ncnn

Форк
0
/
normalize_norm_pack4.comp 
135 строк · 3.9 Кб
1
// Tencent is pleased to support the open source community by making ncnn available.
2
//
3
// Copyright (C) 2019 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
#version 450
16

17
#if NCNN_fp16_storage
18
#extension GL_EXT_shader_16bit_storage: require
19
#endif
20
#if NCNN_fp16_arithmetic
21
#extension GL_EXT_shader_explicit_arithmetic_types_float16: require
22
#endif
23

24
layout (constant_id = 0) const int across_spatial = 0;
25
layout (constant_id = 1) const int across_channel = 0;
26
layout (constant_id = 2) const int channel_shared = 0;
27
layout (constant_id = 3) const int scale_term = 0;
28
layout (constant_id = 4) const float channel_shared_scale = 1.f;
29

30
#define shape_constant_id_offset 5
31
layout (constant_id = shape_constant_id_offset + 0) const int dims = 0;
32
layout (constant_id = shape_constant_id_offset + 1) const int w = 0;
33
layout (constant_id = shape_constant_id_offset + 2) const int h = 0;
34
layout (constant_id = shape_constant_id_offset + 3) const int c = 0;
35
layout (constant_id = shape_constant_id_offset + 4) const int cstep = 0;
36

37
#if NCNN_image_shader
38
layout (binding = 0) uniform unfp sampler3D bottom_blob;
39
layout (binding = 1, imfmtc4) writeonly uniform unfp image3D top_blob;
40
layout (binding = 2) uniform unfp sampler3D coeffs_blob;
41
layout (binding = 3) uniform unfp sampler3D scale_blob;
42
#else
43
layout (binding = 0) buffer bottom_top_blob { sfpvec4 bottom_top_blob_data[]; };
44
layout (binding = 1) readonly buffer coeffs_blob { sfpvec4 coeffs_blob_data[]; };
45
layout (binding = 2) readonly buffer scale_blob { sfpvec4 scale_blob_data[]; };
46
#endif
47

48
layout (push_constant) uniform parameter
49
{
50
    int dims;
51
    int w;
52
    int h;
53
    int c;
54
    int cstep;
55
} p;
56

57
void main()
58
{
59
    int gx = int(gl_GlobalInvocationID.x);
60
    int gy = int(gl_GlobalInvocationID.y);
61
    int gz = int(gl_GlobalInvocationID.z);
62

63
    if (gx >= psc(w) || gy >= psc(h) || gz >= psc(c))
64
        return;
65

66
#if NCNN_image_shader
67
    afpvec4 v = image3d_ld4(bottom_blob, ivec3(gx, gy, gz));
68
#else
69
    const int gi = gz * psc(cstep) + gy * psc(w) + gx;
70

71
    afpvec4 v = buffer_ld4(bottom_top_blob_data, gi);
72
#endif
73

74
    afpvec4 a;
75

76
    if (across_spatial == 1 && across_channel == 1)
77
    {
78
#if NCNN_image_shader
79
        a = image3d_ld4(coeffs_blob, ivec3(0, 0, 0));
80
#else
81
        a = buffer_ld4(coeffs_blob_data, 0);
82
#endif
83
    }
84

85
    if (across_spatial == 1 && across_channel == 0)
86
    {
87
#if NCNN_image_shader
88
        a = image3d_ld4(coeffs_blob, ivec3(gz, 0, 0));
89
#else
90
        a = buffer_ld4(coeffs_blob_data, gz);
91
#endif
92
    }
93

94
    if (across_spatial == 0 && across_channel == 1)
95
    {
96
#if NCNN_image_shader
97
        a = image3d_ld4(coeffs_blob, ivec3(gy * psc(w) + gx, 0, 0));
98
#else
99
        a = buffer_ld4(coeffs_blob_data, gy * psc(w) + gx);
100
#endif
101
    }
102

103
#if !NCNN_image_shader && (NCNN_fp16_packed || NCNN_fp16_storage)
104
    // NOTE coeffs may produce (X, undef, X, undef) on nvidia fp16p/fp16s
105
    // TODO only enable this workaround for some nvidia driver
106
    if (across_channel == 1)
107
    {
108
        a = afpvec4(a.r);
109
    }
110
#endif
111

112
    v = v * a;
113

114
    if (scale_term == 1)
115
    {
116
        if (channel_shared == 1)
117
        {
118
            v = v * afp(channel_shared_scale);
119
        }
120
        else
121
        {
122
#if NCNN_image_shader
123
            v = v * image3d_ld4(scale_blob, ivec3(gz, 0, 0));
124
#else
125
            v = v * buffer_ld4(scale_blob_data, gz);
126
#endif
127
        }
128
    }
129

130
#if NCNN_image_shader
131
    image3d_st4(top_blob, ivec3(gx, gy, gz), v);
132
#else
133
    buffer_st4(bottom_top_blob_data, gi, v);
134
#endif
135
}
136

Использование cookies

Мы используем файлы cookie в соответствии с Политикой конфиденциальности и Политикой использования cookies.

Нажимая кнопку «Принимаю», Вы даете АО «СберТех» согласие на обработку Ваших персональных данных в целях совершенствования нашего веб-сайта и Сервиса GitVerse, а также повышения удобства их использования.

Запретить использование cookies Вы можете самостоятельно в настройках Вашего браузера.