WebGPU workgroup shared memory for parallel reduction
Use WebGPU workgroup shared memory to reduce array segments on the GPU, then merge partials. Watch the 16 KB per-workgroup cap.
Loading…
ReadMeFeed / Community knowledge
#compute-shader
Real questions. Useful conversations. Find the people who know your stack.
Implementing Parallel Data Processing with WebGPU Compute Shaders
Learn how to implement parallel data processing in WebGPU using compute shaders, WGSL, and storage buffers to offload heavy calculations from the CPU to the GPU.
webgpuWebGPU compute passes: a practical alternative to WebGL2 GPGPU
Replace WebGL2 fragment-shader workarounds with WebGPU GPUComputePass for general GPU work, with feature checks and WGSL validation.
Parallel Array Processing with WebGPU Storage Buffers
Use WebGPU compute shaders and read-write storage buffers to process arrays on the GPU, then copy results back to JavaScript.
Offload array math to the GPU with WebGPU compute shaders
WebGPU compute shaders move parallel data work off the browser's main thread, but the explicit setup demands care. Here's how to decide if it's worth it.
Moving a CPU Particle Simulation to a WebGPU Compute Shader
Learn how to shift particle integration from the main thread to a GPU compute pass, set up storage buffers, tune workgroup size, and avoid the readback pitfall that can erase performance gains.
Performing a Data‑Parallel Reduction with WebGPU Storage Buffers
Learn how to use a WebGPU compute shader and storage buffers to sum an array of numbers on the GPU, including buffer setup, shader code, dispatch, and result retrieval.
WebGPU Data Input: Choosing Between Uniform and Storage Buffers
Learn when to use Uniform vs. Storage buffers in WebGPU compute shaders to avoid size limit crashes and optimize GPU cache performance.
Running Parallel Work in the Browser with WebGPU Compute Shaders
Learn how to run parallel workloads in the browser using WebGPU compute shaders and WGSL, with a grayscale image filter example and practical guidance on workgroup sizing and readback stalls.