Method clblasPerfSingle

src/tests/performance/perf-axpy.cpp:234–307

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

    status = clEnqueueWriteBuffer(queue, mobjX_, CL_TRUE, 0,
                                  (lengthX + params_.offBX) * sizeof(ElemType), X_, 0, NULL, &event);
    status |= clEnqueueWriteBuffer(queue, mobjY_, CL_TRUE, 0,
                                  (lengthY + params_.offCY) * sizeof(ElemType), Y_, 0, NULL, &event);
    if (status != CL_SUCCESS)
    {
        cerr << "mobjX_ or mobjY_ 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;
    time = getCurrentTime();

#define TIMING
#ifdef TIMING
    clFinish( queue);
    int iter = 50;
    for ( int i=1; i <= iter; i++)
    {
#endif

        status = (cl_int)clMath::clblas::axpy(params_.N, alpha_, mobjX_, params_.offBX, params_.incx, mobjY_, params_.offCY, params_.incy,
                            1, &queue, 0, NULL, &event);
        if (status != CL_SUCCESS) {
            cerr << "The CLBLAS AXPY 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;