xref: /aosp_15_r20/external/ComputeLibrary/src/core/CL/cl_kernels/common/floor.cl (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1/*
2 * Copyright (c) 2017-2021 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(VEC_SIZE) && defined(VEC_SIZE_LEFTOVER)
27
28/** Perform a floor operation on an input tensor.
29 *
30 * @note Data type can be passed using the -DDATA_TYPE compile flag, e.g. -DDATA_TYPE=float
31 * @note Vector size should be given as a preprocessor argument using -DVEC_SIZE=size. e.g. -DVEC_SIZE=16
32 * @note The leftover size in the X dimension shoud be given as preprocessor argument using -DVEC_SIZE_LEFTOVER is; x_dimension % VEC_SIZE. e.g. -DVEC_SIZE_LEFTOVER=1
33 * @note Can only take floating point data types.
34 *
35 * @param[in]  input_ptr                            Pointer to the source image. Supported data types: F16/F32
36 * @param[in]  input_stride_x                       Stride of the source image in X dimension (in bytes)
37 * @param[in]  input_step_x                         input_stride_x * number of elements along X processed per workitem(in bytes)
38 * @param[in]  input_stride_y                       Stride of the source image in Y dimension (in bytes)
39 * @param[in]  input_step_y                         input_stride_y * number of elements along Y processed per workitem(in bytes)
40 * @param[in]  input_stride_z                       Stride of the source tensor in Z dimension (in bytes)
41 * @param[in]  input_step_z                         input_stride_z * number of elements along Z processed per workitem(in bytes)
42 * @param[in]  input_offset_first_element_in_bytes  The offset of the first element in the source image
43 * @param[out] output_ptr                           Pointer to the destination image. Supported data types: same as @p input_ptr
44 * @param[in]  output_stride_x                      Stride of the destination image in X dimension (in bytes)
45 * @param[in]  output_step_x                        output_stride_x * number of elements along X processed per workitem(in bytes)
46 * @param[in]  output_stride_y                      Stride of the destination image in Y dimension (in bytes)
47 * @param[in]  output_step_y                        output_stride_y * number of elements along Y processed per workitem(in bytes)
48 * @param[in]  output_stride_z                      Stride of the source tensor in Z dimension (in bytes)
49 * @param[in]  output_step_z                        output_stride_z * number of elements along Z processed per workitem(in bytes)
50 * @param[in]  output_offset_first_element_in_bytes The offset of the first element in the destination image
51 */
52__kernel void floor_layer(
53    TENSOR3D_DECLARATION(input),
54    TENSOR3D_DECLARATION(output))
55{
56    // Offset computation
57    const uint x_offs = max((int)(get_global_id(0) * VEC_SIZE - (VEC_SIZE - VEC_SIZE_LEFTOVER) % VEC_SIZE), 0);
58
59    // Address computation
60    __global uchar *input_addr  = input_ptr + input_offset_first_element_in_bytes + x_offs * sizeof(DATA_TYPE) + get_global_id(1) * input_stride_y + get_global_id(2) * input_stride_z;
61    __global uchar *output_addr = output_ptr + output_offset_first_element_in_bytes + x_offs * sizeof(DATA_TYPE) + get_global_id(1) * output_stride_y + get_global_id(2) * output_stride_z;
62
63    VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
64    data0 = floor(VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)input_addr));
65
66    STORE_VECTOR_SELECT(data, DATA_TYPE, output_addr, VEC_SIZE, VEC_SIZE_LEFTOVER, VEC_SIZE_LEFTOVER != 0 && get_global_id(0) == 0);
67}
68#endif // defined(VEC_SIZE) && defined(VEC_SIZE_LEFTOVER)