1 // Tencent is pleased to support the open source community by making ncnn available.
2 //
3 // Copyright (C) 2017 THL A29 Limited, a Tencent company. All rights reserved.
4 //
5 // Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
6 // in compliance with the License. You may obtain a copy of the License at
7 //
8 // https://opensource.org/licenses/BSD-3-Clause
9 //
10 // Unless required by applicable law or agreed to in writing, software distributed
11 // under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
12 // CONDITIONS OF ANY KIND, either express or implied. See the License for the
13 // specific language governing permissions and limitations under the License.
14
15 #include "sigmoid_x86.h"
16
17 #include "x86_activation.h"
18
19 #include <math.h>
20
21 namespace ncnn {
22
Sigmoid_x86()23 Sigmoid_x86::Sigmoid_x86()
24 {
25 #if __SSE2__
26 support_packing = true;
27 #endif // __SSE2__
28 }
29
forward_inplace(Mat & bottom_top_blob,const Option & opt) const30 int Sigmoid_x86::forward_inplace(Mat& bottom_top_blob, const Option& opt) const
31 {
32 int w = bottom_top_blob.w;
33 int h = bottom_top_blob.h;
34 int channels = bottom_top_blob.c;
35 int size = w * h;
36 #if __SSE2__
37 int elempack = bottom_top_blob.elempack;
38
39 #if __AVX__
40 if (elempack == 8)
41 {
42 #pragma omp parallel for num_threads(opt.num_threads)
43 for (int q = 0; q < channels; q++)
44 {
45 float* ptr = bottom_top_blob.channel(q);
46 for (int i = 0; i < size; i++)
47 {
48 __m256 _p = _mm256_loadu_ps(ptr);
49 _mm256_storeu_ps(ptr, sigmoid_avx(_p));
50 ptr += 8;
51 }
52 }
53
54 return 0;
55 }
56 #endif // __AVX__
57
58 if (elempack == 4)
59 {
60 #pragma omp parallel for num_threads(opt.num_threads)
61 for (int q = 0; q < channels; q++)
62 {
63 float* ptr = bottom_top_blob.channel(q);
64 for (int i = 0; i < size; i++)
65 {
66 __m128 _p = _mm_loadu_ps(ptr);
67 _mm_storeu_ps(ptr, sigmoid_sse(_p));
68 ptr += 4;
69 }
70 }
71
72 return 0;
73 }
74 #endif // __SSE2__
75
76 #pragma omp parallel for num_threads(opt.num_threads)
77 for (int q = 0; q < channels; q++)
78 {
79 float* ptr = bottom_top_blob.channel(q);
80
81 int i = 0;
82 #if __SSE2__
83 #if __AVX__
84 for (; i + 7 < size; i += 8)
85 {
86 __m256 _p = _mm256_loadu_ps(ptr);
87 _mm256_storeu_ps(ptr, sigmoid_avx(_p));
88 ptr += 8;
89 }
90 #endif // __AVX__
91 for (; i + 3 < size; i += 4)
92 {
93 __m128 _p = _mm_loadu_ps(ptr);
94 _mm_storeu_ps(ptr, sigmoid_sse(_p));
95 ptr += 4;
96 }
97 #endif // __SSE2__
98 for (; i < size; i++)
99 {
100 *ptr = 1.f / (1.f + exp(-*ptr));
101
102 ptr++;
103 }
104 }
105
106 return 0;
107 }
108
109 } // namespace ncnn
110