resampler 0.5.1

A small audio resampling library
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
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378

use core::arch::x86_64::*;

use crate::fft::{
    Complex32,
    butterflies::ops::{
        complex_mul_avx, complex_mul_i_avx, load_neg_imag_mask_avx, load_scale_avx, v8x_avx,
        w8x_avx,
    },
};

/// Performs a single radix-8 Stockham butterfly stage for stride=1 (out-of-place, AVX+FMA).
///
/// This is a specialized version for the stride=1 case (first stage) that uses chunked streaming
/// to reduce register pressure. Processes outputs in two groups of 4 to minimize spilling.
/// For stride=1, output indices are sequential: j=8*i.
#[target_feature(enable = "avx,fma")]
pub(super) unsafe fn butterfly_radix8_stride1_avx_fma(
    src: &[Complex32],
    dst: &mut [Complex32],
    stage_twiddles: &[Complex32],
) {
    let samples = src.len();
    let eighth_samples = samples >> 3;
    let simd_iters = (eighth_samples >> 2) << 2;

    unsafe {
        let neg_imag_mask = load_neg_imag_mask_avx();
        let scale = load_scale_avx();

        for i in (0..simd_iters).step_by(4) {
            // Load 8 input values from the 8 "eighths" of the input array.
            let z0_ptr = src.as_ptr().add(i) as *const f32;
            let z0 = _mm256_loadu_ps(z0_ptr);

            let z1_ptr = src.as_ptr().add(i + eighth_samples) as *const f32;
            let z1 = _mm256_loadu_ps(z1_ptr);

            let z2_ptr = src.as_ptr().add(i + eighth_samples * 2) as *const f32;
            let z2 = _mm256_loadu_ps(z2_ptr);

            let z3_ptr = src.as_ptr().add(i + eighth_samples * 3) as *const f32;
            let z3 = _mm256_loadu_ps(z3_ptr);

            let z4_ptr = src.as_ptr().add(i + eighth_samples * 4) as *const f32;
            let z4 = _mm256_loadu_ps(z4_ptr);

            let z5_ptr = src.as_ptr().add(i + eighth_samples * 5) as *const f32;
            let z5 = _mm256_loadu_ps(z5_ptr);

            let z6_ptr = src.as_ptr().add(i + eighth_samples * 6) as *const f32;
            let z6 = _mm256_loadu_ps(z6_ptr);

            let z7_ptr = src.as_ptr().add(i + eighth_samples * 7) as *const f32;
            let z7 = _mm256_loadu_ps(z7_ptr);

            // Stride=1 optimization: all twiddles are identity (1+0i), skip multiplication.
            let t1 = z1;
            let t2 = z2;
            let t3 = z3;
            let t4 = z4;
            let t5 = z5;
            let t6 = z6;
            let t7 = z7;

            // Split-radix decomposition: compute even radix-4 on (z0, t2, t4, t6)
            let even_a0 = _mm256_add_ps(z0, t4);
            let even_a1 = _mm256_sub_ps(z0, t4);
            let even_a2 = _mm256_add_ps(t2, t6);
            // even_a3 = i * (t2 - t6): (real, imag) -> (imag, -real)
            let t2_sub_t6 = _mm256_sub_ps(t2, t6);
            let even_a3 = complex_mul_i_avx(t2_sub_t6, neg_imag_mask);

            let x_even_0 = _mm256_add_ps(even_a0, even_a2);
            let x_even_2 = _mm256_sub_ps(even_a0, even_a2);
            let x_even_1 = _mm256_add_ps(even_a1, even_a3);
            let x_even_3 = _mm256_sub_ps(even_a1, even_a3);

            // Compute odd radix-4 on (t1, t3, t5, t7).
            let odd_a0 = _mm256_add_ps(t1, t5);
            let odd_a1 = _mm256_sub_ps(t1, t5);
            let odd_a2 = _mm256_add_ps(t3, t7);
            let t3_sub_t7 = _mm256_sub_ps(t3, t7);
            let odd_a3 = complex_mul_i_avx(t3_sub_t7, neg_imag_mask);

            let x_odd_0 = _mm256_add_ps(odd_a0, odd_a2);
            let x_odd_2 = _mm256_sub_ps(odd_a0, odd_a2);
            let x_odd_1 = _mm256_add_ps(odd_a1, odd_a3);
            let x_odd_3 = _mm256_sub_ps(odd_a1, odd_a3);

            // Apply W_8 twiddles and combine even/odd results.
            // W_8^0 = 1, W_8^1 = (1-i)/√2, W_8^2 = -i, W_8^3 = (-1-i)/√2

            // out0 = x_even_0 + x_odd_0
            let out0 = _mm256_add_ps(x_even_0, x_odd_0);
            // out4 = x_even_0 - x_odd_0
            let out4 = _mm256_sub_ps(x_even_0, x_odd_0);

            // out1 = x_even_1 + W_8^1 * x_odd_1 (W_8^1 = (1-i)/√2)
            let w8_1_x_odd_1 = w8x_avx(x_odd_1, neg_imag_mask, scale);
            let out1 = _mm256_add_ps(x_even_1, w8_1_x_odd_1);
            // out5 = x_even_1 - W_8^1 * x_odd_1
            let out5 = _mm256_sub_ps(x_even_1, w8_1_x_odd_1);

            // out2 = x_even_2 + W_8^2 * x_odd_2
            // W_8^2 = -i = (0, -1), so multiply by -i: (a+bi)*(-i) = (b, -a)
            let w8_2_x_odd_2 = complex_mul_i_avx(x_odd_2, neg_imag_mask);
            let out2 = _mm256_add_ps(x_even_2, w8_2_x_odd_2);
            // out6 = x_even_2 - W_8^2 * x_odd_2
            let out6 = _mm256_sub_ps(x_even_2, w8_2_x_odd_2);

            // out3 = x_even_3 + W_8^3 * x_odd_3 (W_8^3 = (-1-i)/√2)
            let w8_3_x_odd_3 = v8x_avx(x_odd_3, neg_imag_mask, scale);
            let out3 = _mm256_add_ps(x_even_3, w8_3_x_odd_3);
            // out7 = x_even_3 - W_8^3 * x_odd_3
            let out7 = _mm256_sub_ps(x_even_3, w8_3_x_odd_3);

            // Chunked interleaving: process out0-3 first, then out4-7 to reduce register pressure.
            let j = 8 * i;
            let dst_ptr = dst.as_mut_ptr().add(j) as *mut f32;

            // First chunk: Interleave out0-3 using 256-bit operations.
            // Use 256-bit unpacks directly instead of costly 128-bit extract+insert operations.
            let out0_pd = _mm256_castps_pd(out0);
            let out1_pd = _mm256_castps_pd(out1);
            let out2_pd = _mm256_castps_pd(out2);
            let out3_pd = _mm256_castps_pd(out3);

            // 256-bit unpack operations.
            // unpacklo_pd(a,b) = [a[0],b[0],a[2],b[2]] on 4x 64-bit elements
            // unpackhi_pd(a,b) = [a[1],b[1],a[3],b[3]] on 4x 64-bit elements
            let temp_01_lo = _mm256_unpacklo_pd(out0_pd, out1_pd);
            let temp_01_hi = _mm256_unpackhi_pd(out0_pd, out1_pd);
            let temp_23_lo = _mm256_unpacklo_pd(out2_pd, out3_pd);
            let temp_23_hi = _mm256_unpackhi_pd(out2_pd, out3_pd);

            // Use permute2f128_pd to arrange 128-bit lanes to get sequential layout.
            // permute2f128(a,b,0x20) = [lo128(a), lo128(b)]
            // permute2f128(a,b,0x31) = [hi128(a), hi128(b)]
            let result0 = _mm256_castpd_ps(_mm256_permute2f128_pd(temp_01_lo, temp_23_lo, 0x20));
            let result2 = _mm256_castpd_ps(_mm256_permute2f128_pd(temp_01_hi, temp_23_hi, 0x20));
            let result4 = _mm256_castpd_ps(_mm256_permute2f128_pd(temp_01_lo, temp_23_lo, 0x31));
            let result6 = _mm256_castpd_ps(_mm256_permute2f128_pd(temp_01_hi, temp_23_hi, 0x31));

            // Store first chunk.
            _mm256_storeu_ps(dst_ptr, result0);
            _mm256_storeu_ps(dst_ptr.add(16), result2);
            _mm256_storeu_ps(dst_ptr.add(32), result4);
            _mm256_storeu_ps(dst_ptr.add(48), result6);

            // Second chunk: Interleave out4-7 using 256-bit operations.
            let out4_pd = _mm256_castps_pd(out4);
            let out5_pd = _mm256_castps_pd(out5);
            let out6_pd = _mm256_castps_pd(out6);
            let out7_pd = _mm256_castps_pd(out7);

            // 256-bit unpack operations.
            let temp_45_lo = _mm256_unpacklo_pd(out4_pd, out5_pd);
            let temp_45_hi = _mm256_unpackhi_pd(out4_pd, out5_pd);
            let temp_67_lo = _mm256_unpacklo_pd(out6_pd, out7_pd);
            let temp_67_hi = _mm256_unpackhi_pd(out6_pd, out7_pd);

            // Use permute2f128_pd to arrange 128-bit lanes.
            let result1 = _mm256_castpd_ps(_mm256_permute2f128_pd(temp_45_lo, temp_67_lo, 0x20));
            let result3 = _mm256_castpd_ps(_mm256_permute2f128_pd(temp_45_hi, temp_67_hi, 0x20));
            let result5 = _mm256_castpd_ps(_mm256_permute2f128_pd(temp_45_lo, temp_67_lo, 0x31));
            let result7 = _mm256_castpd_ps(_mm256_permute2f128_pd(temp_45_hi, temp_67_hi, 0x31));

            // Store second chunk.
            _mm256_storeu_ps(dst_ptr.add(8), result1);
            _mm256_storeu_ps(dst_ptr.add(24), result3);
            _mm256_storeu_ps(dst_ptr.add(40), result5);
            _mm256_storeu_ps(dst_ptr.add(56), result7);
        }
    }

    super::butterfly_radix8_scalar::<4>(src, dst, stage_twiddles, 1, simd_iters);
}

/// Performs a single radix-8 Stockham butterfly stage for p>1 (out-of-place, AVX+FMA).
///
/// Generic version for p>1 cases. Uses chunked streaming to reduce register pressure
/// while maintaining scattered stores for non-unit strides.
#[target_feature(enable = "avx,fma")]
pub(super) unsafe fn butterfly_radix8_generic_avx_fma(
    src: &[Complex32],
    dst: &mut [Complex32],
    stage_twiddles: &[Complex32],
    stride: usize,
) {
    // We convince the compiler here that stride can't be 0 to optimize better.
    if stride == 0 {
        return;
    }

    let samples = src.len();
    let eighth_samples = samples >> 3;
    let simd_iters = (eighth_samples >> 2) << 2;

    unsafe {
        let neg_imag_mask = load_neg_imag_mask_avx();
        let scale = load_scale_avx();

        for i in (0..simd_iters).step_by(4) {
            let k = i % stride;
            let k0 = k;
            let k1 = k + 1 - ((k + 1 >= stride) as usize) * stride;
            let k2 = k + 2 - ((k + 2 >= stride) as usize) * stride;
            let k3 = k + 3 - ((k + 3 >= stride) as usize) * stride;

            // Load 8 input values from the 8 "eighths" of the input array.
            let z0_ptr = src.as_ptr().add(i) as *const f32;
            let z0 = _mm256_loadu_ps(z0_ptr);

            let z1_ptr = src.as_ptr().add(i + eighth_samples) as *const f32;
            let z1 = _mm256_loadu_ps(z1_ptr);

            let z2_ptr = src.as_ptr().add(i + eighth_samples * 2) as *const f32;
            let z2 = _mm256_loadu_ps(z2_ptr);

            let z3_ptr = src.as_ptr().add(i + eighth_samples * 3) as *const f32;
            let z3 = _mm256_loadu_ps(z3_ptr);

            let z4_ptr = src.as_ptr().add(i + eighth_samples * 4) as *const f32;
            let z4 = _mm256_loadu_ps(z4_ptr);

            let z5_ptr = src.as_ptr().add(i + eighth_samples * 5) as *const f32;
            let z5 = _mm256_loadu_ps(z5_ptr);

            let z6_ptr = src.as_ptr().add(i + eighth_samples * 6) as *const f32;
            let z6 = _mm256_loadu_ps(z6_ptr);

            let z7_ptr = src.as_ptr().add(i + eighth_samples * 7) as *const f32;
            let z7 = _mm256_loadu_ps(z7_ptr);

            // Load prepackaged twiddles in packed format.
            let tw_ptr = stage_twiddles.as_ptr().add(i * 7) as *const f32;
            let w1 = _mm256_loadu_ps(tw_ptr); // w1[i..i+4]
            let w2 = _mm256_loadu_ps(tw_ptr.add(8)); // w2[i..i+4]
            let w3 = _mm256_loadu_ps(tw_ptr.add(16)); // w3[i..i+4]
            let w4 = _mm256_loadu_ps(tw_ptr.add(24)); // w4[i..i+4]
            let w5 = _mm256_loadu_ps(tw_ptr.add(32)); // w5[i..i+4]
            let w6 = _mm256_loadu_ps(tw_ptr.add(40)); // w6[i..i+4]
            let w7 = _mm256_loadu_ps(tw_ptr.add(48)); // w7[i..i+4]

            // Apply twiddle factors
            let t1 = complex_mul_avx(w1, z1);
            let t2 = complex_mul_avx(w2, z2);
            let t3 = complex_mul_avx(w3, z3);
            let t4 = complex_mul_avx(w4, z4);
            let t5 = complex_mul_avx(w5, z5);
            let t6 = complex_mul_avx(w6, z6);
            let t7 = complex_mul_avx(w7, z7);

            // Split-radix decomposition: compute even radix-4 on (z0, t2, t4, t6)
            let even_a0 = _mm256_add_ps(z0, t4);
            let even_a1 = _mm256_sub_ps(z0, t4);
            let even_a2 = _mm256_add_ps(t2, t6);
            let t2_sub_t6 = _mm256_sub_ps(t2, t6);
            let even_a3 = complex_mul_i_avx(t2_sub_t6, neg_imag_mask);

            let x_even_0 = _mm256_add_ps(even_a0, even_a2);
            let x_even_2 = _mm256_sub_ps(even_a0, even_a2);
            let x_even_1 = _mm256_add_ps(even_a1, even_a3);
            let x_even_3 = _mm256_sub_ps(even_a1, even_a3);

            // Compute odd radix-4 on (t1, t3, t5, t7)
            let odd_a0 = _mm256_add_ps(t1, t5);
            let odd_a1 = _mm256_sub_ps(t1, t5);
            let odd_a2 = _mm256_add_ps(t3, t7);
            let t3_sub_t7 = _mm256_sub_ps(t3, t7);
            let odd_a3 = complex_mul_i_avx(t3_sub_t7, neg_imag_mask);

            let x_odd_0 = _mm256_add_ps(odd_a0, odd_a2);
            let x_odd_2 = _mm256_sub_ps(odd_a0, odd_a2);
            let x_odd_1 = _mm256_add_ps(odd_a1, odd_a3);
            let x_odd_3 = _mm256_sub_ps(odd_a1, odd_a3);

            // Apply W_8 twiddles and combine
            let out0 = _mm256_add_ps(x_even_0, x_odd_0);
            let out4 = _mm256_sub_ps(x_even_0, x_odd_0);

            // W_8^1 = (1-i)/√2
            let w8_1_x_odd_1 = w8x_avx(x_odd_1, neg_imag_mask, scale);
            let out1 = _mm256_add_ps(x_even_1, w8_1_x_odd_1);
            let out5 = _mm256_sub_ps(x_even_1, w8_1_x_odd_1);

            // W_8^2 = -i: (a+bi) -> (b, -a)
            let w8_2_x_odd_2 = complex_mul_i_avx(x_odd_2, neg_imag_mask);
            let out2 = _mm256_add_ps(x_even_2, w8_2_x_odd_2);
            let out6 = _mm256_sub_ps(x_even_2, w8_2_x_odd_2);

            // W_8^3 = (-1-i)/√2
            let w8_3_x_odd_3 = v8x_avx(x_odd_3, neg_imag_mask, scale);
            let out3 = _mm256_add_ps(x_even_3, w8_3_x_odd_3);
            let out7 = _mm256_sub_ps(x_even_3, w8_3_x_odd_3);

            // Calculate output indices: j = 8*i - 7*k
            let j0 = 8 * i - 7 * k0;
            let j1 = 8 * (i + 1) - 7 * k1;
            let j2 = 8 * (i + 2) - 7 * k2;
            let j3 = 8 * (i + 3) - 7 * k3;

            let dst_ptr = dst.as_mut_ptr() as *mut f64;

            // First chunk: out0-3
            let out0_pd = _mm256_castps_pd(out0);
            let out1_pd = _mm256_castps_pd(out1);
            let out2_pd = _mm256_castps_pd(out2);
            let out3_pd = _mm256_castps_pd(out3);

            let out0_lo = _mm256_castpd256_pd128(out0_pd);
            let out0_hi = _mm256_extractf128_pd(out0_pd, 1);
            let out1_lo = _mm256_castpd256_pd128(out1_pd);
            let out1_hi = _mm256_extractf128_pd(out1_pd, 1);
            let out2_lo = _mm256_castpd256_pd128(out2_pd);
            let out2_hi = _mm256_extractf128_pd(out2_pd, 1);
            let out3_lo = _mm256_castpd256_pd128(out3_pd);
            let out3_hi = _mm256_extractf128_pd(out3_pd, 1);

            _mm_storel_pd(dst_ptr.add(j0), out0_lo);
            _mm_storel_pd(dst_ptr.add(j0 + stride), out1_lo);
            _mm_storel_pd(dst_ptr.add(j0 + stride * 2), out2_lo);
            _mm_storel_pd(dst_ptr.add(j0 + stride * 3), out3_lo);

            _mm_storeh_pd(dst_ptr.add(j1), out0_lo);
            _mm_storeh_pd(dst_ptr.add(j1 + stride), out1_lo);
            _mm_storeh_pd(dst_ptr.add(j1 + stride * 2), out2_lo);
            _mm_storeh_pd(dst_ptr.add(j1 + stride * 3), out3_lo);

            _mm_storel_pd(dst_ptr.add(j2), out0_hi);
            _mm_storel_pd(dst_ptr.add(j2 + stride), out1_hi);
            _mm_storel_pd(dst_ptr.add(j2 + stride * 2), out2_hi);
            _mm_storel_pd(dst_ptr.add(j2 + stride * 3), out3_hi);

            _mm_storeh_pd(dst_ptr.add(j3), out0_hi);
            _mm_storeh_pd(dst_ptr.add(j3 + stride), out1_hi);
            _mm_storeh_pd(dst_ptr.add(j3 + stride * 2), out2_hi);
            _mm_storeh_pd(dst_ptr.add(j3 + stride * 3), out3_hi);

            // Second chunk: out4-7
            let out4_pd = _mm256_castps_pd(out4);
            let out5_pd = _mm256_castps_pd(out5);
            let out6_pd = _mm256_castps_pd(out6);
            let out7_pd = _mm256_castps_pd(out7);

            let out4_lo = _mm256_castpd256_pd128(out4_pd);
            let out4_hi = _mm256_extractf128_pd(out4_pd, 1);
            let out5_lo = _mm256_castpd256_pd128(out5_pd);
            let out5_hi = _mm256_extractf128_pd(out5_pd, 1);
            let out6_lo = _mm256_castpd256_pd128(out6_pd);
            let out6_hi = _mm256_extractf128_pd(out6_pd, 1);
            let out7_lo = _mm256_castpd256_pd128(out7_pd);
            let out7_hi = _mm256_extractf128_pd(out7_pd, 1);

            _mm_storel_pd(dst_ptr.add(j0 + stride * 4), out4_lo);
            _mm_storel_pd(dst_ptr.add(j0 + stride * 5), out5_lo);
            _mm_storel_pd(dst_ptr.add(j0 + stride * 6), out6_lo);
            _mm_storel_pd(dst_ptr.add(j0 + stride * 7), out7_lo);

            _mm_storeh_pd(dst_ptr.add(j1 + stride * 4), out4_lo);
            _mm_storeh_pd(dst_ptr.add(j1 + stride * 5), out5_lo);
            _mm_storeh_pd(dst_ptr.add(j1 + stride * 6), out6_lo);
            _mm_storeh_pd(dst_ptr.add(j1 + stride * 7), out7_lo);

            _mm_storel_pd(dst_ptr.add(j2 + stride * 4), out4_hi);
            _mm_storel_pd(dst_ptr.add(j2 + stride * 5), out5_hi);
            _mm_storel_pd(dst_ptr.add(j2 + stride * 6), out6_hi);
            _mm_storel_pd(dst_ptr.add(j2 + stride * 7), out7_hi);

            _mm_storeh_pd(dst_ptr.add(j3 + stride * 4), out4_hi);
            _mm_storeh_pd(dst_ptr.add(j3 + stride * 5), out5_hi);
            _mm_storeh_pd(dst_ptr.add(j3 + stride * 6), out6_hi);
            _mm_storeh_pd(dst_ptr.add(j3 + stride * 7), out7_hi);
        }
    }

    super::butterfly_radix8_scalar::<4>(src, dst, stage_twiddles, stride, simd_iters);
}