1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
// SPDX-License-Identifier: AGPL-3.0-or-later
//! Fast Fourier Transform (FFT) Operations
//!
//! **Purpose**: Complex FFT for wave physics and frequency analysis
//!
//! **Architecture**: GPU-resident f64-canonical
//! - All FFT shaders written in f64 (`fft_3d_batched_f64.wgsl`)
//! - f32 path: downcast via `downcast_f64_to_f32` for broad compatibility
//! - f64 path: native when device has `SHADER_F64`, else DF64 fallback
//! - Zero CPU readbacks for f32 pipeline (`Fft1D`, `Fft2D`, `Fft3D`, `Rfft`)
//!
//! **Evolution**: Adapted from FHE NTT (80% structure reuse!)
//! - NTT: Butterfly in integer modular domain
//! - FFT: Butterfly in complex float domain
//! - SAME algorithm (Cooley-Tukey), DIFFERENT arithmetic
//!
//! ## Operations
//!
//! - [`fft_1d`] - 1D Complex FFT (Cooley-Tukey radix-2, f32) — **CRITICAL FOR PPPM**
//! - [`fft_1d_f64`] - 1D Complex FFT (f64) for high-precision MD simulations
//! - [`ifft_1d`] - 1D Inverse FFT (conjugate + normalize)
//! - [`fft_2d`] - 2D FFT via row-column decomposition
//! - [`fft_3d`] - 3D FFT for PPPM long-range forces
//! - [`rfft`] - Real-to-complex FFT (optimized)
//!
//! ## Mathematical Background
//!
//! The Discrete Fourier Transform (DFT):
//! ```text
//! X[k] = Σ(n=0 to N-1) x[n] · exp(-2πikn/N)
//! ```
//!
//! FFT accelerates this from O(N²) to O(N log N) via Cooley-Tukey butterfly.
//!
//! ## Usage Example
//!
//! ```rust,ignore
//! use barracuda::ops::fft::Fft1D;
//! use barracuda::prelude::{WgpuDevice, Tensor};
//! use std::sync::Arc;
//!
//! # async fn example() -> Result<(), Box<dyn std::error::Error>> {
//! let device = Arc::new(WgpuDevice::new().await?);
//!
//! let signal = Tensor::from_data(&signal_data, vec![1024, 2], device.clone())?;
//! let fft = Fft1D::new(signal, 1024)?;
//! let spectrum = fft.execute()?;
//! # Ok(())
//! # }
//! ```
pub use Fft1D;
pub use Fft1DF64;
pub use Fft2D;
pub use Fft3D;
pub use Fft3DF64;
pub use Ifft1D;
pub use Rfft;