src/compiler/translator/wgsl/OutputUniformBlocks.h


Log

Author Commit Date CI Message
Matthew Denton b9cec916 2025-08-12T19:44:18 WGSL: default uniforms gathered in interface block The default uniform struct was being manually output by OutputUniformBlocksAndSamplers(), which did not add the appropriate @align(16) annotations to lay the struct out according to WGSL's uniform address space layout requirements. This CL uses Vulkan's method of gathering the default uniforms into an interface block. The interface block will be output normally by the traverser, including @align() annotations. The variable declaring an instance of the interface block is still output by OutputUniformBlocksAndSamplers() because it needs special @group() and @binding() annotations. Bug: angleproject:376553328 Change-Id: Ib3be7d51ffedefe5ec579a1c9aaf0a535146c694 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6795028 Commit-Queue: Matthew Denton <mpdenton@chromium.org> Reviewed-by: Liza Burakova <liza@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Matthew Denton 7e28089b 2025-07-31T18:44:01 WGSL: support bvecs in uniforms WGSL does not allow booleans in the uniform address space. The last CL uses u32s to represent bools in the uniform address space, and this CL uses vecN<u32> to represent bvecN. Bug: angleproject:376553328 Change-Id: Ibf9f1fbf230ca03dd3d4661ba599e5fb1cb26ac4 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6808955 Reviewed-by: Liza Burakova <liza@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Matthew Denton <mpdenton@chromium.org>
Matthew Denton cac1e824 2025-04-29T18:27:36 WGSL: Output driver uniform and UBO structs This is the WGSL half of the change to implement driver uniforms. Driver uniforms are implemented as a UBO and reuse the default set of driver uniforms. User-provided UBOs don't yet have variables outputted for them. This requires moving MSL's ReduceInterfaceBlocks to the tree_ops dir in order to change interface block definitions into struct definitions. Bug: angleproject:389145696 Change-Id: I27f3837b3d115f2ffac66cc545f3b60ca9f01cb6 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6477564 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Commit-Queue: Matthew Denton <mpdenton@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Matthew Denton 4b69ba93 2025-04-28T16:45:56 WGPU: upload texture bind groups and configured samplers GLSL samplers are split into separate sampler/texture variables in WGSL. Before this CL, the WGSL translator generated shaders that look like: @group(1) @binding(@@@@@@) var ANGLE_sampler_samp2D : sampler; @group(1) @binding(@@@@@@) var ANGLE_texture_samp2D : texture_2d<f32>; @group(1) @binding(@@@@@@) var ANGLE_sampler_sampCube : sampler; @group(1) @binding(@@@@@@) var ANGLE_texture_sampCube : texture_cube<f32>; This CL replaces those with actual binding numbers @group(1) @binding(0) var ANGLE_sampler_samp2D : sampler; @group(1) @binding(1) var ANGLE_texture_samp2D : texture_2d<f32>; ... Such that @binding(n*2) is the WGSL sampler variable corresponding to the n-th GLSL sampler and @binding(n*2+1) is the WGSL texture variable corresponding to the n-th GLSL sampler. This CL then generates binding group layouts matching the above, and uploads textures and configured samplers in bind groups. This makes some of the deqp_gles2 tests 2d texture tests pass, though some fail because they need a flipped y coordinate. Not yet supported: 1. arrays of samplers 2. shadow samplers 3. cube textures Bug: angleproject:389145696 Change-Id: I2ab18ae5ebb4d1289101266bd9451576aa04ce2a Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6382272 Reviewed-by: Liza Burakova <liza@chromium.org> Auto-Submit: Matthew Denton <mpdenton@chromium.org> Commit-Queue: Geoff Lang <geofflang@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Matthew Denton 0cdbc781 2025-03-06T11:22:18 WGSL: Output samplers, including samplers from structs This output two WGSL variables for each GLSL sampler, including samplers in structs. This does not output the correct types for the WGSL variables, yet, nor does it generate texture access function calls. It also can't deal with arrays of samplers, which are not allowed in WGSL. Note that WGSL does not allow structs containing samplers to be arguments to a function, like Vulkan, nor does it allowed arrays of samplers at all, unlike Vulkan. This deals with the former problem the same way as Vulkan and Metal, by monomorphizing functions that take unsupported arguments. Bug: angleproject:389145696 Change-Id: I346688783dd2771c8fe6848b6783d948ed111783 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6253672 Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Matthew Denton <mpdenton@chromium.org>
Matthew Denton 8ee72cc7 2025-01-13T13:44:31 WGSL: support matCx2 in uniforms matCx2 in WGSL does not match std140 layout and so it needs to be translated as array<ANGLE_wrapped_vec2, C> when in a uniform. On use it needs to be converted to a WGSL-native matCx2. This also includes the optimization of an array<matCx2> in a uniform--when indexing into it, only the indexed element will be converted to a native matCx2, instead of converting the entire array and then indexing. Bug: angleproject:376553328 Change-Id: I1d84471234b3d3cf4cf361ae89cb61675d5bf9a8 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6157788 Reviewed-by: Liza Burakova <liza@chromium.org> Commit-Queue: Matthew Denton <mpdenton@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Matthew Denton 53ec86ab 2024-12-17T14:40:31 WGSL: support small stride arrays in uniforms WGSL requires arrays in the uniform address space to have a stride a multiple of 16. This CL makes WGSL translator emit wrapper structs for array element types used in the uniform address space, when the array stride is not a multiple of 16. The exception is for structs that aren't an aligned size of 16n, and for any types matCx2, since they are (or will be) handled in different ways that ensure alignment to 16. This should leave only f32, i32, u32, and vec2. See https://www.w3.org/TR/WGSL/#example-67da5de6 for an example of using a wrapper struct. This requires converting arrays with a wrapper struct element type to arrays with an unwrapped element type when they are first used; this can be "optimized" later for the common case of accessing a single array element, which can then be unwrapped immediately. This CL generates WGSL conversion functions when necessary. After this, the only types that can't yet be used in a uniform are matCx2 and bools. This is #2 in https://docs.google.com/document/d/17Qku1QEbLDhvJS-JJ9lPQAbnuZtLxWhG-ha5eCUhtEY/edit?tab=t.0#bookmark=id.rt3slgehd4te Bug: angleproject:376553328 Change-Id: I1edfa7f481a6cbf5b595643aae8728e67bc4b770 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6092038 Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Liza Burakova <liza@chromium.org> Reviewed-by: Matt Denton <mpdenton@google.com> Commit-Queue: Matt Denton <mpdenton@google.com>
Matthew Denton 473798bf 2024-11-28T00:38:36 WGSL: @align appropriate struct members in uniforms. Structs used in the uniform address space need to have certain members aligned according to the uniform address space layout constraints (substantially similar to std140). This CL adds @align annotations where necessary, in structs used in the uniform address space. Strictly speaking, it's okay to apply @align annotations to all structs used in the WGSL program, but this CL uses a pre-pass AST traverser to records all the structs used in the uniform address space. This is to avoid more unreadable generated code, and when more transformations are applied to these structs in future CLs, less generated code overall. After this, the only types that can't yet be used in a uniform are matCx2, arrays with stride not divisble by 16 (except when the array element type is a struct), and bools. This is #1 in struct translation in https://docs.google.com/document/d/17Qku1QEbLDhvJS-JJ9lPQAbnuZtLxWhG-ha5eCUhtEY/edit?tab=t.0#bookmark=id.rudfrn2o6jv1 Bug: angleproject:376553328 Change-Id: Ibff3414043a6ecb4a01ef8e3e71dad9c1066ddfd Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6056951 Commit-Queue: Matthew Denton <mpdenton@chromium.org> Reviewed-by: Liza Burakova <liza@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Matthew Denton 4a835cf2 2024-11-01T13:31:00 WebGPU: send uniforms to GPU for use in shader Basic uniforms should now be accessible in the shader during a draw call, rather than just via glGetUniform*(). This is unoptimized and creates a new GPU-side copy of all the uniforms, every time any uniform changes. This sets up the bind group layout for a pipeline and sets the bind groups in the renderer pass. Right now these bind groups only contain the default uniform blocks, but in the future will need to contain textures, samplers, and non-default uniform blocks (UBOs). Bug: angleproject:376553328 Change-Id: I50891b81ab2ee374d674213257f76319c0222120 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5980972 Reviewed-by: Geoff Lang <geofflang@chromium.org> Reviewed-by: Liza Burakova <liza@chromium.org> Commit-Queue: Matthew Denton <mpdenton@chromium.org>
Matthew Denton a6ee4641 2024-09-25T11:41:47 WGSL: Output default uniform block and accesses to it Default uniforms are put into a WGSL struct, and all accesses of those uniforms now output struct accesses. Similarly to I/O vars and builtins, these are outputted in a pre-pass, but in the future it might make sense to do what Vulkan does and do an AST transformation to put the default uniforms into a UBO which should be outputted similarly. This does not handle bool, matCx2, or array of element size < 16. Bug: angleproject:42267100 Change-Id: If29e2895a8aba3212b581813316af87273c1515c Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5878759 Reviewed-by: Liza Burakova <liza@chromium.org> Reviewed-by: Geoff Lang <geofflang@chromium.org> Commit-Queue: Matthew Denton <mpdenton@chromium.org>