ncnn

1
// Tencent is pleased to support the open source community by making ncnn available.
2
//
3
// Copyright (C) 2020 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
struct sfpvec8 { f16vec4 abcd; f16vec4 efgh; };
20
#endif
21
#if NCNN_fp16_arithmetic
22
#extension GL_EXT_shader_explicit_arithmetic_types_float16: require
23
#endif
24

25
#extension GL_GOOGLE_include_directive: enable
26
#include "vulkan_activation.comp"
27

28
layout (constant_id = 0) const int bias_term = 0;
29
layout (constant_id = 1) const int activation_type = 0;
30
layout (constant_id = 2) const float activation_param_0 = 0;
31
layout (constant_id = 3) const float activation_param_1 = 0;
32

33
#define shape_constant_id_offset 4
34
layout (constant_id = shape_constant_id_offset + 0) const int w = 0;
35
layout (constant_id = shape_constant_id_offset + 1) const int h = 0;
36
layout (constant_id = shape_constant_id_offset + 2) const int c = 0;
37
layout (constant_id = shape_constant_id_offset + 3) const int cstep = 0;
38

39
layout (constant_id = shape_constant_id_offset + 4) const int outw = 0;
40
layout (constant_id = shape_constant_id_offset + 5) const int outh = 0;
41
layout (constant_id = shape_constant_id_offset + 6) const int outc = 0;
42
layout (constant_id = shape_constant_id_offset + 7) const int outcstep = 0;
43

44
#if NCNN_image_shader
45
layout (binding = 0) uniform unfp sampler3D bottom_blob;
46
layout (binding = 1, imfmtc1) writeonly uniform unfp image3D top_blob;
47
layout (binding = 2) uniform unfp sampler3D weight_blob;
48
layout (binding = 3) uniform unfp sampler3D bias_blob;
49
#else
50
layout (binding = 0) readonly buffer bottom_blob { sfpvec8 bottom_blob_data[]; };
51
layout (binding = 1) writeonly buffer top_blob { sfp top_blob_data[]; };
52
layout (binding = 2) readonly buffer weight_blob { sfpvec8 weight_data[]; };
53
layout (binding = 3) readonly buffer bias_blob { sfp bias_data[]; };
54
#endif
55

56
layout (push_constant) uniform parameter
57
{
58
    int w;
59
    int h;
60
    int c;
61
    int cstep;
62

63
    int outw;
64
    int outh;
65
    int outc;
66
    int outcstep;
67
} p;
68

69
void main()
70
{
71
    int gx = int(gl_GlobalInvocationID.x) * 4;
72
    int gy = int(gl_GlobalInvocationID.y);
73

74
#if NCNN_image_shader
75
    if (gx >= psc(outw) * psc(outh) || gy >= psc(outc))
76
        return;
77
#else
78
    if (gx >= psc(outcstep) || gy >= psc(outc))
79
        return;
80
#endif
81

82
    afp sum0;
83
    afp sum1;
84
    afp sum2;
85
    afp sum3;
86

87
    if (bias_term == 1)
88
    {
89
#if NCNN_image_shader
90
        afp b = image3d_ld1(bias_blob, ivec3(gy, 0, 0));
91
#else
92
        afp b = buffer_ld1(bias_data, gy);
93
#endif
94
        sum0 = b;
95
        sum1 = b;
96
        sum2 = b;
97
        sum3 = b;
98
    }
99
    else
100
    {
101
        sum0 = afp(0.f);
102
        sum1 = afp(0.f);
103
        sum2 = afp(0.f);
104
        sum3 = afp(0.f);
105
    }
106

107
#if NCNN_image_shader
108
    ivec4 gx4 = gx + ivec4(0, 1, 2, 3);
109

110
    ivec4 sy4 = gx4 / psc(w);
111
    ivec4 sx4 = gx4 % psc(w);
112

113
    for (int z = 0; z < psc(c); z++)
114
    {
115
        afpvec8 v0 = image3d_ld8(bottom_blob, ivec3(sx4.r, sy4.r, z));
116
        afpvec8 v1 = image3d_ld8(bottom_blob, ivec3(sx4.g, sy4.g, z));
117
        afpvec8 v2 = image3d_ld8(bottom_blob, ivec3(sx4.b, sy4.b, z));
118
        afpvec8 v3 = image3d_ld8(bottom_blob, ivec3(sx4.a, sy4.a, z));
119

120
        afpvec8 k = image3d_ld8(weight_blob, ivec3(0, z, gy));
121

122
        // sum += dot(v, k);
123
        sum0 += dot(v0[0], k[0]) + dot(v0[1], k[1]);
124
        sum1 += dot(v1[0], k[0]) + dot(v1[1], k[1]);
125
        sum2 += dot(v2[0], k[0]) + dot(v2[1], k[1]);
126
        sum3 += dot(v3[0], k[0]) + dot(v3[1], k[1]);
127
    }
128
#else
129
    int w_offset = gy * psc(c);
130
    int v_offset = gx;
131

132
    for (int z = 0; z < psc(c); z++)
133
    {
134
        afpvec8 v0 = buffer_ld8(bottom_blob_data, v_offset + 0);
135
        afpvec8 v1 = buffer_ld8(bottom_blob_data, v_offset + 1);
136
        afpvec8 v2 = buffer_ld8(bottom_blob_data, v_offset + 2);
137
        afpvec8 v3 = buffer_ld8(bottom_blob_data, v_offset + 3);
138

139
        afpvec8 k = buffer_ld8(weight_data, w_offset);
140

141
        // sum += dot(v, k);
142
        sum0 += dot(v0[0], k[0]) + dot(v0[1], k[1]);
143
        sum1 += dot(v1[0], k[0]) + dot(v1[1], k[1]);
144
        sum2 += dot(v2[0], k[0]) + dot(v2[1], k[1]);
145
        sum3 += dot(v3[0], k[0]) + dot(v3[1], k[1]);
146

147
        w_offset += 1;
148
        v_offset += psc(cstep);
149
    }
150
#endif
151

152
    sum0 = activation_afp(sum0, activation_type, activation_param_0, activation_param_1);
153
    sum1 = activation_afp(sum1, activation_type, activation_param_0, activation_param_1);
154
    sum2 = activation_afp(sum2, activation_type, activation_param_0, activation_param_1);
155
    sum3 = activation_afp(sum3, activation_type, activation_param_0, activation_param_1);
156

157
#if NCNN_image_shader
158
    image3d_st1(top_blob, ivec3(sx4.r, sy4.r, gy), sum0);
159
    image3d_st1(top_blob, ivec3(sx4.g, sy4.g, gy), sum1);
160
    image3d_st1(top_blob, ivec3(sx4.b, sy4.b, gy), sum2);
161
    image3d_st1(top_blob, ivec3(sx4.a, sy4.a, gy), sum3);
162
#else
163
    int gi = gy * psc(outcstep) + gx;
164

165
    buffer_st1(top_blob_data, gi + 0, sum0);
166
    if (gx + 1 < psc(outcstep)) buffer_st1(top_blob_data, gi + 1, sum1);
167
    if (gx + 2 < psc(outcstep)) buffer_st1(top_blob_data, gi + 2, sum2);
168
    if (gx + 3 < psc(outcstep)) buffer_st1(top_blob_data, gi + 3, sum3);
169
#endif
170
}
171