[dawn] Add additional tier2 for maxSampledTexturesPerShaderStage.

- Adds a new tier2 for some limits and increase
  maxSampledTexturesPerShaderStage to 48 for the new tier. Note
  that we need to add a new tier and cannot just increase tier1
  because OpenGL basically never allows more than 16 sampled
  textures, and upgrading tier1 to 48 would result in a lot of
  OpenGL backends to regress to using tier0 and cause some other
  limits to drop as well which may be a breaking change.
- From investigating the different devices per platform, chose
  48 for the next tier value mainly because of Vulkan and Metal.
  Metal's limits at Apple Gen 4 and beyond (96) which is split
  (unequally) between sampled and storage textures would
  comfortably fit 48 because its currently split 54/42. For
  Vulkan, using 48 would only additionally lose the devices
  where it's limit is 31, which according to the database is
  only around 100 devices. For D3D11 and D3D12, the value
  doesn't matter unless we want to increase it past 128.
  Finally, for OpenGL, increasing the value past 16 will already
  lose a huge majority of devices (60%+) so I figured it doesn't
  make sense to try to optimize for that backend's compatibility
  for tier1 at all.

- Detailed breakdown per backend:
  - Vulkan:
    - Relvant limit is maxPerStageDescriptorSampledImages.
    - From the link below, it seems like the increased tier
      will contain a vast majority of devices.
    - https://vulkan.gpuinfo.org/displaydevicelimit.php?name=maxPerStageDescriptorSampledImages&platform=all

  - Metal:
    - Relevant limit is maxTextureArgumentEntriesPerFunc.
    - From our own tables:
      https://source.chromium.org/chromium/chromium/src/+/main:third_party/dawn/src/dawn/native/metal/PhysicalDeviceMTL.mm;l=816;drc=bf712ec1a13783224debb691ba88ad5c15b93194
    - This means we will drop support for tier1 for Apple
      families 1, 2, and 3. This is up to around the iPhone 7.

  - D3D12:
    - Relevant limit is maxSRVsPerStage.
    - From our own tables:
      https://source.chromium.org/chromium/chromium/src/+/main:third_party/dawn/src/dawn/native/d3d12/PhysicalDeviceD3D12.cpp;l=316?q=maxSampledTexturesPerShaderStage&ss=chromium%2Fchromium%2Fsrc:third_party%2Fdawn%2Fsrc%2Fdawn%2Fnative%2F
    - Minimum seems like it was already 128 for all D3D12, so increasing
      to 48 should be no-op.

  - D3D11:
    - Relevant limit is D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT.
    - Documentation of it:
      https://microsoft.github.io/windows-docs-rs/doc/windows/Win32/Graphics/Direct3D11/constant.D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT.html
    - The documentation also implies the min was already 128, so no-op.

  - OpenGL:
    - Relevant limits are GL_MAX_TEXTURE_IMAGE_UNITS and
      GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS.
    - From gpuinfo:
      https://opengles.gpuinfo.org/displaycapability.php?name=GL_MAX_TEXTURE_IMAGE_UNITS&esversion=2
      https://opengles.gpuinfo.org/displaycapability.php?name=GL_MAX_TEXTURE_IMAGE_UNITS&esversion=3
      https://opengles.gpuinfo.org/displaycapability.php?name=GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS&esversion=2
      https://opengles.gpuinfo.org/displaycapability.php?name=GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS&esversion=3
    - For what it's worth, increasing this limit for this backend past
      16, necessarily loses a majority of devices. As a result, there's
      really no point in considering the results for this class.

Bug: 475255737
Change-Id: I079140e7fbe2823b4d6e9b4c212e56a2abda6128
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/287198
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Auto-Submit: Loko Kung <lokokung@google.com>
Commit-Queue: Loko Kung <lokokung@google.com>
4 files changed
tree: a08df2bffc23a81a2036517114a7262ff447b272
  1. .github/
  2. .vscode/
  3. build_overrides/
  4. docs/
  5. generator/
  6. include/
  7. infra/
  8. scripts/
  9. src/
  10. test/
  11. third_party/
  12. tools/
  13. webgpu-cts/
  14. .bazelrc
  15. .clang-format
  16. .clang-format-ignore
  17. .clang-tidy
  18. .git-blame-ignore-revs
  19. .gitattributes
  20. .gitignore
  21. .gitmodules
  22. .gn
  23. .style.yapf
  24. .vpython3
  25. AUTHORS
  26. BUILD.bazel
  27. BUILD.gn
  28. CMakeLists.txt
  29. CMakeSettings.json
  30. CODE_OF_CONDUCT.md
  31. codereview.settings
  32. CONTRIBUTING.md
  33. CPPLINT.cfg
  34. DEPS
  35. DIR_METADATA
  36. go.mod
  37. go.sum
  38. go_presubmit_support.py
  39. LICENSE
  40. OWNERS
  41. PRESUBMIT.py
  42. README.chromium
  43. README.md
  44. test_spec_presubmit_support.py
  45. unsafe_buffers_paths.txt
  46. WATCHLISTS
  47. WORKSPACE.bazel
README.md

Build Status Matrix Space

Dawn, a WebGPU implementation

Dawn is an open-source and cross-platform implementation of the WebGPU standard. More precisely it implements webgpu.h that is a one-to-one mapping with the WebGPU IDL. Dawn is meant to be integrated as part of a larger system and is the underlying implementation of WebGPU in Chromium.

Dawn provides several WebGPU building blocks:

  • WebGPU C/C++ headers that applications and other building blocks use.
    • The webgpu.h version that Dawn implements.
    • A C++ wrapper for the webgpu.h.
  • A “native” implementation of WebGPU using platforms' GPU APIs: D3D12, Metal, Vulkan and OpenGL. See per API support for more details.
  • A client-server implementation of WebGPU for applications that are in a sandbox without access to native drivers
  • Tint is a compiler for the WebGPU Shader Language (WGSL) that can be used in standalone to convert shaders from and to WGSL.

Helpful links:

Documentation table of content

Developer documentation:

User documentation: (TODO, figure out what overlaps with the webgpu.h docs)

License

BSD 3-Clause License, please see LICENSE.

Disclaimer

This is not an officially supported Google product.