void
xSyrk<float>::roundtrip_func()
{
timer.Start(timer_id);
cl_int err;
buffer_.buf_a_ = clCreateBuffer(ctx_, CL_MEM_READ_ONLY,
(buffer_.lda_ * buffer_.a_num_vectors_ +
buffer_.offA_) * sizeof(float),
NULL, &err);
buffer_.buf_c_ = clCreateBuffer(ctx_, CL_MEM_READ_WRITE,
(buffer_.ldc_ * buffer_.c_num_vectors_ +
buffer_.offC_) * sizeof(float),
NULL, &err);
this->initialize_gpu_buffer();
clblasSsyrk(order_, buffer_.uplo_, buffer_.trans_a_, buffer_.n_,
buffer_.k_, buffer_.alpha_, buffer_.buf_a_, buffer_.offA_,
buffer_.lda_, buffer_.beta_, buffer_.buf_c_, buffer_.offC_,
buffer_.ldc_, 1, &queue_, 0, NULL, NULL);
err = clEnqueueReadBuffer(queue_, buffer_.buf_c_, CL_TRUE,
buffer_.offC_*sizeof(float), buffer_.ldc_*buffer_.c_num_vectors_*sizeof(float),
buffer_.c_, 0, NULL, &event_);
clWaitForEvents(1, &event_);
timer.Stop(timer_id);
}
void
xSyrk<float>::
call_func()
{
timer.Start(timer_id);
clblasSsyrk(order_, buffer_.uplo_, buffer_.trans_a_, buffer_.n_,
buffer_.k_, buffer_.alpha_, buffer_.buf_a_, buffer_.offA_,
buffer_.lda_, buffer_.beta_, buffer_.buf_c_, buffer_.offC_,
buffer_.ldc_, 1, &queue_, 0, NULL, &event_);
clWaitForEvents(1, &event_);
timer.Stop(timer_id);
}
ErrorStatus crossprod_clblas(cl_device_id device, void *inMatrix, void *outMatrix, int nrow, int ncol, bool use_float)
{
std::stringstream result;
float *input_matrix_f = (float *)inMatrix;
float *output_matrix_f = (float *)outMatrix;
double *input_matrix_d = (double *)inMatrix;
double *output_matrix_d = (double *)outMatrix;
if (debug) {
result << "crossprod_clblas( " << (use_float ? "FLOAT" : "DOUBLE") <<
", nrow = " << nrow << ", ncol = " << ncol << ")" << std::endl << std::endl;
}
cl_int err = CL_SUCCESS;
clblasStatus status = clblasSetup();
if (status != CL_SUCCESS) {
if (debug) {
result << "clblasSetup: " << clblasErrorToString(status) << std::endl;
}
err = CL_INVALID_OPERATION;
}
// get first platform
cl_platform_id platform = NULL;
if (err == CL_SUCCESS) {
err = clGetPlatformIDs(1, &platform, NULL);
}
if (debug && err == CL_SUCCESS) {
result << "Platform: " << getPlatformInfoString(platform, CL_PLATFORM_NAME) << std::endl;
result << "Device: " << getDeviceInfoString(device, CL_DEVICE_NAME) << std::endl;
}
// context
cl_context context = NULL;
if (err == CL_SUCCESS) {
if (debug) {
result << "clCreateContext:" << std::endl;
}
context = clCreateContext(NULL, 1, &device, NULL, NULL, &err);
}
// queue
cl_command_queue queue = NULL;
if (err == CL_SUCCESS) {
#ifdef CL_VERSION_2_0
if (debug) {
result << "clCreateCommandQueueWithProperties:" << std::endl;
}
queue = clCreateCommandQueueWithProperties(context, device, NULL, &err);
#else
if (debug) {
result << "clCreateCommandQueue:" << std::endl;
}
queue = clCreateCommandQueue(context, device, 0, &err);
#endif
}
// buffers
cl_mem cl_input_matrix = NULL;
if (err == CL_SUCCESS) {
if (debug) {
result << "clCreateBuffer cl_input_matrix:" << std::endl;
}
if (use_float) {
cl_input_matrix = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR,
nrow * ncol * sizeof(float), input_matrix_f, &err);
} else {
cl_input_matrix = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR,
nrow * ncol * sizeof(double), input_matrix_d, &err);
}
}
cl_mem cl_output_matrix = NULL;
if (err == CL_SUCCESS) {
if (debug) {
result << "clCreateBuffer cl_output_vector:" << std::endl;
}
if (use_float) {
cl_output_matrix = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_COPY_HOST_PTR,
ncol * ncol * sizeof(float), output_matrix_f, &err);
} else {
cl_output_matrix = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_COPY_HOST_PTR,
ncol * ncol * sizeof(double), output_matrix_d, &err);
}
}
// ++++++++++++
const clblasOrder order = clblasColumnMajor;
const clblasTranspose transA = clblasTrans;
const size_t lda = nrow;
const size_t ldc = ncol;
const cl_float alpha = 1.0;
clblasUplo uplo = clblasUpper;
cl_event event = NULL;
if (err == CL_SUCCESS) {
if (use_float) {
if (debug) {
result << "clblasSsyrk:" << std::endl;
}
status = clblasSsyrk(order, uplo, transA, ncol, nrow, alpha, cl_input_matrix, 0, lda, 0.0,
cl_output_matrix, 0, ldc, 1, &queue, 0, NULL, &event);
if (status != CL_SUCCESS && debug) {
result << "clblasSgemm error:" << clblasErrorToString(status) << std::endl;
}
} else {
if (debug) {
result << "clblasDsyrk:" << std::endl;
}
status = clblasDsyrk(order, uplo, transA, ncol, nrow, alpha, cl_input_matrix, 0, lda, 0.0,
cl_output_matrix, 0, ldc, 1, &queue, 0, NULL, &event);
if (status != CL_SUCCESS) {
if (debug) {
result << "clblasDgemm error:" << clblasErrorToString(status) << std::endl;
}
err = status;
}
}
}
if (err == CL_SUCCESS) {
/* Wait for calculations to be finished. */
if (debug) {
result << "clWaitForEvents:" << std::endl;
}
err = clWaitForEvents(1, &event);
}
// retrieve result
if (err == CL_SUCCESS) {
if (debug) {
result << "Retrieve result:" << std::endl;
}
if (use_float) {
clEnqueueReadBuffer(queue, cl_output_matrix, CL_TRUE, 0, ncol * ncol * sizeof(float), output_matrix_f, 0, NULL, NULL);
symmetrizeSquare_f(output_matrix_f, ncol);
} else {
clEnqueueReadBuffer(queue, cl_output_matrix, CL_TRUE, 0, ncol * ncol * sizeof(double), output_matrix_d, 0, NULL, NULL);
symmetrizeSquare_d(output_matrix_d, ncol);
}
}
std::string err_str = clErrorToString(err);
result << std::endl << err_str << std::endl;
// cleanup
clReleaseMemObject(cl_output_matrix);
cl_output_matrix = NULL;
clReleaseMemObject(cl_input_matrix);
cl_input_matrix = NULL;
clReleaseCommandQueue(queue);
queue = NULL;
clReleaseContext(context);
context = NULL;
if (debug) {
CERR << result.str();
}
ErrorStatus errorStatus = { err, status };
// return status != CL_SUCCESS ? clblasErrorToString(status) : clErrorToString(err);
return errorStatus;
}
int
main(void)
{
cl_int err;
cl_platform_id platform = 0;
cl_device_id device = 0;
cl_context_properties props[3] = { CL_CONTEXT_PLATFORM, 0, 0 };
cl_context ctx = 0;
cl_command_queue queue = 0;
cl_mem bufA, bufC;
cl_event event = NULL;
int ret = 0;
/* Setup OpenCL environment. */
err = clGetPlatformIDs(1, &platform, NULL);
if (err != CL_SUCCESS) {
printf( "clGetPlatformIDs() failed with %d\n", err );
return 1;
}
err = clGetDeviceIDs(platform, CL_DEVICE_TYPE_GPU, 1, &device, NULL);
if (err != CL_SUCCESS) {
printf( "clGetDeviceIDs() failed with %d\n", err );
return 1;
}
props[1] = (cl_context_properties)platform;
ctx = clCreateContext(props, 1, &device, NULL, NULL, &err);
if (err != CL_SUCCESS) {
printf( "clCreateContext() failed with %d\n", err );
return 1;
}
queue = clCreateCommandQueue(ctx, device, 0, &err);
if (err != CL_SUCCESS) {
printf( "clCreateCommandQueue() failed with %d\n", err );
clReleaseContext(ctx);
return 1;
}
/* Setup clblas. */
err = clblasSetup();
if (err != CL_SUCCESS) {
printf("clblasSetup() failed with %d\n", err);
clReleaseCommandQueue(queue);
clReleaseContext(ctx);
return 1;
}
/* Prepare OpenCL memory objects and place matrices inside them. */
bufA = clCreateBuffer(ctx, CL_MEM_READ_ONLY, N * K * sizeof(*A),
NULL, &err);
bufC = clCreateBuffer(ctx, CL_MEM_READ_WRITE, N * N * sizeof(*C),
NULL, &err);
err = clEnqueueWriteBuffer(queue, bufA, CL_TRUE, 0,
N * K * sizeof(*A), A, 0, NULL, NULL);
err = clEnqueueWriteBuffer(queue, bufC, CL_TRUE, 0,
N * N * sizeof(*C), C, 0, NULL, NULL);
/* Call clblas extended function. Perform SYRK for the lower right sub-matrices */
err = clblasSsyrk(order, uplo, transA, N - off, K - off,
alpha, bufA, offA, lda, beta, bufC, offC, ldc,
1, &queue, 0, NULL, &event);
if (err != CL_SUCCESS) {
printf("clblasSsyrkEx() failed with %d\n", err);
ret = 1;
}
else {
/* Wait for calculations to be finished. */
err = clWaitForEvents(1, &event);
/* Fetch results of calculations from GPU memory. */
err = clEnqueueReadBuffer(queue, bufC, CL_TRUE, 0,
N * N * sizeof(*result),
result, 0, NULL, NULL);
/* At this point you will get the result of SSYRK placed in 'result' array. */
puts("");
printResult("clblasSsyrkEx result");
}
/* Release OpenCL memory objects. */
clReleaseMemObject(bufC);
clReleaseMemObject(bufA);
/* Finalize work with clblas. */
clblasTeardown();
/* Release OpenCL working objects. */
clReleaseCommandQueue(queue);
clReleaseContext(ctx);
return ret;
}
