| 291 | |
| 292 | |
| 293 | template <typename ElemType> nano_time_t |
| 294 | GercPerformanceTest<ElemType>::clblasPerfSingle(void) |
| 295 | { |
| 296 | nano_time_t time; |
| 297 | cl_event event; |
| 298 | cl_int status; |
| 299 | cl_command_queue queue = base_->commandQueues()[0]; |
| 300 | |
| 301 | status = clEnqueueWriteBuffer(queue, mobjA_, CL_TRUE, 0, |
| 302 | (lengthA + params_.offa) * sizeof(ElemType), backA_, 0, NULL, &event); |
| 303 | if (status != CL_SUCCESS) { |
| 304 | cerr << "Matrix A buffer object enqueuing error, status = " << |
| 305 | status << endl; |
| 306 | |
| 307 | return NANOTIME_ERR; |
| 308 | } |
| 309 | |
| 310 | status = clWaitForEvents(1, &event); |
| 311 | if (status != CL_SUCCESS) { |
| 312 | cout << "Wait on event failed, status = " << |
| 313 | status << endl; |
| 314 | |
| 315 | return NANOTIME_ERR; |
| 316 | } |
| 317 | |
| 318 | event = NULL; |
| 319 | time = getCurrentTime(); |
| 320 | |
| 321 | #define TIMING |
| 322 | #ifdef TIMING |
| 323 | clFinish( queue); |
| 324 | |
| 325 | int iter = 20; |
| 326 | for ( int i = 1; i <= iter; i++) |
| 327 | { |
| 328 | #endif |
| 329 | |
| 330 | status = (cl_int)clMath::clblas::gerc(params_.order, params_.M, params_.N, alpha_, |
| 331 | mobjx_, params_.offBX, params_.incx, mobjy_, params_.offCY, params_.incy, mobjA_, params_.offa, params_.lda, 1, |
| 332 | &queue, 0, NULL, &event); |
| 333 | if (status != CL_SUCCESS) { |
| 334 | cerr << "The CLBLAS GERC function failed, status = " << |
| 335 | status << endl; |
| 336 | |
| 337 | return NANOTIME_ERR; |
| 338 | } |
| 339 | #ifdef TIMING |
| 340 | } // iter loop |
| 341 | clFinish( queue); |
| 342 | time = getCurrentTime() - time; |
| 343 | time /= iter; |
| 344 | #else |
| 345 | |
| 346 | status = flushAll(1, &queue); |
| 347 | if (status != CL_SUCCESS) { |
| 348 | cerr << "clFlush() failed, status = " << status << endl; |
| 349 | return NANOTIME_ERR; |
| 350 | } |
nothing calls this directly
no test coverage detected