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
|
/*
* Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
* Copyright (c) 2017 ARM Limited.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* 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.
*/
#include "helpers.h"
#ifdef SATURATE
#define SUB(x, y) sub_sat((x), (y))
#else /* SATURATE */
#define SUB(x, y) (x) - (y)
#endif /* SATURATE */
/** This function subtracts one tensors from another.
*
* @attention The input and output data_types need to be passed at compile time using -DDATA_TYPE_IN1, -DDATA_TYPE_IN2 and -DDATA_TYPE_OUT:
* e.g. -DDATA_TYPE_IN1=uchar -DDATA_TYPE_IN2=uchar -DDATA_TYPE_OUT=short
* @attention To perform saturating operation -DSATURATE has to be passed to the compiler otherwise wrapping policy will be used.
*
* @param[in] in1_ptr Pointer to the source tensor. Supported data types: U8, S16
* @param[in] in1_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] in1_step_x in1_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] in1_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] in1_step_y in1_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] in1_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] in1_step_z in1_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] in1_offset_first_element_in_bytes The offset of the first element in the source tensor
* @param[in] in2_ptr Pointer to the source tensor. Supported data types: U8, S16
* @param[in] in2_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] in2_step_x in2_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] in2_stride_y Stride of the source tensor in Y dimension (in bytes)
* @param[in] in2_step_y in2_stride_y * number of elements along Y processed per workitem(in bytes)
* @param[in] in2_stride_z Stride of the source tensor in Z dimension (in bytes)
* @param[in] in2_step_z in2_stride_z * number of elements along Z processed per workitem(in bytes)
* @param[in] in2_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, S16
* @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 arithmetic_sub_ex(
TENSOR3D_DECLARATION(in1),
TENSOR3D_DECLARATION(in2),
TENSOR3D_DECLARATION(out))
{
// Get pixels pointer
Tensor3D in1 = CONVERT_TO_TENSOR3D_STRUCT(in1);
Tensor3D in2 = CONVERT_TO_TENSOR3D_STRUCT(in2);
Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(out);
// Load values
VEC_DATA_TYPE(DATA_TYPE_OUT, 16)
in_a = CONVERT(vload16(0, (__global DATA_TYPE_IN1 *)in1.ptr), VEC_DATA_TYPE(DATA_TYPE_OUT, 16));
VEC_DATA_TYPE(DATA_TYPE_OUT, 16)
in_b = CONVERT(vload16(0, (__global DATA_TYPE_IN2 *)in2.ptr), VEC_DATA_TYPE(DATA_TYPE_OUT, 16));
// Calculate and store result
vstore16(SUB(in_a, in_b), 0, (__global DATA_TYPE_OUT *)out.ptr);
}
|
