- Setting the depth buffer for the backbuffer is now done as a boolean instead of a bit depth integer, in love.conf or love.window.setMode.
- Error if depth writes are enabled when the active canvas setup or backbuffer does not have a depth buffer (now it matches stencil behaviour).
- Don't allocate a backbuffer depth-stencil buffer if they're not requested. Metal also determines the format for that buffer based on which combination of depth and stencil is requested. OpenGL still has to allocate the depth-stencil buffer for the backbuffer all the time, because changing it requires the context to be recreated.
When usedpiscale=false, love will no longer automatically scale coordinates by the screen's DPI scale factor (so all coordinates will be in pixels), even when highdpi=true. love.window.getDPIScale will return 1, love.graphics.newCanvas will default to a dpi scale of 1 instead of the screen's dpi scale, etc.
love.window.getNativeDPIScale() will return the screen's DPI scale as reported by the OS no matter what usedpiscale is set to.
All enum errors have (hopefully) been changed to a luax_enumerror, which has a
fixed error message. If additionally a list of valid options is passed, it
lists that in the error message. For every enum error with few options I've
implemented this using a getConstants call.
This solution has been designed specifically to reduce the number of template
instantiations (as I've been told that was a concern). All new template code
happens in places where there already was an instantiation of the relevant
StringMap. Of course there is still std::vector<std::string>...
--HG--
branch : minor
With the highdpi window flag enabled on a retina-capable display and OS, content should now appear to the user at the same size and in the same positions as with the flag disabled.
As a result, mouse and touch coordinates, Texture and graphics dimensions, and the graphics coordinate system now use pixel density-scaled units instead of pixels. Raw pixel units should generally only be used for things such as shader algorithms which execute per-pixel and rely on accurate pixel dimensions. love.window.fromPixels and friends typically don’t need to be used anymore.
Images, Canvases, and Fonts can have an optional explicit ‘pixel density’ set when creating them. This allows for easily loading high pixel density content which displays at the same size as regular or low pixel density content.
API changes:
- Added Texture:getPixelWidth/getPixelHeight/getPixelDimensions and Texture:getPixelDensity. Texture:getWidth/getHeight return the pixel density-scaled width and height (as it will appear on the screen when drawn) rather than the number of pixels on each texture dimension.
- Added love.graphics.getPixelWidth/getPixelHeight/getPixelDimensions.
- Added optional ‘pixeldensity’ field to the settings table parameter of love.graphics.newImage. It defaults to 1, or if the file the Image was loaded from has “@2x”, “@3x”, etc. at the end of its name, it uses that number as the pixel density scale by default.
- love.graphics.newCanvas now takes a table as its third parameter, with fields “format”, “msaa”, and “pixeldensity”. pixeldensity defaults to the main screen’s pixel density. The width and height parameters specify the visual size that the Canvas will be drawn at / can be drawn to (pixel density-scaled units).
- love.graphics.newVideo accepts a table as its second parameter, with optional fields “audio” and “pixeldensity”. pixeldensity defaults to 1.
- love.graphics.newFont variants have an optional pixeldensity parameter at the end of the argument list. For TrueType fonts this defaults to the current pixel density scale of the screen, and for BMFonts and ImageFonts this defaults to 1.
- Added Font:getPixelDensity.
- Renamed love.window.getPixelScale to love.window.getPixelDensity.
--HG--
branch : minor
- Removed love.graphics.set/getCanvas, Canvas:renderTo, Canvas:newImageData, love.graphics.clear, love.graphics.discard, and love.graphics.newScreenshot.
- Added love.graphics.beginPass and love.graphics.endPass. All rendering must happen within a pass. love.draw is now called while a render pass to the main screen is active.
- Added love.graphics.renderPass, which is a wrapper for begin/endPass which calls the supplied function argument in between a begin/end. This is similar to the old Canvas:renderTo.
- Added love.graphics.isPassActive, getPassCanvases, getPassWidth, getPassHeight, and getPassDimensions.
- Added love.graphics.captureScreenshot.
- Added a new love.run callback love.drawpasses, which is called after love.update and before love.draw. It is the place to render to Canvases using beginPass or renderPass, since love.draw now always draws to the main screen.
- Renamed the ‘canvasswitches’ field in the table returned by love.graphics.getStats to ‘renderpasses’.
love.graphics.beginPass has the following variants (and renderPass mirrors them with an additional function argument at the end):
- love.graphics.beginPass(), begins rendering to the main screen without clearing it.
- love.graphics.beginPass(r, g, b [, a]), begins rendering to the main screen and clears the screen to the specified color.
- love.graphics.beginPass(canvas [, willstencil]), begins rendering to the specified Canvas without clearing it. love.graphics.stencil can only be used within the Canvas if ‘willstencil’ is true.
- love.graphics.beginPass(canvas, r, g, b [, a] [, willstencil]), beings rendering to the specified Canvas and clears it to the specified color. love.graphics.stencil can only be used within the Canvas if ‘willstencil’ is true.
- love.graphics.beginPass(info), where ‘info’ is a table in the following form:
{
{canvas [, r, g, b, a]}, — A canvas and an optional color to clear the Canvas to when the pass begins.
{canvas2, [r, g, b, a]}, — Additional Canvases and optional clear colors for multi-canvas rendering.
…,
stencil = false, — Optional boolean field to specify whether love.graphics.stencil is allowed within this pass. False by default.
}
The pass info table can be created once up-front and used for multiple beginPass/renderPass calls.
love.graphics.endPass can take an optional callback function argument, which causes endPass to capture the contents of the Canvas drawn to in the render pass to a new ImageData and passes it to the supplied function (this replaces Canvas:newImageData). This does not work on the main screen.
love.graphics.captureScreenshot replaces love.graphics.newScreenshot. It takes a callback function argument which causes love.graphics.present to capture the contents of the screen to a new ImageData and passes it to the supplied function. captureScreenshot can only be called before drawing to the main screen begins in the current frame (e.g. in love.keypressed, love.update, etc.)
--HG--
branch : minor
0 disables vsync, 1 sets it to sync every screen refreshes, 2 sets it to sync every 2 screen refreshes (effectively setting the framerate to half the monitor's refresh rate), etc.
As a special case, -1 also attempts to use adaptive vsync (vsync enabled when FPS is >= monitor's refresh rate, disabled whe not) if the driver supports it.
--HG--
branch : minor
Gamma-correct blending and shader math can be enabled globally via the new 't.gammacorrect' boolean flag in love.conf.
The new function love.graphics.isGammaCorrect will return true if it was requested in love.conf and is supported on the system.
When gamma correct rendering is enabled, colors (including the colors of pixels from images) are automatically converted from sRGB to linear RGB before use. When drawing to the main screen or to a canvas with the 'normal' or 'srgb' format, the final output colors of pixel shaders are automatically converted from linear RGB to sRGB after blending and before the color is stored in the pixel.
This lets the rendering pipeline do math using linear RGB values for colors rather than sRGB values, so the math is correct, without making users of the APIs manually linearize their colors with the love.math.gammaToLinear function (which still exists). The final output of the screen is encoded as sRGB, which is what systems expect. Canvases (except when otherwise requested) store their contents with sRGB encoding for increased precision with darker colors.
the 'srgb' window setting flag has been removed, as well as the 'srgb' image flag. A new image flag 'linear' has been added, which when set to true will cause the colors of the image to always be treated as linear RGB rather than sRGB, when gamma-correct rendering is enabled.
A new function 'Shader:sendColor' has been added, which has the same argument structure as 'Shader:send' but expects colors in the range of [0, 255]. When gamma-correct rendering is enabled it automatically gamma-corrects the given colors (by applying gammaToLinear to them.)
New shader code functions have been added: gammaToLinear, linearToGamma, gammaCorrectColor, and unGammaCorrectColor. When gamma-correct rendering is enabled, the LOVE_GAMMA_CORRECT #define is set and gammaCorrectColor and unGammaCorrectColor are aliases for gammaToLinear and linearToGamma respectively, otherwise the functions do nothing.
The new shader functions have 'precise' and 'fast' variants. If the LOVE_PRECISE_GAMMA define is set, then the normal functions default to the precise variants, otherwise they default to the fast variants. Currently the define is set in vertex shaders and not set in pixel shaders.
The default per-vertex color is automatically gamma-corrected by LÖVE when gamma correct rendering is enabled, but any custom named per-vertex attribute specified with the new custom attribute Mesh functionality won't be, unless 'gammaCorrectColor' or similar functions are explicitly used (preferably inside the vertex shader code that has the custom attribute.)
Sasha Szpakowski
Bart van Strien