feat(core): add WGPU (WebGPU) compute and rendering bindings#1497
Draft
sqrew wants to merge 5 commits intocarp-lang:masterfrom
Draft
feat(core): add WGPU (WebGPU) compute and rendering bindings#1497sqrew wants to merge 5 commits intocarp-lang:masterfrom
sqrew wants to merge 5 commits intocarp-lang:masterfrom
Conversation
Adds GPU compute support to core via a synchronous Carp API over wgpu-native (Vulkan / Metal / DX12 backend). - WGPU.carp: context, pipeline, buffer, bind group, dispatch, readback - GpuArray.carp: ergonomic multi-buffer dispatch abstraction - wgpu_helpers.h: thin C shim that converts async wgpu-native callbacks into blocking synchronous calls via spin-poll on wgpuInstanceProcessEvents All public API functions return (Result T String) for explicit error handling. Requires libwgpu_native and webgpu.h headers. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Extends the existing compute bindings with a full WebGPU rendering layer:
core/WGPURender.carp -- Carp bindings for GPU rendering:
- Surface creation from a GLFW window (X11 + Windows via C helper)
- Render pipelines (fullscreen triangle pass + geometry pass)
- Offscreen render textures and samplers
- Frame management: begin-frame / run-pass / end-frame
- Geometry pipeline: vertex buffers, index buffers, uniform buffers,
depth textures, indexed drawing (run-geom-pass-indexed)
- Vertex layout: position float32x3 + color float32x3, stride 24
- WGPURender.surface-format: queries actual GPU format to avoid
pipeline/swapchain format mismatch
core/wgpu_render_helpers.h -- header-only C shim:
- wgpu_create_surface_from_glfw: GLFW -> X11/Win32 native handles ->
WGPUSurface; queries GPU capabilities for preferred swapchain format
- Full geometry pipeline with depth testing (Depth24Plus, Less)
- wgpu_run_geom_pass / wgpu_run_geom_pass_indexed with depth attachment
- Uniform buffer create/update, index buffer, depth texture helpers
- depthSlice = WGPU_DEPTH_SLICE_UNDEFINED on color attachments (wgpu 0.20+)
- XSetWindowBackground on X11 to prevent screen imprint without compositor
core/GLFW.carp -- added GLFW_CLIENT_API / GLFW_NO_API constants
(required: hint window before creation when using WGPU instead of OpenGL)
core/wgpu_helpers.h -- suppress -Wunused-function for header-only helpers
examples/wgpu-pyramid.carp -- spinning rainbow pyramid demo:
- 5 vertices, 4 triangular faces, 12 indices (no base)
- Per-vertex colors: yellow apex, red/green/blue/purple base corners
- Uniform buffer: time + aspect ratio (16 bytes, GPU-side MVP)
- WGSL computes rotate_y(t) * rotate_x(0.3) + perspective projection
- Bounce: sin(t*2)*0.3 on Y translation
- Detects surface format at runtime for pipeline compatibility
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Three new headless/windowed GPU demos covering the full compute+render surface, plus the small API additions needed to run them. examples/wgpu-compute.carp -- GPU Array Add (headless compute) - Uploads two float arrays, runs WGSL element-wise add, reads back - Uses GpuArray for ergonomic buffer management - Demonstrates full compute path with zero render API surface examples/wgpu-gradient.carp -- Animated Plasma Gradient (windowed render) - Fullscreen triangle from vertex index, no geometry buffers needed - Time uniform drives HSV plasma shader via sine combinations - Demonstrates fullscreen pass pipeline end-to-end examples/wgpu-mandelbrot.carp -- Interactive Mandelbrot (compute + render) - Compute shader writes iteration counts into a storage buffer - Fullscreen fragment shader reads that buffer and maps to vivid colors - W/S zoom, arrow pan, ESC quit -- smooth_i log-normalized coloring - Demonstrates both pipelines in one frame loop core/WGPURender.carp -- two new Carp bind group helpers: - create-render-uniform-bind-group (uniform at binding 0) - create-render-storage-uniform-bind-group (storage at 0, uniform at 1) Both derive layout from the render pipeline; return (Result bg String). core/wgpu_render_helpers.h -- C implementations of the above: - wgpu_create_render_uniform_bind_group - wgpu_create_render_storage_uniform_bind_group Both use wgpuRenderPipelineGetBindGroupLayout to auto-derive layout, so no separate BindGroupLayout creation is needed in Carp. core/GpuArray.carp -- fix load path case: "wgpu.carp" -> "WGPU.carp" (case-sensitive on Linux; was silently broken outside bullet-heaven) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
examples/wgpu-*.carp, examples/wgpu-pyramid.carp: Replace absolute /home/sqrew/Desktop/Carp-fork/core/... paths with bare filenames (GLFW.carp, WGPURender.carp, GpuArray.carp). Carp resolves these via CARP_DIR/core/ at build time, matching the convention used by all other examples (SDL.carp etc.). test/wgpu.carp: GPU compute test suite. Runs element-wise array add through the full pipeline: create-context -> create-pipeline -> GpuArray upload -> dispatch-with -> download -> assert result == [11 22 33 44 55]. If create-context returns Error (no GPU / headless CI), the suite reports one skipped pass and exits 0 rather than failing. Confirmed passing on real GPU: 1 passed, 0 failed. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Contributor
Author
|
I absolutely don't expect this to be merged as is but I wanted to show where I am at with it and see if it's something even interesting to anyone. WGPU would open up the world of cross platform GPU compute on top of cross platform rendering and has a few example test showcases of both. It's currently working locally without issues but there are sure to be concerns such as requiring wgpu-native. I'm totally open to simply contributing this as an external library but I posed it as a part of the core language due to Carp's current game-focused narrative. It would also open up the possibility of replacing the OpenGL api in favor of a more modern alternative that matches Carp's natural functional style more closely. Anyways, I know this is large and unexpected so no rush in looking at this. Thanks for checking it out. |
Collaborator
|
Cool stuff! You make an interesting point with it perhaps being a better alternative for the "bundled graphics api", but need to think about it a bit :) |
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.
This PR introduces first-class support for WebGPU in Carp via the wgpu-native library.
Key Features:
Benefits:
Note: Requires wgpu-native installed on the system (linked via -lwgpu_native).