [BASE+RA*8-8], 0 // Initialize control var. - | mov dword [BASE+RA*8-4], 0xfffe7fff + | mov dword [BASE+RA*8-4], LJ_KEYINDEX |1: | ins_next |5: // Despecialize bytecode if any of the checks fail. | mov PC_OP, BC_JMP | branchPC RD + |.if JIT + | cmp byte [PC], BC_ITERN + | jne >6 + |.endif | mov byte [PC], BC_ITERC | jmp <1 + |.if JIT + |6: // Unpatch JLOOP. + | mov RA, [DISPATCH+DISPATCH_J(trace)] + | movzx RC, word [PC+2] + | mov TRACE:RA, [RA+RC*4] + | mov eax, TRACE:RA->startins + | mov al, BC_ITERC + | mov dword [PC], eax + | jmp <1 + |.endif break; case BC_VARG: diff --git a/love/src/jni/LuaJIT-2.1/src/xb1build.bat b/love/src/jni/LuaJIT-2.1/src/xb1build.bat index 12c73dd6..2eb68171 100644 --- a/love/src/jni/LuaJIT-2.1/src/xb1build.bat +++ b/love/src/jni/LuaJIT-2.1/src/xb1build.bat @@ -14,7 +14,7 @@ @set LJMT=mt /nologo @set DASMDIR=..\dynasm @set DASM=%DASMDIR%\dynasm.lua -@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c +@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c lib_buffer.c %LJCOMPILE% host\minilua.c @if errorlevel 1 goto :BAD diff --git a/love/src/jni/LuaJIT-2.1/src/xedkbuild.bat b/love/src/jni/LuaJIT-2.1/src/xedkbuild.bat index 240ec878..37322d03 100644 --- a/love/src/jni/LuaJIT-2.1/src/xedkbuild.bat +++ b/love/src/jni/LuaJIT-2.1/src/xedkbuild.bat @@ -14,7 +14,7 @@ @set LJMT=mt /nologo @set DASMDIR=..\dynasm @set DASM=%DASMDIR%\dynasm.lua -@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c +@set ALL_LIB=lib_base.c lib_math.c lib_bit.c lib_string.c lib_table.c lib_io.c lib_os.c lib_package.c lib_debug.c lib_jit.c lib_ffi.c lib_buffer.c %LJCOMPILE% host\minilua.c @if errorlevel 1 goto :BAD diff --git a/love/src/jni/love b/love/src/jni/love index 1bb22d2e..0f9e598e 160000 --- a/love/src/jni/love +++ b/love/src/jni/love @@ -1 +1 @@ -Subproject commit 1bb22d2e90eb43146b4e577055064260b4649c2d +Subproject commit 0f9e598ebb5f77a38124ab7ac8c4a92a1403ee47