src/compiler/translator/tree_ops/RewritePixelLocalStorage.cpp


Log

Author Commit Date CI Message
Chris Dalton 9d41585e 2022-08-12T14:20:34 Make PLS coherent on D3D 11.3 Adds a new internal memory qualifier to the compiler called "rasterOrdered", which we set in RewritePixelLocalStorage.cpp when D3D 11.3 Rasterizer Order Views are supported. The HLSL translator then generates RasterizerOrderedTexture2D<> instead of RWTexture2D<> when this qualifier is set. Bug: angleproject:7279 Change-Id: I39b8c3279b7bff93b7e57272e8fb84d9c0312616 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3830288 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Chris Dalton <chris@rive.app>
Chris Dalton 4a636cdd 2022-07-20T22:44:30 Require all PLS formats to consume exactly 4 bytes of storage D3D 11.0 UAVs only support R32_FLOAT, R32_UINT, R32_SINT formats. EXT_shader_pixel_local_storage explicitly states that all PLS variables consume exactly 4 bytes. ESSL images can only have both read and write access if their format is r32f, r32i, r32ui. (We were able to circumvent this via aliasing, but it was a huge source of bugs.) There is a large precedent for only supporting 4 bytes of storage in the capabilities we use for PLS, so this CL removes support for all PLS storage formats that are not 4 bytes. It also implements an "R32" mode for PLS, that does manual packing and unpacking of r32* image formats. If the application wants larger formats, it can always define multiple PLS planes and piece them together. Next up we ought to be able to support rg16* types with more packing/unpacking. With aliasing gone, and with a bit of tweaking, the PLS tests now pass on the Pixel 4 GLES bot. Bug: angleproject:7279 Bug: angleproject:7388 Bug: angleproject:7524 Bug: angleproject:7527 Change-Id: I6b8f62c2428ade6cb5413e33360d734e55dda0eb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3782579 Reviewed-by: Stephen White <senorblanco@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Chris Dalton <chris@rive.app>
Chris Dalton fa3d7d5c 2022-08-05T00:57:40 Make PLS coherent on Vulkan Uses the VK_EXT_fragment_shader_interlock extension to make the shader image implementation of PLS coherent on Vulkan. This extension is supported on AMD, Apple, NVIDIA, and Intel. Bug: angleproject:7279 Change-Id: Ic0253eb20932eb6be0b1f433ba454e48b57be2f5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3813816 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Chris Dalton <chris@rive.app> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Shahbaz Youssefi 6f80f0f0 2022-08-06T02:29:19 Translator: Clean up the compile flag passing interface Historically, compile flags were sent to the translator as a bitmask. Recently, we were getting close to running out of bits. Additionally, direct-to-metal work had started to introduce constants to be passed to the translator, which were misplaced in ShBuiltInResources and Caps. Recent work on Pixel Local Storage adds even more constants, aggravating the situation. In this change, the interface to passing compile flags is reworked. A struct is passed (instead of a bitmask) that has one bit for each flag. This can be indefinitely extended. Additionally, the constants needed by metal and PLS are also placed in this struct. In turn, the backends can set these options directly, and don't have to hack them into Caps to further get hacked into ShBuiltInResources. Bug: angleproject:7559 Change-Id: If93f1e1b8818ad3a0ac708ab04ab93b4b397d114 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3812562 Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com> Commit-Queue: Amirali Abdolrashidi <abdolrashidi@google.com> Auto-Submit: Shahbaz Youssefi <syoussefi@chromium.org>
Chris Dalton 861149c7 2022-08-03T15:43:29 Make PLS coherent on desktop OpenGL Implements ANGLE_shader_pixel_local_storage_coherent using fragment shader synchronization extensions: NV_fragment_shader_interlock INTEL_fragment_shader_ordering ARB_fragment_shader_interlock With these extensions combined, we get coherency all 3 big desktop vendors: NVIDIA, Intel, and AMD. Bug: angleproject:7279 Change-Id: Ie20b251fb772898e89994b799640f1f2806581eb Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3773990 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Chris Dalton d57ce815 2022-07-16T16:33:08 Use "readwrite" PLS images when possible We actually only need readonly/writeonly aliases on ESSL, non r32f/r32ui. For all other cases, this change updates the compiler to emit a single readwrite image. We also optimize this image with the "restrict" qualifier since PLS specifically disallows aliasing. Removing the aliased load and store also eliminates our issue with an Intel driver bug, and all the PLS tests now pass without any workarounds. Bug: angleproject:7279 Bug: angleproject:7398 Change-Id: I350b239793647da33add96509b8f4b1bbef02245 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3767537 Commit-Queue: Chris Dalton <chris@rive.app> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Kenneth Russell <kbr@chromium.org>
Chris Dalton 77aa34ae 2022-07-16T13:08:00 Add support for PLS as function arguments Bug: angleproject:7279 Change-Id: I89d5c02148cbdbbd02dc4840ffada2c96c2a849b Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3767534 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Chris Dalton <chris@rive.app>
Chris Dalton c460c299 2022-06-21T11:20:45 Implement GLSL additions for ANGLE_shader_pixel_local_storage Specs out, implements, and thoroughly tests the GLSL additions for ANGLE_shader_pixel_local_storage. Adds a simple transformation that rewrites PLS directly into shader images. Updates the existing PLS tests to use the newly built-in PLS features and ensures they continue passing. For now, applications call glBindImageTexture to configure their pixel local storage. The OpenGL ES API side of this extension will follow shortly. Bug: angleproject:7279 Change-Id: I141183069b5cbfcca01cbb77b5b36d3e5f834bf5 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3761876 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Chris Dalton <chris@rive.app> Reviewed-by: Kenneth Russell <kbr@chromium.org>