int main(void)
{
int nx = NX;
int ny = NY;
POLYBENCH_2D_ARRAY_DECL(A,DATA_TYPE,NX,NY,nx,ny);
POLYBENCH_1D_ARRAY_DECL(x,DATA_TYPE,NY,ny);
POLYBENCH_1D_ARRAY_DECL(y,DATA_TYPE,NY,ny);
POLYBENCH_1D_ARRAY_DECL(y_outputFromGpu,DATA_TYPE,NY,ny);
POLYBENCH_1D_ARRAY_DECL(tmp,DATA_TYPE,NX,nx);
init_array(nx, ny, POLYBENCH_ARRAY(x), POLYBENCH_ARRAY(A));
read_cl_file();
cl_initialization();
cl_mem_init(POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(x), POLYBENCH_ARRAY(y), POLYBENCH_ARRAY(tmp));
cl_load_prog();
cl_launch_kernel(nx, ny);
errcode = clEnqueueReadBuffer(clCommandQue, y_mem_obj, CL_TRUE, 0, NY*sizeof(DATA_TYPE), POLYBENCH_ARRAY(y_outputFromGpu), 0, NULL, NULL);
if(errcode != CL_SUCCESS) printf("Error in reading GPU mem\n");
#ifdef RUN_ON_CPU
/* Start timer. */
polybench_start_instruments;
atax_cpu(nx, ny, POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(x), POLYBENCH_ARRAY(y), POLYBENCH_ARRAY(tmp));
/* Stop and print timer. */
printf("CPU Time in seconds:\n");
polybench_stop_instruments;
polybench_print_instruments;
compareResults(ny, POLYBENCH_ARRAY(y), POLYBENCH_ARRAY(y_outputFromGpu));
#else
print_array(ny, POLYBENCH_ARRAY(y_outputFromGpu));
#endif //RUN_ON_CPU
cl_clean_up();
POLYBENCH_FREE_ARRAY(A);
POLYBENCH_FREE_ARRAY(x);
POLYBENCH_FREE_ARRAY(y);
POLYBENCH_FREE_ARRAY(y_outputFromGpu);
POLYBENCH_FREE_ARRAY(tmp);
return 0;
}
int main(void) {
DATA_TYPE *A;
DATA_TYPE *y;
DATA_TYPE *tmp;
/////////////////////////
size_t oldSizes[1] = { NY };
size_t newSizes[1];
getNewSizes(oldSizes, NULL, newSizes, NULL, "atax_kernel2", 1);
NY = newSizes[0];
/////////////////////////
A = (DATA_TYPE *)malloc(NX_DEFAULT * NY * sizeof(DATA_TYPE));
y = (DATA_TYPE *)malloc(NY * sizeof(DATA_TYPE));
tmp = (DATA_TYPE *)malloc(NX_DEFAULT * sizeof(DATA_TYPE));
init_array(tmp, A);
platform = new Platform(PLATFORM_ID);
context = platform->getContext();
Device device = platform->getDevice(DEVICE_ID);
Queue queue(*context,device,Queue::EnableProfiling);
cl_mem_init(A, y, tmp,queue);
Program program(context,KERNEL_DIRECTORY KERNEL_FILE_NAME);
if(!program.build(device)){
std::cout << "Error building the program: \n";
std::cout <<program.getBuildLog(device);
}
kernel2=program.createKernel(kernel2Name.c_str());
cl_launch_kernel(queue);
queue.readBuffer(*y_mem_obj,NY * sizeof(DATA_TYPE), y);
queue.finish();
atax_cpu(A, tmp, y);
cl_clean_up();
free(A);
free(y);
free(tmp);
return 0;
}
int main(int argc, char** argv)
{
double t_start, t_end;
DATA_TYPE* A;
DATA_TYPE* x;
DATA_TYPE* y;
DATA_TYPE* y_outputFromGpu;
DATA_TYPE* tmp;
A = (DATA_TYPE*)malloc(NX*NY*sizeof(DATA_TYPE));
x = (DATA_TYPE*)malloc(NY*sizeof(DATA_TYPE));
y = (DATA_TYPE*)malloc(NY*sizeof(DATA_TYPE));
y_outputFromGpu = (DATA_TYPE*)malloc(NY*sizeof(DATA_TYPE));
tmp = (DATA_TYPE*)malloc(NX*sizeof(DATA_TYPE));
fprintf(stdout, "<< Matrix Transpose and Vector Multiplication >>\n");
init_array(x, A);
t_start = rtclock();
atax_OMP(A, x, y_outputFromGpu, tmp);
t_end = rtclock();
fprintf(stdout, "GPU Runtime: %0.6lfs\n", t_end - t_start);
t_start = rtclock();
atax_cpu(A, x, y, tmp);
t_end = rtclock();
fprintf(stdout, "CPU Runtime: %0.6lfs\n", t_end - t_start);
compareResults(y, y_outputFromGpu);
free(A);
free(x);
free(y);
free(y_outputFromGpu);
free(tmp);
return 0;
}