2 * Copyright (c) 2017-2018 ARM Limited.
4 * SPDX-License-Identifier: MIT
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:
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
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
26 /** This kernel concatenates the input tensor into the output tensor along the first dimension
28 * @param[in] src_ptr Pointer to the source tensor. Supported data types: QS8, QASYMM8, QS16, F16, F32
29 * @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
30 * @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
31 * @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
32 * @param[in] src_step_y src_stride_y * number of elements along Y processed per workitem(in bytes)
33 * @param[in] src_stride_z Stride of the source tensor in Z dimension (in bytes)
34 * @param[in] src_step_z src_stride_z * number of elements along Z processed per workitem(in bytes)
35 * @param[in] src_offset_first_element_in_bytes The offset of the first element in the source tensor
36 * @param[out] dst_ptr Pointer to the destination tensor. Supported data types: same as @p src_ptr
37 * @param[in] dst_stride_x Stride of the destination tensor in X dimension (in bytes)
38 * @param[in] dst_step_x dst_stride_x * number of elements along X processed per workitem(in bytes)
39 * @param[in] dst_stride_y Stride of the destination tensor in Y dimension (in bytes)
40 * @param[in] dst_step_y dst_stride_y * number of elements along Y processed per workitem(in bytes)
41 * @param[in] dst_stride_z Stride of the source tensor in Z dimension (in bytes)
42 * @param[in] dst_step_z dst_stride_z * number of elements along Z processed per workitem(in bytes)
43 * @param[in] dst_offset_first_element_in_bytes The offset of the first element in the destination tensor
44 * @param[in] offset The offset to the first valid element of the output tensor in bytes
46 __kernel void concatenate_width(
47 TENSOR3D_DECLARATION(src),
48 TENSOR3D_DECLARATION(dst),
51 Tensor3D src = CONVERT_TO_TENSOR3D_STRUCT(src);
52 Tensor3D dst = CONVERT_TO_TENSOR3D_STRUCT(dst);
54 VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
55 source_values = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)src.ptr);
58 (source_values, 0, (__global DATA_TYPE *)(dst.ptr + offset));
61 /** This kernel concatenates the input tensor into the output tensor along the third dimension
63 * @param[in] src_ptr Pointer to the source tensor. Supported data types: QS8, QS16, F16, F32
64 * @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
65 * @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
66 * @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
67 * @param[in] src_step_y src_stride_y * number of elements along Y processed per workitem(in bytes)
68 * @param[in] src_stride_z Stride of the source tensor in Z dimension (in bytes)
69 * @param[in] src_step_z src_stride_z * number of elements along Z processed per workitem(in bytes)
70 * @param[in] src_offset_first_element_in_bytes The offset of the first element in the source tensor
71 * @param[out] dst_ptr Pointer to the destination tensor. Supported data types: same as @p src_ptr
72 * @param[in] dst_stride_x Stride of the destination tensor in X dimension (in bytes)
73 * @param[in] dst_step_x dst_stride_x * number of elements along X processed per workitem(in bytes)
74 * @param[in] dst_stride_y Stride of the destination tensor in Y dimension (in bytes)
75 * @param[in] dst_step_y dst_stride_y * number of elements along Y processed per workitem(in bytes)
76 * @param[in] dst_stride_z Stride of the source tensor in Z dimension (in bytes)
77 * @param[in] dst_step_z dst_stride_z * number of elements along Z processed per workitem(in bytes)
78 * @param[in] dst_offset_first_element_in_bytes The offset of the first element in the destination tensor
79 * @param[in] offsets The offsets to the first valid element of the output tensor in bytes
81 __kernel void concatenate_depth(
82 TENSOR3D_DECLARATION(src),
83 TENSOR3D_DECLARATION(dst),
86 Tensor3D src = CONVERT_TO_TENSOR3D_STRUCT(src);
87 Tensor3D dst = CONVERT_TO_TENSOR3D_STRUCT(dst);
89 VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE)
90 source_values = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)tensor3D_offset(&src, -offsets.x, -offsets.y, 0));
93 (source_values, 0, (__global DATA_TYPE *)(dst.ptr + offsets.z));