Update LuaJIT to LuaJIT/LuaJIT@43d0a19.

2026-08-12 00:20:57 +02:00 · 2023-11-25 21:41:59 +08:00
parent abd98a1378
commit 044d5b5e34
218 changed files with 1999 additions and 1417 deletions
@@ -0,0 +1 @@
+/.relver export-subst
@@ -0,0 +1 @@
+1700008891
@@ -1,7 +1,7 @@
 ===============================================================================
 LuaJIT -- a Just-In-Time Compiler for Lua. https://luajit.org/

-Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+Copyright (C) 2005-2023 Mike Pall. All rights reserved.

 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal
@@ -10,16 +10,21 @@
 # For MSVC, please follow the instructions given in src/msvcbuild.bat.
 # For MinGW and Cygwin, cd to src and run make with the Makefile there.
 #
-# Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+# Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 ##############################################################################

 MAJVER=  2
 MINVER=  1
-RELVER=  0
-PREREL=  -beta3
-VERSION= $(MAJVER).$(MINVER).$(RELVER)$(PREREL)
 ABIVER=  5.1

+# LuaJIT uses rolling releases. The release version is based on the time of
+# the latest git commit. The 'git' command must be available during the build.
+RELVER= $(shell cat src/luajit_relver.txt 2>/dev/null || : )
+# Note: setting it with := doesn't work, since it will change during the build.
+
+MMVERSION= $(MAJVER).$(MINVER)
+VERSION= $(MMVERSION).$(RELVER)
+
 ##############################################################################
 #
 # Change the installation path as needed. This automatically adjusts
@@ -33,10 +38,10 @@ DPREFIX= $(DESTDIR)$(PREFIX)
 INSTALL_BIN=   $(DPREFIX)/bin
 INSTALL_LIB=   $(DPREFIX)/$(MULTILIB)
 INSTALL_SHARE= $(DPREFIX)/share
-INSTALL_DEFINC= $(DPREFIX)/include/luajit-$(MAJVER).$(MINVER)
+INSTALL_DEFINC= $(DPREFIX)/include/luajit-$(MMVERSION)
 INSTALL_INC=   $(INSTALL_DEFINC)

-INSTALL_LJLIBD= $(INSTALL_SHARE)/luajit-$(VERSION)
+INSTALL_LJLIBD= $(INSTALL_SHARE)/luajit-$(MMVERSION)
 INSTALL_JITLIB= $(INSTALL_LJLIBD)/jit
 INSTALL_LMODD= $(INSTALL_SHARE)/lua
 INSTALL_LMOD= $(INSTALL_LMODD)/$(ABIVER)
@@ -50,10 +55,10 @@ INSTALL_TSYMNAME= luajit
 INSTALL_ANAME= libluajit-$(ABIVER).a
 INSTALL_SOSHORT1= libluajit-$(ABIVER).so
 INSTALL_SOSHORT2= libluajit-$(ABIVER).so.$(MAJVER)
-INSTALL_SONAME= $(INSTALL_SOSHORT2).$(MINVER).$(RELVER)
+INSTALL_SONAME= libluajit-$(ABIVER).so.$(VERSION)
 INSTALL_DYLIBSHORT1= libluajit-$(ABIVER).dylib
 INSTALL_DYLIBSHORT2= libluajit-$(ABIVER).$(MAJVER).dylib
-INSTALL_DYLIBNAME= libluajit-$(ABIVER).$(MAJVER).$(MINVER).$(RELVER).dylib
+INSTALL_DYLIBNAME= libluajit-$(ABIVER).$(VERSION).dylib
 INSTALL_PCNAME= luajit.pc

 INSTALL_STATIC= $(INSTALL_LIB)/$(INSTALL_ANAME)
@@ -78,7 +83,8 @@ INSTALL_F= install -m 0644
 UNINSTALL= $(RM)
 LDCONFIG= ldconfig -n 2>/dev/null
 SED_PC= sed -e "s|^prefix=.*|prefix=$(PREFIX)|" \
-            -e "s|^multilib=.*|multilib=$(MULTILIB)|"
+	    -e "s|^multilib=.*|multilib=$(MULTILIB)|" \
+	    -e "s|^relver=.*|relver=$(RELVER)|"
 ifneq ($(INSTALL_DEFINC),$(INSTALL_INC))
  SED_PC+= -e "s|^includedir=.*|includedir=$(INSTALL_INC)|"
 endif
@@ -92,7 +98,9 @@ FILES_INC= lua.h lualib.h lauxlib.h luaconf.h lua.hpp luajit.h
 FILES_JITLIB= bc.lua bcsave.lua dump.lua p.lua v.lua zone.lua \
 	      dis_x86.lua dis_x64.lua dis_arm.lua dis_arm64.lua \
 	      dis_arm64be.lua dis_ppc.lua dis_mips.lua dis_mipsel.lua \
-	      dis_mips64.lua dis_mips64el.lua vmdef.lua
+	      dis_mips64.lua dis_mips64el.lua \
+	      dis_mips64r6.lua dis_mips64r6el.lua \
+	      vmdef.lua

 ifeq (,$(findstring Windows,$(OS)))
  HOST_SYS:= $(shell uname -s)
@@ -113,9 +121,9 @@ endif
 INSTALL_DEP= src/luajit

 default all $(INSTALL_DEP):
-	@echo "==== Building LuaJIT $(VERSION) ===="
+	@echo "==== Building LuaJIT $(MMVERSION) ===="
 	$(MAKE) -C src
-	@echo "==== Successfully built LuaJIT $(VERSION) ===="
+	@echo "==== Successfully built LuaJIT $(MMVERSION) ===="

 install: $(INSTALL_DEP)
 	@echo "==== Installing LuaJIT $(VERSION) to $(PREFIX) ===="
@@ -134,18 +142,12 @@ install: $(INSTALL_DEP)
 	  $(RM) $(FILE_PC).tmp
 	cd src && $(INSTALL_F) $(FILES_INC) $(INSTALL_INC)
 	cd src/jit && $(INSTALL_F) $(FILES_JITLIB) $(INSTALL_JITLIB)
+	$(SYMLINK) $(INSTALL_TNAME) $(INSTALL_TSYM)
 	@echo "==== Successfully installed LuaJIT $(VERSION) to $(PREFIX) ===="
-	@echo ""
-	@echo "Note: the development releases deliberately do NOT install a symlink for luajit"
-	@echo "You can do this now by running this command (with sudo):"
-	@echo ""
-	@echo "  $(SYMLINK) $(INSTALL_TNAME) $(INSTALL_TSYM)"
-	@echo ""
-

 uninstall:
 	@echo "==== Uninstalling LuaJIT $(VERSION) from $(PREFIX) ===="
-	$(UNINSTALL) $(INSTALL_T) $(INSTALL_STATIC) $(INSTALL_DYN) $(INSTALL_SHORT1) $(INSTALL_SHORT2) $(INSTALL_MAN)/$(FILE_MAN) $(INSTALL_PC)
+	$(UNINSTALL) $(INSTALL_TSYM) $(INSTALL_T) $(INSTALL_STATIC) $(INSTALL_DYN) $(INSTALL_SHORT1) $(INSTALL_SHORT2) $(INSTALL_MAN)/$(FILE_MAN) $(INSTALL_PC)
 	for file in $(FILES_JITLIB); do \
 	  $(UNINSTALL) $(INSTALL_JITLIB)/$$file; \
 	  done
@@ -159,8 +161,9 @@ uninstall:
 ##############################################################################

 amalg:
-	@echo "Building LuaJIT $(VERSION)"
+	@echo "==== Building LuaJIT $(MMVERSION) (amalgamation) ===="
 	$(MAKE) -C src amalg
+	@echo "==== Successfully built LuaJIT $(MMVERSION) (amalgamation) ===="

 clean:
 	$(MAKE) -C src clean
@@ -1,11 +1,11 @@
-README for LuaJIT 2.1.0-beta3
-----------------------------
+README for LuaJIT 2.1
+---------------------

 LuaJIT is a Just-In-Time (JIT) compiler for the Lua programming language.

 Project Homepage: https://luajit.org/

-LuaJIT is Copyright (C) 2005-2022 Mike Pall.
+LuaJIT is Copyright (C) 2005-2023 Mike Pall.
 LuaJIT is free software, released under the MIT license.
 See full Copyright Notice in the COPYRIGHT file or in luajit.h.

@@ -1,4 +1,4 @@
-/* Copyright (C) 2004-2022 Mike Pall.
+/* Copyright (C) 2004-2023 Mike Pall.
 *
 * You are welcome to use the general ideas of this design for your own sites.
 * But please do not steal the stylesheet, the layout or the color scheme.
@@ -1,4 +1,4 @@
-/* Copyright (C) 2004-2022 Mike Pall.
+/* Copyright (C) 2004-2023 Mike Pall.
 *
 * You are welcome to use the general ideas of this design for your own sites.
 * But please do not steal the stylesheet, the layout or the color scheme.
@@ -206,11 +206,9 @@ img.right {
 .ext {
  color: #ff8000;
 }
-.new {
-  font-size: 6pt;
-  vertical-align: middle;
-  background: #ff8000;
-  color: #ffffff;
+.note {
+  padding: 0.5em 1em;
+  border-left: 3px solid #bfcfff;
 }
 #site {
  clear: both;
@@ -3,7 +3,7 @@
 <head>
 <title>Contact</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -46,9 +46,9 @@
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -94,7 +94,7 @@ don't like that, please complain to Google or Microsoft, not me.
 <h2>Copyright</h2>
 <p>
 All documentation is
-Copyright &copy; 2005-2022 Mike Pall.
+Copyright &copy; 2005-2023 Mike Pall.
 </p>


@@ -102,7 +102,7 @@ Copyright &copy; 2005-2022 Mike Pall.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>String Buffer Library</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -57,9 +57,9 @@
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -88,12 +88,6 @@ The string buffer library also includes a high-performance
 <a href="serialize">serializer</a> for Lua objects.
 </p>

-<h2 id="wip" style="color:#ff0000">Work in Progress</h2>
-<p>
-<b style="color:#ff0000">This library is a work in progress. More
-functionality will be added soon.</b>
-</p>
-
 <h2 id="use">Using the String Buffer Library</h2>
 <p>
 The string buffer library is built into LuaJIT by default, but it's not
@@ -685,7 +679,7 @@ mappings of files are OK, but only if the file does not change.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>Lua/C API Extensions</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -46,9 +46,9 @@
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -173,7 +173,7 @@ Also note that this mechanism is not without overhead.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>FFI Library</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -46,9 +46,9 @@
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -316,7 +316,7 @@ without undue conversion penalties.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>ffi.* API Functions</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -51,9 +51,9 @@ td.abiparam { font-weight: bold; width: 6em; }
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -463,8 +463,10 @@ otherwise. The following parameters are currently defined:
 <tr class="odd">
 <td class="abiparam">win</td><td class="abidesc">Windows variant of the standard ABI</td></tr>
 <tr class="even">
-<td class="abiparam">uwp</td><td class="abidesc">Universal Windows Platform</td></tr>
+<td class="abiparam">pauth</td><td class="abidesc">Pointer authentication ABI</td></tr>
 <tr class="odd">
+<td class="abiparam">uwp</td><td class="abidesc">Universal Windows Platform</td></tr>
+<tr class="even">
 <td class="abiparam">gc64</td><td class="abidesc">64 bit GC references</td></tr>
 </table>

@@ -556,7 +558,7 @@ named <tt>i</tt>.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>FFI Semantics</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -51,9 +51,9 @@ td.convop { font-style: italic; width: 40%; }
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -1246,7 +1246,7 @@ compiled.</li>
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>FFI Tutorial</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -53,9 +53,9 @@ td.idiomlua b { font-weight: normal; color: #2142bf; }
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -587,7 +587,7 @@ it to a local variable in the function scope is unnecessary.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>jit.* Library</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -46,9 +46,9 @@
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -135,7 +135,9 @@ Contains the LuaJIT version string.
 <h3 id="jit_version_num"><tt>jit.version_num</tt></h3>
 <p>
 Contains the version number of the LuaJIT core. Version xx.yy.zz
-is represented by the decimal number xxyyzz.
+is represented by the decimal number xxyyzz.<br>
+<b>DEPRECATED after the switch to
+<a href="https://luajit.org/status.html#release"><span class="ext">&raquo;</span>&nbsp;rolling releases</a>. zz is frozen at 99.</b>
 </p>

 <h3 id="jit_os"><tt>jit.os</tt></h3>
@@ -185,7 +187,7 @@ if you want to know more.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>Profiler</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -46,9 +46,9 @@
 <a class="current" href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -349,7 +349,7 @@ use.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>Extensions</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -63,14 +63,15 @@ td.excinterop {
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
 </div>
 <div id="main">
+
 <p>
 LuaJIT is fully upwards-compatible with Lua 5.1. It supports all
 <a href="https://www.lua.org/manual/5.1/manual.html#5"><span class="ext">&raquo;</span>&nbsp;standard Lua
@@ -425,9 +426,7 @@ the toolchain used to compile LuaJIT:
 on the C&nbsp;stack. The contents of the C++&nbsp;exception object
 pass through unmodified.</li>
 <li>Lua errors can be caught on the C++ side with <tt>catch(...)</tt>.
-The corresponding Lua error message can be retrieved from the Lua stack.<br>
-For MSVC for Windows 64 bit this requires compilation of your C++ code
-with <tt>/EHa</tt>.</li>
+The corresponding Lua error message can be retrieved from the Lua stack.</li>
 <li>Throwing Lua errors across C++ frames is safe. C++ destructors
 will be called.</li>
 </ul>
@@ -462,7 +461,7 @@ C++ destructors.</li>
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>Installation</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -20,7 +20,7 @@ table.compat tr.compathead td {
  font-weight: bold;
  border-bottom: 2px solid #bfcfff;
 }
-td.compatname, td.compatver {
+td.compatname {
  width: 10%;
 }
 td.compatbits {
@@ -69,18 +69,18 @@ td.compatx {
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
 </div>
 <div id="main">
 <p>
-LuaJIT is only distributed as a source package. This page explains
-how to build and install LuaJIT with different operating systems
-and C&nbsp;compilers.
+LuaJIT is only distributed as source code &mdash; get it from the
+<a href="https://luajit.org/download.html"><span class="ext">&raquo;</span>&nbsp;git repository</a>. This page explains how to build
+and install the LuaJIT binary and library for different operating systems.
 </p>
 <p>
 For the impatient (on POSIX systems):
@@ -90,201 +90,21 @@ make &amp;&amp; sudo make install
 </pre>

 <h2 id="req">Requirements</h2>
-<h3 id="systems">Systems</h3>
 <p>
-LuaJIT currently builds out-of-the box on most systems:
+LuaJIT currently builds out-of-the box on most systems. Please check the
+supported operating systems and CPU architectures on the
+<a href="https://luajit.org/status.html"><span class="ext">&raquo;</span>&nbsp;status page</a>.
 </p>
-<table class="compat">
-<tr class="compathead">
-<td class="compatname">OS</td>
-<td class="compatver">Min. Version</td>
-<td class="compatx">Requirements</td>
-<td class="compatx">LuaJIT Versions</td>
-</tr>
-<tr class="odd separate">
-<td class="compatname"><a href="#windows">Windows</a></td>
-<td class="compatver">7</td>
-<td class="compatx">x86 or x64, ARM64: TBA</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="even">
-<td class="compatname"><a href="#posix">Linux</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="odd">
-<td class="compatname"><a href="#posix">*BSD</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="even">
-<td class="compatname"><a href="#posix">macOS (OSX)</a></td>
-<td class="compatver">10.4</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">v2.1 &ndash;</td>
-</tr>
-<tr class="odd">
-<td class="compatname"><a href="#posix">POSIX</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">mmap, dlopen</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="even separate">
-<td class="compatname"><a href="#android">Android</a></td>
-<td class="compatver">4.0</td>
-<td class="compatx">Recent Android NDK</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="odd">
-<td class="compatname"><a href="#ios">iOS</a></td>
-<td class="compatver">3.0</td>
-<td class="compatx">Xcode iOS SDK</td>
-<td class="compatx">v2.1 &ndash;</td>
-</tr>
-<tr class="even separate">
-<td class="compatname"><a href="#consoles">PS3</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">PS3 SDK</td>
-<td class="compatx">v2.0 &ndash; v2.1 EOL</td>
-</tr>
-<tr class="odd">
-<td class="compatname"><a href="#consoles">PS4</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">PS4 SDK (ORBIS)</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="even">
-<td class="compatname"><a href="#consoles">PS5</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">PS5 SDK (PROSPERO)</td>
-<td class="compatx">v2.1 &ndash;</td>
-</tr>
-<tr class="odd">
-<td class="compatname"><a href="#consoles">PS Vita</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">PS Vita SDK (PSP2)</td>
-<td class="compatx">v2.0 &ndash; v2.1 EOL</td>
-</tr>
-<tr class="even">
-<td class="compatname"><a href="#consoles">Xbox 360</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">Xbox 360 SDK (XEDK)</td>
-<td class="compatx">v2.0 &ndash; v2.1 EOL</td>
-</tr>
-<tr class="odd">
-<td class="compatname"><a href="#consoles">Xbox One</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">Xbox One SDK (DURANGO)</td>
-<td class="compatx">v2.1 &ndash;</td>
-</tr>
-<tr class="even">
-<td class="compatname"><a href="#consoles">Nintendo Switch</a></td>
-<td class="compatver">&nbsp;</td>
-<td class="compatx">NintendoSDK + NX Addon</td>
-<td class="compatx">v2.1 &ndash;</td>
-</tr>
-</table>
-<p>
-The codebase has compatibility defines for some more systems, but
-without official support.
-</p>
-<h3 id="toolchains">Toolchains</h3>
 <p>
 Building LuaJIT requires a recent toolchain based on GCC, Clang/LLVM or
 MSVC++.
 </p>
 <p>
 The Makefile-based build system requires GNU Make and supports
-cross-builds. Batch files are provided for MSVC++ builds and console
 cross-builds.
 </p>
-<h3 id="architectures">CPU Architectures</h3>
-<table class="compat">
-<tr class="compathead">
-<td class="compatname">CPU</td>
-<td class="compatbits">Bits</td>
-<td class="compatx">Requirements</td>
-<td class="compatx">Variants</td>
-<td class="compatx">LuaJIT Versions</td>
-</tr>
-<tr class="odd separate">
-<td class="compatname">x86</td>
-<td class="compatbits">32</td>
-<td class="compatx">v2.1+: SSE2</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="even">
-<td class="compatname">x64</td>
-<td class="compatbits">64</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="odd">
-<td class="compatname">ARM</td>
-<td class="compatbits">32</td>
-<td class="compatx">ARMv5+, ARM9E+</td>
-<td class="compatx">hard-fp + soft-fp</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="even">
-<td class="compatname">ARM64</td>
-<td class="compatbits">64</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">ARM64le + ARM64be</td>
-<td class="compatx">v2.1 &ndash;</td>
-</tr>
-<tr class="odd">
-<td class="compatname">PPC32</td>
-<td class="compatbits">32</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">hard-fp + soft-fp</td>
-<td class="compatx">v2.0 &ndash; v2.1 EOL</td>
-</tr>
-<tr class="even">
-<td class="compatname">PPC/e500</td>
-<td class="compatbits">32</td>
-<td class="compatx">e500v2</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">v2.0 EOL</td>
-</tr>
-<tr class="odd">
-<td class="compatname">MIPS32</td>
-<td class="compatbits">32</td>
-<td class="compatx">MIPS32r1 &ndash; r5</td>
-<td class="compatx">hard-fp + soft-fp</td>
-<td class="compatx">v2.0 &ndash;</td>
-</tr>
-<tr class="even">
-<td class="compatname">MIPS64</td>
-<td class="compatbits">64</td>
-<td class="compatx">MIPS64r1 &ndash; r5</td>
-<td class="compatx">hard-fp + soft-fp</td>
-<td class="compatx">v2.1 &ndash;</td>
-</tr>
-<tr class="odd">
-<td class="compatname">MIPS64</td>
-<td class="compatbits">64</td>
-<td class="compatx">MIPS64r6</td>
-<td class="compatx">hard-fp + soft-fp</td>
-<td class="compatx">v2.1 EOL</td>
-</tr>
-<tr class="even">
-<td class="compatname">RISC-V</td>
-<td class="compatbits">64</td>
-<td class="compatx">RVA22+</td>
-<td class="compatx">&nbsp;</td>
-<td class="compatx">TBA</td>
-</tr>
-</table>
 <p>
-There are no plans to add historic architectures or to continue support
-for end-of-life (EOL) architectures, for which no new CPUs are commonly
-available anymore. Likewise, there are no plans to support marginal
-and/or de-facto-dead architectures.
+Batch files are provided for MSVC++ builds and console cross-builds.
 </p>

 <h2>Configuring LuaJIT</h2>
@@ -294,7 +114,6 @@ Usually there is no need to tweak the settings. The following files
 hold all user-configurable settings:
 </p>
 <ul>
-<li><tt>src/luaconf.h</tt> sets some configuration variables.</li>
 <li><tt>Makefile</tt> has settings for <b>installing</b> LuaJIT (POSIX
 only).</li>
 <li><tt>src/Makefile</tt> has settings for <b>compiling</b> LuaJIT
@@ -317,23 +136,11 @@ Please check the note about the
 <h2 id="posix">POSIX Systems (Linux, macOS, *BSD etc.)</h2>
 <h3>Prerequisites</h3>
 <p>
-Depending on your distribution, you may need to install a package for
-GCC, the development headers and/or a complete SDK. E.g. on a current
-Debian/Ubuntu, install <tt>libc6-dev</tt> with the package manager.
+Depending on your distribution, you may need to install a package for a
+compiler (GCC or Clang/LLVM), the development headers and/or a complete SDK.
+E.g. on a current Debian/Ubuntu, install <tt>build-essential</tt> with the
+package manager.
 </p>
-<p>
-The recommended way to fetch the latest version is to do a pull from
-the git repository.
-</p>
-<p>
-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):
-</p>
-<pre class="code">
-tar zxf LuaJIT-XX.YY.ZZ.tar.gz
-cd LuaJIT-XX.YY.ZZ
 </pre>
 <h3>Building LuaJIT</h3>
 <p>
@@ -394,15 +201,10 @@ Either install one of the open source SDKs
 GCC plus the required development headers.
 Or install Microsoft's Visual Studio (MSVC).
 </p>
-<p>
-Next, pull from the git repository or download the source package and
-unpack it using an archive manager (e.g. the Windows Explorer) to
-a directory of your choice.
-</p>
 <h3>Building with MSVC</h3>
 <p>
-Open a "Visual Studio Command Prompt" (either x86 or x64), <tt>cd</tt> to the
-directory where you've unpacked the sources and run these commands:
+Open a "Visual Studio Command Prompt" (x86, x64 or ARM64), <tt>cd</tt> to the
+directory with the source code and run these commands:
 </p>
 <pre class="code">
 cd src
@@ -412,11 +214,14 @@ msvcbuild
 Check the <tt>msvcbuild.bat</tt> file for more options.
 Then follow the installation instructions below.
 </p>
+<p>
+For an x64 to ARM64 cross-build run this first: <tt>vcvarsall.bat x64_arm64</tt>
+</p>
 <h3>Building with MinGW or Cygwin</h3>
 <p>
 Open a command prompt window and make sure the MinGW or Cygwin programs
-are in your path. Then <tt>cd</tt> to the directory of the git repository
-or where you've unpacked the sources. Then run this command for MinGW:
+are in your path. Then <tt>cd</tt> to the directory of the git repository.
+Then run this command for MinGW:
 </p>
 <pre class="code">
 mingw32-make
@@ -518,15 +323,15 @@ make HOST_CC="gcc -m32" CROSS=arm-linux-gnueabi- \
 make HOST_CC="gcc -m32" CROSS=arm-linux-gnueabihf-

 # ARM64
-make CROSS=aarch64-linux-
+make CROSS=aarch64-linux-gnu-

 # PPC
 make HOST_CC="gcc -m32" CROSS=powerpc-linux-gnu-

 # MIPS32 big-endian
-make HOST_CC="gcc -m32" CROSS=mips-linux-
+make HOST_CC="gcc -m32" CROSS=mips-linux-gnu-
 # MIPS32 little-endian
-make HOST_CC="gcc -m32" CROSS=mipsel-linux-
+make HOST_CC="gcc -m32" CROSS=mipsel-linux-gnu-

 # MIPS64 big-endian
 make CROSS=mips-linux- TARGET_CFLAGS="-mips64r2 -mabi=64"
@@ -605,8 +410,8 @@ make HOST_CC="gcc -m32" CROSS=ppu-lv2-
 To cross-compile for the other consoles from a Windows host, open a
 "Native Tools Command Prompt for VS". You need to choose either the 32
 or the 64&nbsp;bit version of the host compiler to match the target.
-Then <tt>cd</tt> to the <tt>src</tt> directory below where you've
-unpacked the sources and run the build command given in the table:
+Then <tt>cd</tt> to the <tt>src</tt> directory below the source code
+and run the build command given in the table:
 </p>
 <table class="compat">
 <tr class="compathead">
@@ -766,7 +571,7 @@ to me (the upstream) and not you (the package maintainer), anyway.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>LuaJIT</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -67,30 +67,6 @@ table.fcompat td {
  background-image: -o-linear-gradient(#41bfbf 10%, #b0ffff 95%);
  background-image: -ms-linear-gradient(#41bfbf 10%, #b0ffff 95%);
 }
-table.stats td {
-  color: #ffffff;
-  background: #a0a0a0;
-  background-image: linear-gradient(#808080 10%, #d0d0d0 95%);
-  background-image: -moz-linear-gradient(#808080 10%, #d0d0d0 95%);
-  background-image: -webkit-linear-gradient(#808080 10%, #d0d0d0 95%);
-  background-image: -o-linear-gradient(#808080 10%, #d0d0d0 95%);
-  background-image: -ms-linear-gradient(#808080 10%, #d0d0d0 95%);
-}
-table.stats td.speed {
-  color: #ff4020;
-}
-table.stats td.kb {
-  color: #ffff80;
-  background: #808080;
-  background-image: linear-gradient(#606060 10%, #c0c0c0 95%);
-  background-image: -moz-linear-gradient(#606060 10%, #c0c0c0 95%);
-  background-image: -webkit-linear-gradient(#606060 10%, #c0c0c0 95%);
-  background-image: -o-linear-gradient(#606060 10%, #c0c0c0 95%);
-  background-image: -ms-linear-gradient(#606060 10%, #c0c0c0 95%);
-}
-table.feature small {
-  font-size: 50%;
-}
 </style>
 </head>
 <body>
@@ -131,9 +107,9 @@ table.feature small {
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
@@ -146,7 +122,7 @@ Lua is a powerful, dynamic and light-weight programming language.
 It may be embedded or used as a general-purpose, stand-alone language.
 </p>
 <p>
-LuaJIT is Copyright &copy; 2005-2022 Mike Pall, released under the
+LuaJIT is Copyright &copy; 2005-2023 Mike Pall, released under the
 <a href="https://www.opensource.org/licenses/mit-license.php"><span class="ext">&raquo;</span>&nbsp;MIT open source license</a>.
 </p>
 <p>
@@ -173,23 +149,20 @@ LuaJIT is Copyright &copy; 2005-2022 Mike Pall, released under the
 </table>

 <h2>Overview</h2>
-<table class="feature stats">
-<tr>
-<td class="speed">3x<br>-&nbsp;&nbsp;100x</td>
-<td class="kb">115&nbsp;<small>KB</small><br>VM</td>
-<td class="kb">90&nbsp;<small>KB</small><br>JIT</td>
-<td class="kloc">63&nbsp;<small>KLOC</small><br>C</td>
-<td class="kloc">24&nbsp;<small>KLOC</small><br>ASM</td>
-<td class="kloc">11&nbsp;<small>KLOC</small><br>Lua</td>
-</tr>
-</table>
 <p style="margin-top: 1em;">
 LuaJIT has been successfully used as a <b>scripting middleware</b> in
 games, appliances, network and graphics apps, numerical simulations,
-trading platforms and many other specialty applications. It scales from
-embedded devices, smartphones, desktops up to server farms. It combines
-high flexibility with high performance
-and an unmatched <b>low memory footprint</b>.
+trading platforms and many other specialty applications.
+</p>
+<p>
+LuaJIT is part of a hundred million web sites, huge SaaS installations,
+network switches, set-top boxes and other embedded devices. You've probably
+already used LuaJIT without knowing about it.
+</p>
+<p>
+LuaJIT scales from embedded devices, smartphones, desktops up to server
+farms. It combines high flexibility with high performance and an unmatched
+<b>low memory footprint</b>.
 </p>
 <p>
 LuaJIT has been in continuous development since 2005. It's widely
@@ -220,7 +193,7 @@ Please select a sub-topic in the navigation bar to learn more about LuaJIT.
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -3,7 +3,7 @@
 <head>
 <title>Running LuaJIT</title>
 <meta charset="utf-8">
-<meta name="Copyright" content="Copyright (C) 2005-2022">
+<meta name="Copyright" content="Copyright (C) 2005-2023">
 <meta name="Language" content="en">
 <link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
 <link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
@@ -68,14 +68,15 @@ td.param_default {
 <a href="ext_profiler.html">Profiler</a>
 </li></ul>
 </li><li>
-<a href="status.html">Status</a>
+<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
 </li><li>
-<a href="faq.html">FAQ</a>
+<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
 </li><li>
 <a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
 </li></ul>
 </div>
 <div id="main">
+
 <p>
 LuaJIT has only a single stand-alone executable, called <tt>luajit</tt> on
 POSIX systems or <tt>luajit.exe</tt> on Windows. It can be used to run simple
@@ -119,7 +120,8 @@ file name:
 </p>
 <ul>
 <li><tt>c</tt> &mdash; C source file, exported bytecode data.</li>
-<li><tt>h</tt> &mdash; C header file, static bytecode data.</li>
+<li><tt>cc</tt> &mdash; C++ source file, exported bytecode data.</li>
+<li><tt>h</tt> &mdash; C/C++ header file, static bytecode data.</li>
 <li><tt>obj</tt> or <tt>o</tt> &mdash; Object file, exported bytecode data
 (OS- and architecture-specific).</li>
 <li><tt>raw</tt> or any other extension &mdash; Raw bytecode file (portable).
@@ -302,7 +304,7 @@ Here are the parameters and their default settings:
 </div>
 <div id="foot">
 <hr class="hide">
-Copyright &copy; 2005-2022
+Copyright &copy; 2005-2023
 <span class="noprint">
 &middot;
 <a href="contact.html">Contact</a>
@@ -1,6 +1,6 @@
 /*
 ** DynASM ARM encoding engine.
-** Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+** Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 ** Released under the MIT license. See dynasm.lua for full copyright notice.
 */

@@ -70,7 +70,7 @@ struct dasm_State {
  size_t lgsize;
  int *pclabels;		/* PC label chains/pos ptrs. */
  size_t pcsize;
-  void **globals;		/* Array of globals (bias -10). */
+  void **globals;		/* Array of globals. */
  dasm_Section *section;	/* Pointer to active section. */
  size_t codesize;		/* Total size of all code sections. */
  int maxsection;		/* 0 <= sectionidx < maxsection. */
@@ -87,7 +87,6 @@ void dasm_init(Dst_DECL, int maxsection)
 {
  dasm_State *D;
  size_t psz = 0;
-  int i;
  Dst_REF = NULL;
  DASM_M_GROW(Dst, struct dasm_State, Dst_REF, psz, DASM_PSZ(maxsection));
  D = Dst_REF;
@@ -98,12 +97,7 @@ void dasm_init(Dst_DECL, int maxsection)
  D->pcsize = 0;
  D->globals = NULL;
  D->maxsection = maxsection;
-  for (i = 0; i < maxsection; i++) {
-    D->sections[i].buf = NULL;  /* Need this for pass3. */
-    D->sections[i].rbuf = D->sections[i].buf - DASM_SEC2POS(i);
-    D->sections[i].bsize = 0;
-    D->sections[i].epos = 0;  /* Wrong, but is recalculated after resize. */
-  }
+  memset((void *)D->sections, 0, maxsection * sizeof(dasm_Section));
 }

 /* Free DynASM state. */
@@ -123,7 +117,7 @@ void dasm_free(Dst_DECL)
 void dasm_setupglobal(Dst_DECL, void **gl, unsigned int maxgl)
 {
  dasm_State *D = Dst_REF;
-  D->globals = gl - 10;  /* Negative bias to compensate for locals. */
+  D->globals = gl;
  DASM_M_GROW(Dst, int, D->lglabels, D->lgsize, (10+maxgl)*sizeof(int));
 }

@@ -148,6 +142,7 @@ void dasm_setup(Dst_DECL, const void *actionlist)
  if (D->pclabels) memset((void *)D->pclabels, 0, D->pcsize);
  for (i = 0; i < D->maxsection; i++) {
    D->sections[i].pos = DASM_SEC2POS(i);
+    D->sections[i].rbuf = D->sections[i].buf - D->sections[i].pos;
    D->sections[i].ofs = 0;
  }
 }
@@ -372,7 +367,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  break;
 	case DASM_REL_LG:
 	  if (n < 0) {
-	    n = (int)((ptrdiff_t)D->globals[-n] - (ptrdiff_t)cp - 4);
+	    n = (int)((ptrdiff_t)D->globals[-n-10] - (ptrdiff_t)cp - 4);
 	    goto patchrel;
 	  }
 	  /* fallthrough */
@@ -396,7 +391,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  }
 	  break;
 	case DASM_LABEL_LG:
-	  ins &= 2047; if (ins >= 20) D->globals[ins-10] = (void *)(base + n);
+	  ins &= 2047; if (ins >= 20) D->globals[ins-20] = (void *)(base + n);
 	  break;
 	case DASM_LABEL_PC: break;
 	case DASM_IMM:
@@ -1,7 +1,7 @@
 ------------------------------------------------------------------------------
 -- DynASM ARM module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- See dynasm.lua for full copyright notice.
 ------------------------------------------------------------------------------

@@ -1,6 +1,6 @@
 /*
 ** DynASM ARM64 encoding engine.
-** Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+** Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 ** Released under the MIT license. See dynasm.lua for full copyright notice.
 */

@@ -72,7 +72,7 @@ struct dasm_State {
  size_t lgsize;
  int *pclabels;		/* PC label chains/pos ptrs. */
  size_t pcsize;
-  void **globals;		/* Array of globals (bias -10). */
+  void **globals;		/* Array of globals. */
  dasm_Section *section;	/* Pointer to active section. */
  size_t codesize;		/* Total size of all code sections. */
  int maxsection;		/* 0 <= sectionidx < maxsection. */
@@ -89,7 +89,6 @@ void dasm_init(Dst_DECL, int maxsection)
 {
  dasm_State *D;
  size_t psz = 0;
-  int i;
  Dst_REF = NULL;
  DASM_M_GROW(Dst, struct dasm_State, Dst_REF, psz, DASM_PSZ(maxsection));
  D = Dst_REF;
@@ -100,12 +99,7 @@ void dasm_init(Dst_DECL, int maxsection)
  D->pcsize = 0;
  D->globals = NULL;
  D->maxsection = maxsection;
-  for (i = 0; i < maxsection; i++) {
-    D->sections[i].buf = NULL;  /* Need this for pass3. */
-    D->sections[i].rbuf = D->sections[i].buf - DASM_SEC2POS(i);
-    D->sections[i].bsize = 0;
-    D->sections[i].epos = 0;  /* Wrong, but is recalculated after resize. */
-  }
+  memset((void *)D->sections, 0, maxsection * sizeof(dasm_Section));
 }

 /* Free DynASM state. */
@@ -125,7 +119,7 @@ void dasm_free(Dst_DECL)
 void dasm_setupglobal(Dst_DECL, void **gl, unsigned int maxgl)
 {
  dasm_State *D = Dst_REF;
-  D->globals = gl - 10;  /* Negative bias to compensate for locals. */
+  D->globals = gl;
  DASM_M_GROW(Dst, int, D->lglabels, D->lgsize, (10+maxgl)*sizeof(int));
 }

@@ -150,6 +144,7 @@ void dasm_setup(Dst_DECL, const void *actionlist)
  if (D->pclabels) memset((void *)D->pclabels, 0, D->pcsize);
  for (i = 0; i < D->maxsection; i++) {
    D->sections[i].pos = DASM_SEC2POS(i);
+    D->sections[i].rbuf = D->sections[i].buf - D->sections[i].pos;
    D->sections[i].ofs = 0;
  }
 }
@@ -444,7 +439,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  break;
 	case DASM_REL_LG:
 	  if (n < 0) {
-	    ptrdiff_t na = (ptrdiff_t)D->globals[-n] - (ptrdiff_t)cp + 4;
+	    ptrdiff_t na = (ptrdiff_t)D->globals[-n-10] - (ptrdiff_t)cp + 4;
 	    n = (int)na;
 	    CK((ptrdiff_t)n == na, RANGE_REL);
 	    goto patchrel;
@@ -487,7 +482,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  goto patchrel;
 	}
 	case DASM_LABEL_LG:
-	  ins &= 2047; if (ins >= 20) D->globals[ins-10] = (void *)(base + n);
+	  ins &= 2047; if (ins >= 20) D->globals[ins-20] = (void *)(base + n);
 	  break;
 	case DASM_LABEL_PC: break;
 	case DASM_IMM:
@@ -1,7 +1,7 @@
 ------------------------------------------------------------------------------
 -- DynASM ARM64 module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- See dynasm.lua for full copyright notice.
 ------------------------------------------------------------------------------

@@ -549,7 +549,7 @@ end
 local function parse_load_pair(params, nparams, n, op)
  if params[n+2] then werror("too many operands") end
  local pn, p2 = params[n], params[n+1]
-  local scale = shr(op, 30) == 0 and 2 or 3
+  local scale = 2 + shr(op, 31 - band(shr(op, 26), 1))
  local p1, wb = match(pn, "^%[%s*(.-)%s*%](!?)$")
  if not p1 then
    if not p2 then
@@ -806,8 +806,8 @@ map_op = {
  ["ldrsw_*"] = "98000000DxB|b8800000DxL",
  -- NOTE: ldur etc. are handled by ldr et al.

-  ["stp_*"]   = "28000000DAwP|a8000000DAxP|2c000000DAsP|6c000000DAdP",
-  ["ldp_*"]   = "28400000DAwP|a8400000DAxP|2c400000DAsP|6c400000DAdP",
+  ["stp_*"]   = "28000000DAwP|a8000000DAxP|2c000000DAsP|6c000000DAdP|ac000000DAqP",
+  ["ldp_*"]   = "28400000DAwP|a8400000DAxP|2c400000DAsP|6c400000DAdP|ac400000DAqP",
  ["ldpsw_*"] = "68400000DAxP",

  -- Branches.
@@ -823,6 +823,13 @@ map_op = {
  tbz_3  = "36000000DTBw|36000000DTBx",
  tbnz_3 = "37000000DTBw|37000000DTBx",

+  -- ARM64e: Pointer authentication codes (PAC).
+  blraaz_1  = "d63f081fNx",
+  braa_2    = "d71f0800NDx",
+  braaz_1   = "d61f081fNx",
+  pacibsp_0 = "d503237f",
+  retab_0   = "d65f0fff",
+
  -- Miscellaneous instructions.
  -- TODO: hlt, hvc, smc, svc, eret, dcps[123], drps, mrs, msr
  -- TODO: sys, sysl, ic, dc, at, tlbi
@@ -935,7 +942,7 @@ local function parse_template(params, template, nparams, pos)
 	werror("bad register type")
      end
      parse_reg_type = false
-    elseif p == "x" or p == "w" or p == "d" or p == "s" then
+    elseif p == "x" or p == "w" or p == "d" or p == "s" or p == "q" then
      if parse_reg_type ~= p then
 	werror("register size mismatch")
      end
@@ -1,6 +1,6 @@
 /*
 ** DynASM MIPS encoding engine.
-** Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+** Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 ** Released under the MIT license. See dynasm.lua for full copyright notice.
 */

@@ -69,7 +69,7 @@ struct dasm_State {
  size_t lgsize;
  int *pclabels;		/* PC label chains/pos ptrs. */
  size_t pcsize;
-  void **globals;		/* Array of globals (bias -10). */
+  void **globals;		/* Array of globals. */
  dasm_Section *section;	/* Pointer to active section. */
  size_t codesize;		/* Total size of all code sections. */
  int maxsection;		/* 0 <= sectionidx < maxsection. */
@@ -86,7 +86,6 @@ void dasm_init(Dst_DECL, int maxsection)
 {
  dasm_State *D;
  size_t psz = 0;
-  int i;
  Dst_REF = NULL;
  DASM_M_GROW(Dst, struct dasm_State, Dst_REF, psz, DASM_PSZ(maxsection));
  D = Dst_REF;
@@ -97,12 +96,7 @@ void dasm_init(Dst_DECL, int maxsection)
  D->pcsize = 0;
  D->globals = NULL;
  D->maxsection = maxsection;
-  for (i = 0; i < maxsection; i++) {
-    D->sections[i].buf = NULL;  /* Need this for pass3. */
-    D->sections[i].rbuf = D->sections[i].buf - DASM_SEC2POS(i);
-    D->sections[i].bsize = 0;
-    D->sections[i].epos = 0;  /* Wrong, but is recalculated after resize. */
-  }
+  memset((void *)D->sections, 0, maxsection * sizeof(dasm_Section));
 }

 /* Free DynASM state. */
@@ -122,7 +116,7 @@ void dasm_free(Dst_DECL)
 void dasm_setupglobal(Dst_DECL, void **gl, unsigned int maxgl)
 {
  dasm_State *D = Dst_REF;
-  D->globals = gl - 10;  /* Negative bias to compensate for locals. */
+  D->globals = gl;
  DASM_M_GROW(Dst, int, D->lglabels, D->lgsize, (10+maxgl)*sizeof(int));
 }

@@ -147,6 +141,7 @@ void dasm_setup(Dst_DECL, const void *actionlist)
  if (D->pclabels) memset((void *)D->pclabels, 0, D->pcsize);
  for (i = 0; i < D->maxsection; i++) {
    D->sections[i].pos = DASM_SEC2POS(i);
+    D->sections[i].rbuf = D->sections[i].buf - D->sections[i].pos;
    D->sections[i].ofs = 0;
  }
 }
@@ -350,7 +345,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  break;
 	case DASM_REL_LG:
 	  if (n < 0) {
-	    n = (int)((ptrdiff_t)D->globals[-n] - (ptrdiff_t)cp);
+	    n = (int)((ptrdiff_t)D->globals[-n-10] - (ptrdiff_t)cp);
 	    goto patchrel;
 	  }
 	  /* fallthrough */
@@ -369,7 +364,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  }
 	  break;
 	case DASM_LABEL_LG:
-	  ins &= 2047; if (ins >= 20) D->globals[ins-10] = (void *)(base + n);
+	  ins &= 2047; if (ins >= 20) D->globals[ins-20] = (void *)(base + n);
 	  break;
 	case DASM_LABEL_PC: break;
 	case DASM_IMMS:
@@ -1,7 +1,7 @@
 ------------------------------------------------------------------------------
 -- DynASM MIPS32/MIPS64 module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- See dynasm.lua for full copyright notice.
 ------------------------------------------------------------------------------

@@ -1,7 +1,7 @@
 ------------------------------------------------------------------------------
 -- DynASM MIPS64 module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- See dynasm.lua for full copyright notice.
 ------------------------------------------------------------------------------
 -- This module just sets 64 bit mode for the combined MIPS/MIPS64 module.
@@ -1,6 +1,6 @@
 /*
 ** DynASM PPC/PPC64 encoding engine.
-** Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+** Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 ** Released under the MIT license. See dynasm.lua for full copyright notice.
 */

@@ -69,7 +69,7 @@ struct dasm_State {
  size_t lgsize;
  int *pclabels;		/* PC label chains/pos ptrs. */
  size_t pcsize;
-  void **globals;		/* Array of globals (bias -10). */
+  void **globals;		/* Array of globals. */
  dasm_Section *section;	/* Pointer to active section. */
  size_t codesize;		/* Total size of all code sections. */
  int maxsection;		/* 0 <= sectionidx < maxsection. */
@@ -86,7 +86,6 @@ void dasm_init(Dst_DECL, int maxsection)
 {
  dasm_State *D;
  size_t psz = 0;
-  int i;
  Dst_REF = NULL;
  DASM_M_GROW(Dst, struct dasm_State, Dst_REF, psz, DASM_PSZ(maxsection));
  D = Dst_REF;
@@ -97,12 +96,7 @@ void dasm_init(Dst_DECL, int maxsection)
  D->pcsize = 0;
  D->globals = NULL;
  D->maxsection = maxsection;
-  for (i = 0; i < maxsection; i++) {
-    D->sections[i].buf = NULL;  /* Need this for pass3. */
-    D->sections[i].rbuf = D->sections[i].buf - DASM_SEC2POS(i);
-    D->sections[i].bsize = 0;
-    D->sections[i].epos = 0;  /* Wrong, but is recalculated after resize. */
-  }
+  memset((void *)D->sections, 0, maxsection * sizeof(dasm_Section));
 }

 /* Free DynASM state. */
@@ -122,7 +116,7 @@ void dasm_free(Dst_DECL)
 void dasm_setupglobal(Dst_DECL, void **gl, unsigned int maxgl)
 {
  dasm_State *D = Dst_REF;
-  D->globals = gl - 10;  /* Negative bias to compensate for locals. */
+  D->globals = gl;
  DASM_M_GROW(Dst, int, D->lglabels, D->lgsize, (10+maxgl)*sizeof(int));
 }

@@ -147,6 +141,7 @@ void dasm_setup(Dst_DECL, const void *actionlist)
  if (D->pclabels) memset((void *)D->pclabels, 0, D->pcsize);
  for (i = 0; i < D->maxsection; i++) {
    D->sections[i].pos = DASM_SEC2POS(i);
+    D->sections[i].rbuf = D->sections[i].buf - D->sections[i].pos;
    D->sections[i].ofs = 0;
  }
 }
@@ -354,7 +349,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  break;
 	case DASM_REL_LG:
 	  if (n < 0) {
-	    n = (int)((ptrdiff_t)D->globals[-n] - (ptrdiff_t)cp);
+	    n = (int)((ptrdiff_t)D->globals[-n-10] - (ptrdiff_t)cp);
 	    goto patchrel;
 	  }
 	  /* fallthrough */
@@ -368,7 +363,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  cp[-1] |= ((n+4) & ((ins & 2048) ? 0x0000fffc: 0x03fffffc));
 	  break;
 	case DASM_LABEL_LG:
-	  ins &= 2047; if (ins >= 20) D->globals[ins-10] = (void *)(base + n);
+	  ins &= 2047; if (ins >= 20) D->globals[ins-20] = (void *)(base + n);
 	  break;
 	case DASM_LABEL_PC: break;
 	case DASM_IMM:
@@ -1,7 +1,7 @@
 ------------------------------------------------------------------------------
 -- DynASM PPC/PPC64 module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- See dynasm.lua for full copyright notice.
 --
 -- Support for various extensions contributed by Caio Souza Oliveira.
@@ -1,6 +1,6 @@
 /*
 ** DynASM encoding engine prototypes.
-** Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+** Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 ** Released under the MIT license. See dynasm.lua for full copyright notice.
 */

@@ -1,7 +1,7 @@
 ------------------------------------------------------------------------------
 -- DynASM x64 module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- See dynasm.lua for full copyright notice.
 ------------------------------------------------------------------------------
 -- This module just sets 64 bit mode for the combined x86/x64 module.
@@ -1,6 +1,6 @@
 /*
 ** DynASM x86 encoding engine.
-** Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+** Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 ** Released under the MIT license. See dynasm.lua for full copyright notice.
 */

@@ -68,7 +68,7 @@ struct dasm_State {
  size_t lgsize;
  int *pclabels;		/* PC label chains/pos ptrs. */
  size_t pcsize;
-  void **globals;		/* Array of globals (bias -10). */
+  void **globals;		/* Array of globals. */
  dasm_Section *section;	/* Pointer to active section. */
  size_t codesize;		/* Total size of all code sections. */
  int maxsection;		/* 0 <= sectionidx < maxsection. */
@@ -85,7 +85,6 @@ void dasm_init(Dst_DECL, int maxsection)
 {
  dasm_State *D;
  size_t psz = 0;
-  int i;
  Dst_REF = NULL;
  DASM_M_GROW(Dst, struct dasm_State, Dst_REF, psz, DASM_PSZ(maxsection));
  D = Dst_REF;
@@ -96,12 +95,7 @@ void dasm_init(Dst_DECL, int maxsection)
  D->pcsize = 0;
  D->globals = NULL;
  D->maxsection = maxsection;
-  for (i = 0; i < maxsection; i++) {
-    D->sections[i].buf = NULL;  /* Need this for pass3. */
-    D->sections[i].rbuf = D->sections[i].buf - DASM_SEC2POS(i);
-    D->sections[i].bsize = 0;
-    D->sections[i].epos = 0;  /* Wrong, but is recalculated after resize. */
-  }
+  memset((void *)D->sections, 0, maxsection * sizeof(dasm_Section));
 }

 /* Free DynASM state. */
@@ -121,7 +115,7 @@ void dasm_free(Dst_DECL)
 void dasm_setupglobal(Dst_DECL, void **gl, unsigned int maxgl)
 {
  dasm_State *D = Dst_REF;
-  D->globals = gl - 10;  /* Negative bias to compensate for locals. */
+  D->globals = gl;
  DASM_M_GROW(Dst, int, D->lglabels, D->lgsize, (10+maxgl)*sizeof(int));
 }

@@ -146,6 +140,7 @@ void dasm_setup(Dst_DECL, const void *actionlist)
  if (D->pclabels) memset((void *)D->pclabels, 0, D->pcsize);
  for (i = 0; i < D->maxsection; i++) {
    D->sections[i].pos = DASM_SEC2POS(i);
+    D->sections[i].rbuf = D->sections[i].buf - D->sections[i].pos;
    D->sections[i].ofs = 0;
  }
 }
@@ -445,7 +440,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	  break;
 	}
 	case DASM_REL_LG: p++; if (n >= 0) goto rel_pc;
-	  b++; n = (int)(ptrdiff_t)D->globals[-n];
+	  b++; n = (int)(ptrdiff_t)D->globals[-n-10];
 	  /* fallthrough */
 	case DASM_REL_A: rel_a:
 	  n -= (unsigned int)(ptrdiff_t)(cp+4); goto wd; /* !x64 */
@@ -459,7 +454,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	}
 	case DASM_IMM_LG:
 	  p++;
-	  if (n < 0) { dasma((ptrdiff_t)D->globals[-n]); break; }
+	  if (n < 0) { dasma((ptrdiff_t)D->globals[-n-10]); break; }
 	  /* fallthrough */
 	case DASM_IMM_PC: {
 	  int *pb = DASM_POS2PTR(D, n);
@@ -469,7 +464,7 @@ int dasm_encode(Dst_DECL, void *buffer)
 	case DASM_LABEL_LG: {
 	  int idx = *p++;
 	  if (idx >= 10)
-	    D->globals[idx] = (void *)(base + (*p == DASM_SETLABEL ? *b : n));
+	    D->globals[idx-10] = (void *)(base + (*p == DASM_SETLABEL ? *b : n));
 	  break;
 	}
 	case DASM_LABEL_PC: case DASM_SETLABEL: break;
@@ -1,7 +1,7 @@
 ------------------------------------------------------------------------------
 -- DynASM x86/x64 module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- See dynasm.lua for full copyright notice.
 ------------------------------------------------------------------------------

@@ -2,7 +2,7 @@
 -- DynASM. A dynamic assembler for code generation engines.
 -- Originally designed and implemented for LuaJIT.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- See below for full copyright notice.
 ------------------------------------------------------------------------------

@@ -17,7 +17,7 @@ local _info = {
  url =		"https://luajit.org/dynasm.html",
  license =	"MIT",
  copyright =	[[
-Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+Copyright (C) 2005-2023 Mike Pall. All rights reserved.

 Permission is hereby granted, free of charge, to any person obtaining
 a copy of this software and associated documentation files (the
@@ -74,7 +74,7 @@ luajit \-jv \-e "for i=1,10 do for j=1,10 do for k=1,100 do end end end"
 Runs some nested loops and shows the resulting traces.
 .SH COPYRIGHT
 .PP
-\fBLuaJIT\fR is Copyright \(co 2005-2022 Mike Pall.
+\fBLuaJIT\fR is Copyright \(co 2005-2023 Mike Pall.
 .br
 \fBLuaJIT\fR is open source software, released under the MIT license.
 .SH SEE ALSO
@@ -1,8 +1,8 @@
 # Package information for LuaJIT to be used by pkg-config.
 majver=2
 minver=1
-relver=0
-version=${majver}.${minver}.${relver}-beta3
+relver=ROLLING
+version=${majver}.${minver}.${relver}
 abiver=5.1

 prefix=/usr/local
@@ -1,4 +1,6 @@
 luajit
+luajit.h
+luajit_relver.txt
 lj_bcdef.h
 lj_ffdef.h
 lj_libdef.h
@@ -7,12 +7,11 @@
 # Also works with MinGW and Cygwin on Windows.
 # Please check msvcbuild.bat for building with MSVC on Windows.
 #
-# Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+# Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 ##############################################################################

 MAJVER=  2
 MINVER=  1
-RELVER=  0
 ABIVER=  5.1
 NODOTABIVER= 51

@@ -234,7 +233,7 @@ TARGET_ALDFLAGS= $(LDOPTIONS) $(TARGET_XLDFLAGS) $(TARGET_FLAGS) $(TARGET_LDFLAG
 TARGET_ASHLDFLAGS= $(LDOPTIONS) $(TARGET_XSHLDFLAGS) $(TARGET_FLAGS) $(TARGET_SHLDFLAGS)
 TARGET_ALIBS= $(TARGET_XLIBS) $(LIBS) $(TARGET_LIBS)

-TARGET_TESTARCH=$(shell $(TARGET_CC) $(TARGET_TCFLAGS) -E lj_arch.h -dM)
+TARGET_TESTARCH:=$(shell $(TARGET_CC) $(TARGET_TCFLAGS) -E lj_arch.h -dM)
 ifneq (,$(findstring LJ_TARGET_X64 ,$(TARGET_TESTARCH)))
  TARGET_LJARCH= x64
 else
@@ -323,13 +322,13 @@ ifeq (Darwin,$(TARGET_SYS))
  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)
+  TARGET_XSHLDFLAGS+= -install_name $(TARGET_DYLIBPATH) -compatibility_version $(MAJVER).$(MINVER) -current_version $(MAJVER).$(MINVER).255
 else
 ifeq (iOS,$(TARGET_SYS))
  TARGET_STRIP+= -x
  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)
+  TARGET_XSHLDFLAGS+= -install_name $(TARGET_DYLIBPATH) -compatibility_version $(MAJVER).$(MINVER) -current_version $(MAJVER).$(MINVER).255
  ifeq (arm64,$(TARGET_LJARCH))
    TARGET_XCFLAGS+= -fno-omit-frame-pointer
  endif
@@ -388,10 +387,11 @@ MINILUA_O= host/minilua.o
 MINILUA_LIBS= -lm
 MINILUA_T= host/minilua
 MINILUA_X= $(MINILUA_T)
+MINILUA_DEP=

 ifeq (,$(HOST_LUA))
  HOST_LUA= $(MINILUA_X)
-  DASM_DEP= $(MINILUA_T)
+  MINILUA_DEP= $(MINILUA_T)
 endif

 DASM_DIR= ../dynasm
@@ -433,6 +433,10 @@ ifneq (,$(findstring LJ_NO_UNWIND 1,$(TARGET_TESTARCH)))
  DASM_AFLAGS+= -D NO_UNWIND
  TARGET_ARCH+= -DLUAJIT_NO_UNWIND
 endif
+ifneq (,$(findstring LJ_ABI_PAUTH 1,$(TARGET_TESTARCH)))
+  DASM_AFLAGS+= -D PAUTH
+  TARGET_ARCH+= -DLJ_ABI_PAUTH=1
+endif
 DASM_AFLAGS+= -D VER=$(subst LJ_ARCH_VERSION_,,$(filter LJ_ARCH_VERSION_%,$(subst LJ_ARCH_VERSION ,LJ_ARCH_VERSION_,$(TARGET_TESTARCH))))
 ifeq (Windows,$(TARGET_SYS))
  DASM_AFLAGS+= -D WIN
@@ -470,6 +474,14 @@ endif
 DASM_FLAGS= $(DASM_XFLAGS) $(DASM_AFLAGS)
 DASM_DASC= vm_$(DASM_ARCH).dasc

+GIT= git
+ifeq (Windows,$(HOST_SYS)$(HOST_MSYS))
+  GIT_RELVER= if exist ..\.git ( $(GIT) show -s --format=%%ct >luajit_relver.txt ) else ( type ..\.relver >luajit_relver.txt )
+else
+  GIT_RELVER= [ -e ../.git ] && $(GIT) show -s --format=%ct >luajit_relver.txt 2>/dev/null || cat ../.relver >luajit_relver.txt 2>/dev/null || :
+endif
+GIT_DEP= $(wildcard ../.git/HEAD ../.git/refs/heads/*)
+
 BUILDVM_O= host/buildvm.o host/buildvm_asm.o host/buildvm_peobj.o \
 	   host/buildvm_lib.o host/buildvm_fold.o
 BUILDVM_T= host/buildvm
@@ -516,8 +528,8 @@ LUAJIT_T= luajit

 ALL_T= $(LUAJIT_T) $(LUAJIT_A) $(LUAJIT_SO) $(HOST_T)
 ALL_HDRGEN= lj_bcdef.h lj_ffdef.h lj_libdef.h lj_recdef.h lj_folddef.h \
-	    host/buildvm_arch.h
-ALL_GEN= $(LJVM_S) $(ALL_HDRGEN) $(LIB_VMDEFP)
+	    host/buildvm_arch.h luajit.h
+ALL_GEN= $(LJVM_S) $(ALL_HDRGEN) luajit_relver.txt $(LIB_VMDEFP)
 WIN_RM= *.obj *.lib *.exp *.dll *.exe *.manifest *.pdb *.ilk
 ALL_RM= $(ALL_T) $(ALL_GEN) *.o host/*.o $(WIN_RM)

@@ -641,7 +653,12 @@ $(MINILUA_T): $(MINILUA_O)
 	$(E) "HOSTLINK  $@"
 	$(Q)$(HOST_CC) $(HOST_ALDFLAGS) -o $@ $(MINILUA_O) $(MINILUA_LIBS) $(HOST_ALIBS)

-host/buildvm_arch.h: $(DASM_DASC) $(DASM_DEP) $(DASM_DIR)/*.lua lj_arch.h lua.h luaconf.h
+luajit.h: $(MINILUA_DEP) $(GIT_DEP) luajit_rolling.h
+	$(E) "VERSION   $@"
+	$(Q)$(GIT_RELVER)
+	$(Q)$(HOST_LUA) host/genversion.lua
+
+host/buildvm_arch.h: $(DASM_DASC) $(MINILUA_DEP) lj_arch.h lua.h luaconf.h
 	$(E) "DYNASM    $@"
 	$(Q)$(DASM) $(DASM_FLAGS) -o $@ $(DASM_DASC)

@@ -55,7 +55,7 @@ lj_asm.o: lj_asm.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_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_prng.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
@@ -1,6 +1,6 @@
 /*
 ** LuaJIT VM builder.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** This is a tool to build the hand-tuned assembler code required for
 ** LuaJIT's bytecode interpreter. It supports a variety of output formats
@@ -329,6 +329,7 @@ static void emit_vmdef(BuildCtx *ctx)
 #endif
  int i;
  fprintf(ctx->fp, "-- This is a generated file. DO NOT EDIT!\n\n");
+  fprintf(ctx->fp, "assert(require(\"jit\").version == \"%s\", \"LuaJIT core/library version mismatch\")\n\n", LUAJIT_VERSION);
  fprintf(ctx->fp, "return {\n\n");

  fprintf(ctx->fp, "bcnames = \"");
@@ -1,6 +1,6 @@
 /*
 ** LuaJIT VM builder.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #ifndef _BUILDVM_H
@@ -1,6 +1,6 @@
 /*
 ** LuaJIT VM builder: Assembler source code emitter.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #include "buildvm.h"
@@ -243,6 +243,12 @@ void emit_asm(BuildCtx *ctx)

  fprintf(ctx->fp, "\t.file \"buildvm_%s.dasc\"\n", ctx->dasm_arch);
  fprintf(ctx->fp, "\t.text\n");
+#if LJ_TARGET_MIPS32 && !LJ_ABI_SOFTFP
+  fprintf(ctx->fp, "\t.module fp=32\n");
+#endif
+#if LJ_TARGET_MIPS
+  fprintf(ctx->fp, "\t.set nomips16\n\t.abicalls\n\t.set noreorder\n\t.set nomacro\n");
+#endif
  emit_asm_align(ctx, 4);

 #if LJ_TARGET_PS3
@@ -268,9 +274,6 @@ void emit_asm(BuildCtx *ctx)
 	  ".save {r4, r5, r6, r7, r8, r9, r10, r11, lr}\n"
 	  ".pad #28\n");
 #endif
-#endif
-#if LJ_TARGET_MIPS
-  fprintf(ctx->fp, ".set nomips16\n.abicalls\n.set noreorder\n.set nomacro\n");
 #endif

  for (i = rel = 0; i < ctx->nsym; i++) {
@@ -1,6 +1,6 @@
 /*
 ** LuaJIT VM builder: IR folding hash table generator.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #include "buildvm.h"
@@ -1,6 +1,6 @@
 /*
 ** LuaJIT VM builder: library definition compiler.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #include "buildvm.h"
@@ -378,7 +378,8 @@ void emit_lib(BuildCtx *ctx)
      char *p;
      /* Simplistic pre-processor. Only handles top-level #if/#endif. */
      if (buf[0] == '#' && buf[1] == 'i' && buf[2] == 'f') {
-	int ok = 1, len = strlen(buf);
+	int ok = 1;
+	size_t len = strlen(buf);
 	if (buf[len-1] == '\n') {
 	  buf[len-1] = 0;
 	  if (buf[len-2] == '\r') {
@@ -1,6 +1,6 @@
 /*
 ** LuaJIT VM builder: PE object emitter.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Only used for building on Windows, since we cannot assume the presence
 ** of a suitable assembler. The host and target byte order must match.
@@ -9,7 +9,7 @@
 #include "buildvm.h"
 #include "lj_bc.h"

-#if LJ_TARGET_X86ORX64
+#if LJ_TARGET_WINDOWS || LJ_TARGET_CYGWIN

 /* Context for PE object emitter. */
 static char *strtab;
@@ -93,6 +93,17 @@ typedef struct PEsymaux {
 #define PEOBJ_RELOC_ADDR32NB	0x03
 #define PEOBJ_RELOC_OFS		0
 #define PEOBJ_TEXT_FLAGS	0x60500020  /* 60=r+x, 50=align16, 20=code. */
+#define PEOBJ_PDATA_NRELOC	6
+#define PEOBJ_XDATA_SIZE	(8*2+4+6*2)
+#elif LJ_TARGET_ARM64
+#define PEOBJ_ARCH_TARGET	0xaa64
+#define PEOBJ_RELOC_REL32	0x03  /* MS: BRANCH26. */
+#define PEOBJ_RELOC_DIR32	0x01
+#define PEOBJ_RELOC_ADDR32NB	0x02
+#define PEOBJ_RELOC_OFS		(-4)
+#define PEOBJ_TEXT_FLAGS	0x60500020  /* 60=r+x, 50=align16, 20=code. */
+#define PEOBJ_PDATA_NRELOC	4
+#define PEOBJ_XDATA_SIZE	(4+24+4 +4+8)
 #endif

 /* Section numbers (0-based). */
@@ -100,7 +111,7 @@ enum {
  PEOBJ_SECT_ABS = -2,
  PEOBJ_SECT_UNDEF = -1,
  PEOBJ_SECT_TEXT,
-#if LJ_TARGET_X64
+#ifdef PEOBJ_PDATA_NRELOC
  PEOBJ_SECT_PDATA,
  PEOBJ_SECT_XDATA,
 #elif LJ_TARGET_X86
@@ -175,6 +186,9 @@ void emit_peobj(BuildCtx *ctx)
  uint32_t sofs;
  int i, nrsym;
  union { uint8_t b; uint32_t u; } host_endian;
+#ifdef PEOBJ_PDATA_NRELOC
+  uint32_t fcofs = (uint32_t)ctx->sym[ctx->nsym-1].ofs;
+#endif

  sofs = sizeof(PEheader) + PEOBJ_NSECTIONS*sizeof(PEsection);

@@ -188,18 +202,18 @@ void emit_peobj(BuildCtx *ctx)
  /* Flags: 60 = read+execute, 50 = align16, 20 = code. */
  pesect[PEOBJ_SECT_TEXT].flags = PEOBJ_TEXT_FLAGS;

-#if LJ_TARGET_X64
+#ifdef PEOBJ_PDATA_NRELOC
  memcpy(pesect[PEOBJ_SECT_PDATA].name, ".pdata", sizeof(".pdata")-1);
  pesect[PEOBJ_SECT_PDATA].ofs = sofs;
-  sofs += (pesect[PEOBJ_SECT_PDATA].size = 6*4);
+  sofs += (pesect[PEOBJ_SECT_PDATA].size = PEOBJ_PDATA_NRELOC*4);
  pesect[PEOBJ_SECT_PDATA].relocofs = sofs;
-  sofs += (pesect[PEOBJ_SECT_PDATA].nreloc = 6) * PEOBJ_RELOC_SIZE;
+  sofs += (pesect[PEOBJ_SECT_PDATA].nreloc = PEOBJ_PDATA_NRELOC) * PEOBJ_RELOC_SIZE;
  /* Flags: 40 = read, 30 = align4, 40 = initialized data. */
  pesect[PEOBJ_SECT_PDATA].flags = 0x40300040;

  memcpy(pesect[PEOBJ_SECT_XDATA].name, ".xdata", sizeof(".xdata")-1);
  pesect[PEOBJ_SECT_XDATA].ofs = sofs;
-  sofs += (pesect[PEOBJ_SECT_XDATA].size = 8*2+4+6*2);  /* See below. */
+  sofs += (pesect[PEOBJ_SECT_XDATA].size = PEOBJ_XDATA_SIZE);  /* See below. */
  pesect[PEOBJ_SECT_XDATA].relocofs = sofs;
  sofs += (pesect[PEOBJ_SECT_XDATA].nreloc = 1) * PEOBJ_RELOC_SIZE;
  /* Flags: 40 = read, 30 = align4, 40 = initialized data. */
@@ -234,7 +248,7 @@ void emit_peobj(BuildCtx *ctx)
  */
  nrsym = ctx->nrelocsym;
  pehdr.nsyms = 1+PEOBJ_NSECTIONS*2 + 1+ctx->nsym + nrsym;
-#if LJ_TARGET_X64
+#ifdef PEOBJ_PDATA_NRELOC
  pehdr.nsyms += 1;  /* Symbol for lj_err_unwind_win. */
 #endif

@@ -259,7 +273,6 @@ void emit_peobj(BuildCtx *ctx)

 #if LJ_TARGET_X64
  { /* Write .pdata section. */
-    uint32_t fcofs = (uint32_t)ctx->sym[ctx->nsym-1].ofs;
    uint32_t pdata[3];  /* Start of .text, end of .text and .xdata. */
    PEreloc reloc;
    pdata[0] = 0; pdata[1] = fcofs; pdata[2] = 0;
@@ -308,6 +321,87 @@ void emit_peobj(BuildCtx *ctx)
    reloc.type = PEOBJ_RELOC_ADDR32NB;
    owrite(ctx, &reloc, PEOBJ_RELOC_SIZE);
  }
+#elif LJ_TARGET_ARM64
+  /* https://learn.microsoft.com/en-us/cpp/build/arm64-exception-handling */
+  { /* Write .pdata section. */
+    uint32_t pdata[4];
+    PEreloc reloc;
+    pdata[0] = 0;
+    pdata[1] = 0;
+    pdata[2] = fcofs;
+    pdata[3] = 4+24+4;
+    owrite(ctx, &pdata, sizeof(pdata));
+    /* Start of .text and start of .xdata. */
+    reloc.vaddr = 0; reloc.symidx = 1+2+nrsym+2+2+1;
+    reloc.type = PEOBJ_RELOC_ADDR32NB;
+    owrite(ctx, &reloc, PEOBJ_RELOC_SIZE);
+    reloc.vaddr = 4; reloc.symidx = 1+2+nrsym+2;
+    reloc.type = PEOBJ_RELOC_ADDR32NB;
+    owrite(ctx, &reloc, PEOBJ_RELOC_SIZE);
+    /* Start of vm_ffi_call and start of second part of .xdata. */
+    reloc.vaddr = 8; reloc.symidx = 1+2+nrsym+2+2+1;
+    reloc.type = PEOBJ_RELOC_ADDR32NB;
+    owrite(ctx, &reloc, PEOBJ_RELOC_SIZE);
+    reloc.vaddr = 12; reloc.symidx = 1+2+nrsym+2;
+    reloc.type = PEOBJ_RELOC_ADDR32NB;
+    owrite(ctx, &reloc, PEOBJ_RELOC_SIZE);
+  }
+  { /* Write .xdata section. */
+    uint32_t u32;
+    uint8_t *p, uwc[24];
+    PEreloc reloc;
+
+#define CBE16(x)	(*p = ((x) >> 8) & 0xff, p[1] = (x) & 0xff, p += 2)
+#define CALLOC_S(s)	(*p++ = ((s) >> 4))  /* s < 512 */
+#define CSAVE_FPLR(o)	(*p++ = 0x40 | ((o) >> 3))  /* o <= 504 */
+#define CSAVE_REGP(r,o)	CBE16(0xc800 | (((r) - 19) << 6) | ((o) >> 3))
+#define CSAVE_REGS(r1,r2,o1) do { \
+  int r, o; for (r = r1, o = o1; r <= r2; r += 2, o -= 16) CSAVE_REGP(r, o); \
+} while (0)
+#define CSAVE_REGPX(r,o) CBE16(0xcc00 | (((r) - 19) << 6) | (~(o) >> 3))
+#define CSAVE_FREGP(r,o) CBE16(0xd800 | (((r) - 8) << 6) | ((o) >> 3))
+#define CSAVE_FREGS(r1,r2,o1) do { \
+  int r, o; for (r = r1, o = o1; r <= r2; r += 2, o -= 16) CSAVE_FREGP(r, o); \
+} while (0)
+#define CADD_FP(s)	CBE16(0xe200 | ((s) >> 3))  /* s < 8*256 */
+#define CODE_NOP	0xe3
+#define CODE_END	0xe4
+#define CEND_ALIGN	do { \
+  *p++ = CODE_END; \
+  while ((p - uwc) & 3) *p++ = CODE_NOP; \
+} while (0)
+
+    /* Unwind codes for .text section with handler. */
+    p = uwc;
+    CADD_FP(192);		/* +2 */
+    CSAVE_REGS(19, 28, 176);	/* +5*2 */
+    CSAVE_FREGS(8, 15, 96);	/* +4*2 */
+    CSAVE_FPLR(192);		/* +1 */
+    CALLOC_S(208);		/* +1 */
+    CEND_ALIGN;			/* +1 +1 -> 24 */
+
+    u32 = ((24u >> 2) << 27) | (1u << 20) | (fcofs >> 2);
+    owrite(ctx, &u32, 4);
+    owrite(ctx, &uwc, 24);
+
+    u32 = 0;  /* Handler RVA to be relocated at 4 + 24. */
+    owrite(ctx, &u32, 4);
+
+    /* Unwind codes for vm_ffi_call without handler. */
+    p = uwc;
+    CADD_FP(16);		/* +2 */
+    CSAVE_FPLR(16);		/* +1 */
+    CSAVE_REGPX(19, -32);	/* +2 */
+    CEND_ALIGN;			/* +1 +2 -> 8 */
+
+    u32 = ((8u >> 2) << 27) | (((uint32_t)ctx->codesz - fcofs) >> 2);
+    owrite(ctx, &u32, 4);
+    owrite(ctx, &uwc, 8);
+
+    reloc.vaddr = 4 + 24; reloc.symidx = 1+2+nrsym+2+2;
+    reloc.type = PEOBJ_RELOC_ADDR32NB;
+    owrite(ctx, &reloc, PEOBJ_RELOC_SIZE);
+  }
 #elif LJ_TARGET_X86
  /* Write .sxdata section. */
  for (i = 0; i < nrsym; i++) {
@@ -339,7 +433,7 @@ void emit_peobj(BuildCtx *ctx)
      emit_peobj_sym(ctx, ctx->relocsym[i], 0,
 		     PEOBJ_SECT_UNDEF, PEOBJ_TYPE_FUNC, PEOBJ_SCL_EXTERN);

-#if LJ_TARGET_X64
+#ifdef PEOBJ_PDATA_NRELOC
    emit_peobj_sym_sect(ctx, pesect, PEOBJ_SECT_PDATA);
    emit_peobj_sym_sect(ctx, pesect, PEOBJ_SECT_XDATA);
    emit_peobj_sym(ctx, "lj_err_unwind_win", 0,
@@ -2,7 +2,7 @@
 -- Lua script to dump the bytecode of the library functions written in Lua.
 -- The resulting 'buildvm_libbc.h' is used for the build process of LuaJIT.
 ----------------------------------------------------------------------------
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- Released under the MIT license. See Copyright Notice in luajit.h
 ----------------------------------------------------------------------------

@@ -2,7 +2,7 @@
 -- Lua script to generate a customized, minified version of Lua.
 -- The resulting 'minilua' is used for the build process of LuaJIT.
 ----------------------------------------------------------------------------
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- Released under the MIT license. See Copyright Notice in luajit.h
 ----------------------------------------------------------------------------

@@ -0,0 +1,45 @@
+----------------------------------------------------------------------------
+-- Lua script to embed the rolling release version in luajit.h.
+----------------------------------------------------------------------------
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
+-- Released under the MIT license. See Copyright Notice in luajit.h
+----------------------------------------------------------------------------
+
+local arg = {...}
+local FILE_ROLLING_H = arg[1] or "luajit_rolling.h"
+local FILE_RELVER_TXT = arg[2] or "luajit_relver.txt"
+local FILE_LUAJIT_H = arg[3] or "luajit.h"
+
+local function file_read(file)
+  local fp = assert(io.open(file, "rb"), "run from the wrong directory")
+  local data = assert(fp:read("*a"))
+  fp:close()
+  return data
+end
+
+local function file_write_mod(file, data)
+  local fp = io.open(file, "rb")
+  if fp then
+    local odata = assert(fp:read("*a"))
+    fp:close()
+    if odata == data then return end
+  end
+  fp = assert(io.open(file, "wb"))
+  assert(fp:write(data))
+  assert(fp:close())
+end
+
+local text = file_read(FILE_ROLLING_H)
+local relver = file_read(FILE_RELVER_TXT):match("(%d+)")
+
+if relver then
+  text = text:gsub("ROLLING", relver)
+else
+  io.stderr:write([[
+**** WARNING Cannot determine rolling release version from git log.
+**** WARNING The 'git' command must be available during the build.
+]])
+  file_write_mod(FILE_RELVER_TXT, "ROLLING\n") -- Fallback for install target.
+end
+
+file_write_mod(FILE_LUAJIT_H, text)
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT bytecode listing module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- Released under the MIT license. See Copyright Notice in luajit.h
 ----------------------------------------------------------------------------
 --
@@ -41,7 +41,6 @@

 -- 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")
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT module to save/list bytecode.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- Released under the MIT license. See Copyright Notice in luajit.h
 ----------------------------------------------------------------------------
 --
@@ -11,7 +11,7 @@
 ------------------------------------------------------------------------------

 local jit = require("jit")
-assert(jit.version_num == 20100, "LuaJIT core/library version mismatch")
+assert(jit.version_num == 20199, "LuaJIT core/library version mismatch")
 local bit = require("bit")

 -- Symbol name prefix for LuaJIT bytecode.
@@ -38,7 +38,7 @@ Save LuaJIT bytecode: luajit -b[options] input output
  --        Stop handling options.
  -         Use stdin as input and/or stdout as output.

-File types: c h obj o raw (default)
+File types: c cc h obj o raw (default)
 ]]
  os.exit(1)
 end
@@ -81,7 +81,7 @@ end
 ------------------------------------------------------------------------------

 local map_type = {
-  raw = "raw", c = "c", h = "h", o = "obj", obj = "obj",
+  raw = "raw", c = "c", cc = "c", h = "h", o = "obj", obj = "obj",
 }

 local map_arch = {
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT ARM disassembler module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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.
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT ARM64 disassembler module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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.
@@ -107,24 +107,20 @@ local map_logsr = { -- Logical, shifted register.
    [0] = {
      shift = 29, mask = 3,
      [0] = {
-	shift = 21, mask = 7,
-	[0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg",
-	"andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg"
+	shift = 21, mask = 1,
+	[0] = "andDNMSg", "bicDNMSg"
      },
      {
-	shift = 21, mask = 7,
-	[0] ="orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg",
-	     "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg"
+	shift = 21, mask = 1,
+	[0] = "orr|movDN0MSg", "orn|mvnDN0MSg"
      },
      {
-	shift = 21, mask = 7,
-	[0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg",
-	"eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg"
+	shift = 21, mask = 1,
+	[0] = "eorDNMSg", "eonDNMSg"
      },
      {
-	shift = 21, mask = 7,
-	[0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg",
-	"ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg"
+	shift = 21, mask = 1,
+	[0] = "ands|tstD0NMSg", "bicsDNMSg"
      }
    },
    false -- unallocated
@@ -132,24 +128,20 @@ local map_logsr = { -- Logical, shifted register.
  {
    shift = 29, mask = 3,
    [0] = {
-      shift = 21, mask = 7,
-      [0] = "andDNMSg", "bicDNMSg", "andDNMSg", "bicDNMSg",
-      "andDNMSg", "bicDNMSg", "andDNMg", "bicDNMg"
+      shift = 21, mask = 1,
+      [0] = "andDNMSg", "bicDNMSg"
    },
    {
-      shift = 21, mask = 7,
-      [0] = "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0MSg", "orn|mvnDN0MSg",
-      "orr|movDN0MSg", "orn|mvnDN0MSg", "orr|movDN0Mg", "orn|mvnDN0Mg"
+      shift = 21, mask = 1,
+      [0] = "orr|movDN0MSg", "orn|mvnDN0MSg"
    },
    {
-      shift = 21, mask = 7,
-      [0] = "eorDNMSg", "eonDNMSg", "eorDNMSg", "eonDNMSg",
-      "eorDNMSg", "eonDNMSg", "eorDNMg", "eonDNMg"
+      shift = 21, mask = 1,
+      [0] = "eorDNMSg", "eonDNMSg"
    },
    {
-      shift = 21, mask = 7,
-      [0] = "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMSg", "bicsDNMSg",
-      "ands|tstD0NMSg", "bicsDNMSg", "ands|tstD0NMg", "bicsDNMg"
+      shift = 21, mask = 1,
+      [0] = "ands|tstD0NMSg", "bicsDNMSg"
    }
  }
 }
@@ -735,7 +727,7 @@ local map_cond = {
  "hi", "ls", "ge", "lt", "gt", "le", "al",
 }

-local map_shift = { [0] = "lsl", "lsr", "asr", }
+local map_shift = { [0] = "lsl", "lsr", "asr", "ror"}

 local map_extend = {
  [0] = "uxtb", "uxth", "uxtw", "uxtx", "sxtb", "sxth", "sxtw", "sxtx",
@@ -956,7 +948,7 @@ local function disass_ins(ctx)
    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)
+      local imm12 = lshift(rshift(lshift(op, 10), 20), sz)
      if imm12 ~= 0 then
 	x = "["..rn..", #"..imm12.."]"
      else
@@ -993,8 +985,7 @@ local function disass_ins(ctx)
 	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 sh = 2 + rshift(op, 31 - band(rshift(op, 26), 1))
      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)
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT ARM64BE disassembler wrapper module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT MIPS disassembler module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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.
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT MIPS64 disassembler wrapper module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT MIPS64EL disassembler wrapper module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT MIPS64R6 disassembler wrapper module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT MIPS64R6EL disassembler wrapper module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT MIPSEL disassembler wrapper module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT PPC disassembler module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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.
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT x64 disassembler wrapper module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT x86/x64 disassembler module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 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.
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT compiler dump module.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- Released under the MIT license. See Copyright Notice in luajit.h
 ----------------------------------------------------------------------------
 --
@@ -55,7 +55,6 @@

 -- 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT profiler.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- Released under the MIT license. See Copyright Notice in luajit.h
 ----------------------------------------------------------------------------
 --
@@ -41,7 +41,6 @@

 -- 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- Verbose mode of the LuaJIT compiler.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- Released under the MIT license. See Copyright Notice in luajit.h
 ----------------------------------------------------------------------------
 --
@@ -59,7 +59,6 @@

 -- 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
@@ -1,7 +1,7 @@
 ----------------------------------------------------------------------------
 -- LuaJIT profiler zones.
 --
-- Copyright (C) 2005-2022 Mike Pall. All rights reserved.
+-- Copyright (C) 2005-2023 Mike Pall. All rights reserved.
 -- Released under the MIT license. See Copyright Notice in luajit.h
 ----------------------------------------------------------------------------
 --
@@ -1,6 +1,6 @@
 /*
 ** Auxiliary library for the Lua/C API.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major parts taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2008 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -1,6 +1,6 @@
 /*
 ** Base and coroutine library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major portions taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2011 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -616,7 +616,10 @@ static int ffh_resume(lua_State *L, lua_State *co, int wrap)
    setstrV(L, L->base-LJ_FR2, lj_err_str(L, em));
    return FFH_RES(2);
  }
-  lj_state_growstack(co, (MSize)(L->top - L->base));
+  if (lj_state_cpgrowstack(co, (MSize)(L->top - L->base)) != LUA_OK) {
+    cTValue *msg = --co->top;
+    lj_err_callermsg(L, strVdata(msg));
+  }
  return FFH_RETRY;
 }

@@ -1,6 +1,6 @@
 /*
 ** Bit manipulation library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lib_bit_c
@@ -1,6 +1,6 @@
 /*
 ** Buffer library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lib_buffer_c
@@ -1,6 +1,6 @@
 /*
 ** Debug library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major portions taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2008 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -1,6 +1,6 @@
 /*
 ** FFI library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lib_ffi_c
@@ -745,6 +745,9 @@ LJLIB_CF(ffi_abi)	LJLIB_REC(.)
 #if LJ_ABI_WIN
    "\003win"
 #endif
+#if LJ_ABI_PAUTH
+    "\005pauth"
+#endif
 #if LJ_TARGET_UWP
    "\003uwp"
 #endif
@@ -776,7 +779,7 @@ LJLIB_CF(ffi_metatype)
  if (!(ctype_isstruct(ct->info) || ctype_iscomplex(ct->info) ||
 	ctype_isvector(ct->info)))
    lj_err_arg(L, 1, LJ_ERR_FFI_INVTYPE);
-  tv = lj_tab_setinth(L, t, -(int32_t)id);
+  tv = lj_tab_setinth(L, t, -(int32_t)ctype_typeid(cts, ct));
  if (!tvisnil(tv))
    lj_err_caller(L, LJ_ERR_PROTMT);
  settabV(L, tv, mt);
@@ -1,6 +1,6 @@
 /*
 ** Library initialization.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major parts taken verbatim from the Lua interpreter.
 ** Copyright (C) 1994-2008 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -1,6 +1,6 @@
 /*
 ** I/O library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major portions taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2011 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -1,6 +1,6 @@
 /*
 ** JIT library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lib_jit_c
@@ -422,7 +422,8 @@ LJLIB_CF(jit_util_ircalladdr)
 {
  uint32_t idx = (uint32_t)lj_lib_checkint(L, 1);
  if (idx < IRCALL__MAX) {
-    setintptrV(L->top-1, (intptr_t)(void *)lj_ir_callinfo[idx].func);
+    ASMFunction func = lj_ir_callinfo[idx].func;
+    setintptrV(L->top-1, (intptr_t)(void *)lj_ptr_strip(func));
    return 1;
  }
  return 0;
@@ -742,7 +743,7 @@ LUALIB_API int luaopen_jit(lua_State *L)
 #endif
  lua_pushliteral(L, LJ_OS_NAME);
  lua_pushliteral(L, LJ_ARCH_NAME);
-  lua_pushinteger(L, LUAJIT_VERSION_NUM);
+  lua_pushinteger(L, LUAJIT_VERSION_NUM);  /* Deprecated. */
  lua_pushliteral(L, LUAJIT_VERSION);
  LJ_LIB_REG(L, LUA_JITLIBNAME, jit);
 #if LJ_HASPROFILE
@@ -1,6 +1,6 @@
 /*
 ** Math library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #include <math.h>
@@ -1,6 +1,6 @@
 /*
 ** OS library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major portions taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2008 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -1,6 +1,6 @@
 /*
 ** Package library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major portions taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2012 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -1,6 +1,6 @@
 /*
 ** String library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major portions taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2008 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -1,6 +1,6 @@
 /*
 ** Table library.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major portions taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2008 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -1,6 +1,6 @@
 /*
 ** Public Lua/C API.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Major portions taken verbatim or adapted from the Lua interpreter.
 ** Copyright (C) 1994-2008 Lua.org, PUC-Rio. See Copyright Notice in lua.h
@@ -104,7 +104,12 @@ LUA_API int lua_checkstack(lua_State *L, int size)
  if (size > LUAI_MAXCSTACK || (L->top - L->base + size) > LUAI_MAXCSTACK) {
    return 0;  /* Stack overflow. */
  } else if (size > 0) {
-    lj_state_checkstack(L, (MSize)size);
+    int avail = (int)(mref(L->maxstack, TValue) - L->top);
+    if (size > avail &&
+	lj_state_cpgrowstack(L, (MSize)(size - avail)) != LUA_OK) {
+      L->top--;
+      return 0;  /* Out of memory. */
+    }
  }
  return 1;
 }
@@ -1,6 +1,6 @@
 /*
 ** Target architecture selection.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #ifndef _LJ_ARCH_H
@@ -57,7 +57,7 @@
 #define LUAJIT_TARGET	LUAJIT_ARCH_X64
 #elif defined(__arm__) || defined(__arm) || defined(__ARM__) || defined(__ARM)
 #define LUAJIT_TARGET	LUAJIT_ARCH_ARM
-#elif defined(__aarch64__)
+#elif defined(__aarch64__) || defined(_M_ARM64)
 #define LUAJIT_TARGET	LUAJIT_ARCH_ARM64
 #elif defined(__ppc__) || defined(__ppc) || defined(__PPC__) || defined(__PPC) || defined(__powerpc__) || defined(__powerpc) || defined(__POWERPC__) || defined(__POWERPC) || defined(_M_PPC)
 #define LUAJIT_TARGET	LUAJIT_ARCH_PPC
@@ -66,7 +66,7 @@
 #elif defined(__mips__) || defined(__mips) || defined(__MIPS__) || defined(__MIPS)
 #define LUAJIT_TARGET	LUAJIT_ARCH_MIPS32
 #else
-#error "No support for this architecture (yet)"
+#error "Architecture not supported (in this version), see: https://luajit.org/status.html#architectures"
 #endif

 #endif
@@ -237,7 +237,7 @@
 #define LJ_TARGET_UNIFYROT	2	/* Want only IR_BROR. */
 #define LJ_ARCH_NUMMODE		LJ_NUMMODE_DUAL

-#if __ARM_ARCH == 8 || __ARM_ARCH_8__ || __ARM_ARCH_8A__
+#if __ARM_ARCH >= 8 || __ARM_ARCH_8__ || __ARM_ARCH_8A__
 #define LJ_ARCH_VERSION		80
 #elif __ARM_ARCH == 7 || __ARM_ARCH_7__ || __ARM_ARCH_7A__ || __ARM_ARCH_7R__ || __ARM_ARCH_7S__ || __ARM_ARCH_7VE__
 #define LJ_ARCH_VERSION		70
@@ -259,6 +259,9 @@
 #define LJ_ARCH_NAME		"arm64"
 #define LJ_ARCH_ENDIAN		LUAJIT_LE
 #endif
+#if !defined(LJ_ABI_PAUTH) && defined(__arm64e__)
+#define LJ_ABI_PAUTH		1
+#endif
 #define LJ_TARGET_ARM64		1
 #define LJ_TARGET_EHRETREG	0
 #define LJ_TARGET_EHRAREG	30
@@ -328,6 +331,7 @@
 #define LJ_ARCH_NOFFI		1
 #elif LJ_ARCH_BITS == 64
 #error "No support for PPC64"
+#undef LJ_TARGET_PPC
 #endif

 #if _ARCH_PWR7
@@ -487,36 +491,45 @@
 #elif LJ_TARGET_ARM
 #if defined(__ARMEB__)
 #error "No support for big-endian ARM"
+#undef LJ_TARGET_ARM
 #endif
 #if __ARM_ARCH_6M__ || __ARM_ARCH_7M__ || __ARM_ARCH_7EM__
 #error "No support for Cortex-M CPUs"
+#undef LJ_TARGET_ARM
 #endif
 #if !(__ARM_EABI__ || LJ_TARGET_IOS)
 #error "Only ARM EABI or iOS 3.0+ ABI is supported"
+#undef LJ_TARGET_ARM
 #endif
 #elif LJ_TARGET_ARM64
 #if defined(_ILP32)
 #error "No support for ILP32 model on ARM64"
+#undef LJ_TARGET_ARM64
 #endif
 #elif LJ_TARGET_PPC
 #if defined(_LITTLE_ENDIAN) && (!defined(_BYTE_ORDER) || (_BYTE_ORDER == _LITTLE_ENDIAN))
 #error "No support for little-endian PPC32"
+#undef LJ_TARGET_PPC
 #endif
 #if defined(__NO_FPRS__) && !defined(_SOFT_FLOAT)
-#error "No support for PPC/e500 anymore (use LuaJIT 2.0)"
+#error "No support for PPC/e500, use LuaJIT 2.0"
+#undef LJ_TARGET_PPC
 #endif
 #elif LJ_TARGET_MIPS32
 #if !((defined(_MIPS_SIM_ABI32) && _MIPS_SIM == _MIPS_SIM_ABI32) || (defined(_ABIO32) && _MIPS_SIM == _ABIO32))
 #error "Only o32 ABI supported for MIPS32"
+#undef LJ_TARGET_MIPS
 #endif
 #if LJ_TARGET_MIPSR6
 /* Not that useful, since most available r6 CPUs are 64 bit. */
 #error "No support for MIPS32R6"
+#undef LJ_TARGET_MIPS
 #endif
 #elif LJ_TARGET_MIPS64
 #if !((defined(_MIPS_SIM_ABI64) && _MIPS_SIM == _MIPS_SIM_ABI64) || (defined(_ABI64) && _MIPS_SIM == _ABI64))
 /* MIPS32ON64 aka n32 ABI support might be desirable, but difficult. */
 #error "Only n64 ABI supported for MIPS64"
+#undef LJ_TARGET_MIPS
 #endif
 #endif
 #endif
@@ -603,6 +616,10 @@
 #define LJ_SOFTFP		(!LJ_ARCH_HASFPU)
 #define LJ_SOFTFP32		(LJ_SOFTFP && LJ_32)

+#ifndef LJ_ABI_PAUTH
+#define LJ_ABI_PAUTH		0
+#endif
+
 #if LJ_ARCH_ENDIAN == LUAJIT_BE
 #define LJ_LE			0
 #define LJ_BE			1
@@ -1,6 +1,6 @@
 /*
 ** IR assembler (SSA IR -> machine code).
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lj_asm_c
@@ -29,6 +29,7 @@
 #include "lj_dispatch.h"
 #include "lj_vm.h"
 #include "lj_target.h"
+#include "lj_prng.h"

 #ifdef LUA_USE_ASSERT
 #include <stdio.h>
@@ -93,6 +94,12 @@ typedef struct ASMState {
  MCode *flagmcp;	/* Pending opportunity to merge flag setting ins. */
  MCode *realign;	/* Realign loop if not NULL. */

+#ifdef LUAJIT_RANDOM_RA
+  /* Randomize register allocation. OK for fuzz testing, not for production. */
+  uint64_t prngbits;
+  PRNGState prngstate;
+#endif
+
 #ifdef RID_NUM_KREF
  intptr_t krefk[RID_NUM_KREF];
 #endif
@@ -173,6 +180,41 @@ IRFLDEF(FLOFS)
  0
 };

+#ifdef LUAJIT_RANDOM_RA
+/* Return a fixed number of random bits from the local PRNG state. */
+static uint32_t ra_random_bits(ASMState *as, uint32_t nbits) {
+  uint64_t b = as->prngbits;
+  uint32_t res = (1u << nbits) - 1u;
+  if (b <= res) b = lj_prng_u64(&as->prngstate) | (1ull << 63);
+  res &= (uint32_t)b;
+  as->prngbits = b >> nbits;
+  return res;
+}
+
+/* Pick a random register from a register set. */
+static Reg rset_pickrandom(ASMState *as, RegSet rs)
+{
+  Reg r = rset_pickbot_(rs);
+  rs >>= r;
+  if (rs > 1) {  /* More than one bit set? */
+    while (1) {
+      /* We need to sample max. the GPR or FPR half of the set. */
+      uint32_t d = ra_random_bits(as, RSET_BITS-1);
+      if ((rs >> d) & 1) {
+	r += d;
+	break;
+      }
+    }
+  }
+  return r;
+}
+#define rset_picktop(rs)	rset_pickrandom(as, rs)
+#define rset_pickbot(rs)	rset_pickrandom(as, rs)
+#else
+#define rset_picktop(rs)	rset_picktop_(rs)
+#define rset_pickbot(rs)	rset_pickbot_(rs)
+#endif
+
 /* -- Target-specific instruction emitter --------------------------------- */

 #if LJ_TARGET_X86ORX64
@@ -564,7 +606,11 @@ static Reg ra_allock(ASMState *as, intptr_t k, RegSet allow)
 	IRIns *ir = IR(ref);
 	if ((ir->o == IR_KINT64 && k == (int64_t)ir_kint64(ir)->u64) ||
 #if LJ_GC64
+#if LJ_TARGET_ARM64
+	    (ir->o == IR_KINT && (uint64_t)k == (uint32_t)ir->i) ||
+#else
 	    (ir->o == IR_KINT && k == ir->i) ||
+#endif
 	    (ir->o == IR_KGC && k == (intptr_t)ir_kgc(ir)) ||
 	    ((ir->o == IR_KPTR || ir->o == IR_KKPTR) &&
 	     k == (intptr_t)ir_kptr(ir))
@@ -1888,6 +1934,8 @@ static void asm_head_side(ASMState *as)
  IRRef1 sloadins[RID_MAX];
  RegSet allow = RSET_ALL;  /* Inverse of all coalesced registers. */
  RegSet live = RSET_EMPTY;  /* Live parent registers. */
+  RegSet pallow = RSET_GPR;  /* Registers needed by the parent stack check. */
+  Reg pbase;
  IRIns *irp = &as->parent->ir[REF_BASE];  /* Parent base. */
  int32_t spadj, spdelta;
  int pass2 = 0;
@@ -1898,7 +1946,11 @@ static void asm_head_side(ASMState *as)
    /* Force snap #0 alloc to prevent register overwrite in stack check. */
    asm_snap_alloc(as, 0);
  }
-  allow = asm_head_side_base(as, irp, allow);
+  pbase = asm_head_side_base(as, irp);
+  if (pbase != RID_NONE) {
+    rset_clear(allow, pbase);
+    rset_clear(pallow, pbase);
+  }

  /* Scan all parent SLOADs and collect register dependencies. */
  for (i = as->stopins; i > REF_BASE; i--) {
@@ -1928,6 +1980,7 @@ static void asm_head_side(ASMState *as)
      sloadins[rs] = (IRRef1)i;
      rset_set(live, rs);  /* Block live parent register. */
    }
+    if (!ra_hasspill(regsp_spill(rs))) rset_clear(pallow, regsp_reg(rs));
  }

  /* Calculate stack frame adjustment. */
@@ -2044,7 +2097,7 @@ static void asm_head_side(ASMState *as)
    ExitNo exitno = as->J->exitno;
 #endif
    as->T->topslot = (uint8_t)as->topslot;  /* Remember for child traces. */
-    asm_stack_check(as, as->topslot, irp, allow & RSET_GPR, exitno);
+    asm_stack_check(as, as->topslot, irp, pallow, exitno);
  }
 }

@@ -2435,6 +2488,9 @@ void lj_asm_trace(jit_State *J, GCtrace *T)
  as->realign = NULL;
  as->loopinv = 0;
  as->parent = J->parent ? traceref(J, J->parent) : NULL;
+#ifdef LUAJIT_RANDOM_RA
+  (void)lj_prng_u64(&J2G(J)->prng);  /* Ensure PRNG step between traces. */
+#endif

  /* Reserve MCode memory. */
  as->mctop = as->mctoporig = lj_mcode_reserve(J, &as->mcbot);
@@ -2476,6 +2532,10 @@ void lj_asm_trace(jit_State *J, GCtrace *T)
 #endif
    as->ir = J->curfinal->ir;  /* Use the copied IR. */
    as->curins = J->cur.nins = as->orignins;
+#ifdef LUAJIT_RANDOM_RA
+    as->prngstate = J2G(J)->prng;  /* Must (re)start from identical state. */
+    as->prngbits = 0;
+#endif

    RA_DBG_START();
    RA_DBGX((as, "===== STOP ====="));
@@ -1,6 +1,6 @@
 /*
 ** IR assembler (SSA IR -> machine code).
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #ifndef _LJ_ASM_H
@@ -1,6 +1,6 @@
 /*
 ** ARM IR assembler (SSA IR -> machine code).
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 /* -- Register allocator extensions --------------------------------------- */
@@ -969,24 +969,32 @@ static void asm_hrefk(ASMState *as, IRIns *ir)
 static void asm_uref(ASMState *as, IRIns *ir)
 {
  Reg dest = ra_dest(as, ir, RSET_GPR);
-  if (irref_isk(ir->op1)) {
+  int guarded = (irt_t(ir->t) & (IRT_GUARD|IRT_TYPE)) == (IRT_GUARD|IRT_PGC);
+  if (irref_isk(ir->op1) && !guarded) {
    GCfunc *fn = ir_kfunc(IR(ir->op1));
    MRef *v = &gcref(fn->l.uvptr[(ir->op2 >> 8)])->uv.v;
    emit_lsptr(as, ARMI_LDR, dest, v);
  } else {
-    Reg uv = ra_scratch(as, RSET_GPR);
-    Reg func = ra_alloc1(as, ir->op1, RSET_GPR);
-    if (ir->o == IR_UREFC) {
-      asm_guardcc(as, CC_NE);
+    if (guarded) {
+      asm_guardcc(as, ir->o == IR_UREFC ? CC_NE : CC_EQ);
      emit_n(as, ARMI_CMP|ARMI_K12|1, RID_TMP);
-      emit_opk(as, ARMI_ADD, dest, uv,
-	       (int32_t)offsetof(GCupval, tv), RSET_GPR);
-      emit_lso(as, ARMI_LDRB, RID_TMP, uv, (int32_t)offsetof(GCupval, closed));
-    } else {
-      emit_lso(as, ARMI_LDR, dest, uv, (int32_t)offsetof(GCupval, v));
    }
-    emit_lso(as, ARMI_LDR, uv, func,
-	     (int32_t)offsetof(GCfuncL, uvptr) + 4*(int32_t)(ir->op2 >> 8));
+    if (ir->o == IR_UREFC)
+      emit_opk(as, ARMI_ADD, dest, dest,
+	       (int32_t)offsetof(GCupval, tv), RSET_GPR);
+    else
+      emit_lso(as, ARMI_LDR, dest, dest, (int32_t)offsetof(GCupval, v));
+    if (guarded)
+      emit_lso(as, ARMI_LDRB, RID_TMP, dest,
+	       (int32_t)offsetof(GCupval, closed));
+    if (irref_isk(ir->op1)) {
+      GCfunc *fn = ir_kfunc(IR(ir->op1));
+      int32_t k = (int32_t)gcrefu(fn->l.uvptr[(ir->op2 >> 8)]);
+      emit_loadi(as, dest, k);
+    } else {
+      emit_lso(as, ARMI_LDR, dest, ra_alloc1(as, ir->op1, RSET_GPR),
+	       (int32_t)offsetof(GCfuncL, uvptr) + 4*(int32_t)(ir->op2 >> 8));
+    }
  }
 }

@@ -1990,6 +1998,7 @@ static void asm_prof(ASMState *as, IRIns *ir)
 static void asm_stack_check(ASMState *as, BCReg topslot,
 			    IRIns *irp, RegSet allow, ExitNo exitno)
 {
+  int savereg = 0;
  Reg pbase;
  uint32_t k;
  if (irp) {
@@ -2000,12 +2009,14 @@ static void asm_stack_check(ASMState *as, BCReg topslot,
      pbase = rset_pickbot(allow);
    } else {
      pbase = RID_RET;
-      emit_lso(as, ARMI_LDR, RID_RET, RID_SP, 0);  /* Restore temp. register. */
+      savereg = 1;
    }
  } else {
    pbase = RID_BASE;
  }
  emit_branch(as, ARMF_CC(ARMI_BL, CC_LS), exitstub_addr(as->J, exitno));
+  if (savereg)
+    emit_lso(as, ARMI_LDR, RID_RET, RID_SP, 0);  /* Restore temp. register. */
  k = emit_isk12(0, (int32_t)(8*topslot));
  lj_assertA(k, "slot offset %d does not fit in K12", 8*topslot);
  emit_n(as, ARMI_CMP^k, RID_TMP);
@@ -2017,7 +2028,7 @@ static void asm_stack_check(ASMState *as, BCReg topslot,
    if (ra_hasspill(irp->s))
      emit_lso(as, ARMI_LDR, pbase, RID_SP, sps_scale(irp->s));
    emit_lso(as, ARMI_LDR, RID_TMP, RID_TMP, (i & 4095));
-    if (ra_hasspill(irp->s) && !allow)
+    if (savereg)
      emit_lso(as, ARMI_STR, RID_RET, RID_SP, 0);  /* Save temp. register. */
    emit_loadi(as, RID_TMP, (i & ~4095));
  } else {
@@ -2167,7 +2178,7 @@ static void asm_head_root_base(ASMState *as)
 }

 /* Coalesce BASE register for a side trace. */
-static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
+static Reg asm_head_side_base(ASMState *as, IRIns *irp)
 {
  IRIns *ir;
  asm_head_lreg(as);
@@ -2175,16 +2186,15 @@ static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
  if (ra_hasreg(ir->r) && (rset_test(as->modset, ir->r) || irt_ismarked(ir->t)))
    ra_spill(as, ir);
  if (ra_hasspill(irp->s)) {
-    rset_clear(allow, ra_dest(as, ir, allow));
+    return ra_dest(as, ir, RSET_GPR);
  } else {
    Reg r = irp->r;
    lj_assertA(ra_hasreg(r), "base reg lost");
-    rset_clear(allow, r);
    if (r != ir->r && !rset_test(as->freeset, r))
      ra_restore(as, regcost_ref(as->cost[r]));
    ra_destreg(as, ir, r);
+    return r;
  }
-  return allow;
 }

 /* -- Tail of trace ------------------------------------------------------- */
@@ -2253,7 +2263,7 @@ static Reg asm_setup_call_slots(ASMState *as, IRIns *ir, const CCallInfo *ci)
  }
  if (nslots > as->evenspill)  /* Leave room for args in stack slots. */
    as->evenspill = nslots;
-  return REGSP_HINT(RID_RET);
+  return REGSP_HINT(irt_isfp(ir->t) ? RID_FPRET : RID_RET);
 }

 static void asm_setup_target(ASMState *as)
@@ -1,6 +1,6 @@
 /*
 ** ARM64 IR assembler (SSA IR -> machine code).
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 **
 ** Contributed by Djordje Kovacevic and Stefan Pejic from RT-RK.com.
 ** Sponsored by Cisco Systems, Inc.
@@ -84,18 +84,23 @@ static void asm_guardcc(ASMState *as, A64CC cc)
  emit_cond_branch(as, cc, target);
 }

-/* Emit test and branch instruction to exit for guard. */
-static void asm_guardtnb(ASMState *as, A64Ins ai, Reg r, uint32_t bit)
+/* Emit test and branch instruction to exit for guard, if in range. */
+static int asm_guardtnb(ASMState *as, A64Ins ai, Reg r, uint32_t bit)
 {
  MCode *target = asm_exitstub_addr(as, as->snapno);
  MCode *p = as->mcp;
+  ptrdiff_t delta = target - p;
  if (LJ_UNLIKELY(p == as->invmcp)) {
+    if (as->orignins > 1023) return 0;  /* Delta might end up too large. */
    as->loopinv = 1;
-    *p = A64I_B | A64F_S26(target-p);
-    emit_tnb(as, ai^0x01000000u, r, bit, p-1);
-    return;
+    *p = A64I_B | A64F_S26(delta);
+    ai ^= 0x01000000u;
+    target = p-1;
+  } else if (LJ_UNLIKELY(delta >= 0x1fff)) {
+    return 0;
  }
  emit_tnb(as, ai, r, bit, target);
+  return 1;
 }

 /* Emit compare and branch instruction to exit for guard. */
@@ -211,16 +216,14 @@ static Reg asm_fuseahuref(ASMState *as, IRRef ref, int32_t *ofsp, RegSet allow,
 static uint32_t asm_fuseopm(ASMState *as, A64Ins ai, IRRef ref, RegSet allow)
 {
  IRIns *ir = IR(ref);
+  int logical = (ai & 0x1f000000) == 0x0a000000;
  if (ra_hasreg(ir->r)) {
    ra_noweak(as, ir->r);
    return A64F_M(ir->r);
  } else if (irref_isk(ref)) {
-    uint32_t m;
    int64_t k = get_k64val(as, ref);
-    if ((ai & 0x1f000000) == 0x0a000000)
-      m = emit_isk13(k, irt_is64(ir->t));
-    else
-      m = emit_isk12(k);
+    uint32_t m = logical ? emit_isk13(k, irt_is64(ir->t)) :
+			   emit_isk12(irt_is64(ir->t) ? k : (int32_t)k);
    if (m)
      return m;
  } else if (mayfuse(as, ref)) {
@@ -232,7 +235,7 @@ static uint32_t asm_fuseopm(ASMState *as, A64Ins ai, IRRef ref, RegSet allow)
 		    (IR(ir->op2)->i & (irt_is64(ir->t) ? 63 : 31));
      IRIns *irl = IR(ir->op1);
      if (sh == A64SH_LSL &&
-	  irl->o == IR_CONV &&
+	  irl->o == IR_CONV && !logical &&
 	  irl->op2 == ((IRT_I64<<IRCONV_DSH)|IRT_INT|IRCONV_SEXT) &&
 	  shift <= 4 &&
 	  canfuse(as, irl)) {
@@ -242,7 +245,11 @@ static uint32_t asm_fuseopm(ASMState *as, A64Ins ai, IRRef ref, RegSet allow)
 	Reg m = ra_alloc1(as, ir->op1, allow);
 	return A64F_M(m) | A64F_SH(sh, shift);
      }
-    } else if (ir->o == IR_CONV &&
+    } else if (ir->o == IR_BROR && logical && irref_isk(ir->op2)) {
+      Reg m = ra_alloc1(as, ir->op1, allow);
+      int shift = (IR(ir->op2)->i & (irt_is64(ir->t) ? 63 : 31));
+      return A64F_M(m) | A64F_SH(A64SH_ROR, shift);
+    } else if (ir->o == IR_CONV && !logical &&
 	       ir->op2 == ((IRT_I64<<IRCONV_DSH)|IRT_INT|IRCONV_SEXT)) {
      Reg m = ra_alloc1(as, ir->op1, allow);
      return A64F_M(m) | A64F_EX(A64EX_SXTW);
@@ -419,13 +426,18 @@ static int asm_fuseorshift(ASMState *as, IRIns *ir)
 static void asm_gencall(ASMState *as, const CCallInfo *ci, IRRef *args)
 {
  uint32_t n, nargs = CCI_XNARGS(ci);
-  int32_t ofs = 0;
+  int32_t spofs = 0, spalign = LJ_HASFFI && LJ_TARGET_OSX ? 0 : 7;
  Reg gpr, fpr = REGARG_FIRSTFPR;
-  if ((void *)ci->func)
-    emit_call(as, (void *)ci->func);
+  if (ci->func)
+    emit_call(as, ci->func);
  for (gpr = REGARG_FIRSTGPR; gpr <= REGARG_LASTGPR; gpr++)
    as->cost[gpr] = REGCOST(~0u, ASMREF_L);
  gpr = REGARG_FIRSTGPR;
+#if LJ_HASFFI && LJ_ABI_WIN
+  if ((ci->flags & CCI_VARARG)) {
+    fpr = REGARG_LASTFPR+1;
+  }
+#endif
  for (n = 0; n < nargs; n++) { /* Setup args. */
    IRRef ref = args[n];
    IRIns *ir = IR(ref);
@@ -436,10 +448,21 @@ static void asm_gencall(ASMState *as, const CCallInfo *ci, IRRef *args)
 		     "reg %d not free", fpr);  /* Must have been evicted. */
 	  ra_leftov(as, fpr, ref);
 	  fpr++;
+#if LJ_HASFFI && LJ_ABI_WIN
+	} else if ((ci->flags & CCI_VARARG) && (gpr <= REGARG_LASTGPR)) {
+	  Reg rf = ra_alloc1(as, ref, RSET_FPR);
+	  emit_dn(as, A64I_FMOV_R_D, gpr++, rf & 31);
+#endif
 	} else {
 	  Reg r = ra_alloc1(as, ref, RSET_FPR);
-	  emit_spstore(as, ir, r, ofs + ((LJ_BE && !irt_isnum(ir->t)) ? 4 : 0));
-	  ofs += 8;
+	  int32_t al = spalign;
+#if LJ_HASFFI && LJ_TARGET_OSX
+	  al |= irt_isnum(ir->t) ? 7 : 3;
+#endif
+	  spofs = (spofs + al) & ~al;
+	  if (LJ_BE && al >= 7 && !irt_isnum(ir->t)) spofs += 4, al -= 4;
+	  emit_spstore(as, ir, r, spofs);
+	  spofs += al + 1;
 	}
      } else {
 	if (gpr <= REGARG_LASTGPR) {
@@ -449,10 +472,27 @@ static void asm_gencall(ASMState *as, const CCallInfo *ci, IRRef *args)
 	  gpr++;
 	} else {
 	  Reg r = ra_alloc1(as, ref, RSET_GPR);
-	  emit_spstore(as, ir, r, ofs + ((LJ_BE && !irt_is64(ir->t)) ? 4 : 0));
-	  ofs += 8;
+	  int32_t al = spalign;
+#if LJ_HASFFI && LJ_TARGET_OSX
+	  al |= irt_size(ir->t) - 1;
+#endif
+	  spofs = (spofs + al) & ~al;
+	  if (al >= 3) {
+	    if (LJ_BE && al >= 7 && !irt_is64(ir->t)) spofs += 4, al -= 4;
+	    emit_spstore(as, ir, r, spofs);
+	  } else {
+	    lj_assertA(al == 0 || al == 1, "size %d unexpected", al + 1);
+	    emit_lso(as, al ? A64I_STRH : A64I_STRB, r, RID_SP, spofs);
+	  }
+	  spofs += al + 1;
 	}
      }
+#if LJ_HASFFI && LJ_TARGET_OSX
+    } else {  /* Marker for start of varargs. */
+      gpr = REGARG_LASTGPR+1;
+      fpr = REGARG_LASTFPR+1;
+      spalign = 7;
+#endif
    }
  }
 }
@@ -501,7 +541,7 @@ static void asm_callx(ASMState *as, IRIns *ir)
    ci.func = (ASMFunction)(ir_k64(irf)->u64);
  } else {  /* Need a non-argument register for indirect calls. */
    Reg freg = ra_alloc1(as, func, RSET_RANGE(RID_X8, RID_MAX_GPR)-RSET_FIXED);
-    emit_n(as, A64I_BLR, freg);
+    emit_n(as, A64I_BLR_AUTH, freg);
    ci.func = (ASMFunction)(void *)0;
  }
  asm_gencall(as, &ci, args);
@@ -518,8 +558,6 @@ static void asm_retf(ASMState *as, IRIns *ir)
  as->topslot -= (BCReg)delta;
  if ((int32_t)as->topslot < 0) as->topslot = 0;
  irt_setmark(IR(REF_BASE)->t);  /* Children must not coalesce with BASE reg. */
-  /* Need to force a spill on REF_BASE now to update the stack slot. */
-  emit_lso(as, A64I_STRx, base, RID_SP, ra_spill(as, IR(REF_BASE)));
  emit_setgl(as, base, jit_base);
  emit_addptr(as, base, -8*delta);
  asm_guardcc(as, CC_NE);
@@ -643,25 +681,22 @@ static void asm_strto(ASMState *as, IRIns *ir)
 {
  const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_strscan_num];
  IRRef args[2];
-  Reg dest = 0, tmp;
-  int destused = ra_used(ir);
+  Reg tmp;
  int32_t ofs = 0;
  ra_evictset(as, RSET_SCRATCH);
-  if (destused) {
+  if (ra_used(ir)) {
    if (ra_hasspill(ir->s)) {
      ofs = sps_scale(ir->s);
-      destused = 0;
      if (ra_hasreg(ir->r)) {
 	ra_free(as, ir->r);
 	ra_modified(as, ir->r);
 	emit_spload(as, ir, ir->r, ofs);
      }
    } else {
-      dest = ra_dest(as, ir, RSET_FPR);
+      Reg dest = ra_dest(as, ir, RSET_FPR);
+      emit_lso(as, A64I_LDRd, (dest & 31), RID_SP, 0);
    }
  }
-  if (destused)
-    emit_lso(as, A64I_LDRd, (dest & 31), RID_SP, 0);
  asm_guardcnb(as, A64I_CBZ, RID_RET);
  args[0] = ir->op1; /* GCstr *str */
  args[1] = ASMREF_TMP1; /* TValue *n  */
@@ -752,113 +787,75 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge)
  int destused = ra_used(ir);
  Reg dest = ra_dest(as, ir, allow);
  Reg tab = ra_alloc1(as, ir->op1, rset_clear(allow, dest));
-  Reg key = 0, tmp = RID_TMP;
-  Reg ftmp = RID_NONE, type = RID_NONE, scr = RID_NONE, tisnum = RID_NONE;
+  Reg tmp = RID_TMP, type = RID_NONE, key, tkey;
  IRRef refkey = ir->op2;
  IRIns *irkey = IR(refkey);
-  int isk = irref_isk(ir->op2);
+  int isk = irref_isk(refkey);
  IRType1 kt = irkey->t;
  uint32_t k = 0;
  uint32_t khash;
-  MCLabel l_end, l_loop, l_next;
+  MCLabel l_end, l_loop;
  rset_clear(allow, tab);

-  if (!isk) {
-    key = ra_alloc1(as, ir->op2, irt_isnum(kt) ? RSET_FPR : allow);
-    rset_clear(allow, key);
-    if (!irt_isstr(kt)) {
-      tmp = ra_scratch(as, allow);
-      rset_clear(allow, tmp);
-    }
-  } else if (irt_isnum(kt)) {
-    int64_t val = (int64_t)ir_knum(irkey)->u64;
-    if (!(k = emit_isk12(val))) {
-      key = ra_allock(as, val, allow);
-      rset_clear(allow, key);
-    }
-  } else if (!irt_ispri(kt)) {
-    if (!(k = emit_isk12(irkey->i))) {
-      key = ra_alloc1(as, refkey, allow);
-      rset_clear(allow, key);
-    }
-  }
-
-  /* Allocate constants early. */
-  if (irt_isnum(kt)) {
-    if (!isk) {
-      tisnum = ra_allock(as, LJ_TISNUM << 15, allow);
-      ftmp = ra_scratch(as, rset_exclude(RSET_FPR, key));
-      rset_clear(allow, tisnum);
-    }
-  } else if (irt_isaddr(kt)) {
-    if (isk) {
-      int64_t kk = ((int64_t)irt_toitype(kt) << 47) | irkey[1].tv.u64;
-      scr = ra_allock(as, kk, allow);
+  /* Allocate register for tkey outside of the loop. */
+  if (isk) {
+    int64_t kk;
+    if (irt_isaddr(kt)) {
+      kk = ((int64_t)irt_toitype(kt) << 47) | irkey[1].tv.u64;
+    } else if (irt_isnum(kt)) {
+      kk = (int64_t)ir_knum(irkey)->u64;
+      /* Assumes -0.0 is already canonicalized to +0.0. */
    } else {
-      scr = ra_scratch(as, allow);
+      lj_assertA(irt_ispri(kt) && !irt_isnil(kt), "bad HREF key type");
+      kk = ~((int64_t)~irt_toitype(kt) << 47);
    }
-    rset_clear(allow, scr);
+    k = emit_isk12(kk);
+    tkey = k ? 0 : ra_allock(as, kk, allow);
  } else {
-    lj_assertA(irt_ispri(kt) && !irt_isnil(kt), "bad HREF key type");
-    type = ra_allock(as, ~((int64_t)~irt_toitype(kt) << 47), allow);
-    scr = ra_scratch(as, rset_clear(allow, type));
-    rset_clear(allow, scr);
+    tkey = ra_scratch(as, allow);
  }

  /* Key not found in chain: jump to exit (if merged) or load niltv. */
  l_end = emit_label(as);
  as->invmcp = NULL;
-  if (merge == IR_NE)
+  if (merge == IR_NE) {
    asm_guardcc(as, CC_AL);
-  else if (destused)
-    emit_loada(as, dest, niltvg(J2G(as->J)));
+  } else if (destused) {
+    uint32_t k12 = emit_isk12(offsetof(global_State, nilnode.val));
+    lj_assertA(k12 != 0, "Cannot k12 encode niltv(L)");
+    emit_dn(as, A64I_ADDx^k12, dest, RID_GL);
+  }

  /* Follow hash chain until the end. */
  l_loop = --as->mcp;
-  emit_n(as, A64I_CMPx^A64I_K12^0, dest);
-  emit_lso(as, A64I_LDRx, dest, dest, offsetof(Node, next));
-  l_next = emit_label(as);
+  if (destused)
+    emit_lso(as, A64I_LDRx, dest, dest, offsetof(Node, next));

  /* Type and value comparison. */
  if (merge == IR_EQ)
    asm_guardcc(as, CC_EQ);
  else
    emit_cond_branch(as, CC_EQ, l_end);
+  emit_nm(as, A64I_CMPx^k, tmp, tkey);
+  if (!destused)
+    emit_lso(as, A64I_LDRx, dest, dest, offsetof(Node, next));
+  emit_lso(as, A64I_LDRx, tmp, dest, offsetof(Node, key));
+  *l_loop = A64I_X | A64I_CBNZ | A64F_S19(as->mcp - l_loop) | dest;

-  if (irt_isnum(kt)) {
-    if (isk) {
-      /* Assumes -0.0 is already canonicalized to +0.0. */
-      if (k)
-	emit_n(as, A64I_CMPx^k, tmp);
-      else
-	emit_nm(as, A64I_CMPx, key, tmp);
-      emit_lso(as, A64I_LDRx, tmp, dest, offsetof(Node, key.u64));
+  /* Construct tkey as canonicalized or tagged key. */
+  if (!isk) {
+    if (irt_isnum(kt)) {
+      key = ra_alloc1(as, refkey, RSET_FPR);
+      emit_dnm(as, A64I_CSELx | A64F_CC(CC_EQ), tkey, RID_ZERO, tkey);
+      /* A64I_FMOV_R_D from key to tkey done below. */
    } else {
-      emit_nm(as, A64I_FCMPd, key, ftmp);
-      emit_dn(as, A64I_FMOV_D_R, (ftmp & 31), (tmp & 31));
-      emit_cond_branch(as, CC_LO, l_next);
-      emit_nm(as, A64I_CMPx | A64F_SH(A64SH_LSR, 32), tisnum, tmp);
-      emit_lso(as, A64I_LDRx, tmp, dest, offsetof(Node, key.n));
+      lj_assertA(irt_isaddr(kt), "bad HREF key type");
+      key = ra_alloc1(as, refkey, allow);
+      type = ra_allock(as, irt_toitype(kt) << 15, rset_clear(allow, key));
+      emit_dnm(as, A64I_ADDx | A64F_SH(A64SH_LSL, 32), tkey, key, type);
    }
-  } else if (irt_isaddr(kt)) {
-    if (isk) {
-      emit_nm(as, A64I_CMPx, scr, tmp);
-      emit_lso(as, A64I_LDRx, tmp, dest, offsetof(Node, key.u64));
-    } else {
-      emit_nm(as, A64I_CMPx, tmp, scr);
-      emit_lso(as, A64I_LDRx, scr, dest, offsetof(Node, key.u64));
-    }
-  } else {
-    emit_nm(as, A64I_CMPx, scr, type);
-    emit_lso(as, A64I_LDRx, scr, dest, offsetof(Node, key));
  }

-  *l_loop = A64I_BCC | A64F_S19(as->mcp - l_loop) | CC_NE;
-  if (!isk && irt_isaddr(kt)) {
-    type = ra_allock(as, (int32_t)irt_toitype(kt), allow);
-    emit_dnm(as, A64I_ADDx | A64F_SH(A64SH_LSL, 47), tmp, key, type);
-    rset_clear(allow, type);
-  }
  /* Load main position relative to tab->node into dest. */
  khash = isk ? ir_khash(as, irkey) : 1;
  if (khash == 0) {
@@ -872,7 +869,6 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge)
      emit_dnm(as, A64I_ANDw, dest, dest, tmphash);
      emit_lso(as, A64I_LDRw, dest, tab, offsetof(GCtab, hmask));
    } else if (irt_isstr(kt)) {
-      /* Fetch of str->sid is cheaper than ra_allock. */
      emit_dnm(as, A64I_ANDw, dest, dest, tmp);
      emit_lso(as, A64I_LDRw, tmp, key, offsetof(GCstr, sid));
      emit_lso(as, A64I_LDRw, dest, tab, offsetof(GCtab, hmask));
@@ -881,23 +877,18 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge)
      emit_lso(as, A64I_LDRw, tmp, tab, offsetof(GCtab, hmask));
      emit_dnm(as, A64I_SUBw, dest, dest, tmp);
      emit_dnm(as, A64I_EXTRw | (A64F_IMMS(32-HASH_ROT3)), tmp, tmp, tmp);
-      emit_dnm(as, A64I_EORw, dest, dest, tmp);
-      emit_dnm(as, A64I_EXTRw | (A64F_IMMS(32-HASH_ROT2)), dest, dest, dest);
+      emit_dnm(as, A64I_EORw | A64F_SH(A64SH_ROR, 32-HASH_ROT2), dest, tmp, dest);
      emit_dnm(as, A64I_SUBw, tmp, tmp, dest);
      emit_dnm(as, A64I_EXTRw | (A64F_IMMS(32-HASH_ROT1)), dest, dest, dest);
-      emit_dnm(as, A64I_EORw, tmp, tmp, dest);
      if (irt_isnum(kt)) {
+	emit_dnm(as, A64I_EORw, tmp, tkey, dest);
 	emit_dnm(as, A64I_ADDw, dest, dest, dest);
-	emit_dn(as, A64I_LSRx | A64F_IMMR(32)|A64F_IMMS(32), dest, dest);
-	emit_dm(as, A64I_MOVw, tmp, dest);
-	emit_dn(as, A64I_FMOV_R_D, dest, (key & 31));
+	emit_dn(as, A64I_LSRx | A64F_IMMR(32)|A64F_IMMS(32), dest, tkey);
+	emit_nm(as, A64I_FCMPZd, (key & 31), 0);
+	emit_dn(as, A64I_FMOV_R_D, tkey, (key & 31));
      } else {
-	checkmclim(as);
-	emit_dm(as, A64I_MOVw, tmp, key);
-	emit_dnm(as, A64I_EORw, dest, dest,
-		 ra_allock(as, irt_toitype(kt) << 15, allow));
-	emit_dn(as, A64I_LSRx | A64F_IMMR(32)|A64F_IMMS(32), dest, dest);
-	emit_dm(as, A64I_MOVx, dest, key);
+	emit_dnm(as, A64I_EORw, tmp, key, dest);
+	emit_dnm(as, A64I_EORx | A64F_SH(A64SH_LSR, 32), dest, type, key);
      }
    }
  }
@@ -909,10 +900,10 @@ static void asm_hrefk(ASMState *as, IRIns *ir)
  IRIns *irkey = IR(kslot->op1);
  int32_t ofs = (int32_t)(kslot->op2 * sizeof(Node));
  int32_t kofs = ofs + (int32_t)offsetof(Node, key);
-  int bigofs = !emit_checkofs(A64I_LDRx, ofs);
+  int bigofs = !emit_checkofs(A64I_LDRx, kofs);
  Reg dest = (ra_used(ir) || bigofs) ? ra_dest(as, ir, RSET_GPR) : RID_NONE;
  Reg node = ra_alloc1(as, ir->op1, RSET_GPR);
-  Reg key, idx = node;
+  Reg idx = node;
  RegSet allow = rset_exclude(RSET_GPR, node);
  uint64_t k;
  lj_assertA(ofs % sizeof(Node) == 0, "unaligned HREFK slot");
@@ -931,34 +922,39 @@ static void asm_hrefk(ASMState *as, IRIns *ir)
  } else {
    k = ((uint64_t)irt_toitype(irkey->t) << 47) | (uint64_t)ir_kgc(irkey);
  }
-  key = ra_scratch(as, allow);
-  emit_nm(as, A64I_CMPx, key, ra_allock(as, k, rset_exclude(allow, key)));
-  emit_lso(as, A64I_LDRx, key, idx, kofs);
+  emit_nm(as, A64I_CMPx, RID_TMP, ra_allock(as, k, allow));
+  emit_lso(as, A64I_LDRx, RID_TMP, idx, kofs);
  if (bigofs)
-    emit_opk(as, A64I_ADDx, dest, node, ofs, RSET_GPR);
+    emit_opk(as, A64I_ADDx, dest, node, ofs, rset_exclude(RSET_GPR, node));
 }

 static void asm_uref(ASMState *as, IRIns *ir)
 {
  Reg dest = ra_dest(as, ir, RSET_GPR);
-  if (irref_isk(ir->op1)) {
+  int guarded = (irt_t(ir->t) & (IRT_GUARD|IRT_TYPE)) == (IRT_GUARD|IRT_PGC);
+  if (irref_isk(ir->op1) && !guarded) {
    GCfunc *fn = ir_kfunc(IR(ir->op1));
    MRef *v = &gcref(fn->l.uvptr[(ir->op2 >> 8)])->uv.v;
    emit_lsptr(as, A64I_LDRx, dest, v);
  } else {
-    Reg uv = ra_scratch(as, RSET_GPR);
-    Reg func = ra_alloc1(as, ir->op1, RSET_GPR);
-    if (ir->o == IR_UREFC) {
-      asm_guardcc(as, CC_NE);
-      emit_n(as, (A64I_CMPx^A64I_K12) | A64F_U12(1), RID_TMP);
-      emit_opk(as, A64I_ADDx, dest, uv,
+    if (guarded)
+      asm_guardcnb(as, ir->o == IR_UREFC ? A64I_CBZ : A64I_CBNZ, RID_TMP);
+    if (ir->o == IR_UREFC)
+      emit_opk(as, A64I_ADDx, dest, dest,
 	       (int32_t)offsetof(GCupval, tv), RSET_GPR);
-      emit_lso(as, A64I_LDRB, RID_TMP, uv, (int32_t)offsetof(GCupval, closed));
+    else
+      emit_lso(as, A64I_LDRx, dest, dest, (int32_t)offsetof(GCupval, v));
+    if (guarded)
+      emit_lso(as, A64I_LDRB, RID_TMP, dest,
+	       (int32_t)offsetof(GCupval, closed));
+    if (irref_isk(ir->op1)) {
+      GCfunc *fn = ir_kfunc(IR(ir->op1));
+      uint64_t k = gcrefu(fn->l.uvptr[(ir->op2 >> 8)]);
+      emit_loadu64(as, dest, k);
    } else {
-      emit_lso(as, A64I_LDRx, dest, uv, (int32_t)offsetof(GCupval, v));
+      emit_lso(as, A64I_LDRx, dest, ra_alloc1(as, ir->op1, RSET_GPR),
+	       (int32_t)offsetof(GCfuncL, uvptr) + 8*(int32_t)(ir->op2 >> 8));
    }
-    emit_lso(as, A64I_LDRx, uv, func,
-	     (int32_t)offsetof(GCfuncL, uvptr) + 8*(int32_t)(ir->op2 >> 8));
  }
 }

@@ -1063,7 +1059,7 @@ static void asm_xstore(ASMState *as, IRIns *ir)

 static void asm_ahuvload(ASMState *as, IRIns *ir)
 {
-  Reg idx, tmp, type;
+  Reg idx, tmp;
  int32_t ofs = 0;
  RegSet gpr = RSET_GPR, allow = irt_isnum(ir->t) ? RSET_FPR : RSET_GPR;
  lj_assertA(irt_isnum(ir->t) || irt_ispri(ir->t) || irt_isaddr(ir->t) ||
@@ -1082,8 +1078,9 @@ static void asm_ahuvload(ASMState *as, IRIns *ir)
  } else {
    tmp = ra_scratch(as, gpr);
  }
-  type = ra_scratch(as, rset_clear(gpr, tmp));
-  idx = asm_fuseahuref(as, ir->op1, &ofs, rset_clear(gpr, type), A64I_LDRx);
+  idx = asm_fuseahuref(as, ir->op1, &ofs, rset_clear(gpr, tmp), A64I_LDRx);
+  rset_clear(gpr, idx);
+  if (ofs & FUSE_REG) rset_clear(gpr, ofs & 31);
  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);
@@ -1091,10 +1088,10 @@ static void asm_ahuvload(ASMState *as, IRIns *ir)
    lj_assertA(irt_isinteger(ir->t) || irt_isnum(ir->t),
 	       "bad load type %d", irt_type(ir->t));
    emit_nm(as, A64I_CMPx | A64F_SH(A64SH_LSR, 32),
-	    ra_allock(as, LJ_TISNUM << 15, rset_exclude(gpr, idx)), tmp);
+	    ra_allock(as, LJ_TISNUM << 15, gpr), tmp);
  } else if (irt_isaddr(ir->t)) {
-    emit_n(as, (A64I_CMNx^A64I_K12) | A64F_U12(-irt_toitype(ir->t)), type);
-    emit_dn(as, A64I_ASRx | A64F_IMMR(47), type, tmp);
+    emit_n(as, (A64I_CMNx^A64I_K12) | A64F_U12(-irt_toitype(ir->t)), RID_TMP);
+    emit_dn(as, A64I_ASRx | A64F_IMMR(47), RID_TMP, tmp);
  } else if (irt_isnil(ir->t)) {
    emit_n(as, (A64I_CMNx^A64I_K12) | A64F_U12(1), tmp);
  } else {
@@ -1217,9 +1214,8 @@ dotypecheck:
      emit_nm(as, A64I_CMPx,
 	      ra_allock(as, ~((int64_t)~irt_toitype(t) << 47) , allow), tmp);
    } else {
-      Reg type = ra_scratch(as, allow);
-      emit_n(as, (A64I_CMNx^A64I_K12) | A64F_U12(-irt_toitype(t)), type);
-      emit_dn(as, A64I_ASRx | A64F_IMMR(47), type, tmp);
+      emit_n(as, (A64I_CMNx^A64I_K12) | A64F_U12(-irt_toitype(t)), RID_TMP);
+      emit_dn(as, A64I_ASRx | A64F_IMMR(47), RID_TMP, tmp);
    }
    emit_lso(as, A64I_LDRx, tmp, base, ofs);
    return;
@@ -1289,8 +1285,9 @@ static void asm_tbar(ASMState *as, IRIns *ir)
  Reg link = ra_scratch(as, rset_exclude(RSET_GPR, tab));
  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));
+  /* Keep STRx in the middle to avoid LDP/STP fusion with surrounding code. */
+  emit_lso(as, A64I_STRx, link, tab, (int32_t)offsetof(GCtab, gclist));
  emit_setgl(as, tab, gc.grayagain);
  emit_dn(as, A64I_ANDw^emit_isk13(~LJ_GC_BLACK, 0), mark, mark);
  emit_getgl(as, link, gc.grayagain);
@@ -1304,7 +1301,6 @@ static void asm_obar(ASMState *as, IRIns *ir)
  const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_gc_barrieruv];
  IRRef args[2];
  MCLabel l_end;
-  RegSet allow = RSET_GPR;
  Reg obj, val, tmp;
  /* No need for other object barriers (yet). */
  lj_assertA(IR(ir->op1)->o == IR_UREFC, "bad OBAR type");
@@ -1315,14 +1311,13 @@ static void asm_obar(ASMState *as, IRIns *ir)
  asm_gencall(as, ci, args);
  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);
-  emit_n(as, A64I_TSTw^emit_isk13(LJ_GC_BLACK, 0), tmp);
+  tmp = ra_scratch(as, rset_exclude(RSET_GPR, obj));
+  emit_tnb(as, A64I_TBZ, tmp, lj_ffs(LJ_GC_BLACK), l_end);
  emit_cond_branch(as, CC_EQ, l_end);
  emit_n(as, A64I_TSTw^emit_isk13(LJ_GC_WHITES, 0), RID_TMP);
  val = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, obj));
  emit_lso(as, A64I_LDRB, tmp, obj,
-     (int32_t)offsetof(GCupval, marked)-(int32_t)offsetof(GCupval, tv));
+	   (int32_t)offsetof(GCupval, marked)-(int32_t)offsetof(GCupval, tv));
  emit_lso(as, A64I_LDRB, RID_TMP, val, (int32_t)offsetof(GChead, marked));
 }

@@ -1364,12 +1359,12 @@ static int asm_swapops(ASMState *as, IRRef lref, IRRef rref)
  if (irref_isk(lref))
    return 1;  /* But swap constants to the right. */
  ir = IR(rref);
-  if ((ir->o >= IR_BSHL && ir->o <= IR_BSAR) ||
+  if ((ir->o >= IR_BSHL && ir->o <= IR_BROR) ||
      (ir->o == IR_ADD && ir->op1 == ir->op2) ||
      (ir->o == IR_CONV && ir->op2 == ((IRT_I64<<IRCONV_DSH)|IRT_INT|IRCONV_SEXT)))
    return 0;  /* Don't swap fusable operands to the left. */
  ir = IR(lref);
-  if ((ir->o >= IR_BSHL && ir->o <= IR_BSAR) ||
+  if ((ir->o >= IR_BSHL && ir->o <= IR_BROR) ||
      (ir->o == IR_ADD && ir->op1 == ir->op2) ||
      (ir->o == IR_CONV && ir->op2 == ((IRT_I64<<IRCONV_DSH)|IRT_INT|IRCONV_SEXT)))
    return 1;  /* But swap fusable operands to the right. */
@@ -1415,13 +1410,12 @@ static void asm_intneg(ASMState *as, IRIns *ir)
 static void asm_intmul(ASMState *as, IRIns *ir)
 {
  Reg dest = ra_dest(as, ir, RSET_GPR);
-  Reg left = ra_alloc1(as, ir->op1, rset_exclude(RSET_GPR, dest));
+  Reg left = ra_alloc1(as, ir->op1, RSET_GPR);
  Reg right = ra_alloc1(as, ir->op2, rset_exclude(RSET_GPR, left));
  if (irt_isguard(ir->t)) {  /* IR_MULOV */
    asm_guardcc(as, CC_NE);
    emit_dm(as, A64I_MOVw, dest, dest);  /* Zero-extend. */
-    emit_nm(as, A64I_CMPw | A64F_SH(A64SH_ASR, 31), RID_TMP, dest);
-    emit_dn(as, A64I_ASRx | A64F_IMMR(32), RID_TMP, dest);
+    emit_nm(as, A64I_CMPx | A64F_EX(A64EX_SXTW), dest, dest);
    emit_dnm(as, A64I_SMULL, dest, right, left);
  } else {
    emit_dnm(as, irt_is64(ir->t) ? A64I_MULx : A64I_MULw, dest, left, right);
@@ -1681,16 +1675,15 @@ static void asm_intcomp(ASMState *as, IRIns *ir)
      if (asm_swapops(as, blref, brref)) {
 	Reg tmp = blref; blref = brref; brref = tmp;
      }
+      bleft = ra_alloc1(as, blref, RSET_GPR);
      if (irref_isk(brref)) {
 	uint64_t k = get_k64val(as, brref);
-	if (k && !(k & (k-1)) && (cc == CC_EQ || cc == CC_NE)) {
-	  asm_guardtnb(as, cc == CC_EQ ? A64I_TBZ : A64I_TBNZ,
-		       ra_alloc1(as, blref, RSET_GPR), emit_ctz64(k));
+	if (k && !(k & (k-1)) && (cc == CC_EQ || cc == CC_NE) &&
+	    asm_guardtnb(as, cc == CC_EQ ? A64I_TBZ : A64I_TBNZ, bleft,
+			 emit_ctz64(k)))
 	  return;
-	}
 	m2 = emit_isk13(k, irt_is64(irl->t));
      }
-      bleft = ra_alloc1(as, blref, RSET_GPR);
      ai = (irt_is64(irl->t) ? A64I_TSTx : A64I_TSTw);
      if (!m2)
 	m2 = asm_fuseopm(as, ai, brref, rset_exclude(RSET_GPR, bleft));
@@ -1765,37 +1758,28 @@ static void asm_prof(ASMState *as, IRIns *ir)
 static void asm_stack_check(ASMState *as, BCReg topslot,
 			    IRIns *irp, RegSet allow, ExitNo exitno)
 {
-  Reg pbase;
  uint32_t k;
+  Reg pbase = RID_BASE;
  if (irp) {
-    if (!ra_hasspill(irp->s)) {
-      pbase = irp->r;
-      lj_assertA(ra_hasreg(pbase), "base reg lost");
-    } else if (allow) {
-      pbase = rset_pickbot(allow);
-    } else {
-      pbase = RID_RET;
-      emit_lso(as, A64I_LDRx, RID_RET, RID_SP, 0);  /* Restore temp register. */
-    }
-  } else {
-    pbase = RID_BASE;
+    pbase = irp->r;
+    if (!ra_hasreg(pbase))
+      pbase = allow ? (0x40 | rset_pickbot(allow)) : (0xC0 | RID_RET);
  }
  emit_cond_branch(as, CC_LS, asm_exitstub_addr(as, exitno));
+  if (pbase & 0x80)  /* Restore temp. register. */
+    emit_lso(as, A64I_LDRx, (pbase & 31), RID_SP, 0);
  k = emit_isk12((8*topslot));
  lj_assertA(k, "slot offset %d does not fit in K12", 8*topslot);
  emit_n(as, A64I_CMPx^k, RID_TMP);
-  emit_dnm(as, A64I_SUBx, RID_TMP, RID_TMP, pbase);
+  emit_dnm(as, A64I_SUBx, RID_TMP, RID_TMP, (pbase & 31));
  emit_lso(as, A64I_LDRx, RID_TMP, RID_TMP,
 	   (int32_t)offsetof(lua_State, maxstack));
-  if (irp) {  /* Must not spill arbitrary registers in head of side trace. */
-    if (ra_hasspill(irp->s))
-      emit_lso(as, A64I_LDRx, pbase, RID_SP, sps_scale(irp->s));
-    emit_lso(as, A64I_LDRx, RID_TMP, RID_GL, glofs(as, &J2G(as->J)->cur_L));
-    if (ra_hasspill(irp->s) && !allow)
-      emit_lso(as, A64I_STRx, RID_RET, RID_SP, 0);  /* Save temp register. */
-  } else {
-    emit_getgl(as, RID_TMP, cur_L);
+  if (pbase & 0x40) {
+    emit_getgl(as, (pbase & 31), jit_base);
+    if (pbase & 0x80)  /* Save temp register. */
+      emit_lso(as, A64I_STRx, (pbase & 31), RID_SP, 0);
  }
+  emit_getgl(as, RID_TMP, cur_L);
 }

 /* Restore Lua stack from on-trace state. */
@@ -1837,7 +1821,7 @@ static void asm_stack_restore(ASMState *as, SnapShot *snap)

 /* Marker to prevent patching the GC check exit. */
 #define ARM64_NOPATCH_GC_CHECK \
-  (A64I_ORRx|A64F_D(RID_TMP)|A64F_M(RID_TMP)|A64F_N(RID_TMP))
+  (A64I_ORRx|A64F_D(RID_ZERO)|A64F_M(RID_ZERO)|A64F_N(RID_ZERO))

 /* Check GC threshold and do one or more GC steps. */
 static void asm_gc_check(ASMState *as)
@@ -1892,47 +1876,40 @@ static void asm_loop_tail_fixup(ASMState *as)

 /* -- Head of trace ------------------------------------------------------- */

-/* Reload L register from g->cur_L. */
-static void asm_head_lreg(ASMState *as)
-{
-  IRIns *ir = IR(ASMREF_L);
-  if (ra_used(ir)) {
-    Reg r = ra_dest(as, ir, RSET_GPR);
-    emit_getgl(as, r, cur_L);
-    ra_evictk(as);
-  }
-}
-
 /* Coalesce BASE register for a root trace. */
 static void asm_head_root_base(ASMState *as)
 {
-  IRIns *ir;
-  asm_head_lreg(as);
-  ir = IR(REF_BASE);
-  if (ra_hasreg(ir->r) && (rset_test(as->modset, ir->r) || irt_ismarked(ir->t)))
-    ra_spill(as, ir);
-  ra_destreg(as, ir, RID_BASE);
+  IRIns *ir = IR(REF_BASE);
+  Reg r = ir->r;
+  if (ra_hasreg(r)) {
+    ra_free(as, r);
+    if (rset_test(as->modset, r) || irt_ismarked(ir->t))
+      ir->r = RID_INIT;  /* No inheritance for modified BASE register. */
+    if (r != RID_BASE)
+      emit_movrr(as, ir, r, RID_BASE);
+  }
 }

 /* Coalesce BASE register for a side trace. */
-static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
+static Reg asm_head_side_base(ASMState *as, IRIns *irp)
 {
-  IRIns *ir;
-  asm_head_lreg(as);
-  ir = IR(REF_BASE);
-  if (ra_hasreg(ir->r) && (rset_test(as->modset, ir->r) || irt_ismarked(ir->t)))
-    ra_spill(as, ir);
-  if (ra_hasspill(irp->s)) {
-    rset_clear(allow, ra_dest(as, ir, allow));
-  } else {
-    Reg r = irp->r;
-    lj_assertA(ra_hasreg(r), "base reg lost");
-    rset_clear(allow, r);
-    if (r != ir->r && !rset_test(as->freeset, r))
-      ra_restore(as, regcost_ref(as->cost[r]));
-    ra_destreg(as, ir, r);
+  IRIns *ir = IR(REF_BASE);
+  Reg r = ir->r;
+  if (ra_hasreg(r)) {
+    ra_free(as, r);
+    if (rset_test(as->modset, r) || irt_ismarked(ir->t))
+      ir->r = RID_INIT;  /* No inheritance for modified BASE register. */
+    if (irp->r == r) {
+      return r;  /* Same BASE register already coalesced. */
+    } else if (ra_hasreg(irp->r) && rset_test(as->freeset, irp->r)) {
+      /* Move from coalesced parent reg. */
+      emit_movrr(as, ir, r, irp->r);
+      return irp->r;
+    } else {
+      emit_getgl(as, r, jit_base);  /* Otherwise reload BASE. */
+    }
  }
-  return allow;
+  return RID_NONE;
 }

 /* -- Tail of trace ------------------------------------------------------- */
@@ -1976,20 +1953,47 @@ static void asm_tail_prep(ASMState *as)
 /* Ensure there are enough stack slots for call arguments. */
 static Reg asm_setup_call_slots(ASMState *as, IRIns *ir, const CCallInfo *ci)
 {
-  IRRef args[CCI_NARGS_MAX*2];
+#if LJ_HASFFI
  uint32_t i, nargs = CCI_XNARGS(ci);
-  int nslots = 0, ngpr = REGARG_NUMGPR, nfpr = REGARG_NUMFPR;
-  asm_collectargs(as, ir, ci, args);
-  for (i = 0; i < nargs; i++) {
-    if (args[i] && irt_isfp(IR(args[i])->t)) {
-      if (nfpr > 0) nfpr--; else nslots += 2;
-    } else {
-      if (ngpr > 0) ngpr--; else nslots += 2;
+  if (nargs > (REGARG_NUMGPR < REGARG_NUMFPR ? REGARG_NUMGPR : REGARG_NUMFPR) ||
+      (LJ_TARGET_OSX && (ci->flags & CCI_VARARG))) {
+    IRRef args[CCI_NARGS_MAX*2];
+    int ngpr = REGARG_NUMGPR, nfpr = REGARG_NUMFPR;
+    int spofs = 0, spalign = LJ_TARGET_OSX ? 0 : 7, nslots;
+    asm_collectargs(as, ir, ci, args);
+#if LJ_ABI_WIN
+    if ((ci->flags & CCI_VARARG)) nfpr = 0;
+#endif
+    for (i = 0; i < nargs; i++) {
+      int al = spalign;
+      if (!args[i]) {
+#if LJ_TARGET_OSX
+	/* Marker for start of varaargs. */
+	nfpr = 0;
+	ngpr = 0;
+	spalign = 7;
+#endif
+      } else if (irt_isfp(IR(args[i])->t)) {
+	if (nfpr > 0) { nfpr--; continue; }
+#if LJ_ABI_WIN
+	if ((ci->flags & CCI_VARARG) && ngpr > 0) { ngpr--; continue; }
+#elif LJ_TARGET_OSX
+	al |= irt_isnum(IR(args[i])->t) ? 7 : 3;
+#endif
+      } else {
+	if (ngpr > 0) { ngpr--; continue; }
+#if LJ_TARGET_OSX
+	al |= irt_size(IR(args[i])->t) - 1;
+#endif
+      }
+      spofs = (spofs + 2*al+1) & ~al;  /* Align and bump stack pointer. */
    }
+    nslots = (spofs + 3) >> 2;
+    if (nslots > as->evenspill)  /* Leave room for args in stack slots. */
+      as->evenspill = nslots;
  }
-  if (nslots > as->evenspill)  /* Leave room for args in stack slots. */
-    as->evenspill = nslots;
-  return REGSP_HINT(RID_RET);
+#endif
+  return REGSP_HINT(irt_isfp(ir->t) ? RID_FPRET : RID_RET);
 }

 static void asm_setup_target(ASMState *as)
@@ -1,6 +1,6 @@
 /*
 ** MIPS IR assembler (SSA IR -> machine code).
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 /* -- Register allocator extensions --------------------------------------- */
@@ -1207,22 +1207,29 @@ nolo:
 static void asm_uref(ASMState *as, IRIns *ir)
 {
  Reg dest = ra_dest(as, ir, RSET_GPR);
-  if (irref_isk(ir->op1)) {
+  int guarded = (irt_t(ir->t) & (IRT_GUARD|IRT_TYPE)) == (IRT_GUARD|IRT_PGC);
+  if (irref_isk(ir->op1) && !guarded) {
    GCfunc *fn = ir_kfunc(IR(ir->op1));
    MRef *v = &gcref(fn->l.uvptr[(ir->op2 >> 8)])->uv.v;
    emit_lsptr(as, MIPSI_AL, dest, v, RSET_GPR);
  } else {
-    Reg uv = ra_scratch(as, RSET_GPR);
-    Reg func = ra_alloc1(as, ir->op1, RSET_GPR);
-    if (ir->o == IR_UREFC) {
-      asm_guard(as, MIPSI_BEQ, RID_TMP, RID_ZERO);
-      emit_tsi(as, MIPSI_AADDIU, dest, uv, (int32_t)offsetof(GCupval, tv));
-      emit_tsi(as, MIPSI_LBU, RID_TMP, uv, (int32_t)offsetof(GCupval, closed));
+    if (guarded)
+      asm_guard(as, ir->o == IR_UREFC ? MIPSI_BEQ : MIPSI_BNE, RID_TMP, RID_ZERO);
+    if (ir->o == IR_UREFC)
+      emit_tsi(as, MIPSI_AADDIU, dest, dest, (int32_t)offsetof(GCupval, tv));
+    else
+      emit_tsi(as, MIPSI_AL, dest, dest, (int32_t)offsetof(GCupval, v));
+    if (guarded)
+      emit_tsi(as, MIPSI_LBU, RID_TMP, dest, (int32_t)offsetof(GCupval, closed));
+    if (irref_isk(ir->op1)) {
+      GCfunc *fn = ir_kfunc(IR(ir->op1));
+      GCobj *o = gcref(fn->l.uvptr[(ir->op2 >> 8)]);
+      emit_loada(as, dest, o);
    } else {
-      emit_tsi(as, MIPSI_AL, dest, uv, (int32_t)offsetof(GCupval, v));
+      emit_tsi(as, MIPSI_AL, dest, ra_alloc1(as, ir->op1, RSET_GPR),
+	       (int32_t)offsetof(GCfuncL, uvptr) +
+	       (int32_t)sizeof(MRef) * (int32_t)(ir->op2 >> 8));
    }
-    emit_tsi(as, MIPSI_AL, uv, func, (int32_t)offsetof(GCfuncL, uvptr) +
-	     (int32_t)sizeof(MRef) * (int32_t)(ir->op2 >> 8));
  }
 }

@@ -1337,8 +1344,8 @@ static void asm_fload(ASMState *as, IRIns *ir)
      }
    }
    ofs = field_ofs[ir->op2];
+    lj_assertA(!irt_isfp(ir->t), "bad FP FLOAD");
  }
-  lj_assertA(!irt_isfp(ir->t), "bad FP FLOAD");
  emit_tsi(as, mi, dest, idx, ofs);
 }

@@ -2667,7 +2674,7 @@ static void asm_head_root_base(ASMState *as)
 }

 /* Coalesce BASE register for a side trace. */
-static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
+static Reg asm_head_side_base(ASMState *as, IRIns *irp)
 {
  IRIns *ir = IR(REF_BASE);
  Reg r = ir->r;
@@ -2676,15 +2683,15 @@ static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
    if (rset_test(as->modset, r) || irt_ismarked(ir->t))
      ir->r = RID_INIT;  /* No inheritance for modified BASE register. */
    if (irp->r == r) {
-      rset_clear(allow, r);  /* Mark same BASE register as coalesced. */
+      return r;  /* Same BASE register already coalesced. */
    } else if (ra_hasreg(irp->r) && rset_test(as->freeset, irp->r)) {
-      rset_clear(allow, irp->r);
      emit_move(as, r, irp->r);  /* Move from coalesced parent reg. */
+      return irp->r;
    } else {
      emit_getgl(as, r, jit_base);  /* Otherwise reload BASE. */
    }
  }
-  return allow;
+  return RID_NONE;
 }

 /* -- Tail of trace ------------------------------------------------------- */
@@ -1,6 +1,6 @@
 /*
 ** PPC IR assembler (SSA IR -> machine code).
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 /* -- Register allocator extensions --------------------------------------- */
@@ -840,23 +840,30 @@ static void asm_hrefk(ASMState *as, IRIns *ir)
 static void asm_uref(ASMState *as, IRIns *ir)
 {
  Reg dest = ra_dest(as, ir, RSET_GPR);
-  if (irref_isk(ir->op1)) {
+  int guarded = (irt_t(ir->t) & (IRT_GUARD|IRT_TYPE)) == (IRT_GUARD|IRT_PGC);
+  if (irref_isk(ir->op1) && !guarded) {
    GCfunc *fn = ir_kfunc(IR(ir->op1));
    MRef *v = &gcref(fn->l.uvptr[(ir->op2 >> 8)])->uv.v;
    emit_lsptr(as, PPCI_LWZ, dest, v, RSET_GPR);
  } else {
-    Reg uv = ra_scratch(as, RSET_GPR);
-    Reg func = ra_alloc1(as, ir->op1, RSET_GPR);
-    if (ir->o == IR_UREFC) {
-      asm_guardcc(as, CC_NE);
+    if (guarded) {
+      asm_guardcc(as, ir->o == IR_UREFC ? CC_NE : CC_EQ);
      emit_ai(as, PPCI_CMPWI, RID_TMP, 1);
-      emit_tai(as, PPCI_ADDI, dest, uv, (int32_t)offsetof(GCupval, tv));
-      emit_tai(as, PPCI_LBZ, RID_TMP, uv, (int32_t)offsetof(GCupval, closed));
-    } else {
-      emit_tai(as, PPCI_LWZ, dest, uv, (int32_t)offsetof(GCupval, v));
    }
-    emit_tai(as, PPCI_LWZ, uv, func,
-	     (int32_t)offsetof(GCfuncL, uvptr) + 4*(int32_t)(ir->op2 >> 8));
+    if (ir->o == IR_UREFC)
+      emit_tai(as, PPCI_ADDI, dest, dest, (int32_t)offsetof(GCupval, tv));
+    else
+      emit_tai(as, PPCI_LWZ, dest, dest, (int32_t)offsetof(GCupval, v));
+    if (guarded)
+      emit_tai(as, PPCI_LBZ, RID_TMP, dest, (int32_t)offsetof(GCupval, closed));
+    if (irref_isk(ir->op1)) {
+      GCfunc *fn = ir_kfunc(IR(ir->op1));
+      int32_t k = (int32_t)gcrefu(fn->l.uvptr[(ir->op2 >> 8)]);
+      emit_loadi(as, dest, k);
+    } else {
+      emit_tai(as, PPCI_LWZ, dest, ra_alloc1(as, ir->op1, RSET_GPR),
+	       (int32_t)offsetof(GCfuncL, uvptr) + 4*(int32_t)(ir->op2 >> 8));
+    }
  }
 }

@@ -2186,7 +2193,7 @@ static void asm_head_root_base(ASMState *as)
 }

 /* Coalesce BASE register for a side trace. */
-static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
+static Reg asm_head_side_base(ASMState *as, IRIns *irp)
 {
  IRIns *ir = IR(REF_BASE);
  Reg r = ir->r;
@@ -2195,15 +2202,15 @@ static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
    if (rset_test(as->modset, r) || irt_ismarked(ir->t))
      ir->r = RID_INIT;  /* No inheritance for modified BASE register. */
    if (irp->r == r) {
-      rset_clear(allow, r);  /* Mark same BASE register as coalesced. */
+      return r;  /* Same BASE register already coalesced. */
    } else if (ra_hasreg(irp->r) && rset_test(as->freeset, irp->r)) {
-      rset_clear(allow, irp->r);
      emit_mr(as, r, irp->r);  /* Move from coalesced parent reg. */
+      return irp->r;
    } else {
      emit_getgl(as, r, jit_base);  /* Otherwise reload BASE. */
    }
  }
-  return allow;
+  return RID_NONE;
 }

 /* -- Tail of trace ------------------------------------------------------- */
@@ -1,6 +1,6 @@
 /*
 ** x86/x64 IR assembler (SSA IR -> machine code).
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 /* -- Guard handling ------------------------------------------------------ */
@@ -109,7 +109,7 @@ static int asm_isk32(ASMState *as, IRRef ref, int32_t *k)
 /* Check if there's no conflicting instruction between curins and ref.
 ** Also avoid fusing loads if there are multiple references.
 */
-static int noconflict(ASMState *as, IRRef ref, IROp conflict, int noload)
+static int noconflict(ASMState *as, IRRef ref, IROp conflict, int check)
 {
  IRIns *ir = as->ir;
  IRRef i = as->curins;
@@ -118,7 +118,9 @@ static int noconflict(ASMState *as, IRRef ref, IROp conflict, int noload)
  while (--i > ref) {
    if (ir[i].o == conflict)
      return 0;  /* Conflict found. */
-    else if (!noload && (ir[i].op1 == ref || ir[i].op2 == ref))
+    else if ((check & 1) && (ir[i].o == IR_NEWREF || ir[i].o == IR_CALLS))
+      return 0;
+    else if ((check & 2) && (ir[i].op1 == ref || ir[i].op2 == ref))
      return 0;
  }
  return 1;  /* Ok, no conflict. */
@@ -134,13 +136,14 @@ static IRRef asm_fuseabase(ASMState *as, IRRef ref)
    lj_assertA(irb->op2 == IRFL_TAB_ARRAY, "expected FLOAD TAB_ARRAY");
    /* We can avoid the FLOAD of t->array for colocated arrays. */
    if (ira->o == IR_TNEW && ira->op1 <= LJ_MAX_COLOSIZE &&
-	!neverfuse(as) && noconflict(as, irb->op1, IR_NEWREF, 1)) {
+	!neverfuse(as) && noconflict(as, irb->op1, IR_NEWREF, 0)) {
      as->mrm.ofs = (int32_t)sizeof(GCtab);  /* Ofs to colocated array. */
      return irb->op1;  /* Table obj. */
    }
  } else if (irb->o == IR_ADD && irref_isk(irb->op2)) {
    /* Fuse base offset (vararg load). */
-    as->mrm.ofs = IR(irb->op2)->i;
+    IRIns *irk = IR(irb->op2);
+    as->mrm.ofs = irk->o == IR_KINT ? irk->i : (int32_t)ir_kint64(irk)->u64;
    return irb->op1;
  }
  return ref;  /* Otherwise use the given array base. */
@@ -455,7 +458,7 @@ static Reg asm_fuseload(ASMState *as, IRRef ref, RegSet allow)
    RegSet xallow = (allow & RSET_GPR) ? allow : RSET_GPR;
    if (ir->o == IR_SLOAD) {
      if (!(ir->op2 & (IRSLOAD_PARENT|IRSLOAD_CONVERT)) &&
-	  noconflict(as, ref, IR_RETF, 0) &&
+	  noconflict(as, ref, IR_RETF, 2) &&
 	  !(LJ_GC64 && irt_isaddr(ir->t))) {
 	as->mrm.base = (uint8_t)ra_alloc1(as, REF_BASE, xallow);
 	as->mrm.ofs = 8*((int32_t)ir->op1-1-LJ_FR2) +
@@ -466,12 +469,12 @@ static Reg asm_fuseload(ASMState *as, IRRef ref, RegSet allow)
    } else if (ir->o == IR_FLOAD) {
      /* Generic fusion is only ok for 32 bit operand (but see asm_comp). */
      if ((irt_isint(ir->t) || irt_isu32(ir->t) || irt_isaddr(ir->t)) &&
-	  noconflict(as, ref, IR_FSTORE, 0)) {
+	  noconflict(as, ref, IR_FSTORE, 2)) {
 	asm_fusefref(as, ir, xallow);
 	return RID_MRM;
      }
    } else if (ir->o == IR_ALOAD || ir->o == IR_HLOAD || ir->o == IR_ULOAD) {
-      if (noconflict(as, ref, ir->o + IRDELTA_L2S, 0) &&
+      if (noconflict(as, ref, ir->o + IRDELTA_L2S, 2+(ir->o != IR_ULOAD)) &&
 	  !(LJ_GC64 && irt_isaddr(ir->t))) {
 	asm_fuseahuref(as, ir->op1, xallow);
 	return RID_MRM;
@@ -481,7 +484,7 @@ static Reg asm_fuseload(ASMState *as, IRRef ref, RegSet allow)
      ** Fusing unaligned memory operands is ok on x86 (except for SIMD types).
      */
      if ((!irt_typerange(ir->t, IRT_I8, IRT_U16)) &&
-	  noconflict(as, ref, IR_XSTORE, 0)) {
+	  noconflict(as, ref, IR_XSTORE, 2)) {
 	asm_fusexref(as, ir->op1, xallow);
 	return RID_MRM;
      }
@@ -814,6 +817,7 @@ static void asm_tointg(ASMState *as, IRIns *ir, Reg left)
  emit_rr(as, XO_UCOMISD, left, tmp);
  emit_rr(as, XO_CVTSI2SD, tmp, dest);
  emit_rr(as, XO_XORPS, tmp, tmp);  /* Avoid partial register stall. */
+  checkmclim(as);
  emit_rr(as, XO_CVTTSD2SI, dest, left);
  /* Can't fuse since left is needed twice. */
 }
@@ -856,6 +860,7 @@ static void asm_conv(ASMState *as, IRIns *ir)
      emit_rr(as, XO_SUBSD, dest, bias);  /* Subtract 2^52+2^51 bias. */
      emit_rr(as, XO_XORPS, dest, bias);  /* Merge bias and integer. */
      emit_rma(as, XO_MOVSD, bias, k);
+      checkmclim(as);
      emit_mrm(as, XO_MOVD, dest, asm_fuseload(as, lref, RSET_GPR));
      return;
    } else {  /* Integer to FP conversion. */
@@ -1172,6 +1177,7 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge)
    asm_guardcc(as, CC_E);
  else
    emit_sjcc(as, CC_E, l_end);
+  checkmclim(as);
  if (irt_isnum(kt)) {
    if (isk) {
      /* Assumes -0.0 is already canonicalized to +0.0. */
@@ -1231,7 +1237,6 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge)
 #endif
  }
  emit_sfixup(as, l_loop);
-  checkmclim(as);
 #if LJ_GC64
  if (!isk && irt_isaddr(kt)) {
    emit_rr(as, XO_OR, tmp|REX_64, key);
@@ -1258,6 +1263,7 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge)
      emit_rr(as, XO_ARITH(XOg_SUB), dest, tmp);
      emit_shifti(as, XOg_ROL, tmp, HASH_ROT3);
      emit_rr(as, XO_ARITH(XOg_XOR), dest, tmp);
+      checkmclim(as);
      emit_shifti(as, XOg_ROL, dest, HASH_ROT2);
      emit_rr(as, XO_ARITH(XOg_SUB), tmp, dest);
      emit_shifti(as, XOg_ROL, dest, HASH_ROT1);
@@ -1275,7 +1281,6 @@ static void asm_href(ASMState *as, IRIns *ir, IROp merge)
      } else {
 	emit_rr(as, XO_MOV, tmp, key);
 #if LJ_GC64
-	checkmclim(as);
 	emit_gri(as, XG_ARITHi(XOg_XOR), dest, irt_toitype(kt) << 15);
 	if ((as->flags & JIT_F_BMI2)) {
 	  emit_i8(as, 32);
@@ -1372,24 +1377,31 @@ static void asm_hrefk(ASMState *as, IRIns *ir)
 static void asm_uref(ASMState *as, IRIns *ir)
 {
  Reg dest = ra_dest(as, ir, RSET_GPR);
-  if (irref_isk(ir->op1)) {
+  int guarded = (irt_t(ir->t) & (IRT_GUARD|IRT_TYPE)) == (IRT_GUARD|IRT_PGC);
+  if (irref_isk(ir->op1) && !guarded) {
    GCfunc *fn = ir_kfunc(IR(ir->op1));
    MRef *v = &gcref(fn->l.uvptr[(ir->op2 >> 8)])->uv.v;
    emit_rma(as, XO_MOV, dest|REX_GC64, v);
  } else {
    Reg uv = ra_scratch(as, RSET_GPR);
-    Reg func = ra_alloc1(as, ir->op1, RSET_GPR);
-    if (ir->o == IR_UREFC) {
+    if (ir->o == IR_UREFC)
      emit_rmro(as, XO_LEA, dest|REX_GC64, uv, offsetof(GCupval, tv));
-      asm_guardcc(as, CC_NE);
-      emit_i8(as, 1);
-      emit_rmro(as, XO_ARITHib, XOg_CMP, uv, offsetof(GCupval, closed));
-    } else {
+    else
      emit_rmro(as, XO_MOV, dest|REX_GC64, uv, offsetof(GCupval, v));
+    if (guarded) {
+      asm_guardcc(as, ir->o == IR_UREFC ? CC_E : CC_NE);
+      emit_i8(as, 0);
+      emit_rmro(as, XO_ARITHib, XOg_CMP, uv, offsetof(GCupval, closed));
+    }
+    if (irref_isk(ir->op1)) {
+      GCfunc *fn = ir_kfunc(IR(ir->op1));
+      GCobj *o = gcref(fn->l.uvptr[(ir->op2 >> 8)]);
+      emit_loada(as, uv, o);
+    } else {
+      emit_rmro(as, XO_MOV, uv|REX_GC64, ra_alloc1(as, ir->op1, RSET_GPR),
+	        (int32_t)offsetof(GCfuncL, uvptr) +
+	        (int32_t)sizeof(MRef) * (int32_t)(ir->op2 >> 8));
    }
-    emit_rmro(as, XO_MOV, uv|REX_GC64, func,
-	      (int32_t)offsetof(GCfuncL, uvptr) +
-	      (int32_t)sizeof(MRef) * (int32_t)(ir->op2 >> 8));
  }
 }

@@ -1546,6 +1558,7 @@ static void asm_ahuvload(ASMState *as, IRIns *ir)
  if (irt_islightud(ir->t)) {
    Reg dest = asm_load_lightud64(as, ir, 1);
    if (ra_hasreg(dest)) {
+      checkmclim(as);
      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);
@@ -1593,6 +1606,7 @@ static void asm_ahuvload(ASMState *as, IRIns *ir)
  if (LJ_64 && irt_type(ir->t) >= IRT_NUM) {
    lj_assertA(irt_isinteger(ir->t) || irt_isnum(ir->t),
 	       "bad load type %d", irt_type(ir->t));
+    checkmclim(as);
 #if LJ_GC64
    emit_u32(as, LJ_TISNUM << 15);
 #else
@@ -2877,7 +2891,7 @@ static void asm_head_root_base(ASMState *as)
 }

 /* Coalesce or reload BASE register for a side trace. */
-static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
+static Reg asm_head_side_base(ASMState *as, IRIns *irp)
 {
  IRIns *ir = IR(REF_BASE);
  Reg r = ir->r;
@@ -2886,16 +2900,16 @@ static RegSet asm_head_side_base(ASMState *as, IRIns *irp, RegSet allow)
    if (rset_test(as->modset, r) || irt_ismarked(ir->t))
      ir->r = RID_INIT;  /* No inheritance for modified BASE register. */
    if (irp->r == r) {
-      rset_clear(allow, r);  /* Mark same BASE register as coalesced. */
+      return r;  /* Same BASE register already coalesced. */
    } else if (ra_hasreg(irp->r) && rset_test(as->freeset, irp->r)) {
      /* Move from coalesced parent reg. */
-      rset_clear(allow, irp->r);
      emit_rr(as, XO_MOV, r|REX_GC64, irp->r);
+      return irp->r;
    } else {
      emit_getgl(as, r, jit_base);  /* Otherwise reload BASE. */
    }
  }
-  return allow;
+  return RID_NONE;
 }

 /* -- Tail of trace ------------------------------------------------------- */
@@ -1,6 +1,6 @@
 /*
 ** Internal assertions.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lj_assert_c
@@ -1,6 +1,6 @@
 /*
 ** Bytecode instruction modes.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lj_bc_c
@@ -1,6 +1,6 @@
 /*
 ** Bytecode instruction format.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #ifndef _LJ_BC_H
@@ -1,6 +1,6 @@
 /*
 ** Bytecode dump definitions.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #ifndef _LJ_BCDUMP_H
@@ -1,6 +1,6 @@
 /*
 ** Bytecode reader.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lj_bcread_c
@@ -1,6 +1,6 @@
 /*
 ** Bytecode writer.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lj_bcwrite_c
@@ -189,7 +189,8 @@ static void bcwrite_knum(BCWriteCtx *ctx, GCproto *pt)
      goto save_int;
    } else {
      /* Write a 33 bit ULEB128 for the int (lsb=0) or loword (lsb=1). */
-      if (!LJ_DUALNUM) {  /* Narrow number constants to integers. */
+      if (!LJ_DUALNUM && o->u32.hi != LJ_KEYINDEX) {
+	/* Narrow number constants to integers. */
 	lua_Number num = numV(o);
 	k = lj_num2int(num);
 	if (num == (lua_Number)k) {  /* -0 is never a constant. */
@@ -1,6 +1,6 @@
 /*
 ** Buffer handling.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #define lj_buf_c
@@ -1,6 +1,6 @@
 /*
 ** Buffer handling.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #ifndef _LJ_BUF_H
@@ -1,6 +1,6 @@
 /*
 ** C data arithmetic.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #include "lj_obj.h"
@@ -44,9 +44,13 @@ static int carith_checkarg(lua_State *L, CTState *cts, CDArith *ca)
 	p = (uint8_t *)cdata_getptr(p, ct->size);
 	if (ctype_isref(ct->info)) ct = ctype_rawchild(cts, ct);
      } else if (ctype_isfunc(ct->info)) {
+	CTypeID id0 = i ? ctype_typeid(cts, ca->ct[0]) : 0;
 	p = (uint8_t *)*(void **)p;
 	ct = ctype_get(cts,
 	  lj_ctype_intern(cts, CTINFO(CT_PTR, CTALIGN_PTR|id), CTSIZE_PTR));
+	if (i) {  /* cts->tab may have been reallocated. */
+	  ca->ct[0] = ctype_get(cts, id0);
+	}
      }
      if (ctype_isenum(ct->info)) ct = ctype_child(cts, ct);
      ca->ct[i] = ct;
@@ -1,6 +1,6 @@
 /*
 ** C data arithmetic.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #ifndef _LJ_CARITH_H
@@ -1,6 +1,6 @@
 /*
 ** FFI C call handling.
-** Copyright (C) 2005-2022 Mike Pall. See Copyright Notice in luajit.h
+** Copyright (C) 2005-2023 Mike Pall. See Copyright Notice in luajit.h
 */

 #include "lj_obj.h"
@@ -20,12 +20,15 @@
 #if LJ_TARGET_X86
 /* -- x86 calling conventions --------------------------------------------- */

+#define CCALL_PUSH(arg) \
+  *(GPRArg *)((uint8_t *)cc->stack + nsp) = (GPRArg)(arg), nsp += CTSIZE_PTR
+
 #if LJ_ABI_WIN

 #define CCALL_HANDLE_STRUCTRET \
  /* Return structs bigger than 8 by reference (on stack only). */ \
  cc->retref = (sz > 8); \
-  if (cc->retref) cc->stack[nsp++] = (GPRArg)dp;
+  if (cc->retref) CCALL_PUSH(dp);

 #define CCALL_HANDLE_COMPLEXRET CCALL_HANDLE_STRUCTRET

@@ -40,7 +43,7 @@
    if (ngpr < maxgpr) \
      cc->gpr[ngpr++] = (GPRArg)dp; \
    else \
-      cc->stack[nsp++] = (GPRArg)dp; \
+      CCALL_PUSH(dp); \
  } else {  /* Struct with single FP field ends up in FPR. */ \
    cc->resx87 = ccall_classify_struct(cts, ctr); \
  }
@@ -56,7 +59,7 @@
  if (ngpr < maxgpr) \
    cc->gpr[ngpr++] = (GPRArg)dp; \
  else \
-    cc->stack[nsp++] = (GPRArg)dp;
+    CCALL_PUSH(dp);

 #endif

@@ -67,7 +70,7 @@
    if (ngpr < maxgpr) \
      cc->gpr[ngpr++] = (GPRArg)dp; \
    else \
-      cc->stack[nsp++] = (GPRArg)dp; \
+      CCALL_PUSH(dp); \
  }

 #endif
@@ -278,8 +281,8 @@
  if (ngpr < maxgpr) { \
    dp = &cc->gpr[ngpr]; \
    if (ngpr + n > maxgpr) { \
-      nsp += ngpr + n - maxgpr;  /* Assumes contiguous gpr/stack fields. */ \
-      if (nsp > CCALL_MAXSTACK) goto err_nyi;  /* Too many arguments. */ \
+      nsp += (ngpr + n - maxgpr) * CTSIZE_PTR;  /* Assumes contiguous gpr/stack fields. */ \
+      if (nsp > CCALL_SIZE_STACK) goto err_nyi;  /* Too many arguments. */ \
      ngpr = maxgpr; \
    } else { \
      ngpr += n; \
@@ -345,7 +348,6 @@
      goto done; \
    } else { \
      nfpr = CCALL_NARG_FPR;  /* Prevent reordering. */ \
-      if (LJ_TARGET_OSX && d->size < 8) goto err_nyi; \
    } \
  } else {  /* Try to pass argument in GPRs. */ \
    if (!LJ_TARGET_OSX && (d->info & CTF_ALIGN) > CTALIGN_PTR) \
@@ -356,7 +358,6 @@
      goto done; \
    } else { \
      ngpr = maxgpr;  /* Prevent reordering. */ \
-      if (LJ_TARGET_OSX && d->size < 8) goto err_nyi; \
    } \
  }

@@ -471,8 +472,8 @@
  if (ngpr < maxgpr) { \
    dp = &cc->gpr[ngpr]; \
    if (ngpr + n > maxgpr) { \
-     nsp += ngpr + n - maxgpr;  /* Assumes contiguous gpr/stack fields. */ \
-     if (nsp > CCALL_MAXSTACK) goto err_nyi;  /* Too many arguments. */ \
+     nsp += (ngpr + n - maxgpr) * CTSIZE_PTR;  /* Assumes contiguous gpr/stack fields. */ \
+     if (nsp > CCALL_SIZE_STACK) goto err_nyi;  /* Too many arguments. */ \
     ngpr = maxgpr; \
    } else { \
     ngpr += n; \
@@ -565,8 +566,8 @@
  if (ngpr < maxgpr) { \
    dp = &cc->gpr[ngpr]; \
    if (ngpr + n > maxgpr) { \
-      nsp += ngpr + n - maxgpr;  /* Assumes contiguous gpr/stack fields. */ \
-      if (nsp > CCALL_MAXSTACK) goto err_nyi;  /* Too many arguments. */ \
+      nsp += (ngpr + n - maxgpr) * CTSIZE_PTR;  /* Assumes contiguous gpr/stack fields. */ \
+      if (nsp > CCALL_SIZE_STACK) goto err_nyi;  /* Too many arguments. */ \
      ngpr = maxgpr; \
    } else { \
      ngpr += n; \
@@ -698,10 +699,11 @@ static int ccall_struct_arg(CCallState *cc, CTState *cts, CType *d, int *rcl,
  lj_cconv_ct_tv(cts, d, (uint8_t *)dp, o, CCF_ARG(narg));
  if (ccall_struct_reg(cc, cts, dp, rcl)) {
    /* Register overflow? Pass on stack. */
-    MSize nsp = cc->nsp, n = rcl[1] ? 2 : 1;
-    if (nsp + n > CCALL_MAXSTACK) return 1;  /* Too many arguments. */
-    cc->nsp = nsp + n;
-    memcpy(&cc->stack[nsp], dp, n*CTSIZE_PTR);
+    MSize nsp = cc->nsp, sz = rcl[1] ? 2*CTSIZE_PTR : CTSIZE_PTR;
+    if (nsp + sz > CCALL_SIZE_STACK)
+      return 1;  /* Too many arguments. */
+    cc->nsp = nsp + sz;
+    memcpy((uint8_t *)cc->stack + nsp, dp, sz);
  }
  return 0;  /* Ok. */
 }
@@ -983,6 +985,14 @@ static int ccall_set_args(lua_State *L, CTState *cts, CType *ct,
    fid = ctf->sib;
  }

+#if LJ_TARGET_ARM64 && LJ_ABI_WIN
+  if ((ct->info & CTF_VARARG)) {
+    nsp -= maxgpr * CTSIZE_PTR;  /* May end up with negative nsp. */
+    ngpr = maxgpr;
+    nfpr = CCALL_NARG_FPR;
+  }
+#endif
+
  /* Walk through all passed arguments. */
  for (o = L->base+1, narg = 1; o < top; o++, narg++) {
    CTypeID did;
@@ -1019,25 +1029,31 @@ static int ccall_set_args(lua_State *L, CTState *cts, CType *ct,
      CCALL_HANDLE_STRUCTARG
    } else if (ctype_iscomplex(d->info)) {
      CCALL_HANDLE_COMPLEXARG
-    } else {
+    } else if (!(CCALL_PACK_STACKARG && ctype_isenum(d->info))) {
      sz = CTSIZE_PTR;
    }
-    sz = (sz + CTSIZE_PTR-1) & ~(CTSIZE_PTR-1);
-    n = sz / CTSIZE_PTR;  /* Number of GPRs or stack slots needed. */
+    n = (sz + CTSIZE_PTR-1) / CTSIZE_PTR;  /* Number of GPRs or stack slots needed. */

    CCALL_HANDLE_REGARG  /* Handle register arguments. */

    /* Otherwise pass argument on stack. */
-    if (CCALL_ALIGN_STACKARG && !rp && (d->info & CTF_ALIGN) > CTALIGN_PTR) {
-      MSize align = (1u << ctype_align(d->info-CTALIGN_PTR)) -1;
-      nsp = (nsp + align) & ~align;  /* Align argument on stack. */
+    if (CCALL_ALIGN_STACKARG) {  /* Align argument on stack. */
+      MSize align = (1u << ctype_align(d->info)) - 1;
+      if (rp || (CCALL_PACK_STACKARG && isva && align < CTSIZE_PTR-1))
+	align = CTSIZE_PTR-1;
+      nsp = (nsp + align) & ~align;
    }
-    if (nsp + n > CCALL_MAXSTACK) {  /* Too many arguments. */
+#if LJ_TARGET_ARM64 && LJ_ABI_WIN
+    /* A negative nsp points into cc->gpr. Blame MS for their messy ABI. */
+    dp = ((uint8_t *)cc->stack) + (int32_t)nsp;
+#else
+    dp = ((uint8_t *)cc->stack) + nsp;
+#endif
+    nsp += CCALL_PACK_STACKARG ? sz : n * CTSIZE_PTR;
+    if ((int32_t)nsp > CCALL_SIZE_STACK) {  /* Too many arguments. */
    err_nyi:
      lj_err_caller(L, LJ_ERR_FFI_NYICALL);
    }
-    dp = &cc->stack[nsp];
-    nsp += n;
    isva = 0;

  done:
@@ -1048,7 +1064,8 @@ static int ccall_set_args(lua_State *L, CTState *cts, CType *ct,
    }
    lj_cconv_ct_tv(cts, d, (uint8_t *)dp, o, CCF_ARG(narg));
    /* Extend passed integers to 32 bits at least. */
-    if (ctype_isinteger_or_bool(d->info) && d->size < 4) {
+    if (ctype_isinteger_or_bool(d->info) && d->size < 4 &&
+	(!CCALL_PACK_STACKARG || !((uintptr_t)dp & 3))) {  /* Assumes LJ_LE. */
      if (d->info & CTF_UNSIGNED)
 	*(uint32_t *)dp = d->size == 1 ? (uint32_t)*(uint8_t *)dp :
 					 (uint32_t)*(uint16_t *)dp;
@@ -1095,14 +1112,17 @@ static int ccall_set_args(lua_State *L, CTState *cts, CType *ct,
 #endif
  }
  if (fid) lj_err_caller(L, LJ_ERR_FFI_NUMARG);  /* Too few arguments. */
+#if LJ_TARGET_ARM64 && LJ_ABI_WIN
+  if ((int32_t)nsp < 0) nsp = 0;
+#endif

 #if LJ_TARGET_X64 || (LJ_TARGET_PPC && !LJ_ABI_SOFTFP)
  cc->nfpr = nfpr;  /* Required for vararg functions. */
 #endif
-  cc->nsp = nsp;
-  cc->spadj = (CCALL_SPS_FREE + CCALL_SPS_EXTRA)*CTSIZE_PTR;
-  if (nsp > CCALL_SPS_FREE)
-    cc->spadj += (((nsp-CCALL_SPS_FREE)*CTSIZE_PTR + 15u) & ~15u);
+  cc->nsp = (nsp + CTSIZE_PTR-1) & ~(CTSIZE_PTR-1);
+  cc->spadj = (CCALL_SPS_FREE + CCALL_SPS_EXTRA) * CTSIZE_PTR;
+  if (cc->nsp > CCALL_SPS_FREE * CTSIZE_PTR)
+    cc->spadj += (((cc->nsp - CCALL_SPS_FREE * CTSIZE_PTR) + 15u) & ~15u);
  return gcsteps;
 }

--- a/Show More
+++ b/Show More