compute.kernels.extractScaleValuesF32¶

Summary¶

compute.kernels.extractScaleValuesF32 extracts scalar values and finite-value flags from packed float sources. It supports component-based or magnitude-based extraction with configurable vector stride/offset. The output contains two buffers: extracted values (f32) and validity flags (u32). Use this as a preprocessing stage for scale-domain estimation and histogram/remap pipelines. Finite extracted values receive flag 1; NaN and infinite results are replaced with 0 and receive flag 0. Component count is clamped to 1–4, component index to 0–3, stride to at least the component count, and offset to at least zero. The source must support a read-only storage binding. This method rejects opts.encoder and submits its own work. Omitted result buffers are caller-owned and readback-capable; supplied buffers need at least count * 4 bytes.

Syntax¶

WasmGPU.compute.kernels.extractScaleValuesF32(src: BufferResource, opts: ScaleExtractOptions): ScaleExtractResult
const result = wgpu.compute.kernels.extractScaleValuesF32(src, opts);

Parameters¶

Name	Type	Required	Description
`src`	`BufferResource`	Yes	Source buffer containing float attributes.
`opts`	`ScaleExtractOptions`	Yes	Extraction config including `count` and optional component/stride/value-mode settings.

Returns¶

ScaleExtractResult - Object containing values and flags storage buffers.

Type Details¶

type BufferResource = GPUBuffer | StorageBuffer | UniformBuffer;

type ScaleExtractOptions = {
    encoder?: GPUCommandEncoder;
    label?: string;
    validateLimits?: boolean;
    count: number;
    componentCount?: number;
    componentIndex?: number;
    valueMode?: "component" | "magnitude";
    stride?: number;
    offset?: number;
    values?: StorageBuffer;
    flags?: StorageBuffer;
};

type ScaleExtractResult = {
    values: StorageBuffer;
    flags: StorageBuffer;
};

Example¶

const canvas = document.querySelector("canvas");
const wgpu = await WasmGPU.create(canvas);

const packed = wgpu.compute.createStorageBuffer({ data: new Float32Array([1, 2, 3, 4, 5, 6]), copySrc: true });
const result = wgpu.compute.kernels.extractScaleValuesF32(packed, {
    count: 2,
    componentCount: 3,
    componentIndex: 0,
    stride: 3,
    valueMode: "magnitude"
});

console.log(await wgpu.compute.readback.readF32(result.values, 0, 2));