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
|
/*
* Copyright (c) 2018 Arm Limited.
*
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include "helpers.h"
#if defined(DATA_TYPE) && defined(SELECT_DATA_TYPE) && defined(VEC_SIZE)
/** This function perform a select operation between two tensors when condition tensor has the same rank.
*
* @attention The data_type need to be passed at compile time using -DDATA_TYPE: e.g. -DDATA_TYPE=uchar
* @attention The select operation data_type need to be passed at compile time using -DSELECT_DATA_TYPE: e.g. -DSELECT_DATA_TYPE=uchar
* @attention Vector size should be given as a preprocessor argument using -DVEC_SIZE=size. e.g. -DVEC_SIZE=16
*
* @param[in] c_ptr Pointer to the source tensor. Supported data types: U8
* @param[in] c_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] c_step_x c_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] c_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] c_step_y c_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] c_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] c_step_z c_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] c_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[in] x_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] x_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] x_step_x x_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] x_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] x_step_y x_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] x_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] x_step_z x_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] x_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[in] y_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] y_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] y_step_x y_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] y_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] y_step_y y_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] y_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] y_step_z y_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] y_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[out] out_ptr Pointer to the destination tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] out_stride_x Stride of the destination tensor in X dimension (in bytes)
* @param[in] out_step_x out_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] out_stride_y Stride of the destination tensor in Y dimension (in bytes)
* @param[in] out_step_y out_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] out_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] out_step_z out_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] out_offset_first_element_in_bytes The offset of the first element in the destination tensor
*/
__kernel void select_same_rank(
TENSOR3D_DECLARATION(c),
TENSOR3D_DECLARATION(x),
TENSOR3D_DECLARATION(y),
TENSOR3D_DECLARATION(out))
{
// Get pixels pointer
Tensor3D c_t = CONVERT_TO_TENSOR3D_STRUCT(c);
Tensor3D x_t = CONVERT_TO_TENSOR3D_STRUCT(x);
Tensor3D y_t = CONVERT_TO_TENSOR3D_STRUCT(y);
Tensor3D out_t = CONVERT_TO_TENSOR3D_STRUCT(out);
// Load values
VEC_DATA_TYPE(SELECT_DATA_TYPE, VEC_SIZE)
in_c = CONVERT((VLOAD(VEC_SIZE)(0, (__global uchar *)c_t.ptr)), VEC_DATA_TYPE(SELECT_DATA_TYPE, VEC_SIZE));
VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
in_x = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)x_t.ptr);
VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
in_y = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)y_t.ptr);
// Calculate and store result
VSTORE(VEC_SIZE)
(select(in_y, in_x, in_c > (SELECT_DATA_TYPE)0), 0, (__global DATA_TYPE *)out_t.ptr);
}
/** This function perform a select operation between two tensors when condition tensor has a different rank.
*
* @attention The data_type need to be passed at compile time using -DDATA_TYPE: e.g. -DDATA_TYPE=uchar
* @attention The select operation data_type need to be passed at compile time using -DSELECT_DATA_TYPE: e.g. -DSELECT_DATA_TYPE=uchar
* @attention Vector size should be given as a preprocessor argument using -DVEC_SIZE=size. e.g. -DVEC_SIZE=16
*
* @param[in] c_ptr Pointer to the source tensor. Supported data types: U8
* @param[in] c_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] c_step_x c_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] c_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[in] x_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] x_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] x_step_x x_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] x_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] x_step_y x_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] x_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] x_step_z x_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] x_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[in] y_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] y_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] y_step_x y_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] y_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] y_step_y y_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] y_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] y_step_z y_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] y_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[out] out_ptr Pointer to the destination tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] out_stride_x Stride of the destination tensor in X dimension (in bytes)
* @param[in] out_step_x out_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] out_stride_y Stride of the destination tensor in Y dimension (in bytes)
* @param[in] out_step_y out_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] out_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] out_step_z out_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] out_offset_first_element_in_bytes The offset of the first element in the destination tensor
*/
__kernel void select_different_rank_2(
VECTOR_DECLARATION(c),
TENSOR3D_DECLARATION(x),
TENSOR3D_DECLARATION(y),
TENSOR3D_DECLARATION(out))
{
const int c_idx = get_global_id(1);
// Get pixels pointer
Vector c_t = CONVERT_TO_VECTOR_STRUCT_NO_STEP(c);
Tensor3D x_t = CONVERT_TO_TENSOR3D_STRUCT(x);
Tensor3D y_t = CONVERT_TO_TENSOR3D_STRUCT(y);
Tensor3D out_t = CONVERT_TO_TENSOR3D_STRUCT(out);
// Load values
VEC_DATA_TYPE(SELECT_DATA_TYPE, VEC_SIZE)
in_c = *((__global uchar *)(c_t.ptr + c_idx * c_t.stride_x));
VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
in_x = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)x_t.ptr);
VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
in_y = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)y_t.ptr);
// Calculate and store result
VSTORE(VEC_SIZE)
(select(in_y, in_x, in_c > (SELECT_DATA_TYPE)0), 0, (__global DATA_TYPE *)out_t.ptr);
}
#endif /* defined(DATA_TYPE) && defined(SELECT_DATA_TYPE) && defined(VEC_SIZE) */
#if defined(DATA_TYPE) && defined(SELECT_DATA_TYPE) && defined(VEC_SIZE) && defined(DEPTH_SIZE)
/** This function perform a select operation between two tensors when condition tensor has a different rank.
*
* @attention The data_type need to be passed at compile time using -DDATA_TYPE: e.g. -DDATA_TYPE=uchar
* @attention The select operation data_type need to be passed at compile time using -DSELECT_DATA_TYPE: e.g. -DSELECT_DATA_TYPE=uchar
* @attention Vector size should be given as a preprocessor argument using -DVEC_SIZE=size. e.g. -DVEC_SIZE=16
*
* @param[in] c_ptr Pointer to the source tensor. Supported data types: U8
* @param[in] c_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] c_step_x c_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] c_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[in] x_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] x_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] x_step_x x_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] x_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] x_step_y x_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] x_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] x_step_z x_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] x_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[in] y_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] y_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] y_step_x y_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] y_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] y_step_y y_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] y_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] y_step_z y_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] y_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[out] out_ptr Pointer to the destination tensor. Supported data types: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
* @param[in] out_stride_x Stride of the destination tensor in X dimension (in bytes)
* @param[in] out_step_x out_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] out_stride_y Stride of the destination tensor in Y dimension (in bytes)
* @param[in] out_step_y out_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] out_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] out_step_z out_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] out_offset_first_element_in_bytes The offset of the first element in the destination tensor
*/
__kernel void select_different_rank_n(
VECTOR_DECLARATION(c),
TENSOR3D_DECLARATION(x),
TENSOR3D_DECLARATION(y),
TENSOR3D_DECLARATION(out))
{
const int c_idx = get_global_id(2) / DEPTH_SIZE;
// Get pixels pointer
Vector c_t = CONVERT_TO_VECTOR_STRUCT_NO_STEP(c);
Tensor3D x_t = CONVERT_TO_TENSOR3D_STRUCT(x);
Tensor3D y_t = CONVERT_TO_TENSOR3D_STRUCT(y);
Tensor3D out_t = CONVERT_TO_TENSOR3D_STRUCT(out);
// Load values
VEC_DATA_TYPE(SELECT_DATA_TYPE, VEC_SIZE)
in_c = *((__global uchar *)(c_t.ptr + c_idx * c_t.stride_x));
VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
in_x = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)x_t.ptr);
VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
in_y = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)y_t.ptr);
// Calculate and store result
VSTORE(VEC_SIZE)
(select(in_y, in_x, in_c > (SELECT_DATA_TYPE)0), 0, (__global DATA_TYPE *)out_t.ptr);
}
#endif /* defined(DATA_TYPE) && defined(SELECT_DATA_TYPE) && defined(VEC_SIZE) && defined(DEPTH_SIZE) */
|
