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
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
|
/*
* Copyright (c) 2019-2020 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.
*/
#ifndef ARM_COMPUTE_CL /* Needed by Utils.cpp to handle OpenCL exceptions properly */
#error "This example needs to be built with -DARM_COMPUTE_CL"
#endif /* ARM_COMPUTE_CL */
#include "CommonGemmExampleOptions.h"
#include "arm_compute/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.h"
#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/KernelDescriptors.h"
#include "arm_compute/core/Types.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "arm_compute/runtime/CL/CLFunctions.h"
#include "arm_compute/runtime/CL/CLScheduler.h"
#include "arm_compute/runtime/CL/CLTuner.h"
#include "tests/CL/Helper.h"
#include "utils/Utils.h"
#include "utils/command_line/CommandLineOptions.h"
#include "utils/command_line/CommandLineParser.h"
#include <cstdlib>
using namespace arm_compute;
using namespace utils;
using namespace arm_compute::misc::shape_calculator;
using namespace gemm_tuner;
namespace
{
/** Structure holding all tunable gemm configs specific to this example/strategy */
struct GemmConfigs
{
size_t m0{ 4 }; /**< Number of rows processed by the matrix multiplication */
size_t n0{ 4 }; /**< Number of columns processed by the matrix multiplication */
size_t k0{ 4 }; /**< Number of partial accumulations performed by the matrix multiplication */
};
/** Formatted output of the GemmConfigs type
*
* @param[out] os Output stream.
* @param[in] configs Tunable configurations to output
*
* @return Modified output stream.
*/
::std::ostream &operator<<(::std::ostream &os, const GemmConfigs &configs)
{
std::string false_str = std::string("false");
std::string true_str = std::string("true");
os << "m0 : " << configs.m0 << std::endl;
os << "n0 : " << configs.n0 << std::endl;
os << "k0 : " << configs.k0 << std::endl;
return os;
}
/** Command line options for gemm configs */
class GemmConfigOptions
{
public:
/** Constructor
*
* @param[in,out] parser A parser on which "parse()" hasn't been called yet.
*/
GemmConfigOptions(CommandLineParser &parser)
: m0(parser.add_positional_option<SimpleOption<size_t>>("m0", 4)),
n0(parser.add_positional_option<SimpleOption<size_t>>("n0", 4)),
k0(parser.add_positional_option<SimpleOption<size_t>>("k0", 4))
{
m0->set_help("Number of rows processed by the matrix multiplication");
n0->set_help("Number of columns processed by the matrix multiplication");
k0->set_help("Number of partial accumulations performed by the matrix multiplication");
}
/** Prevent instances of this class from being copied (As this class contains pointers) */
GemmConfigOptions(const GemmConfigOptions &) = delete;
/** Prevent instances of this class from being copied (As this class contains pointers) */
GemmConfigOptions &operator=(const GemmConfigOptions &) = delete;
/** Allow instances of this class to be moved */
GemmConfigOptions(GemmConfigOptions &&) = default;
/** Allow instances of this class to be moved */
GemmConfigOptions &operator=(GemmConfigOptions &&) = default;
/** Default destructor */
~GemmConfigOptions() = default;
SimpleOption<size_t> *m0; /**< Number of rows processed by the matrix multiplication option */
SimpleOption<size_t> *n0; /**< Number of columns processed by the matrix multiplication option */
SimpleOption<size_t> *k0; /**< Number of partial accumulations performed by the matrix multiplication option */
};
/** Consumes the gemm configuration options and creates a structure containing all information
*
* @param[in] options Options to consume
*
* @return Structure containing the gemm configurations
*/
GemmConfigs consume_gemm_configs(const GemmConfigOptions &options)
{
GemmConfigs configs;
configs.m0 = options.m0->value();
configs.n0 = options.n0->value();
configs.k0 = options.k0->value();
return configs;
}
} // namespace
// Create function for CLGEMMMatrixMultiplyNativeKernel
using CLGEMMMatrixMultiplyNative = test::CLSynthetizeFunction<CLGEMMMatrixMultiplyNativeKernel>;
class CLGEMMMatrixMultiplyNativeExample : public Example
{
public:
bool do_setup(int argc, char **argv) override
{
// Default parameters
const float alpha = 1.0f;
const float beta = 0.0f;
const ActivationLayerInfo act_info = ActivationLayerInfo();
CommonGemmExampleParams params;
GemmConfigs configs;
// Set up command line parser and options
CommandLineParser parser;
CommonGemmExampleOptions param_options(parser);
GemmConfigOptions config_options(parser);
// Parse command line options
parser.parse(argc, argv);
if(param_options.help->is_set() && param_options.help->value())
{
// Print help message
parser.print_help(argv[0]);
return false;
}
if(!parser.validate())
{
// Invalid arguments. Use default parameters and configs
std::cerr << "Invalid arguments." << std::endl;
parser.print_help(argv[0]);
std::cerr << "Falling back to default parameters and configs" << std::endl;
}
else
{
// Get parameters and configs from command-line options
params = consume_common_gemm_example_parameters(param_options);
configs = consume_gemm_configs(config_options);
}
// Print gemm parameters and configurations
std::cerr << "Gemm parameters:" << std::endl;
std::cerr << params << std::endl;
std::cerr << "Gemm configurations:" << std::endl;
std::cerr << configs << std::endl;
CLScheduler::get().default_init(&tuner);
lhs.allocator()->init(TensorInfo(TensorShape(params.K, params.M, params.B), 1, params.data_type));
rhs.allocator()->init(TensorInfo(TensorShape(params.N, params.K, params.B), 1, params.data_type));
bias.allocator()->init(TensorInfo(TensorShape(params.N, 1, params.B), 1, params.data_type));
GEMMLHSMatrixInfo lhs_info;
lhs_info.m0 = configs.m0;
lhs_info.k0 = configs.k0;
GEMMRHSMatrixInfo rhs_info;
rhs_info.n0 = configs.n0;
rhs_info.k0 = configs.k0;
GEMMKernelInfo kernel_info;
kernel_info.m = params.M;
kernel_info.n = params.N;
kernel_info.k = params.K;
kernel_info.depth_output_gemm3d = 0;
kernel_info.reinterpret_input_as_3d = false;
kernel_info.broadcast_bias = true;
kernel_info.activation_info = act_info;
// Configure function
gemm.configure(&lhs, &rhs, &bias, &dst, alpha, beta, lhs_info, rhs_info, kernel_info);
// Allocate tensors
lhs.allocator()->allocate();
rhs.allocator()->allocate();
bias.allocator()->allocate();
dst.allocator()->allocate();
return true;
}
void do_run() override
{
// Execute the function
gemm.run();
// Make sure all the OpenCL jobs are done executing:
CLScheduler::get().sync();
}
void do_teardown() override
{
}
private:
CLTensor lhs{};
CLTensor rhs{};
CLTensor bias{};
CLTensor dst{};
CLTuner tuner{};
CLGEMMMatrixMultiplyNative gemm{};
};
/** Main program for gemm native test
*
* @param[in] argc Number of arguments
* @param[in] argv Arguments ( [optional] M, [optional] N, [optional] K, [optional] B, [optional] m0, [optional] n0, [optional] k0 )
*/
int main(int argc, char **argv)
{
return run_example<CLGEMMMatrixMultiplyNativeExample>(argc, argv);
}
|
