CPU FFT API

CPU-based FFT fallback for environments without WebGPU.

cpuFFT()

Performs a 1D forward FFT on the CPU.

Signature

function cpuFFT(data: Float32Array): Float32Array;

Parameters

Parameter	Type	Description
data	Float32Array	Interleaved complex data [Re, Im, ...]. Must be power of 2.

Returns

Float32Array — Frequency domain data.

Example

import { cpuFFT } from 'webgpu-fft';

const signal = new Float32Array([1, 0, 2, 0, 3, 0, 4, 0]);
const spectrum = cpuFFT(signal);

cpuIFFT()

Performs a 1D inverse FFT.

Signature

function cpuIFFT(data: Float32Array): Float32Array;

Example

import { cpuIFFT } from 'webgpu-fft';

const reconstructed = cpuIFFT(spectrum);

cpuFFT2D()

Performs a 2D FFT on image data.

Signature

function cpuFFT2D(
  data: Float32Array,
  width: number,
  height: number
): Float32Array;

Example

import { cpuFFT2D } from 'webgpu-fft';

const freqData = cpuFFT2D(imageData, 256, 256);

cpuIFFT2D()

Performs a 2D inverse FFT.

Signature

function cpuIFFT2D(
  data: Float32Array,
  width: number,
  height: number
): Float32Array;

validateFFTInput()

Validates input data for 1D FFT.

Signature

function validateFFTInput(data: Float32Array): void;

Throws FFTError if input is invalid.

validateFFT2DInput()

Validates input data for 2D FFT.

Signature

function validateFFT2DInput(
  data: Float32Array,
  width: number,
  height: number
): void;

Throws FFTError if input is invalid.

Performance Notes

• CPU FFT is significantly slower than GPU for large sizes
• Best used for testing or small transforms (< 256 elements)
• See CPU Fallback Architecture for implementation details

Related

• FFTEngine — GPU-accelerated FFT
• CPU Fallback — Architecture deep dive