Method clblasPerfSingle

src/tests/performance/perf-hpr2.cpp:260–332

template <typename ElemType> nano_time_t
Hpr2PerformanceTest<ElemType>::clblasPerfSingle(void)
{
    nano_time_t time;
    cl_event event;
    cl_int status;
    cl_command_queue queue = base_->commandQueues()[0];

    status = clEnqueueWriteBuffer(queue, mobjAP_, CL_TRUE, 0,
                                  (((params_.N * (params_.N + 1))/2) + params_.offa) *
                                    sizeof(ElemType), backAP_, 0, NULL, &event);
    if (status != CL_SUCCESS) {
        cerr << "Matrix A buffer object enqueuing error, status = " <<
                 status << endl;

        return NANOTIME_ERR;
    }

    status = clWaitForEvents(1, &event);
    if (status != CL_SUCCESS) {
        cout << "Wait on event failed, status = " <<
                status << endl;

        return NANOTIME_ERR;
    }

    event = NULL;

#define TIMING
#ifdef TIMING
    clFinish( queue);
    time = getCurrentTime();

    int iter = 20;
    for ( int i = 1; i <= iter; i++)
    {
#endif
    status = (cl_int)clMath::clblas::hpr2(params_.order, params_.uplo, params_.N, alpha_, mobjX_, params_.offBX, params_.incx,
				mobjY_, params_.offCY, params_.incy, mobjAP_, params_.offa, 1, &queue, 0, NULL, &event);

    if (status != CL_SUCCESS) {
        cerr << "The CLBLAS HPR2 function failed, status = " <<
                status << endl;

        return NANOTIME_ERR;
    }

#ifdef TIMING
    } // iter loop
    clFinish( queue);
    time = getCurrentTime() - time;
    time /= iter;
#else
    status = flushAll(1, &queue);
    if (status != CL_SUCCESS) {
        cerr << "clFlush() failed, status = " << status << endl;
        return NANOTIME_ERR;
    }