xref: /aosp_15_r20/external/ComputeLibrary/src/core/CL/cl_kernels/common/elementwise_unary.cl (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1/*
2 * Copyright (c) 2018-2021 Arm Limited.
3 *
4 * SPDX-License-Identifier: MIT
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24#include "helpers.h"
25
26#if defined(DATA_TYPE) && defined(OPERATION)
27
28// Calculate exponential
29#define exp_op(input) exp(input)
30// Calculate reverse square root
31#define rsqrt_op(input) rsqrt(input)
32// Calculate negative
33#define neg_op(input) (-input)
34// Calculate sine
35#define sin_op(input) sin(input)
36// Calculate abs for floating point values
37#define fabs_op(input) fabs(input)
38// Calculate natural_log
39#define natural_log_op(input) log(input)
40// Calculate round (Cannot use round function as it rounds halfway cases away from zero).
41#if defined(VEC_SIZE)
42#define VEC_TYPE VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
43#define round_op(input) CONVERT(CONVERT_SAT_ROUND(input, VEC_DATA_TYPE(int, VEC_SIZE), rte), VEC_TYPE)
44#define logical_not_op(input) CONVERT(CONVERT(!input, VEC_TYPE) & ((VEC_TYPE)0x1), VEC_TYPE)
45#else // defined(VEC_SIZE)
46#define round_op(input) CONVERT(CONVERT_SAT_ROUND(input, int, rte), DATA_TYPE)
47#define logical_not_op(input) ((!input) & 0x1)
48#endif // defined(VEC_SIZE)
49
50/** Applies element wise unary operator in a tensor.
51 *
52 * @param[in]  in_ptr                            Pointer to the source image. Supported data types: F16/32.
53 * @param[in]  in_stride_x                       Stride of the source tensor in X dimension (in bytes)
54 * @param[in]  in_step_x                         in_stride_x * number of elements along X processed per workitem(in bytes)
55 * @param[in]  in_stride_y                       Stride of the source tensor in Y dimension (in bytes)
56 * @param[in]  in_step_y                         in_stride_y * number of elements along Y processed per workitem(in bytes)
57 * @param[in]  in_stride_z                       Stride of the source tensor in Z dimension (in bytes)
58 * @param[in]  in_step_z                         in_stride_z * number of elements along Z processed per workitem(in bytes)
59 * @param[in]  in_offset_first_element_in_bytes  Offset of the first element in the source image
60 * @param[out] out_ptr                           Pointer to the destination image. Supported data types: F16/32.
61 * @param[in]  out_stride_x                      Stride of the destination image in X dimension (in bytes)
62 * @param[in]  out_step_x                        out_stride_x * number of elements along X processed per workitem(in bytes)
63 * @param[in]  out_step_y                        Stride of the destination tensor in Y dimension (in bytes)
64 * @param[in]  out_step_y                        out_stride_y * number of elements along Y processed per workitem(in bytes)
65 * @param[in]  out_stride_z                      Stride of the destination tensor in Z dimension (in bytes)
66 * @param[in]  out_step_z                        out_stride_z * number of elements along Z processed per workitem(in bytes)
67 * @param[in]  out_offset_first_element_in_bytes Offset of the first element in the destination image
68 */
69__kernel void elementwise_unary(
70    TENSOR3D_DECLARATION(in),
71    TENSOR3D_DECLARATION(out))
72{
73    Tensor3D in  = CONVERT_TO_TENSOR3D_STRUCT(in);
74    Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(out);
75
76#if defined(VEC_SIZE) && defined(LAST_ACCESSED_X)
77    // Check if access on width gets out of bounds
78    // If it does shift access vector to access elements within bounds
79    const int xi = (int)(get_global_id(0) * VEC_SIZE);
80    in.ptr -= max(xi - (int)LAST_ACCESSED_X, 0) * in_stride_x;
81    out.ptr -= max(xi - (int)LAST_ACCESSED_X, 0) * out_stride_x;
82
83    VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
84    data = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)in.ptr);
85
86    VSTORE(VEC_SIZE)
87    (OPERATION(data), 0, (__global DATA_TYPE *)out.ptr);
88#else  // !defined(VEC_SIZE) || !defined(LAST_ACCESSED_X)
89    *((__global DATA_TYPE *)(out.ptr)) = (DATA_TYPE)(OPERATION(*((__global DATA_TYPE *)in.ptr)));
90#endif // defined(VEC_SIZE) && defined(LAST_ACCESSED_X)
91}
92#endif // defined(DATA_TYPE) && defined(OPERATION)
93