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
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
|
#include "common.h"
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#define TOL 1e-8
#define MAX_ERRORS 8
const char* KERNEL_SOURCE = "kernel void vecadd(global float *a, \n"
" global float *b, \n"
" global float *c) \n"
"{ \n"
" int i = get_global_id(0); \n"
" c[i] = a[i] + b[i]; \n"
"} \n";
unsigned checkResults(size_t N, float* a, float* b, float* results);
// Run everything as normal
unsigned run1(Context cl, cl_kernel kernel, cl_mem d_a, cl_mem d_b, cl_mem d_c,
size_t N)
{
cl_int err;
float *h_a, *h_b, *h_c;
size_t dataSize = N * sizeof(cl_float);
// Initialise data
srand(0);
h_a =
clEnqueueMapBuffer(cl.queue, d_a, CL_TRUE, CL_MAP_WRITE_INVALIDATE_REGION,
0, dataSize, 0, NULL, NULL, &err);
checkError(err, "mapping d_a buffer");
h_b =
clEnqueueMapBuffer(cl.queue, d_b, CL_TRUE, CL_MAP_WRITE_INVALIDATE_REGION,
0, dataSize, 0, NULL, NULL, &err);
checkError(err, "mapping d_b buffer");
h_c = clEnqueueMapBuffer(cl.queue, d_c, CL_FALSE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_c buffer");
for (unsigned i = 0; i < N; i++)
{
h_a[i] = rand() / (float)RAND_MAX;
h_b[i] = rand() / (float)RAND_MAX;
h_c[i] = 0;
}
err = clEnqueueUnmapMemObject(cl.queue, d_a, h_a, 0, NULL, NULL);
checkError(err, "unmapping d_a");
err = clEnqueueUnmapMemObject(cl.queue, d_b, h_b, 0, NULL, NULL);
checkError(err, "unmapping d_b");
err = clEnqueueUnmapMemObject(cl.queue, d_c, h_c, 0, NULL, NULL);
checkError(err, "unmapping d_c");
err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &d_a);
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_b);
err |= clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_c);
checkError(err, "setting kernel args");
err =
clEnqueueNDRangeKernel(cl.queue, kernel, 1, NULL, &N, NULL, 0, NULL, NULL);
checkError(err, "enqueuing kernel");
h_c = clEnqueueMapBuffer(cl.queue, d_c, CL_FALSE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_c buffer");
err = clFinish(cl.queue);
checkError(err, "running kernel");
unsigned errors = checkResults(N, h_a, h_b, h_c);
err = clEnqueueUnmapMemObject(cl.queue, d_c, h_c, 0, NULL, NULL);
checkError(err, "unmapping d_c");
return errors;
}
// Don't unmap input buffers before running kernel
// Should result in "Invalid read from buffer mapped for writing" error
unsigned run2(Context cl, cl_kernel kernel, cl_mem d_a, cl_mem d_b, cl_mem d_c,
size_t N)
{
cl_int err;
float *h_a, *h_b, *h_c;
size_t dataSize = N * sizeof(cl_float);
// Initialise data
srand(0);
h_a =
clEnqueueMapBuffer(cl.queue, d_a, CL_TRUE, CL_MAP_WRITE_INVALIDATE_REGION,
0, dataSize, 0, NULL, NULL, &err);
checkError(err, "mapping d_a buffer");
h_b =
clEnqueueMapBuffer(cl.queue, d_b, CL_TRUE, CL_MAP_WRITE_INVALIDATE_REGION,
0, dataSize, 0, NULL, NULL, &err);
checkError(err, "mapping d_b buffer");
h_c = clEnqueueMapBuffer(cl.queue, d_c, CL_FALSE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_c buffer");
for (unsigned i = 0; i < N; i++)
{
h_a[i] = rand() / (float)RAND_MAX;
h_b[i] = rand() / (float)RAND_MAX;
h_c[i] = 0;
}
err = clEnqueueUnmapMemObject(cl.queue, d_c, h_c, 0, NULL, NULL);
checkError(err, "unmapping d_c");
err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &d_a);
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_b);
err |= clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_c);
checkError(err, "setting kernel args");
err =
clEnqueueNDRangeKernel(cl.queue, kernel, 1, NULL, &N, NULL, 0, NULL, NULL);
checkError(err, "enqueuing kernel");
h_c = clEnqueueMapBuffer(cl.queue, d_c, CL_FALSE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_c buffer");
err = clFinish(cl.queue);
checkError(err, "running kernel");
unsigned errors = checkResults(N, h_a, h_b, h_c);
err = clEnqueueUnmapMemObject(cl.queue, d_a, h_a, 0, NULL, NULL);
checkError(err, "unmapping d_a");
err = clEnqueueUnmapMemObject(cl.queue, d_b, h_b, 0, NULL, NULL);
checkError(err, "unmapping d_b");
err = clEnqueueUnmapMemObject(cl.queue, d_c, h_c, 0, NULL, NULL);
checkError(err, "unmapping d_c");
return errors;
}
// Don't unmap output buffer before running kernel
// Should result in "Invalid write to mapped buffer" error
unsigned run3(Context cl, cl_kernel kernel, cl_mem d_a, cl_mem d_b, cl_mem d_c,
size_t N)
{
cl_int err;
float *h_a, *h_b, *h_c;
size_t dataSize = N * sizeof(cl_float);
// Initialise data
srand(0);
h_a =
clEnqueueMapBuffer(cl.queue, d_a, CL_TRUE, CL_MAP_WRITE_INVALIDATE_REGION,
0, dataSize, 0, NULL, NULL, &err);
checkError(err, "mapping d_a buffer");
h_b =
clEnqueueMapBuffer(cl.queue, d_b, CL_TRUE, CL_MAP_WRITE_INVALIDATE_REGION,
0, dataSize, 0, NULL, NULL, &err);
checkError(err, "mapping d_b buffer");
h_c = clEnqueueMapBuffer(cl.queue, d_c, CL_FALSE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_c buffer");
for (unsigned i = 0; i < N; i++)
{
h_a[i] = rand() / (float)RAND_MAX;
h_b[i] = rand() / (float)RAND_MAX;
h_c[i] = 0;
}
err = clEnqueueUnmapMemObject(cl.queue, d_a, h_a, 0, NULL, NULL);
checkError(err, "unmapping d_a");
err = clEnqueueUnmapMemObject(cl.queue, d_b, h_b, 0, NULL, NULL);
checkError(err, "unmapping d_b");
err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &d_a);
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_b);
err |= clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_c);
checkError(err, "setting kernel args");
err =
clEnqueueNDRangeKernel(cl.queue, kernel, 1, NULL, &N, NULL, 0, NULL, NULL);
checkError(err, "enqueuing kernel");
err = clFinish(cl.queue);
checkError(err, "running kernel");
unsigned errors = checkResults(N, h_a, h_b, h_c);
err = clEnqueueUnmapMemObject(cl.queue, d_c, h_c, 0, NULL, NULL);
checkError(err, "unmapping d_c");
return errors;
}
// Re-map input buffers for reading
// Should not result in any error
unsigned run4(Context cl, cl_kernel kernel, cl_mem d_a, cl_mem d_b, cl_mem d_c,
size_t N)
{
cl_int err;
float *h_a, *h_b, *h_c;
size_t dataSize = N * sizeof(cl_float);
// Initialise data
srand(0);
h_a =
clEnqueueMapBuffer(cl.queue, d_a, CL_TRUE, CL_MAP_WRITE_INVALIDATE_REGION,
0, dataSize, 0, NULL, NULL, &err);
checkError(err, "mapping d_a buffer");
h_b =
clEnqueueMapBuffer(cl.queue, d_b, CL_TRUE, CL_MAP_WRITE_INVALIDATE_REGION,
0, dataSize, 0, NULL, NULL, &err);
checkError(err, "mapping d_b buffer");
h_c = clEnqueueMapBuffer(cl.queue, d_c, CL_FALSE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_c buffer");
for (unsigned i = 0; i < N; i++)
{
h_a[i] = rand() / (float)RAND_MAX;
h_b[i] = rand() / (float)RAND_MAX;
h_c[i] = 0;
}
h_a = clEnqueueMapBuffer(cl.queue, d_a, CL_TRUE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_a buffer");
h_b = clEnqueueMapBuffer(cl.queue, d_b, CL_TRUE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_b buffer");
err = clEnqueueUnmapMemObject(cl.queue, d_a, h_a, 0, NULL, NULL);
checkError(err, "unmapping d_a");
err = clEnqueueUnmapMemObject(cl.queue, d_b, h_b, 0, NULL, NULL);
checkError(err, "unmapping d_b");
err = clEnqueueUnmapMemObject(cl.queue, d_c, h_c, 0, NULL, NULL);
checkError(err, "unmapping d_c");
err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &d_a);
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_b);
err |= clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_c);
checkError(err, "setting kernel args");
err =
clEnqueueNDRangeKernel(cl.queue, kernel, 1, NULL, &N, NULL, 0, NULL, NULL);
checkError(err, "enqueuing kernel");
h_c = clEnqueueMapBuffer(cl.queue, d_c, CL_FALSE, CL_MAP_READ, 0, dataSize, 0,
NULL, NULL, &err);
checkError(err, "mapping d_c buffer");
err = clFinish(cl.queue);
checkError(err, "running kernel");
unsigned errors = checkResults(N, h_a, h_b, h_c);
err = clEnqueueUnmapMemObject(cl.queue, d_c, h_c, 0, NULL, NULL);
checkError(err, "unmapping d_c");
return errors;
}
int main(int argc, char* argv[])
{
cl_int err;
cl_kernel kernel;
cl_mem d_a, d_b, d_c;
size_t N = 1;
if (argc > 1)
{
N = atoi(argv[1]);
}
Context cl = createContext(KERNEL_SOURCE, "");
kernel = clCreateKernel(cl.program, "vecadd", &err);
checkError(err, "creating kernel");
size_t dataSize = N * sizeof(cl_float);
d_a = clCreateBuffer(cl.context, CL_MEM_READ_ONLY, dataSize, NULL, &err);
checkError(err, "creating d_a buffer");
d_b = clCreateBuffer(cl.context, CL_MEM_READ_ONLY, dataSize, NULL, &err);
checkError(err, "creating d_b buffer");
d_c = clCreateBuffer(cl.context, CL_MEM_WRITE_ONLY, dataSize, NULL, &err);
checkError(err, "creating d_c buffer");
unsigned errors = 0;
errors += run1(cl, kernel, d_a, d_b, d_c, N);
errors += run2(cl, kernel, d_a, d_b, d_c, N);
errors += run3(cl, kernel, d_a, d_b, d_c, N);
errors += run4(cl, kernel, d_a, d_b, d_c, N);
clReleaseMemObject(d_a);
clReleaseMemObject(d_b);
clReleaseMemObject(d_c);
clReleaseKernel(kernel);
releaseContext(cl);
return (errors != 0);
}
unsigned checkResults(size_t N, float* a, float* b, float* results)
{
// Check results
unsigned errors = 0;
for (unsigned i = 0; i < N; i++)
{
float ref = a[i] + b[i];
if (fabs(ref - results[i]) > TOL)
{
if (errors < MAX_ERRORS)
{
fprintf(stderr, "%4d: %.4f != %.4f\n", i, results[i], ref);
}
errors++;
}
}
if (errors)
printf("%d errors detected\n", errors);
return errors;
}
|
