numr 0.5.2

High-performance numerical computing with multi-backend GPU acceleration (CPU/CUDA/WebGPU)
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

//! AVX2 activation kernels
//!
//! Vectorized implementations using 256-bit registers.

#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;

use super::super::math::avx2::{exp_f32, exp_f64, tanh_f32, tanh_f64};
use super::{
    elu_scalar_f32, elu_scalar_f64, gelu_scalar_f32, gelu_scalar_f64, leaky_relu_scalar_f32,
    leaky_relu_scalar_f64, sigmoid_scalar_f32, sigmoid_scalar_f64, silu_scalar_f32,
    silu_scalar_f64,
};

const F32_LANES: usize = 8;
const F64_LANES: usize = 4;

/// AVX2 sigmoid for f32
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn sigmoid_f32(a: *const f32, out: *mut f32, len: usize) {
    let chunks = len / F32_LANES;
    let one = _mm256_set1_ps(1.0);

    for c in 0..chunks {
        let offset = c * F32_LANES;
        let x = _mm256_loadu_ps(a.add(offset));
        let neg_x = _mm256_sub_ps(_mm256_setzero_ps(), x);
        let exp_neg_x = exp_f32(neg_x);
        let result = _mm256_div_ps(one, _mm256_add_ps(one, exp_neg_x));
        _mm256_storeu_ps(out.add(offset), result);
    }

    let processed = chunks * F32_LANES;
    if processed < len {
        sigmoid_scalar_f32(a.add(processed), out.add(processed), len - processed);
    }
}

/// AVX2 sigmoid for f64
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn sigmoid_f64(a: *const f64, out: *mut f64, len: usize) {
    let chunks = len / F64_LANES;
    let one = _mm256_set1_pd(1.0);

    for c in 0..chunks {
        let offset = c * F64_LANES;
        let x = _mm256_loadu_pd(a.add(offset));
        let neg_x = _mm256_sub_pd(_mm256_setzero_pd(), x);
        let exp_neg_x = exp_f64(neg_x);
        let result = _mm256_div_pd(one, _mm256_add_pd(one, exp_neg_x));
        _mm256_storeu_pd(out.add(offset), result);
    }

    let processed = chunks * F64_LANES;
    if processed < len {
        sigmoid_scalar_f64(a.add(processed), out.add(processed), len - processed);
    }
}

/// AVX2 SiLU for f32
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn silu_f32(a: *const f32, out: *mut f32, len: usize) {
    let chunks = len / F32_LANES;
    let one = _mm256_set1_ps(1.0);

    for c in 0..chunks {
        let offset = c * F32_LANES;
        let x = _mm256_loadu_ps(a.add(offset));
        let neg_x = _mm256_sub_ps(_mm256_setzero_ps(), x);
        let exp_neg_x = exp_f32(neg_x);
        let result = _mm256_div_ps(x, _mm256_add_ps(one, exp_neg_x));
        _mm256_storeu_ps(out.add(offset), result);
    }

    let processed = chunks * F32_LANES;
    if processed < len {
        silu_scalar_f32(a.add(processed), out.add(processed), len - processed);
    }
}

/// AVX2 SiLU for f64
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn silu_f64(a: *const f64, out: *mut f64, len: usize) {
    let chunks = len / F64_LANES;
    let one = _mm256_set1_pd(1.0);

    for c in 0..chunks {
        let offset = c * F64_LANES;
        let x = _mm256_loadu_pd(a.add(offset));
        let neg_x = _mm256_sub_pd(_mm256_setzero_pd(), x);
        let exp_neg_x = exp_f64(neg_x);
        let result = _mm256_div_pd(x, _mm256_add_pd(one, exp_neg_x));
        _mm256_storeu_pd(out.add(offset), result);
    }

    let processed = chunks * F64_LANES;
    if processed < len {
        silu_scalar_f64(a.add(processed), out.add(processed), len - processed);
    }
}

/// AVX2 GELU for f32
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn gelu_f32(a: *const f32, out: *mut f32, len: usize) {
    let chunks = len / F32_LANES;
    let half = _mm256_set1_ps(0.5);
    let one = _mm256_set1_ps(1.0);
    let sqrt_2_over_pi = _mm256_set1_ps(0.7978845608);
    let tanh_coef = _mm256_set1_ps(0.044715);

    for c in 0..chunks {
        let offset = c * F32_LANES;
        let x = _mm256_loadu_ps(a.add(offset));

        let x_cubed = _mm256_mul_ps(_mm256_mul_ps(x, x), x);
        let inner = _mm256_mul_ps(sqrt_2_over_pi, _mm256_fmadd_ps(tanh_coef, x_cubed, x));

        let tanh_inner = tanh_f32(inner);
        let result = _mm256_mul_ps(half, _mm256_mul_ps(x, _mm256_add_ps(one, tanh_inner)));

        _mm256_storeu_ps(out.add(offset), result);
    }

    let processed = chunks * F32_LANES;
    if processed < len {
        gelu_scalar_f32(a.add(processed), out.add(processed), len - processed);
    }
}

/// AVX2 GELU for f64
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn gelu_f64(a: *const f64, out: *mut f64, len: usize) {
    let chunks = len / F64_LANES;
    let half = _mm256_set1_pd(0.5);
    let one = _mm256_set1_pd(1.0);
    let sqrt_2_over_pi = _mm256_set1_pd(0.7978845608028654);
    let tanh_coef = _mm256_set1_pd(0.044715);

    for c in 0..chunks {
        let offset = c * F64_LANES;
        let x = _mm256_loadu_pd(a.add(offset));

        let x_cubed = _mm256_mul_pd(_mm256_mul_pd(x, x), x);
        let inner = _mm256_mul_pd(sqrt_2_over_pi, _mm256_fmadd_pd(tanh_coef, x_cubed, x));

        let tanh_inner = tanh_f64(inner);
        let result = _mm256_mul_pd(half, _mm256_mul_pd(x, _mm256_add_pd(one, tanh_inner)));

        _mm256_storeu_pd(out.add(offset), result);
    }

    let processed = chunks * F64_LANES;
    if processed < len {
        gelu_scalar_f64(a.add(processed), out.add(processed), len - processed);
    }
}

/// AVX2 Leaky ReLU for f32
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn leaky_relu_f32(a: *const f32, out: *mut f32, len: usize, negative_slope: f32) {
    let chunks = len / F32_LANES;
    let v_slope = _mm256_set1_ps(negative_slope);
    let zero = _mm256_setzero_ps();

    for c in 0..chunks {
        let offset = c * F32_LANES;
        let x = _mm256_loadu_ps(a.add(offset));
        let mask = _mm256_cmp_ps(x, zero, _CMP_GT_OQ);
        let scaled = _mm256_mul_ps(v_slope, x);
        let result = _mm256_blendv_ps(scaled, x, mask);
        _mm256_storeu_ps(out.add(offset), result);
    }

    let processed = chunks * F32_LANES;
    if processed < len {
        leaky_relu_scalar_f32(
            a.add(processed),
            out.add(processed),
            len - processed,
            negative_slope,
        );
    }
}

/// AVX2 Leaky ReLU for f64
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn leaky_relu_f64(a: *const f64, out: *mut f64, len: usize, negative_slope: f64) {
    let chunks = len / F64_LANES;
    let v_slope = _mm256_set1_pd(negative_slope);
    let zero = _mm256_setzero_pd();

    for c in 0..chunks {
        let offset = c * F64_LANES;
        let x = _mm256_loadu_pd(a.add(offset));
        let mask = _mm256_cmp_pd(x, zero, _CMP_GT_OQ);
        let scaled = _mm256_mul_pd(v_slope, x);
        let result = _mm256_blendv_pd(scaled, x, mask);
        _mm256_storeu_pd(out.add(offset), result);
    }

    let processed = chunks * F64_LANES;
    if processed < len {
        leaky_relu_scalar_f64(
            a.add(processed),
            out.add(processed),
            len - processed,
            negative_slope,
        );
    }
}

/// AVX2 ELU for f32
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn elu_f32(a: *const f32, out: *mut f32, len: usize, alpha: f32) {
    let chunks = len / F32_LANES;
    let v_alpha = _mm256_set1_ps(alpha);
    let one = _mm256_set1_ps(1.0);
    let zero = _mm256_setzero_ps();

    for c in 0..chunks {
        let offset = c * F32_LANES;
        let x = _mm256_loadu_ps(a.add(offset));
        let mask = _mm256_cmp_ps(x, zero, _CMP_GT_OQ);
        let exp_x = exp_f32(x);
        let neg_result = _mm256_mul_ps(v_alpha, _mm256_sub_ps(exp_x, one));
        let result = _mm256_blendv_ps(neg_result, x, mask);
        _mm256_storeu_ps(out.add(offset), result);
    }

    let processed = chunks * F32_LANES;
    if processed < len {
        elu_scalar_f32(a.add(processed), out.add(processed), len - processed, alpha);
    }
}

/// AVX2 ELU for f64
#[target_feature(enable = "avx2", enable = "fma")]
pub unsafe fn elu_f64(a: *const f64, out: *mut f64, len: usize, alpha: f64) {
    let chunks = len / F64_LANES;
    let v_alpha = _mm256_set1_pd(alpha);
    let one = _mm256_set1_pd(1.0);
    let zero = _mm256_setzero_pd();

    for c in 0..chunks {
        let offset = c * F64_LANES;
        let x = _mm256_loadu_pd(a.add(offset));
        let mask = _mm256_cmp_pd(x, zero, _CMP_GT_OQ);
        let exp_x = exp_f64(x);
        let neg_result = _mm256_mul_pd(v_alpha, _mm256_sub_pd(exp_x, one));
        let result = _mm256_blendv_pd(neg_result, x, mask);
        _mm256_storeu_pd(out.add(offset), result);
    }

    let processed = chunks * F64_LANES;
    if processed < len {
        elu_scalar_f64(a.add(processed), out.add(processed), len - processed, alpha);
    }
}