oxifft 0.3.1

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
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293

//! Tests for `pruned::partial` (PartialFft / PartialStrategy).

use crate::api::{Direction, Flags, Plan};
use crate::kernel::Complex;
use crate::prelude::*;

use super::partial::{PartialFft, PartialStrategy};

// ─── Helper ──────────────────────────────────────────────────────────────────

/// Run a full forward FFT and return the complete output vector.
fn full_fft(input: &[Complex<f64>]) -> Vec<Complex<f64>> {
    let n = input.len();
    let plan = Plan::dft_1d(n, Direction::Forward, Flags::ESTIMATE).unwrap();
    let mut output = vec![Complex::new(0.0_f64, 0.0); n];
    plan.execute(input, &mut output);
    output
}

/// Absolute difference between two complex numbers.
fn cdiff(a: Complex<f64>, b: Complex<f64>) -> f64 {
    let dr = a.re - b.re;
    let di = a.im - b.im;
    (dr * dr + di * di).sqrt()
}

// ─── Test 1: Goertzel single bin ─────────────────────────────────────────────

#[test]
fn test_goertzel_single_bin_vs_full_fft() {
    let n = 64;
    let input: Vec<Complex<f64>> = (0..n)
        .map(|i| Complex::new((i as f64) / (n as f64), 0.0))
        .collect();

    let full = full_fft(&input);

    let bin = 5_usize;
    let pf = PartialFft::<f64>::new_sparse(n, &[bin]);
    let mut out = vec![Complex::new(0.0_f64, 0.0); 1];
    pf.execute(&input, &mut out);

    assert!(
        cdiff(out[0], full[bin]) < 1e-10,
        "bin {bin}: got {:?}, expected {:?}",
        out[0],
        full[bin]
    );
}

// ─── Test 2: Goertzel multiple bins ──────────────────────────────────────────

#[test]
fn test_goertzel_multi_bins_vs_full_fft() {
    let n = 256;
    let input: Vec<Complex<f64>> = (0..n)
        .map(|i| {
            let t = (i as f64) / (n as f64);
            Complex::new((3.0 * t).sin(), (7.0 * t).cos())
        })
        .collect();

    let full = full_fft(&input);

    // K=4 bins; 4 < log2(256)=8, so Goertzel is chosen
    let bins = [3_usize, 17, 50, 200];
    let pf = PartialFft::<f64>::new_sparse(n, &bins);
    let mut out = vec![Complex::new(0.0_f64, 0.0); bins.len()];
    pf.execute(&input, &mut out);

    for (i, &bin) in bins.iter().enumerate() {
        assert!(
            cdiff(out[i], full[bin]) < 1e-10,
            "bin {bin}: got {:?}, expected {:?}",
            out[i],
            full[bin]
        );
    }
}

// ─── Test 3: Goertzel edge cases ─────────────────────────────────────────────

#[test]
fn test_goertzel_edge_cases() {
    let n = 64_usize;
    let input: Vec<Complex<f64>> = (0..n)
        .map(|i| Complex::new(1.0 + (i as f64) * 0.1, (i as f64) * 0.05))
        .collect();

    let full = full_fft(&input);

    // DC bin (k=0)
    let pf_dc = PartialFft::<f64>::new_sparse(n, &[0]);
    let mut out_dc = vec![Complex::new(0.0_f64, 0.0); 1];
    pf_dc.execute(&input, &mut out_dc);
    assert!(
        cdiff(out_dc[0], full[0]) < 1e-10,
        "DC bin mismatch: got {:?}, expected {:?}",
        out_dc[0],
        full[0]
    );

    // Nyquist bin (k = N/2)
    let nyquist = n / 2;
    let pf_ny = PartialFft::<f64>::new_sparse(n, &[nyquist]);
    let mut out_ny = vec![Complex::new(0.0_f64, 0.0); 1];
    pf_ny.execute(&input, &mut out_ny);
    assert!(
        cdiff(out_ny[0], full[nyquist]) < 1e-10,
        "Nyquist bin mismatch: got {:?}, expected {:?}",
        out_ny[0],
        full[nyquist]
    );

    // Last bin (k = N-1)
    let last = n - 1;
    let pf_last = PartialFft::<f64>::new_sparse(n, &[last]);
    let mut out_last = vec![Complex::new(0.0_f64, 0.0); 1];
    pf_last.execute(&input, &mut out_last);
    assert!(
        cdiff(out_last[0], full[last]) < 1e-10,
        "Last bin mismatch: got {:?}, expected {:?}",
        out_last[0],
        full[last]
    );
}

// ─── Test 4: OutputPruned prefix ─────────────────────────────────────────────

#[test]
fn test_output_pruned_prefix_vs_full_fft() {
    let n = 1024_usize;
    let input: Vec<Complex<f64>> = (0..n)
        .map(|i| {
            let t = (i as f64) / (n as f64);
            Complex::new(t.sin() + (5.0 * t).cos(), t * 0.1)
        })
        .collect();

    let full = full_fft(&input);

    // m=16, n=1024: log2(1024)=10, 16 > 10 so strategy is actually FullThenSlice
    // Use m=8 to get OutputPruned: 8 < 10 and 8 is power-of-two and 8 <= 512
    let m = 8_usize;
    let pf = PartialFft::<f64>::new_prefix(n, m);

    // Verify OutputPruned strategy was chosen
    assert!(
        matches!(pf.strategy(), PartialStrategy::OutputPruned { m: 8 }),
        "Expected OutputPruned, got {:?}",
        pf.strategy()
    );

    let mut out = vec![Complex::new(0.0_f64, 0.0); m];
    pf.execute(&input, &mut out);

    // Goertzel on N=1024 accumulates O(N·eps) ≈ 1e-9 error; use 2e-8 tolerance.
    // MIRI soft-float accumulates more rounding error than hardware, so we use 2e-8.
    for k in 0..m {
        assert!(
            cdiff(out[k], full[k]) < 2e-8,
            "prefix bin {k}: got {:?}, expected {:?}",
            out[k],
            full[k]
        );
    }
}

// ─── Test 5: FullThenSlice ───────────────────────────────────────────────────

#[test]
fn test_full_then_slice_vs_full_fft() {
    let n = 1024_usize;
    let input: Vec<Complex<f64>> = (0..n)
        .map(|i| Complex::new((i as f64).sin(), (i as f64 * 1.3).cos()))
        .collect();

    let full = full_fft(&input);

    // m=900: 900 >= log2(1024)=10, so FullThenSlice is chosen
    let m = 900_usize;
    let pf = PartialFft::<f64>::new_prefix(n, m);

    assert!(
        matches!(pf.strategy(), PartialStrategy::FullThenSlice { .. }),
        "Expected FullThenSlice, got {:?}",
        pf.strategy()
    );

    let mut out = vec![Complex::new(0.0_f64, 0.0); m];
    pf.execute(&input, &mut out);

    for k in 0..m {
        assert!(
            cdiff(out[k], full[k]) < 1e-10,
            "slice bin {k}: got {:?}, expected {:?}",
            out[k],
            full[k]
        );
    }
}

// ─── Test 6: Strategy crossover assertions ───────────────────────────────────

#[test]
fn test_strategy_crossover_sparse_goertzel() {
    // K=1 << log2(1024)=10: must pick Goertzel
    let pf = PartialFft::<f64>::new_sparse(1024, &[5]);
    assert!(
        matches!(pf.strategy(), PartialStrategy::Goertzel { .. }),
        "Expected Goertzel for K=1, got {:?}",
        pf.strategy()
    );
}

#[test]
fn test_strategy_crossover_prefix_output_pruned() {
    // n=1024, m=8: 8 < 10, power-of-two, 8 <= 512 → OutputPruned
    let pf = PartialFft::<f64>::new_prefix(1024, 8);
    assert!(
        matches!(pf.strategy(), PartialStrategy::OutputPruned { m: 8 }),
        "Expected OutputPruned for m=8, n=1024, got {:?}",
        pf.strategy()
    );
}

#[test]
fn test_strategy_crossover_prefix_full_then_slice() {
    // n=1024, m=900: 900 >= log2(1024)=10 → FullThenSlice
    let pf = PartialFft::<f64>::new_prefix(1024, 900);
    assert!(
        matches!(pf.strategy(), PartialStrategy::FullThenSlice { .. }),
        "Expected FullThenSlice for m=900, n=1024, got {:?}",
        pf.strategy()
    );
}

// ─── Test 7: Numerical stability at edge frequencies ─────────────────────────

#[test]
fn test_numerical_stability_dc_bin() {
    // bin 0: omega = 0.0 — potential cancellation in naive implementations
    let n = 256_usize;
    let input: Vec<Complex<f64>> = vec![Complex::new(1.0, 0.0); n];
    let full = full_fft(&input);

    let pf = PartialFft::<f64>::new_sparse(n, &[0]);
    let mut out = vec![Complex::new(0.0_f64, 0.0); 1];
    pf.execute(&input, &mut out);

    // DC of constant 1 signal = N
    assert!(
        (out[0].re - n as f64).abs() < 1e-10,
        "DC real: expected {}, got {}",
        n,
        out[0].re
    );
    assert!(
        out[0].im.abs() < 1e-10,
        "DC imag should be ~0, got {}",
        out[0].im
    );
    assert!(
        cdiff(out[0], full[0]) < 1e-10,
        "DC stability: got {:?}, expected {:?}",
        out[0],
        full[0]
    );
}

#[test]
fn test_numerical_stability_nyquist_bin() {
    // bin N/2: omega = pi — another potential catastrophic cancellation point
    let n = 256_usize;
    let input: Vec<Complex<f64>> = (0..n)
        .map(|i| Complex::new(if i % 2 == 0 { 1.0 } else { -1.0 }, 0.0))
        .collect();

    let full = full_fft(&input);
    let nyquist = n / 2;

    let pf = PartialFft::<f64>::new_sparse(n, &[nyquist]);
    let mut out = vec![Complex::new(0.0_f64, 0.0); 1];
    pf.execute(&input, &mut out);

    // MIRI soft-float accumulates more rounding error than hardware; use 2e-9 here.
    assert!(
        cdiff(out[0], full[nyquist]) < 2e-9,
        "Nyquist stability: got {:?}, expected {:?}",
        out[0],
        full[nyquist]
    );
}