File size: 2,765 Bytes
be903e2 | 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 | // Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2017 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.
#include "clip_x86.h"
#if __SSE2__
#include <emmintrin.h>
#if __AVX__
#include <immintrin.h>
#endif // __AVX__
#endif // __SSE2__
namespace ncnn {
Clip_x86::Clip_x86()
{
#if __SSE2__
support_packing = true;
#endif // __SSE2__
}
int Clip_x86::forward_inplace(Mat& bottom_top_blob, const Option& opt) const
{
int w = bottom_top_blob.w;
int h = bottom_top_blob.h;
int d = bottom_top_blob.d;
int channels = bottom_top_blob.c;
int elempack = bottom_top_blob.elempack;
int size = w * h * d * elempack;
#pragma omp parallel for num_threads(opt.num_threads)
for (int q = 0; q < channels; q++)
{
float* ptr = bottom_top_blob.channel(q);
int i = 0;
#if __SSE2__
#if __AVX__
#if __AVX512F__
__m512 _min_avx512 = _mm512_set1_ps(min);
__m512 _max_avx512 = _mm512_set1_ps(max);
for (; i + 15 < size; i += 16)
{
__m512 _p = _mm512_loadu_ps(ptr);
_p = _mm512_max_ps(_p, _min_avx512);
_p = _mm512_min_ps(_p, _max_avx512);
_mm512_storeu_ps(ptr, _p);
ptr += 16;
}
#endif // __AVX512F__
__m256 _min_avx = _mm256_set1_ps(min);
__m256 _max_avx = _mm256_set1_ps(max);
for (; i + 7 < size; i += 8)
{
__m256 _p = _mm256_loadu_ps(ptr);
_p = _mm256_max_ps(_p, _min_avx);
_p = _mm256_min_ps(_p, _max_avx);
_mm256_storeu_ps(ptr, _p);
ptr += 8;
}
#endif // __AVX__
__m128 _min = _mm_set1_ps(min);
__m128 _max = _mm_set1_ps(max);
for (; i + 3 < size; i += 4)
{
__m128 _p = _mm_load_ps(ptr);
_p = _mm_max_ps(_p, _min);
_p = _mm_min_ps(_p, _max);
_mm_store_ps(ptr, _p);
ptr += 4;
}
#endif // __SSE2__
for (; i < size; i++)
{
if (*ptr < min)
*ptr = min;
if (*ptr > max)
*ptr = max;
ptr++;
}
}
return 0;
}
} //namespace ncnn
|