Branch: refs/heads/main
Home: https://github.com/WebKit/WebKit
Commit: 49ef3002488aa8b83b2f1975816d4172863f1554
https://github.com/WebKit/WebKit/commit/49ef3002488aa8b83b2f1975816d4172863f1554
Author: Nikolas Zimmermann <nzimmerm...@igalia.com>
Date: 2025-05-15 (Thu, 15 May 2025)
Changed paths:
M
LayoutTests/imported/w3c/web-platform-tests/css/css-transforms/perspective-transforms-equivalence.html
M Source/WebCore/SourcesGTK.txt
M Source/WebCore/SourcesWPE.txt
M Source/WebCore/platform/TextureMapper.cmake
A Source/WebCore/platform/graphics/gbm/MemoryMappedGPUBuffer.cpp
A Source/WebCore/platform/graphics/gbm/MemoryMappedGPUBuffer.h
M Source/WebCore/platform/graphics/skia/SkiaPaintingEngine.cpp
M Source/WebCore/platform/graphics/skia/SkiaPaintingEngine.h
M Source/WebCore/platform/graphics/texmap/BitmapTexture.cpp
M Source/WebCore/platform/graphics/texmap/BitmapTexture.h
M Source/WebCore/platform/graphics/texmap/BitmapTexturePool.cpp
M
Source/WebCore/platform/graphics/texmap/coordinated/CoordinatedBackingStoreTile.cpp
Log Message:
-----------
[GTK][WPE] Add support for memory-mapped GPU buffers
https://bugs.webkit.org/show_bug.cgi?id=290800
Reviewed by Carlos Garcia Campos.
When using CPU rendering we produce pixel data that has to be uploaded
into a GL texture for composition - usually using glTexSubImage2D().
In order to save precious memory bandwidth on embedded devices, explore an
alternative mechanism, where buffer uploading happens out of the scope of
OpenGL to avoid overhead, such as copying the pixel data into a staging
buffer in Mesa etc.
The idea is to create a GBM buffer object, export it as a dma-buf, wrap
the dma-buf in an EGLImage and use it as backing for a regular OpenGL
texture, using glEGLImageTargetTexture2DOES(). Later on to update the
texture, one can mmap() the dma-buf fd, to gain write access to the
dma-buf. All memory operations have to be fenced using DMA_BUF_SYNC_*
system calls to ensure that the data is properly synchronized between
CPU and GPU (on shared memory systems, that are common in embedded).
Optimize BitmapTexture::updateContents() for the case when we want
to update a linear, mappable texture (backed by a dma-buf). Avoid
using glCopyTexSubImage2D, and use memcpy() instead on the mapped
memory region -- that avoids driver pitfalls, giving more control
about when synchronization happens.
Allocate tile textures always using GL_NEAREST instead of GL_LINEAR
filter -- GL_NEAREST is less expensive than GL_LINEAR, consuming
less bandwidth. However, it can only be used for textures that are
not scaled/filtered/... to avoid quality loss.
Add WEBKIT_SKIA_USE_LINEAR_TILE_TEXTURES environment variable,
only when set to 1 linear textures are used for the tiles. Off by
default, since the benefit is highly platform dependant.
Canonical link: https://commits.webkit.org/294962@main
To unsubscribe from these emails, change your notification settings at
https://github.com/WebKit/WebKit/settings/notifications
_______________________________________________
webkit-changes mailing list
webkit-changes@lists.webkit.org
https://lists.webkit.org/mailman/listinfo/webkit-changes
