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
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
|
/* ************************************************************************
* Copyright 2013 Advanced Micro Devices, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
* ************************************************************************/
// $Id
#ifndef CLBLAS_BENCHMARK_XHEMV_HXX__
#define CLBLAS_BENCHMARK_XHEMV_HXX__
#include "clfunc_common.hpp"
template <typename T>
struct xHemvBuffer
{
clblasOrder order;
clblasUplo uplo;
size_t N;
T alpha;
T* cpuX;
cl_mem X;
size_t offx;
int incx;
T beta;
T* cpuY;
cl_mem Y;
size_t offy;
int incy;
T* cpuA;
cl_mem A;
size_t offa;
size_t lda;
}; // struct buffer
template <typename T>
class xHemv : public clblasFunc
{
public:
xHemv(StatisticalTimer& timer, cl_device_type devType) : clblasFunc(timer, devType)
{
timer.getUniqueID("clHemv", 0);
}
~xHemv()
{
delete buffer.cpuA;
delete buffer.cpuX;
OPENCL_V_THROW( clReleaseMemObject(buffer.A), "releasing buffer A");
OPENCL_V_THROW( clReleaseMemObject(buffer.X), "releasing buffer C");
}
double gflops()
{
return static_cast<double>((2 * buffer.N * buffer.N)/time_in_ns());
}
std::string gflops_formula()
{
return "2*N*N/time";
}
void setup_buffer(int order_option, int side_option, int
uplo_option, int diag_option, int transA_option, int
transB_option, size_t M, size_t N, size_t K,
size_t lda, size_t ldb, size_t ldc,size_t offA,
size_t offB, size_t offC, double alpha,
double beta);
void initialize_cpu_buffer();
void initialize_gpu_buffer();
void reset_gpu_write_buffer();
void call_func();
void read_gpu_buffer()
{
//cl_int err;
//to-do need to fill up
}
void roundtrip_func()
{//to-do need to fill up
}
void roundtrip_setup_buffer(int order_option, int side_option, int uplo_option,
int diag_option, int transA_option, int transB_option,
size_t M, size_t N, size_t K, size_t lda, size_t ldb,
size_t ldc, size_t offA, size_t offBX, size_t offCY,
double alpha, double beta)
{}
void releaseGPUBuffer_deleteCPUBuffer()
{
//this is necessary since we are running a iteration of tests and calculate the average time. (in client.cpp)
//need to do this before we eventually hit the destructor
//to do
}
protected:
protected:
void initialize_scalars(double alpha, double beta)
{
buffer.alpha = makeScalar<T>(alpha);
buffer.beta = makeScalar<T>(beta);
}
private:
xHemvBuffer<T> buffer;
};
template <typename T>
void xHemv<T>::setup_buffer(int order_option, int side_option, int
uplo_option, int diag_option, int transA_option, int
transB_option, size_t M, size_t N, size_t K,
size_t lda, size_t ldb, size_t ldc,size_t offA,
size_t offB, size_t offC, double alpha,
double beta)
{
initialize_scalars(alpha, beta);
buffer.offa = offA;
buffer.offx = offB;
buffer.incx = 1;//If this changes, remember to adjust size of Y in rest of the file
buffer.offy = offC;
buffer.incy = 1;//If this changes, remember to adjust size of Y in rest of the file
buffer.N = M;
if (order_option == 0)
{
buffer.order = clblasRowMajor;
}
else
{
buffer.order = clblasColumnMajor;
}
if (uplo_option == 0)
{
buffer.uplo = clblasUpper;
}
else
{
buffer.uplo = clblasLower;
}
if (lda == 0)
{
buffer.lda = buffer.N;
}
else if (lda < buffer.N)
{
std::cerr << "lda:wrong size\n";
exit(1);
}
else
{
buffer.lda = lda;
}
buffer.cpuX = new T[buffer.N];
buffer.cpuY = new T[buffer.N];
buffer.cpuA = new T[buffer.N * buffer.lda];
cl_int err;
buffer.A = clCreateBuffer(ctx_, CL_MEM_READ_ONLY,
buffer.N * buffer.lda*sizeof(T),
NULL, &err);
buffer.X = clCreateBuffer(ctx_, CL_MEM_READ_WRITE,
buffer.N*sizeof(T),
NULL, &err);
buffer.Y = clCreateBuffer(ctx_, CL_MEM_READ_WRITE,
buffer.N*sizeof(T),
NULL, &err);
}
template <typename T>
void xHemv<T>::initialize_cpu_buffer()
{
srand(10);
for (size_t i = 0; i < buffer.N; ++i)
{
for (size_t j = 0; j < buffer.lda; ++j)
{
buffer.cpuA[i*buffer.lda+j] = random<T>(UPPER_BOUND<T>()) /
randomScale<T>();
}
}
for (size_t i = 0; i < buffer.N; ++i)
{
buffer.cpuX[i] = random<T>(UPPER_BOUND<T>()) /
randomScale<T>();
buffer.cpuY[i] = random<T>(UPPER_BOUND<T>()) /
randomScale<T>();
}
}
template <typename T>
void xHemv<T>::initialize_gpu_buffer()
{
cl_int err;
err = clEnqueueWriteBuffer(queue_, buffer.A, CL_TRUE,
buffer.offa * sizeof(T),
buffer.N * buffer.lda*sizeof(T),
buffer.cpuA, 0, NULL, NULL);
err = clEnqueueWriteBuffer(queue_, buffer.X, CL_TRUE, 0,
buffer.N*sizeof(T),
buffer.cpuX, 0, NULL, NULL);
err = clEnqueueWriteBuffer(queue_, buffer.Y, CL_TRUE, 0,
buffer.N*sizeof(T),
buffer.cpuY, 0, NULL, NULL);
}
template <typename T>
void xHemv<T>::reset_gpu_write_buffer()
{
cl_int err;
err = clEnqueueWriteBuffer(queue_, buffer.A, CL_TRUE,
buffer.offa * sizeof(T),
buffer.N * buffer.lda*sizeof(T),
buffer.cpuA, 0, NULL, NULL);;
}
template <>
void xHemv<cl_float2>::call_func()
{
timer.Start(timer_id);
clblasChemv(buffer.order, buffer.uplo, buffer.N, buffer.alpha, buffer.A,
buffer.offa, buffer.lda, buffer.X, buffer.offx, buffer.incx,
buffer.beta, buffer.Y, buffer.offy, buffer.incy, 1, &queue_, 0, NULL,&event_);
clWaitForEvents(1, &event_);
timer.Stop(timer_id);
}
template <>
void xHemv<cl_double2>::call_func()
{
timer.Start(timer_id);
clblasZhemv(buffer.order, buffer.uplo, buffer.N, buffer.alpha, buffer.A,
buffer.offa, buffer.lda, buffer.X, buffer.offx, buffer.incx,
buffer.beta, buffer.Y, buffer.offy, buffer.incy, 1, &queue_, 0, NULL,&event_);
clWaitForEvents(1, &event_);
timer.Stop(timer_id);
}
template<>
double
xHemv<cl_float2>::
gflops()
{
return static_cast<double>((8 * buffer.N * buffer.N)/time_in_ns());
}
template<>
double
xHemv<cl_double2>::
gflops()
{
return static_cast<double>((8 * buffer.N * buffer.N)/time_in_ns());
}
template<>
std::string
xHemv<cl_float2>::
gflops_formula()
{
return "8*N*N/time";
}
template<>
std::string
xHemv<cl_double2>::
gflops_formula()
{
return "8*N*N/time";
}
#endif // ifndef CLBLAS_BENCHMARK_XSYR_HXX__
|
