oxifft 0.2.0

Pure Rust implementation of FFTW - the Fastest Fourier Transform in the West
Documentation 
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
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261

//! SIMD-optimized codelets.
//!
//! This module provides SIMD-accelerated versions of the small DFT kernels.
//! The codelets use the architecture-specific SIMD backends when available.

// Items after statements are intentional for precomputed twiddle tables
#![allow(clippy::items_after_statements)]
// Large stack arrays are intentional for performance in fixed-size transforms
#![allow(clippy::large_stack_arrays)]

use core::any::TypeId;

use crate::kernel::{Complex, Float};
use crate::simd::{detect_simd_level, SimdLevel};

pub(crate) mod backends;
mod large_sizes;
mod small_sizes;
#[cfg(test)]
mod tests;

pub use large_sizes::*;
pub use small_sizes::*;

/// Detect if SIMD acceleration is available and beneficial.
#[inline]
pub fn simd_available() -> bool {
    let level = detect_simd_level();
    matches!(
        level,
        SimdLevel::Sse2 | SimdLevel::Avx | SimdLevel::Avx2 | SimdLevel::Avx512 | SimdLevel::Neon
    )
}

/// Size-2 DFT with automatic SIMD dispatch.
///
/// This function selects the best implementation based on available CPU features
/// and the float type. For f64, uses SIMD acceleration when available.
#[inline]
pub fn notw_2_dispatch<T: Float>(x: &mut [Complex<T>]) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_2_simd_f64(x_f64);
        return;
    }
    // Fallback to scalar for other types
    super::notw_2(x);
}

/// Size-4 DFT with automatic SIMD dispatch.
///
/// This function selects the best implementation based on available CPU features
/// and the float type. For f64, uses SIMD acceleration when available.
#[inline]
pub fn notw_4_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_4_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to scalar for other types
    super::notw_4(x, sign);
}

/// Size-8 DFT with automatic SIMD dispatch.
///
/// This function selects the best implementation based on available CPU features
/// and the float type. For f64, uses SIMD acceleration when available.
#[inline]
pub fn notw_8_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_8_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to scalar for other types
    super::notw_8(x, sign);
}

/// Size-16 DFT with automatic SIMD dispatch.
///
/// Uses scalar implementation with twiddle recurrence optimization.
#[inline]
pub fn notw_16_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_16_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to scalar for other types
    super::notw_16(x, sign);
}

/// Size-32 DFT with automatic SIMD dispatch.
///
/// Uses scalar implementation with twiddle recurrence optimization.
#[inline]
pub fn notw_32_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_32_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to scalar for other types
    super::notw_32(x, sign);
}

/// Size-64 DFT with automatic SIMD dispatch.
///
/// Uses scalar implementation with twiddle recurrence optimization.
#[inline]
pub fn notw_64_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_64_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to scalar for other types
    super::notw_64(x, sign);
}

/// Size-128 DFT with automatic SIMD dispatch.
///
/// Uses scalar implementation with twiddle recurrence optimization.
#[inline]
pub fn notw_128_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_128_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to scalar for other types
    super::notw_128(x, sign);
}

/// Size-256 DFT with automatic SIMD dispatch.
///
/// Uses scalar implementation with twiddle recurrence optimization.
#[inline]
pub fn notw_256_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_256_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to scalar for other types
    super::notw_256(x, sign);
}

/// Size-512 DFT with automatic SIMD dispatch.
///
/// Uses iterative DIT with SIMD butterflies for optimal performance.
#[inline]
pub fn notw_512_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_512_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to iterative DIT for other types.
    // Note: Must use execute_inplace directly to avoid infinite recursion,
    // since CooleyTukeySolver::execute dispatches back to this codelet.
    use crate::dft::problem::Sign;
    use crate::dft::solvers::CooleyTukeySolver;
    let sign_enum = if sign < 0 {
        Sign::Forward
    } else {
        Sign::Backward
    };
    CooleyTukeySolver::default().execute_dit_inplace(x, sign_enum);
}

/// Size-1024 DFT with automatic SIMD dispatch.
///
/// Uses iterative DIT with SIMD butterflies for optimal performance.
#[inline]
pub fn notw_1024_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_1024_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to iterative DIT for other types.
    // Note: Must use execute_inplace directly to avoid infinite recursion,
    // since CooleyTukeySolver::execute dispatches back to this codelet.
    use crate::dft::problem::Sign;
    use crate::dft::solvers::CooleyTukeySolver;
    let sign_enum = if sign < 0 {
        Sign::Forward
    } else {
        Sign::Backward
    };
    CooleyTukeySolver::default().execute_dit_inplace(x, sign_enum);
}

/// Size-4096 DFT with automatic SIMD dispatch.
///
/// Uses iterative DIT with SIMD butterflies for optimal performance.
#[inline]
pub fn notw_4096_dispatch<T: Float>(x: &mut [Complex<T>], sign: i32) {
    // Check if T is f64 at runtime
    if TypeId::of::<T>() == TypeId::of::<f64>() {
        // Safety: We verified T is f64, so the memory layout is identical
        let x_f64 = unsafe {
            core::slice::from_raw_parts_mut(x.as_mut_ptr().cast::<Complex<f64>>(), x.len())
        };
        notw_4096_simd_f64(x_f64, sign);
        return;
    }
    // Fallback to iterative DIT for other types.
    // Note: Must use execute_inplace directly to avoid infinite recursion,
    // since CooleyTukeySolver::execute dispatches back to this codelet.
    use crate::dft::problem::Sign;
    use crate::dft::solvers::CooleyTukeySolver;
    let sign_enum = if sign < 0 {
        Sign::Forward
    } else {
        Sign::Backward
    };
    CooleyTukeySolver::default().execute_dit_inplace(x, sign_enum);
}