aboutsummaryrefslogtreecommitdiff
path: root/src/core/CL/cl_kernels/common/floor.cl
blob: f6dd4edd2e392cfc5438aa514c1c0deb256427a6 (plain)
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
/*
 * Copyright (c) 2017-2021 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(VEC_SIZE) && defined(VEC_SIZE_LEFTOVER)

/** Perform a floor operation on an input tensor.
 *
 * @note Data type can be passed using the -DDATA_TYPE compile flag, e.g. -DDATA_TYPE=float
 * @note Vector size should be given as a preprocessor argument using -DVEC_SIZE=size. e.g. -DVEC_SIZE=16
 * @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
 * @note Can only take floating point data types.
 *
 * @param[in]  input_ptr                            Pointer to the source image. Supported data types: F16/F32
 * @param[in]  input_stride_x                       Stride of the source image in X dimension (in bytes)
 * @param[in]  input_step_x                         input_stride_x * number of elements along X processed per workitem(in bytes)
 * @param[in]  input_stride_y                       Stride of the source image in Y dimension (in bytes)
 * @param[in]  input_step_y                         input_stride_y * number of elements along Y processed per workitem(in bytes)
 * @param[in]  input_stride_z                       Stride of the source tensor in Z dimension (in bytes)
 * @param[in]  input_step_z                         input_stride_z * number of elements along Z processed per workitem(in bytes)
 * @param[in]  input_offset_first_element_in_bytes  The offset of the first element in the source image
 * @param[out] output_ptr                           Pointer to the destination image. Supported data types: same as @p input_ptr
 * @param[in]  output_stride_x                      Stride of the destination image in X dimension (in bytes)
 * @param[in]  output_step_x                        output_stride_x * number of elements along X processed per workitem(in bytes)
 * @param[in]  output_stride_y                      Stride of the destination image in Y dimension (in bytes)
 * @param[in]  output_step_y                        output_stride_y * number of elements along Y processed per workitem(in bytes)
 * @param[in]  output_stride_z                      Stride of the source tensor in Z dimension (in bytes)
 * @param[in]  output_step_z                        output_stride_z * number of elements along Z processed per workitem(in bytes)
 * @param[in]  output_offset_first_element_in_bytes The offset of the first element in the destination image
 */
__kernel void floor_layer(
    TENSOR3D_DECLARATION(input),
    TENSOR3D_DECLARATION(output))
{
    // Offset computation
    const uint x_offs = max((int)(get_global_id(0) * VEC_SIZE - (VEC_SIZE - VEC_SIZE_LEFTOVER) % VEC_SIZE), 0);

    // Address computation
    __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;
    __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;

    VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
    data0 = floor(VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)input_addr));

    STORE_VECTOR_SELECT(data, DATA_TYPE, output_addr, VEC_SIZE, VEC_SIZE_LEFTOVER, VEC_SIZE_LEFTOVER != 0 && get_global_id(0) == 0);
}
#endif // defined(VEC_SIZE) && defined(VEC_SIZE_LEFTOVER)