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

use core::arch::aarch64::*;

use super::{
    super::ops::{complex_mul, complex_mul_i, load_neg_imag_mask},
    COS_2PI_5, COS_4PI_5, SIN_2PI_5, SIN_4PI_5,
};
use crate::fft::Complex32;

/// Performs a single radix-5 Stockham butterfly stage for stride=1 (out-of-place, NEON).
///
/// This is a specialized version for the stride=1 case (first stage) that uses sequential stores
/// instead of scattered stores. When stride==1, output indices are sequential: j = 5*i, j+1, ..., j+4.
/// This provides significant performance benefits through write-combining and better cache utilization.
#[target_feature(enable = "neon")]
pub(super) unsafe fn butterfly_radix5_stride1_neon(
    src: &[Complex32],
    dst: &mut [Complex32],
    stage_twiddles: &[Complex32],
) {
    let samples = src.len();
    let fifth_samples = samples / 5;
    let simd_iters = (fifth_samples >> 1) << 1;

    unsafe {
        let cos_2pi_5_vec = vdupq_n_f32(COS_2PI_5);
        let sin_2pi_5_vec = vdupq_n_f32(SIN_2PI_5);
        let cos_4pi_5_vec = vdupq_n_f32(COS_4PI_5);
        let sin_4pi_5_vec = vdupq_n_f32(SIN_4PI_5);

        let neg_imag = load_neg_imag_mask();

        for i in (0..simd_iters).step_by(2) {
            // Load z0 from first fifth.
            let z0_ptr = src.as_ptr().add(i) as *const f32;
            let z0 = vld1q_f32(z0_ptr);

            // Load z1, z2, z3, z4 from other fifths using contiguous loads.
            let z1_ptr = src.as_ptr().add(i + fifth_samples) as *const f32;
            let z1 = vld1q_f32(z1_ptr);

            let z2_ptr = src.as_ptr().add(i + fifth_samples * 2) as *const f32;
            let z2 = vld1q_f32(z2_ptr);

            let z3_ptr = src.as_ptr().add(i + fifth_samples * 3) as *const f32;
            let z3 = vld1q_f32(z3_ptr);

            let z4_ptr = src.as_ptr().add(i + fifth_samples * 4) as *const f32;
            let z4 = vld1q_f32(z4_ptr);

            // Stride=1 optimization: Skip twiddle loads and multiplications (identity twiddles).
            // t_k = (1+0i) * z_k = z_k
            let t1 = z1;
            let t2 = z2;
            let t3 = z3;
            let t4 = z4;

            // Radix-5 DFT.
            let sum_all = vaddq_f32(vaddq_f32(vaddq_f32(t1, t2), t3), t4);

            let a1 = vaddq_f32(t1, t4);
            let a2 = vaddq_f32(t2, t3);

            // b1 = i * (t1 - t4), b2 = i * (t2 - t3)
            let t1_sub_t4 = vsubq_f32(t1, t4);
            let t2_sub_t3 = vsubq_f32(t2, t3);
            let b1 = complex_mul_i(t1_sub_t4, neg_imag);
            let b2 = complex_mul_i(t2_sub_t3, neg_imag);

            // c1 = z0 + COS_2PI_5 * a1 + COS_4PI_5 * a2
            let c1 = vaddq_f32(
                z0,
                vaddq_f32(vmulq_f32(cos_2pi_5_vec, a1), vmulq_f32(cos_4pi_5_vec, a2)),
            );

            // c2 = z0 + COS_4PI_5 * a1 + COS_2PI_5 * a2
            let c2 = vaddq_f32(
                z0,
                vaddq_f32(vmulq_f32(cos_4pi_5_vec, a1), vmulq_f32(cos_2pi_5_vec, a2)),
            );

            // d1 = SIN_2PI_5 * b1 + SIN_4PI_5 * b2
            let d1 = vaddq_f32(vmulq_f32(sin_2pi_5_vec, b1), vmulq_f32(sin_4pi_5_vec, b2));

            // d2 = SIN_4PI_5 * b1 - SIN_2PI_5 * b2
            let d2 = vsubq_f32(vmulq_f32(sin_4pi_5_vec, b1), vmulq_f32(sin_2pi_5_vec, b2));

            // Final outputs.
            let out0 = vaddq_f32(z0, sum_all);
            let out1 = vaddq_f32(c1, d1);
            let out4 = vsubq_f32(c1, d1);
            let out2 = vaddq_f32(c2, d2);
            let out3 = vsubq_f32(c2, d2);

            let j = 5 * i;
            let dst_ptr = dst.as_mut_ptr() as *mut f32;

            let out0_0 = vget_low_f32(out0);
            let out1_0 = vget_low_f32(out1);
            let out2_0 = vget_low_f32(out2);
            let out3_0 = vget_low_f32(out3);
            let out4_0 = vget_low_f32(out4);

            vst1_f32(dst_ptr.add(j * 2), out0_0);
            vst1_f32(dst_ptr.add((j + 1) * 2), out1_0);
            vst1_f32(dst_ptr.add((j + 2) * 2), out2_0);
            vst1_f32(dst_ptr.add((j + 3) * 2), out3_0);
            vst1_f32(dst_ptr.add((j + 4) * 2), out4_0);

            let out0_1 = vget_high_f32(out0);
            let out1_1 = vget_high_f32(out1);
            let out2_1 = vget_high_f32(out2);
            let out3_1 = vget_high_f32(out3);
            let out4_1 = vget_high_f32(out4);

            let j1 = 5 * (i + 1);
            vst1_f32(dst_ptr.add(j1 * 2), out0_1);
            vst1_f32(dst_ptr.add((j1 + 1) * 2), out1_1);
            vst1_f32(dst_ptr.add((j1 + 2) * 2), out2_1);
            vst1_f32(dst_ptr.add((j1 + 3) * 2), out3_1);
            vst1_f32(dst_ptr.add((j1 + 4) * 2), out4_1);
        }
    }

    super::butterfly_radix5_scalar::<2>(src, dst, stage_twiddles, 1, simd_iters);
}

/// Performs a single radix-5 Stockham butterfly stage for p>1 (out-of-place, NEON).
///
/// This is the generic version for stride>1 cases. Uses direct SIMD stores,
/// accepting non-sequential stores as the shuffle overhead isn't justified.
#[target_feature(enable = "neon")]
pub(super) unsafe fn butterfly_radix5_generic_neon(
    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 fifth_samples = samples / 5;
    let simd_iters = (fifth_samples >> 1) << 1;

    unsafe {
        let cos_2pi_5_vec = vdupq_n_f32(COS_2PI_5);
        let sin_2pi_5_vec = vdupq_n_f32(SIN_2PI_5);
        let cos_4pi_5_vec = vdupq_n_f32(COS_4PI_5);
        let sin_4pi_5_vec = vdupq_n_f32(SIN_4PI_5);

        let neg_imag = load_neg_imag_mask();

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

            // Load z0 from first fifth.
            let z0_ptr = src.as_ptr().add(i) as *const f32;
            let z0 = vld1q_f32(z0_ptr);

            // Load z1, z2, z3, z4 from other fifths using contiguous loads.
            let z1_ptr = src.as_ptr().add(i + fifth_samples) as *const f32;
            let z1 = vld1q_f32(z1_ptr);

            let z2_ptr = src.as_ptr().add(i + fifth_samples * 2) as *const f32;
            let z2 = vld1q_f32(z2_ptr);

            let z3_ptr = src.as_ptr().add(i + fifth_samples * 3) as *const f32;
            let z3 = vld1q_f32(z3_ptr);

            let z4_ptr = src.as_ptr().add(i + fifth_samples * 4) as *const f32;
            let z4 = vld1q_f32(z4_ptr);

            // Load pre-packaged twiddles: [w1[i..i+2], w2[i..i+2], w3[i..i+2], w4[i..i+2]]
            let tw_ptr = stage_twiddles.as_ptr().add(i * 4) as *const f32;
            let w1 = vld1q_f32(tw_ptr); // w1[i], w1[i+1]
            let w2 = vld1q_f32(tw_ptr.add(4)); // w2[i], w2[i+1]
            let w3 = vld1q_f32(tw_ptr.add(8)); // w3[i], w3[i+1]
            let w4 = vld1q_f32(tw_ptr.add(12)); // w4[i], w4[i+1]

            // Complex multiply using optimized helper function.
            let t1 = complex_mul(z1, w1);
            let t2 = complex_mul(z2, w2);
            let t3 = complex_mul(z3, w3);
            let t4 = complex_mul(z4, w4);

            // Radix-5 DFT.
            let sum_all = vaddq_f32(vaddq_f32(vaddq_f32(t1, t2), t3), t4);

            let a1 = vaddq_f32(t1, t4);
            let a2 = vaddq_f32(t2, t3);

            // b1 = i * (t1 - t4), b2 = i * (t2 - t3)
            let t1_sub_t4 = vsubq_f32(t1, t4);
            let t2_sub_t3 = vsubq_f32(t2, t3);
            let b1 = complex_mul_i(t1_sub_t4, neg_imag);
            let b2 = complex_mul_i(t2_sub_t3, neg_imag);

            // c1 = z0 + COS_2PI_5 * a1 + COS_4PI_5 * a2
            let c1 = vaddq_f32(
                z0,
                vaddq_f32(vmulq_f32(cos_2pi_5_vec, a1), vmulq_f32(cos_4pi_5_vec, a2)),
            );

            // c2 = z0 + COS_4PI_5 * a1 + COS_2PI_5 * a2
            let c2 = vaddq_f32(
                z0,
                vaddq_f32(vmulq_f32(cos_4pi_5_vec, a1), vmulq_f32(cos_2pi_5_vec, a2)),
            );

            // d1 = SIN_2PI_5 * b1 + SIN_4PI_5 * b2
            let d1 = vaddq_f32(vmulq_f32(sin_2pi_5_vec, b1), vmulq_f32(sin_4pi_5_vec, b2));

            // d2 = SIN_4PI_5 * b1 - SIN_2PI_5 * b2
            let d2 = vsubq_f32(vmulq_f32(sin_4pi_5_vec, b1), vmulq_f32(sin_2pi_5_vec, b2));

            // Final outputs.
            let out0 = vaddq_f32(z0, sum_all);
            let out1 = vaddq_f32(c1, d1);
            let out4 = vsubq_f32(c1, d1);
            let out2 = vaddq_f32(c2, d2);
            let out3 = vsubq_f32(c2, d2);

            // Calculate output indices: j = 5*i - 4*k
            let j0 = 5 * i - 4 * k0;
            let j1 = 5 * (i + 1) - 4 * k1;

            // Extract and store each complex number separately using direct f32 stores.
            let dst_ptr = dst.as_mut_ptr() as *mut f32;

            let out0_0 = vget_low_f32(out0);
            let out1_0 = vget_low_f32(out1);
            let out2_0 = vget_low_f32(out2);
            let out3_0 = vget_low_f32(out3);
            let out4_0 = vget_low_f32(out4);

            vst1_f32(dst_ptr.add(j0 << 1), out0_0);
            vst1_f32(dst_ptr.add((j0 + stride) << 1), out1_0);
            vst1_f32(dst_ptr.add((j0 + stride * 2) << 1), out2_0);
            vst1_f32(dst_ptr.add((j0 + stride * 3) << 1), out3_0);
            vst1_f32(dst_ptr.add((j0 + stride * 4) << 1), out4_0);

            let out0_1 = vget_high_f32(out0);
            let out1_1 = vget_high_f32(out1);
            let out2_1 = vget_high_f32(out2);
            let out3_1 = vget_high_f32(out3);
            let out4_1 = vget_high_f32(out4);

            vst1_f32(dst_ptr.add(j1 << 1), out0_1);
            vst1_f32(dst_ptr.add((j1 + stride) << 1), out1_1);
            vst1_f32(dst_ptr.add((j1 + stride * 2) << 1), out2_1);
            vst1_f32(dst_ptr.add((j1 + stride * 3) << 1), out3_1);
            vst1_f32(dst_ptr.add((j1 + stride * 4) << 1), out4_1);
        }
    }

    super::butterfly_radix5_scalar::<2>(src, dst, stage_twiddles, stride, simd_iters);
}