Hugging Face's logo

camenduru
/
ncnn

Model card Files
xet
Community
ncnn
File size: 18,214 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
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
 
// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2022 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.

#version 450

#if NCNN_fp16_storage
#extension GL_EXT_shader_16bit_storage: require
#endif
#if NCNN_fp16_arithmetic
#extension GL_EXT_shader_explicit_arithmetic_types_float16: require
#endif

#define shape_constant_id_offset 0
layout (constant_id = shape_constant_id_offset + 0) const int w = 0;
layout (constant_id = shape_constant_id_offset + 1) const int h = 0;
layout (constant_id = shape_constant_id_offset + 2) const int c = 0;
layout (constant_id = shape_constant_id_offset + 3) const int cstep = 0;

layout (constant_id = shape_constant_id_offset + 4) const int outcstep = 0;

layout (constant_id = shape_constant_id_offset + 5) const int block_x = 0;
layout (constant_id = shape_constant_id_offset + 6) const int block_y = 0;

#if NCNN_image_shader
layout (binding = 0) uniform unfp sampler3D bottom_blob;
layout (binding = 1, imfmtc1) writeonly uniform unfp image3D bottom_tm_blob;
#else
layout (binding = 0) readonly buffer bottom_blob { sfp bottom_blob_data[]; };
layout (binding = 1) writeonly buffer bottom_tm_blob { sfp bottom_tm_blob_data[]; };
#endif

layout (push_constant) uniform parameter
{
    int w;
    int h;
    int c;
    int cstep;

    int outcstep;

    int block_x;
    int block_y;
} p;

void main()
{
    int gx = int(gl_GlobalInvocationID.x);
    int gy = int(gl_GlobalInvocationID.y);
    int gz = int(gl_GlobalInvocationID.z);

    if (gx >= psc(block_x) || gy >= psc(block_y) || gz >= psc(c))
        return;

    // load 6x6
    int sx = gx * 4;
    int sy = gy * 4;

#if NCNN_image_shader
    afp v00 = image3d_ld1(bottom_blob, ivec3(sx + 0, sy + 0, gz));
    afp v01 = image3d_ld1(bottom_blob, ivec3(sx + 1, sy + 0, gz));
    afp v02 = image3d_ld1(bottom_blob, ivec3(sx + 2, sy + 0, gz));
    afp v03 = image3d_ld1(bottom_blob, ivec3(sx + 3, sy + 0, gz));
    afp v04 = image3d_ld1(bottom_blob, ivec3(sx + 4, sy + 0, gz));
    afp v05 = image3d_ld1(bottom_blob, ivec3(sx + 5, sy + 0, gz));

    afp v10 = image3d_ld1(bottom_blob, ivec3(sx + 0, sy + 1, gz));
    afp v11 = image3d_ld1(bottom_blob, ivec3(sx + 1, sy + 1, gz));
    afp v12 = image3d_ld1(bottom_blob, ivec3(sx + 2, sy + 1, gz));
    afp v13 = image3d_ld1(bottom_blob, ivec3(sx + 3, sy + 1, gz));
    afp v14 = image3d_ld1(bottom_blob, ivec3(sx + 4, sy + 1, gz));
    afp v15 = image3d_ld1(bottom_blob, ivec3(sx + 5, sy + 1, gz));

    afp v20 = image3d_ld1(bottom_blob, ivec3(sx + 0, sy + 2, gz));
    afp v21 = image3d_ld1(bottom_blob, ivec3(sx + 1, sy + 2, gz));
    afp v22 = image3d_ld1(bottom_blob, ivec3(sx + 2, sy + 2, gz));
    afp v23 = image3d_ld1(bottom_blob, ivec3(sx + 3, sy + 2, gz));
    afp v24 = image3d_ld1(bottom_blob, ivec3(sx + 4, sy + 2, gz));
    afp v25 = image3d_ld1(bottom_blob, ivec3(sx + 5, sy + 2, gz));

    afp v30 = image3d_ld1(bottom_blob, ivec3(sx + 0, sy + 3, gz));
    afp v31 = image3d_ld1(bottom_blob, ivec3(sx + 1, sy + 3, gz));
    afp v32 = image3d_ld1(bottom_blob, ivec3(sx + 2, sy + 3, gz));
    afp v33 = image3d_ld1(bottom_blob, ivec3(sx + 3, sy + 3, gz));
    afp v34 = image3d_ld1(bottom_blob, ivec3(sx + 4, sy + 3, gz));
    afp v35 = image3d_ld1(bottom_blob, ivec3(sx + 5, sy + 3, gz));

    afp v40 = image3d_ld1(bottom_blob, ivec3(sx + 0, sy + 4, gz));
    afp v41 = image3d_ld1(bottom_blob, ivec3(sx + 1, sy + 4, gz));
    afp v42 = image3d_ld1(bottom_blob, ivec3(sx + 2, sy + 4, gz));
    afp v43 = image3d_ld1(bottom_blob, ivec3(sx + 3, sy + 4, gz));
    afp v44 = image3d_ld1(bottom_blob, ivec3(sx + 4, sy + 4, gz));
    afp v45 = image3d_ld1(bottom_blob, ivec3(sx + 5, sy + 4, gz));

    afp v50 = image3d_ld1(bottom_blob, ivec3(sx + 0, sy + 5, gz));
    afp v51 = image3d_ld1(bottom_blob, ivec3(sx + 1, sy + 5, gz));
    afp v52 = image3d_ld1(bottom_blob, ivec3(sx + 2, sy + 5, gz));
    afp v53 = image3d_ld1(bottom_blob, ivec3(sx + 3, sy + 5, gz));
    afp v54 = image3d_ld1(bottom_blob, ivec3(sx + 4, sy + 5, gz));
    afp v55 = image3d_ld1(bottom_blob, ivec3(sx + 5, sy + 5, gz));
#else
    int v_offset_0 = gz * psc(cstep) + sy * psc(w) + sx;
    ivec4 v_offset = v_offset_0 + ivec4(0, 1, 2, 3) * psc(w);
    ivec2 v_offset45 = v_offset_0 + ivec2(4, 5) * psc(w);

    afp v00 = buffer_ld1(bottom_blob_data, v_offset.r + 0);
    afp v01 = sx + 1 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.r + 1) : afp(0.f);
    afp v02 = sx + 2 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.r + 2) : afp(0.f);
    afp v03 = sx + 3 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.r + 3) : afp(0.f);
    afp v04 = sx + 4 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.r + 4) : afp(0.f);
    afp v05 = sx + 5 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.r + 5) : afp(0.f);

    afp v10 = sy + 1 < psc(h) ? buffer_ld1(bottom_blob_data, v_offset.g + 0) : afp(0.f);
    afp v11 = sy + 1 < psc(h) && sx + 1 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.g + 1) : afp(0.f);
    afp v12 = sy + 1 < psc(h) && sx + 2 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.g + 2) : afp(0.f);
    afp v13 = sy + 1 < psc(h) && sx + 3 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.g + 3) : afp(0.f);
    afp v14 = sy + 1 < psc(h) && sx + 4 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.g + 4) : afp(0.f);
    afp v15 = sy + 1 < psc(h) && sx + 5 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.g + 5) : afp(0.f);

    afp v20 = sy + 2 < psc(h) ? buffer_ld1(bottom_blob_data, v_offset.b + 0) : afp(0.f);
    afp v21 = sy + 2 < psc(h) && sx + 1 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.b + 1) : afp(0.f);
    afp v22 = sy + 2 < psc(h) && sx + 2 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.b + 2) : afp(0.f);
    afp v23 = sy + 2 < psc(h) && sx + 3 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.b + 3) : afp(0.f);
    afp v24 = sy + 2 < psc(h) && sx + 4 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.b + 4) : afp(0.f);
    afp v25 = sy + 2 < psc(h) && sx + 5 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.b + 5) : afp(0.f);

    afp v30 = sy + 3 < psc(h) ? buffer_ld1(bottom_blob_data, v_offset.a + 0) : afp(0.f);
    afp v31 = sy + 3 < psc(h) && sx + 1 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.a + 1) : afp(0.f);
    afp v32 = sy + 3 < psc(h) && sx + 2 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.a + 2) : afp(0.f);
    afp v33 = sy + 3 < psc(h) && sx + 3 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.a + 3) : afp(0.f);
    afp v34 = sy + 3 < psc(h) && sx + 4 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.a + 4) : afp(0.f);
    afp v35 = sy + 3 < psc(h) && sx + 5 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset.a + 5) : afp(0.f);

    afp v40 = sy + 4 < psc(h) ? buffer_ld1(bottom_blob_data, v_offset45.x + 0) : afp(0.f);
    afp v41 = sy + 4 < psc(h) && sx + 1 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.x + 1) : afp(0.f);
    afp v42 = sy + 4 < psc(h) && sx + 2 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.x + 2) : afp(0.f);
    afp v43 = sy + 4 < psc(h) && sx + 3 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.x + 3) : afp(0.f);
    afp v44 = sy + 4 < psc(h) && sx + 4 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.x + 4) : afp(0.f);
    afp v45 = sy + 4 < psc(h) && sx + 5 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.x + 5) : afp(0.f);

    afp v50 = sy + 5 < psc(h) ? buffer_ld1(bottom_blob_data, v_offset45.y + 0) : afp(0.f);
    afp v51 = sy + 5 < psc(h) && sx + 1 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.y + 1) : afp(0.f);
    afp v52 = sy + 5 < psc(h) && sx + 2 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.y + 2) : afp(0.f);
    afp v53 = sy + 5 < psc(h) && sx + 3 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.y + 3) : afp(0.f);
    afp v54 = sy + 5 < psc(h) && sx + 4 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.y + 4) : afp(0.f);
    afp v55 = sy + 5 < psc(h) && sx + 5 < psc(w) ? buffer_ld1(bottom_blob_data, v_offset45.y + 5) : afp(0.f);
#endif

#define sq2 1.41421356237
#define sq2_d2 1.41421356237/2

    // const float itm[6][6] = {
    //     {1.0f,  0.0f,  -2.5f,  0.0f,  1.0f, 0.0f},
    //     {0.0f, -sq2,   -2.0f,  sq2/2, 1.0f, 0.0f},
    //     {0.0f,  sq2,   -2.0f, -sq2/2, 1.0f, 0.0f},
    //     {0.0f, -sq2/2, -0.5f,  sq2,   1.0f, 0.0f},
    //     {0.0f,  sq2/2, -0.5f, -sq2,   1.0f, 0.0f},
    //     {0.0f,  1.0f,   0.0f,  -2.5f, 0.0f, 1.0f}
    // };

    // implicit transpose
    afp m00 = v00 - v02 * afp(2.5) + v04;
    afp m01 = v10 - v12 * afp(2.5) + v14;
    afp m02 = v20 - v22 * afp(2.5) + v24;
    afp m03 = v30 - v32 * afp(2.5) + v34;
    afp m04 = v40 - v42 * afp(2.5) + v44;
    afp m05 = v50 - v52 * afp(2.5) + v54;

    afp m10 = v04 - v02 * afp(2) + v03 * afp(sq2_d2) - v01 * afp(sq2);
    afp m11 = v14 - v12 * afp(2) + v13 * afp(sq2_d2) - v11 * afp(sq2);
    afp m12 = v24 - v22 * afp(2) + v23 * afp(sq2_d2) - v21 * afp(sq2);
    afp m13 = v34 - v32 * afp(2) + v33 * afp(sq2_d2) - v31 * afp(sq2);
    afp m14 = v44 - v42 * afp(2) + v43 * afp(sq2_d2) - v41 * afp(sq2);
    afp m15 = v54 - v52 * afp(2) + v53 * afp(sq2_d2) - v51 * afp(sq2);

    afp m20 = v04 - v02 * afp(2) - v03 * afp(sq2_d2) + v01 * afp(sq2);
    afp m21 = v14 - v12 * afp(2) - v13 * afp(sq2_d2) + v11 * afp(sq2);
    afp m22 = v24 - v22 * afp(2) - v23 * afp(sq2_d2) + v21 * afp(sq2);
    afp m23 = v34 - v32 * afp(2) - v33 * afp(sq2_d2) + v31 * afp(sq2);
    afp m24 = v44 - v42 * afp(2) - v43 * afp(sq2_d2) + v41 * afp(sq2);
    afp m25 = v54 - v52 * afp(2) - v53 * afp(sq2_d2) + v51 * afp(sq2);

    afp m30 = v04 - v02 * afp(0.5) + v03 * afp(sq2) - v01 * afp(sq2_d2);
    afp m31 = v14 - v12 * afp(0.5) + v13 * afp(sq2) - v11 * afp(sq2_d2);
    afp m32 = v24 - v22 * afp(0.5) + v23 * afp(sq2) - v21 * afp(sq2_d2);
    afp m33 = v34 - v32 * afp(0.5) + v33 * afp(sq2) - v31 * afp(sq2_d2);
    afp m34 = v44 - v42 * afp(0.5) + v43 * afp(sq2) - v41 * afp(sq2_d2);
    afp m35 = v54 - v52 * afp(0.5) + v53 * afp(sq2) - v51 * afp(sq2_d2);

    afp m40 = v04 - v02 * afp(0.5) - v03 * afp(sq2) + v01 * afp(sq2_d2);
    afp m41 = v14 - v12 * afp(0.5) - v13 * afp(sq2) + v11 * afp(sq2_d2);
    afp m42 = v24 - v22 * afp(0.5) - v23 * afp(sq2) + v21 * afp(sq2_d2);
    afp m43 = v34 - v32 * afp(0.5) - v33 * afp(sq2) + v31 * afp(sq2_d2);
    afp m44 = v44 - v42 * afp(0.5) - v43 * afp(sq2) + v41 * afp(sq2_d2);
    afp m45 = v54 - v52 * afp(0.5) - v53 * afp(sq2) + v51 * afp(sq2_d2);

    afp m50 = v01 - v03 * afp(2.5) + v05;
    afp m51 = v11 - v13 * afp(2.5) + v15;
    afp m52 = v21 - v23 * afp(2.5) + v25;
    afp m53 = v31 - v33 * afp(2.5) + v35;
    afp m54 = v41 - v43 * afp(2.5) + v45;
    afp m55 = v51 - v53 * afp(2.5) + v55;

    v00 = m00 - m02 * afp(2.5) + m04;
    v10 = m10 - m12 * afp(2.5) + m14;
    v20 = m20 - m22 * afp(2.5) + m24;
    v30 = m30 - m32 * afp(2.5) + m34;
    v40 = m40 - m42 * afp(2.5) + m44;
    v50 = m50 - m52 * afp(2.5) + m54;

    v01 = m04 - m02 * afp(2) + m03 * afp(sq2_d2) - m01 * afp(sq2);
    v11 = m14 - m12 * afp(2) + m13 * afp(sq2_d2) - m11 * afp(sq2);
    v21 = m24 - m22 * afp(2) + m23 * afp(sq2_d2) - m21 * afp(sq2);
    v31 = m34 - m32 * afp(2) + m33 * afp(sq2_d2) - m31 * afp(sq2);
    v41 = m44 - m42 * afp(2) + m43 * afp(sq2_d2) - m41 * afp(sq2);
    v51 = m54 - m52 * afp(2) + m53 * afp(sq2_d2) - m51 * afp(sq2);

    v02 = m04 - m02 * afp(2) - m03 * afp(sq2_d2) + m01 * afp(sq2);
    v12 = m14 - m12 * afp(2) - m13 * afp(sq2_d2) + m11 * afp(sq2);
    v22 = m24 - m22 * afp(2) - m23 * afp(sq2_d2) + m21 * afp(sq2);
    v32 = m34 - m32 * afp(2) - m33 * afp(sq2_d2) + m31 * afp(sq2);
    v42 = m44 - m42 * afp(2) - m43 * afp(sq2_d2) + m41 * afp(sq2);
    v52 = m54 - m52 * afp(2) - m53 * afp(sq2_d2) + m51 * afp(sq2);

    v03 = m04 - m02 * afp(0.5) + m03 * afp(sq2) - m01 * afp(sq2_d2);
    v13 = m14 - m12 * afp(0.5) + m13 * afp(sq2) - m11 * afp(sq2_d2);
    v23 = m24 - m22 * afp(0.5) + m23 * afp(sq2) - m21 * afp(sq2_d2);
    v33 = m34 - m32 * afp(0.5) + m33 * afp(sq2) - m31 * afp(sq2_d2);
    v43 = m44 - m42 * afp(0.5) + m43 * afp(sq2) - m41 * afp(sq2_d2);
    v53 = m54 - m52 * afp(0.5) + m53 * afp(sq2) - m51 * afp(sq2_d2);

    v04 = m04 - m02 * afp(0.5) - m03 * afp(sq2) + m01 * afp(sq2_d2);
    v14 = m14 - m12 * afp(0.5) - m13 * afp(sq2) + m11 * afp(sq2_d2);
    v24 = m24 - m22 * afp(0.5) - m23 * afp(sq2) + m21 * afp(sq2_d2);
    v34 = m34 - m32 * afp(0.5) - m33 * afp(sq2) + m31 * afp(sq2_d2);
    v44 = m44 - m42 * afp(0.5) - m43 * afp(sq2) + m41 * afp(sq2_d2);
    v54 = m54 - m52 * afp(0.5) - m53 * afp(sq2) + m51 * afp(sq2_d2);

    v05 = m01 - m03 * afp(2.5) + m05;
    v15 = m11 - m13 * afp(2.5) + m15;
    v25 = m21 - m23 * afp(2.5) + m25;
    v35 = m31 - m33 * afp(2.5) + m35;
    v45 = m41 - m43 * afp(2.5) + m45;
    v55 = m51 - m53 * afp(2.5) + m55;

    // store 36
#if NCNN_image_shader
    int x = gy * psc(block_x) + gx;

    image3d_st1(bottom_tm_blob, ivec3(x, gz, 0), v00);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 1), v01);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 2), v02);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 3), v03);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 4), v04);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 5), v05);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 6), v10);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 7), v11);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 8), v12);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 9), v13);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 10), v14);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 11), v15);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 12), v20);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 13), v21);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 14), v22);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 15), v23);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 16), v24);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 17), v25);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 18), v30);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 19), v31);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 20), v32);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 21), v33);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 22), v34);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 23), v35);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 24), v40);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 25), v41);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 26), v42);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 27), v43);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 28), v44);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 29), v45);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 30), v50);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 31), v51);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 32), v52);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 33), v53);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 34), v54);
    image3d_st1(bottom_tm_blob, ivec3(x, gz, 35), v55);
#else
    int v_tm_offset = gz * psc(block_x) * psc(block_y) + gy * psc(block_x) + gx;

    buffer_st1(bottom_tm_blob_data, v_tm_offset + 0 * psc(outcstep), v00);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 1 * psc(outcstep), v01);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 2 * psc(outcstep), v02);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 3 * psc(outcstep), v03);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 4 * psc(outcstep), v04);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 5 * psc(outcstep), v05);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 6 * psc(outcstep), v10);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 7 * psc(outcstep), v11);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 8 * psc(outcstep), v12);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 9 * psc(outcstep), v13);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 10 * psc(outcstep), v14);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 11 * psc(outcstep), v15);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 12 * psc(outcstep), v20);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 13 * psc(outcstep), v21);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 14 * psc(outcstep), v22);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 15 * psc(outcstep), v23);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 16 * psc(outcstep), v24);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 17 * psc(outcstep), v25);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 18 * psc(outcstep), v30);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 19 * psc(outcstep), v31);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 20 * psc(outcstep), v32);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 21 * psc(outcstep), v33);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 22 * psc(outcstep), v34);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 23 * psc(outcstep), v35);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 24 * psc(outcstep), v40);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 25 * psc(outcstep), v41);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 26 * psc(outcstep), v42);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 27 * psc(outcstep), v43);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 28 * psc(outcstep), v44);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 29 * psc(outcstep), v45);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 30 * psc(outcstep), v50);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 31 * psc(outcstep), v51);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 32 * psc(outcstep), v52);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 33 * psc(outcstep), v53);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 34 * psc(outcstep), v54);
    buffer_st1(bottom_tm_blob_data, v_tm_offset + 35 * psc(outcstep), v55);
#endif
}