Struct Fft1d 

pub struct Fft1d { /* private fields */ }

GPU-accelerated radix-2 1D FFT (or IFFT) kernel.

A single Fft1d instance is bound to a specific transform size and direction (forward/inverse). Construct it once with Fft1d::new and reuse it for multiple Fft1d::execute or Fft1d::execute_batch calls.

Implementations

impl Fft1d

pub fn new(ctx: &GpuContext, size: u32, inverse: bool) -> GpuResult<Self>

Create a new FFT kernel for the given transform size and direction.

Validation (pure-Rust, no GPU required)

• size must be a power of two.
• size must be in the range [4, 2048].

Errors

Returns GpuError::InvalidConfig if the size fails validation, or a pipeline / shader error if WGSL compilation fails on the GPU device.

pub fn size(&self) -> u32

Transform size (number of complex samples).

pub fn is_inverse(&self) -> bool

true when this kernel computes the inverse FFT.

pub fn execute( &self, ctx: &GpuContext, real_in: &[f32], imag_in: &[f32], ) -> GpuResult<(Vec<f32>, Vec<f32>)>

Execute a single 1D FFT/IFFT.

Arguments

• real_in — real parts of the input (length must equal self.size).
• imag_in — imaginary parts of the input (length must equal self.size).

Returns

A tuple (real_out, imag_out) each of length self.size.

Errors

Returns an error if input lengths are wrong or any GPU operation fails.

pub fn execute_batch( &self, ctx: &GpuContext, batch: &[(Vec<f32>, Vec<f32>)], ) -> GpuResult<Vec<(Vec<f32>, Vec<f32>)>>

Execute a batch of FFT/IFFT transforms in one GPU dispatch.

All transforms in the batch must use the same size (enforced by the shader’s workgroup layout). Each workgroup handles one transform.

Errors

Returns an error if any batch element has the wrong size or if a GPU operation fails.