1
// Tencent is pleased to support the open source community by making ncnn available.
3
// Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
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
8
// https://opensource.org/licenses/BSD-3-Clause
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.
18
#extension GL_EXT_shader_16bit_storage: require
20
#if NCNN_fp16_arithmetic
21
#extension GL_EXT_shader_explicit_arithmetic_types_float16: require
25
layout (binding = 0) uniform highp sampler3D square_blob;
26
layout (binding = 1, rgba32f) writeonly uniform highp image3D sqsum_blob;
28
layout (binding = 0) readonly buffer square_blob { vec4 square_blob_data[]; };
29
layout (binding = 1) writeonly buffer sqsum_blob { vec4 sqsum_blob_data[]; };
32
layout (push_constant) uniform parameter
47
int gx = int(gl_GlobalInvocationID.x);
48
int gy = int(gl_GlobalInvocationID.y);
49
int gz = int(gl_GlobalInvocationID.z);
51
if (gx >= p.outw || gy >= p.outh || gz >= p.outc)
64
vec4 v0 = texelFetch(square_blob, ivec3(sx, sy, gz), 0);
70
vec4 v0 = texelFetch(square_blob, ivec3(sx, sy, gz), 0);
71
vec4 v1 = texelFetch(square_blob, ivec3(sx + 1, sy, gz), 0);
80
vec4 v0 = texelFetch(square_blob, ivec3(sx, sy, gz), 0);
81
vec4 v2 = texelFetch(square_blob, ivec3(sx, sy + 1, gz), 0);
87
vec4 v0 = texelFetch(square_blob, ivec3(sx, sy, gz), 0);
88
vec4 v1 = texelFetch(square_blob, ivec3(sx + 1, sy, gz), 0);
89
vec4 v2 = texelFetch(square_blob, ivec3(sx, sy + 1, gz), 0);
90
vec4 v3 = texelFetch(square_blob, ivec3(sx + 1, sy + 1, gz), 0);
92
sum = v0 + v1 + v2 + v3;
98
int v_offset = gz * p.cstep + sx;
102
sum = square_blob_data[v_offset];
104
else if (sx == p.w - 2)
106
vec4 v0 = square_blob_data[v_offset];
107
vec4 v1 = square_blob_data[v_offset + 1];
111
else if (sx == p.w - 3)
113
vec4 v0 = square_blob_data[v_offset];
114
vec4 v1 = square_blob_data[v_offset + 1];
115
vec4 v2 = square_blob_data[v_offset + 2];
121
vec4 v0 = square_blob_data[v_offset];
122
vec4 v1 = square_blob_data[v_offset + 1];
123
vec4 v2 = square_blob_data[v_offset + 2];
124
vec4 v3 = square_blob_data[v_offset + 3];
126
sum = v0 + v1 + v2 + v3;
131
imageStore(sqsum_blob, ivec3(gx, gy, gz), sum);
133
int gi = gz * p.outcstep + gx;
135
sqsum_blob_data[gi] = sum;