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
|
// =================================================================================================
// This file is part of the CLBlast project. The project is licensed under the MIT license. This
// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
// width of 100 characters per line.
//
// Author(s):
// Cedric Nugteren <www.cedricnugteren.nl>
//
// This file implements a wrapper around the clBLAS library, such that its routines can be called
// in a similar way as the CLBlast routines: using alpha and beta to determine the precision.
//
// =================================================================================================
#ifndef CLBLAST_TEST_WRAPPER_CLBLAS_H_
#define CLBLAST_TEST_WRAPPER_CLBLAS_H_
#include <clBLAS.h>
#include "internal/utilities.h"
namespace clblast {
// =================================================================================================
// BLAS level-1 (vector-vector) routines
// Calls {clblasSaxpy, clblasDaxpy, clblasCaxpy, clblasZaxpy} with the arguments forwarded.
clblasStatus clblasXaxpy(
size_t n, float alpha,
const cl_mem x_vec, size_t x_offset, size_t x_inc,
const cl_mem y_vec, size_t y_offset, size_t y_inc,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasSaxpy(n, alpha,
x_vec, x_offset, static_cast<int>(x_inc),
y_vec, y_offset, static_cast<int>(y_inc),
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXaxpy(
size_t n, double alpha,
const cl_mem x_vec, size_t x_offset, size_t x_inc,
const cl_mem y_vec, size_t y_offset, size_t y_inc,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDaxpy(n, alpha,
x_vec, x_offset, static_cast<int>(x_inc),
y_vec, y_offset, static_cast<int>(y_inc),
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXaxpy(
size_t n, float2 alpha,
const cl_mem x_vec, size_t x_offset, size_t x_inc,
const cl_mem y_vec, size_t y_offset, size_t y_inc,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
return clblasCaxpy(n, cl_alpha,
x_vec, x_offset, static_cast<int>(x_inc),
y_vec, y_offset, static_cast<int>(y_inc),
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXaxpy(
size_t n, double2 alpha,
const cl_mem x_vec, size_t x_offset, size_t x_inc,
const cl_mem y_vec, size_t y_offset, size_t y_inc,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
return clblasZaxpy(n, cl_alpha,
x_vec, x_offset, static_cast<int>(x_inc),
y_vec, y_offset, static_cast<int>(y_inc),
num_queues, queues, num_wait_events, wait_events, events);
}
// =================================================================================================
// BLAS level-2 (matrix-vector) routines
// =================================================================================================
// BLAS level-3 (matrix-matrix) routines
// This calls {clblasSgemm, clblasDgemm, clblasCgemm, clblasZgemm} with the arguments forwarded.
clblasStatus clblasXgemm(
clblasOrder layout, clblasTranspose tran_a, clblasTranspose tran_b,
size_t m, size_t n, size_t k, float alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, float beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasSgemm(layout, tran_a, tran_b,
m, n, k, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemm(
clblasOrder layout, clblasTranspose tran_a, clblasTranspose tran_b,
size_t m, size_t n, size_t k, double alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, double beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDgemm(layout, tran_a, tran_b,
m, n, k, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemm(
clblasOrder layout, clblasTranspose tran_a, clblasTranspose tran_b,
size_t m, size_t n, size_t k, float2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, float2 beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_float2{{beta.real(), beta.imag()}};
return clblasCgemm(layout, tran_a, tran_b,
m, n, k, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, cl_beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemm(
clblasOrder layout, clblasTranspose tran_a, clblasTranspose tran_b,
size_t m, size_t n, size_t k, double2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, double2 beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_double2{{beta.real(), beta.imag()}};
return clblasZgemm(layout, tran_a, tran_b,
m, n, k, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, cl_beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
// This calls {clblasSsymm, clblasDsymm} with the arguments forwarded.
clblasStatus clblasXsymm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
size_t m, size_t n, float alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, float beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasSsymm(layout, side, triangle,
m, n, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsymm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
size_t m, size_t n, double alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, double beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDsymm(layout, side, triangle,
m, n, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsymm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
size_t m, size_t n, float2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, float2 beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_float2{{beta.real(), beta.imag()}};
return clblasCsymm(layout, side, triangle,
m, n, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, cl_beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsymm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
size_t m, size_t n, double2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, double2 beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_double2{{beta.real(), beta.imag()}};
return clblasZsymm(layout, side, triangle,
m, n, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, cl_beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
// =================================================================================================
} // namespace clblast
// CLBLAST_TEST_WRAPPER_CLBLAS_H_
#endif
|
