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
|
/*
* Copyright (c) 2017-2018 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(VEC_SIZE)
#if defined(WIDTH_OFFSET) && defined(DEPTH)
/** This kernel concatenates the input tensor into the output tensor along the first dimension
*
* @note The data type has to be passed at compile time using -DDATA_TYPE. i.e. -DDATA_TYPE=float
* @note Vector size has to be passed at compile time using -DVEC_SIZE. i.e. -DVEC_SIZE=16
* @note The offset for the first spatial dimension has to be passed at compile time using -DWIDTH_OFFSET. i.e. -DWIDTH_OFFSET=128
* @note Tensor depth should be given as a preprocessor argument using -DDEPTH=size. e.g. -DDEPTH16
*
* @param[in] src_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] src_step_y src_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] src_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] src_step_z src_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] src_stride_w Stride of the first source tensor in Z dimension (in bytes)
* @param[in] src_step_w src_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] src_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[out] dst_ptr Pointer to the destination tensor. Supported data types: same as @p src_ptr
* @param[in] dst_stride_x Stride of the destination tensor in X dimension (in bytes)
* @param[in] dst_step_x dst_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] dst_stride_y Stride of the destination tensor in Y dimension (in bytes)
* @param[in] dst_step_y dst_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] dst_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] dst_step_z dst_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] dst_stride_w Stride of the destination tensor in Z dimension (in bytes)
* @param[in] dst_step_w output_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] dst_offset_first_element_in_bytes The offset of the first element in the destination tensor
* @param[in] offset The offset to the first valid element of the output tensor in bytes
*/
__kernel void concatenate_width(
TENSOR4D_DECLARATION(src),
TENSOR4D_DECLARATION(dst))
{
Tensor4D src = CONVERT_TO_TENSOR4D_STRUCT(src, DEPTH);
Tensor4D dst = CONVERT_TO_TENSOR4D_STRUCT(dst, DEPTH);
VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
source_values = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)src.ptr);
VSTORE(VEC_SIZE)
(source_values, 0, (__global DATA_TYPE *)(dst.ptr) + WIDTH_OFFSET);
}
#endif /* defined(WIDTH_OFFSET) && defined(DEPTH) */
/** This kernel concatenates the input tensor into the output tensor along the third dimension
*
* @note The data type has to be passed at compile time using -DDATA_TYPE. i.e. -DDATA_TYPE=float
* @note Vector size has to be passed at compile time using -DVEC_SIZE. i.e. -DVEC_SIZE=16
*
* @param[in] src_ptr Pointer to the source tensor. Supported data types: F16, F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] src_step_y src_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] src_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] src_step_z src_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] src_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[out] dst_ptr Pointer to the destination tensor. Supported data types: same as @p src_ptr
* @param[in] dst_stride_x Stride of the destination tensor in X dimension (in bytes)
* @param[in] dst_step_x dst_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] dst_stride_y Stride of the destination tensor in Y dimension (in bytes)
* @param[in] dst_step_y dst_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] dst_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] dst_step_z dst_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] dst_offset_first_element_in_bytes The offset of the first element in the destination tensor
* @param[in] offsets The offsets to the first valid element of the output tensor in bytes
*/
__kernel void concatenate_depth(
TENSOR3D_DECLARATION(src),
TENSOR3D_DECLARATION(dst),
int3 offsets)
{
Tensor3D src = CONVERT_TO_TENSOR3D_STRUCT(src);
Tensor3D dst = CONVERT_TO_TENSOR3D_STRUCT(dst);
VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
source_values = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)tensor3D_offset(&src, -offsets.x, -offsets.y, 0));
VSTORE(VEC_SIZE)
(source_values, 0, (__global DATA_TYPE *)(dst.ptr + offsets.z));
}
#endif /* defined(DATA_TYPE) && defined(VEC_SIZE) */
|
