WasmGPU.compute.kernels.subU32¶
Summary¶
subU32 computes a - b element by element using WebGPU u32 arithmetic.
Syntax¶
WasmGPU.compute.kernels.subU32(a: StorageBuffer, b: StorageBuffer, opts?: VectorKernelOptions): StorageBuffer
Parameters¶
| Name | Type | Required | Description |
|---|---|---|---|
a |
StorageBuffer |
Yes | Minuend input vector. |
b |
StorageBuffer |
Yes | Subtrahend input vector. |
opts |
VectorKernelOptions |
No | Element count, reusable output, encoder, label, and workgroup-limit validation. |
Returns¶
StorageBuffer - Output containing a - b for each selected element. A newly allocated buffer enables COPY_SRC; a supplied opts.out is returned unchanged by identity.
Type Details¶
type VectorKernelOptions = {
count?: number;
out?: StorageBuffer;
encoder?: GPUCommandEncoder;
label?: string;
validateLimits?: boolean;
};
Values use WGSL u32 arithmetic; overflow wraps modulo 2³². When opts.count is omitted, the input logical lengths must match and that shared length is used. Optional output buffer with at least count * 4 bytes. The output must be distinct from every input. When opts.encoder is supplied, commands are recorded but not submitted by this call.
Example¶
const wgpu = await WasmGPU.create(document.querySelector("canvas"));
const a = wgpu.compute.createStorageBuffer({ data: new Uint32Array([1, 2, 3, 4]) });
const b = wgpu.compute.createStorageBuffer({ data: new Uint32Array([1, 2, 3, 4]) });
const out = wgpu.compute.kernels.subU32(a, b);
const values = await wgpu.compute.readback.readU32(out);
console.log(Array.from(values));