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
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
|
/*
* Copyright (c) 2017 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 "arm_compute/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.h"
#include "arm_compute/core/AccessWindowStatic.h"
#include "arm_compute/core/CL/CLHelpers.h"
#include "arm_compute/core/CL/CLKernelLibrary.h"
#include "arm_compute/core/CL/ICLTensor.h"
#include "arm_compute/core/CL/OpenCL.h"
#include "arm_compute/core/Error.h"
#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/Types.h"
#include "arm_compute/core/Utils.h"
#include "arm_compute/core/Validate.h"
using namespace arm_compute;
CLGEMMMatrixAccumulateBiasesKernel::CLGEMMMatrixAccumulateBiasesKernel()
: _accum(nullptr), _biases(nullptr)
{
}
void CLGEMMMatrixAccumulateBiasesKernel::configure(ICLTensor *accum, const ICLTensor *biases)
{
ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(accum, 1, DataType::F16, DataType::F32);
ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(biases, 1, DataType::F16, DataType::F32);
ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(biases, accum);
ARM_COMPUTE_ERROR_ON(biases->info()->num_dimensions() != 1);
_biases = biases;
_accum = accum;
// Create kernel
std::string data_type_name = lower_string(string_from_data_type(accum->info()->data_type()));
_kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("gemm_accumulate_biases_" + data_type_name));
// Configure kernel window
const unsigned int num_elems_processed_per_iteration = max_cl_vector_width / data_size_from_type(accum->info()->data_type());
Window win = calculate_max_window(*_accum->info(), Steps(num_elems_processed_per_iteration));
AccessWindowStatic biases_access(biases->info(), 0, 0, biases->info()->dimension(0), biases->info()->dimension(1));
AccessWindowHorizontal accum_access(_accum->info(), 0, num_elems_processed_per_iteration);
update_window_and_padding(win, biases_access, accum_access);
ICLKernel::configure(win);
}
void CLGEMMMatrixAccumulateBiasesKernel::run(const Window &window, cl::CommandQueue &queue)
{
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
Window accum_slice = window.first_slice_window_2D();
Window biases_slice(accum_slice);
biases_slice.set(Window::DimY, Window::Dimension(0, 1, 1));
// Run kernel
do
{
// Set arguments
unsigned int idx = 0;
add_2D_tensor_argument(idx, _accum, accum_slice);
add_1D_tensor_argument(idx, _biases, biases_slice);
enqueue(queue, *this, accum_slice);
}
while(window.slide_window_slice_2D(accum_slice));
}
|
