Update WebGPU - #8955
Open
shoaibkamil wants to merge 48 commits into
Open
Conversation
- Replace WGPUShaderModuleWGSLDescriptor with WGPUShaderSourceWGSL - Replace WGPUProgrammableStageDescriptor with WGPUComputeState - Convert all string fields from const char* to WGPUStringView - Update callbacks to use *CallbackInfo structures (WGPUBufferMapCallbackInfo, etc) - Replace WGPU_*_INIT macros with explicit field initialization - Change wgpuBufferReference to wgpuBufferAddRef - Update WGPUConstantEntry.key to use WGPUStringView - Initialize WGPUDeviceDescriptor.defaultQueue and uncapturedErrorCallbackInfo explicitly Co-authored-by: AI Assistant (Cursor/Claude)
- Update callback signatures to use WGPUStringView and userdata2 - Replace WGPURequiredLimits/WGPUSupportedLimits with WGPULimits - Fix WGPUStringView field initialization (data/length) - Update device descriptor to use deviceLostCallbackInfo structure - Convert wgpuAdapterRequestDevice/wgpuInstanceRequestAdapter to use callback info structures - Change wgpuAdapterGetLimits return type check to WGPUStatus Co-authored-by: AI Assistant (Cursor/Claude)
- Define NULL and SIZE_MAX in mini_webgpu.h for freestanding runtime - Replace explicit field initialization with WGPU_*_INIT macros - Replace ~(size_t)0 with WGPU_STRLEN for clarity Co-authored-by: AI Assistant (Cursor/Claude)
Co-authored-by: AI Assistant (Cursor/Claude)
…awn+emscripten+node
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…e_promises # Conflicts: # doc/WebGPU.md # src/LLVM_Runtime_Linker.cpp # src/runtime/CMakeLists.txt # vcpkg-configuration.json
- Add vcpkg overlay port for Dawn that forces SHARED library build (required for runtime dlopen loading in JITModule.cpp) - Add arm64-osx-halide vcpkg triplet with dynamic linkage override - Update FindHalide_WebGPU.cmake to resolve Dawn via vcpkg CMake config with Apple framework dependencies (Metal, IOSurface, etc.) - Fix 32-bit ARM runtime target to use generic i386 to avoid cross-compilation issues
- Move arm64-osx-halide and x64-osx-halide triplets to cmake/vcpkg-triplets/ (the canonical location per vcpkg-configuration.json); remove the shadowing copies from cmake/vcpkg/triplets/ - Fix both triplets to use static linkage for all packages except Dawn, which must remain a shared library for dlopen loading; add missing VCPKG_CMAKE_SYSTEM_NAME and VCPKG_OSX_ARCHITECTURES to arm64-osx-halide - Drop the now-redundant cmake/vcpkg/triplets entry from vcpkg-configuration.json - Fix JITModule.cpp: when HL_WEBGPU_NATIVE_LIB is not set the fallback library search was incorrectly skipped (error was still empty, making the guard false); change condition to `env_libname.empty() || !error.empty()` so the search runs when no env var is provided - doc/WebGPU.md: update Dawn vcpkg port version, fix triplet description, remove duplicate HL_WEBGPU_NATIVE_LIB/HL_WEBGPU_NODE_BINDINGS note Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #8955 +/- ##
==========================================
+ Coverage 70.17% 70.18% +0.01%
==========================================
Files 255 255
Lines 78904 78910 +6
Branches 18866 18870 +4
==========================================
+ Hits 55367 55386 +19
- Misses 17883 17895 +12
+ Partials 5654 5629 -25 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
Apply the same WebGPU cleanup pattern used in generator_gpu_only test: device_free() buffers and halide_device_release() before exit. WebGPU via Node requires all GPU object references to be released or the process hangs. Co-authored-by: OpenCode 1.17.15 [ds4/deepseek-v4-flash]
…alide into shoaibkamil/webgpu_use_promises
…creation errors Several mullapudi2016 autoscheduler tests aborted with an opaque "wgpuDeviceCreateComputePipelineAsync failed (3)" error. The async pipeline-creation callback was discarding the WGPU error message, hiding the real cause: device creation only raised buffer-size and workgroup-storage limits, not maxComputeInvocationsPerWorkgroup or maxComputeWorkgroupSizeX/Y/Z, so it fell back to the default 256 invocations per workgroup even though the adapter supports more. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
1 task
alexreinking
added a commit
that referenced
this pull request
Jul 24, 2026
Drops Metal and non-SSE4.1 testing from Intel macOS. WebGPU will be restored as part of PR #8955 because our backend is broken against the latest Dawn that still builds.
Contributor
Author
|
I think this is now ready for review, but will not pass CI of course. |
Member
|
Unless WebGPU is added to the coverage workflow, we're going to see "0%" The solution isn't to run all the tests under coverage, it's just to |
shoaibkamil
marked this pull request as ready for review
July 27, 2026 13:39
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Update to latest WebGPU, Node, Dawn, and emscripten versions. Use a promise-based API instead of old-style tick-based APIs. This unifies the backend, but requires specific support for timed waits on promises. Dawn is currently the only implementation that has this.