yfft 0.1.0

Simple fast Fourier transform library optimized for small, power-of-two, single precision transforms and x86 systems 
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

pub use packed_simd::{f32x4, f32x8, i32x4, i32x8, u32x4, u32x8, u64x4};
#[cfg(target_arch = "x86")]
use std::arch::x86 as vendor;
#[cfg(target_arch = "x86_64")]
use std::arch::x86_64 as vendor;
use std::mem;

#[cfg(test)]
use num_complex::Complex;

#[allow(dead_code)]
#[inline]
pub fn f32x4_bitxor(lhs: f32x4, rhs: f32x4) -> f32x4 {
    let x2: i32x4 = unsafe { mem::transmute(lhs) };
    let y2: i32x4 = unsafe { mem::transmute(rhs) };
    let z = x2 ^ y2;
    unsafe { mem::transmute(z) }
}

#[allow(dead_code)]
pub fn f32x4_to_array(x: f32x4) -> [f32; 4] {
    [x.extract(0), x.extract(1), x.extract(2), x.extract(3)]
}

/// `neg_mask` must be `[0x80000000, 0x80000000, 0, 0]`
#[inline]
pub fn f32x4_complex_mul_rrii(x: f32x4, y: f32x4, neg_mask: f32x4) -> f32x4 {
    let y_iirr = shuffle!(y, y, [2, 3, 4, 5]);

    // (y1a.r * ta.r, y1b.r * tb.r, y1a.i * ta.i, y1b.i * tb.i)
    let t2 = x * y;

    // (y1a.r * ta.i, y1b.r * tb.i, y1a.i * ta.r, y1b.i * tb.r)
    let t3 = x * y_iirr;

    // (y1a.r * ta.r, y1b.r * tb.r, y1a.r * ta.i, y1b.r * tb.i)
    let t4 = shuffle!(t2, t3, [0, 1, 4, 5]);

    // (y1a.i * ta.i, y1b.i * tb.i, y1a.i * ta.r, y1b.i * tb.r)
    let t5 = shuffle!(t2, t3, [2, 3, 6, 7]);

    // (-y1a.i * ta.i, -y1b.i * tb.i, y1a.i * ta.r, y1b.i * tb.r)
    let t6 = f32x4_bitxor(t5, neg_mask);

    // (y3a.r, y3b.r, y3a.i, y3b.i) =
    // (y1a.r * ta.r - y1a.i * ta.i, y1b.r * tb.r - y1b.i * tb.i,
    //  y1a.r * ta.i + y1a.i * ta.r, y1b.r * tb.i + y1b.i * tb.r)
    t4 + t6
}

#[test]
fn test_f32x4_complex_mul_rrii() {
    let neg_mask = unsafe { mem::transmute(u32x4::new(0x80000000, 0x80000000, 0, 0)) };

    let c1: Complex<f32> = Complex::new(123f32, 456f32);
    let c2: Complex<f32> = Complex::new(789f32, 135f32);
    let c3: Complex<f32> = Complex::new(114f32, 514f32);
    let c4: Complex<f32> = Complex::new(987f32, 654f32);

    let d1 = c1 * c3;
    let d2 = c2 * c4;

    let x = f32x4::new(c1.re, c2.re, c1.im, c2.im);
    let y = f32x4::new(c3.re, c4.re, c3.im, c4.im);
    let z = f32x4_complex_mul_rrii(x, y, neg_mask);

    assert_eq!(f32x4_to_array(z), [d1.re, d2.re, d1.im, d2.im]);
}

#[cfg(all(target_feature = "sse3", not(target_feature = "fma")))]
#[allow(dead_code)]
pub fn sse3_fma_f32x4_fmaddsub(x: f32x4, y: f32x4, z: f32x4) -> f32x4 {
    let xy = unsafe { mem::transmute(x * y) };
    let z = unsafe { mem::transmute(z) };
    let w = unsafe { vendor::_mm_addsub_ps(xy, z) };
    unsafe { mem::transmute(w) }
}

#[cfg(all(target_feature = "sse3", target_feature = "fma"))]
#[allow(dead_code)]
pub fn sse3_fma_f32x4_fmaddsub(x: f32x4, y: f32x4, z: f32x4) -> f32x4 {
    let x = unsafe { mem::transmute(x) };
    let y = unsafe { mem::transmute(y) };
    let z = unsafe { mem::transmute(z) };
    let w = unsafe { vendor::_mm_fmaddsub_ps(x, y, z) };
    unsafe { mem::transmute(w) }
}

#[cfg(target_feature = "sse3")]
#[inline]
#[allow(dead_code)]
pub fn sse3_f32x4_complex_mul_riri(x: f32x4, y: f32x4) -> f32x4 {
    // (r1, i1, ...) * (r3, i3, ...)
    //   --> ((r1 * r3) - (i1 * i3), (r1 * i3) + (i1 * r3), ...)
    let x1: f32x4 = shuffle!(x, x, [0, 0, 2, 2]); // movsldup
    let x2: f32x4 = shuffle!(x, x, [1, 1, 3, 3]); // movshdup
    let y1: f32x4 = y;
    let y2: f32x4 = shuffle!(y, y, [1, 0, 3, 2]); // shufps
    let z = sse3_fma_f32x4_fmaddsub(x1, y1, x2 * y2); // vaddsubps/vfmaddsubXXXps
    return z;
}

#[cfg(target_feature = "sse3")]
#[test]
#[allow(dead_code)]
fn test_sse3_f32x4_complex_mul_riri() {
    let c1: Complex<f32> = Complex::new(123f32, 456f32);
    let c2: Complex<f32> = Complex::new(789f32, 135f32);
    let c3: Complex<f32> = Complex::new(114f32, 514f32);
    let c4: Complex<f32> = Complex::new(987f32, 654f32);

    let d1 = c1 * c3;
    let d2 = c2 * c4;

    let x = f32x4::new(c1.re, c1.im, c2.re, c2.im);
    let y = f32x4::new(c3.re, c3.im, c4.re, c4.im);
    let z = sse3_f32x4_complex_mul_riri(x, y);

    assert_eq!(f32x4_to_array(z), [d1.re, d1.im, d2.re, d2.im]);
}

#[cfg(all(target_feature = "avx", not(target_feature = "fma")))]
#[allow(dead_code)]
pub fn avx_fma_f32x8_fmadd(x: f32x8, y: f32x8, z: f32x8) -> f32x8 {
    x * y + z
}

#[cfg(all(target_feature = "avx", target_feature = "fma"))]
#[allow(dead_code)]
pub fn avx_fma_f32x8_fmadd(x: f32x8, y: f32x8, z: f32x8) -> f32x8 {
    let x: vendor::__m256 = unsafe { mem::transmute(x) };
    let y: vendor::__m256 = unsafe { mem::transmute(y) };
    let z: vendor::__m256 = unsafe { mem::transmute(z) };
    unsafe { mem::transmute(vendor::_mm256_fmadd_ps(x, y, z)) }
}

#[cfg(all(target_feature = "avx", not(target_feature = "fma")))]
#[allow(dead_code)]
pub fn avx_fma_f32x8_fmsub(x: f32x8, y: f32x8, z: f32x8) -> f32x8 {
    x * y - z
}

#[cfg(all(target_feature = "avx", target_feature = "fma"))]
#[allow(dead_code)]
pub fn avx_fma_f32x8_fmsub(x: f32x8, y: f32x8, z: f32x8) -> f32x8 {
    let x: vendor::__m256 = unsafe { mem::transmute(x) };
    let y: vendor::__m256 = unsafe { mem::transmute(y) };
    let z: vendor::__m256 = unsafe { mem::transmute(z) };
    unsafe { mem::transmute(vendor::_mm256_fmsub_ps(x, y, z)) }
}

#[cfg(all(target_feature = "avx", not(target_feature = "fma")))]
#[allow(dead_code)]
pub fn avx_fma_f32x8_fmaddsub(x: f32x8, y: f32x8, z: f32x8) -> f32x8 {
    let xy: vendor::__m256 = unsafe { mem::transmute(x * y) };
    let z: vendor::__m256 = unsafe { mem::transmute(z) };
    unsafe { mem::transmute(vendor::_mm256_addsub_ps(xy, z)) }
}

#[cfg(all(target_feature = "avx", target_feature = "fma"))]
#[allow(dead_code)]
pub fn avx_fma_f32x8_fmaddsub(x: f32x8, y: f32x8, z: f32x8) -> f32x8 {
    let x: vendor::__m256 = unsafe { mem::transmute(x) };
    let y: vendor::__m256 = unsafe { mem::transmute(y) };
    let z: vendor::__m256 = unsafe { mem::transmute(z) };
    unsafe { mem::transmute(vendor::_mm256_fmaddsub_ps(x, y, z)) }
}

#[cfg(target_feature = "avx")]
#[inline]
#[allow(dead_code)]
pub fn avx_f32x8_complex_mul_riri(x: f32x8, y: f32x8) -> f32x8 {
    // (r1, i1, ...) * (r3, i3, ...)
    //   --> ((r1 * r3) - (i1 * i3), (r1 * i3) + (i1 * r3), ...)
    let x1: f32x8 = shuffle!(x, x, [0, 0, 2, 2, 4, 4, 6, 6]); // vmovsldup
    let x2: f32x8 = shuffle!(x, x, [1, 1, 3, 3, 5, 5, 7, 7]); // vmovshdup
    let y1: f32x8 = y;
    let y2: f32x8 = shuffle!(y, y, [1, 0, 3, 2, 5, 4, 7, 6]); // vpermilps
    let z = avx_fma_f32x8_fmaddsub(x1, y1, x2 * y2); // vaddsubps/vfmaddsubXXXps
    return z;
}

#[cfg(target_feature = "avx")]
#[test]
#[allow(dead_code)]
fn test_avx_f32x8_complex_mul_riri() {
    let c1: Complex<f32> = Complex::new(123f32, 456f32);
    let c2: Complex<f32> = Complex::new(789f32, 135f32);
    let c3: Complex<f32> = Complex::new(114f32, 514f32);
    let c4: Complex<f32> = Complex::new(987f32, 654f32);
    let c5: Complex<f32> = Complex::new(12f32, 46f32);
    let c6: Complex<f32> = Complex::new(78f32, 15f32);
    let c7: Complex<f32> = Complex::new(11f32, 54f32);
    let c8: Complex<f32> = Complex::new(98f32, 64f32);

    let d1 = c1 * c3;
    let d2 = c2 * c4;
    let d3 = c5 * c7;
    let d4 = c6 * c8;

    let x = f32x8::new(c1.re, c1.im, c2.re, c2.im, c5.re, c5.im, c6.re, c6.im);
    let y = f32x8::new(c3.re, c3.im, c4.re, c4.im, c7.re, c7.im, c8.re, c8.im);
    let z = avx_f32x8_complex_mul_riri(x, y);

    assert_eq!(
        f32x8_to_array(z),
        [d1.re, d1.im, d2.re, d2.im, d3.re, d3.im, d4.re, d4.im]
    );
}

#[cfg(target_feature = "avx")]
#[allow(dead_code)]
#[inline]
pub fn avx_f32x8_bitxor(lhs: f32x8, rhs: f32x8) -> f32x8 {
    let x2: i32x8 = unsafe { mem::transmute(lhs) };
    let y2: i32x8 = unsafe { mem::transmute(rhs) };
    let z = x2 ^ y2;
    unsafe { mem::transmute(z) }
}

#[cfg(target_feature = "avx")]
#[allow(dead_code)]
pub fn f32x8_to_array(x: f32x8) -> [f32; 8] {
    [
        x.extract(0),
        x.extract(1),
        x.extract(2),
        x.extract(3),
        x.extract(4),
        x.extract(5),
        x.extract(6),
        x.extract(7),
    ]
}