int main(int argc, char *argv[])
{
double t_start, t_end;
DATA_TYPE A[NI][NJ][NK];
DATA_TYPE B[NI][NJ][NK]; // CPU target results
DATA_TYPE B_outputFromGpu[NI][NJ][NK]; // GPU exec results
//initialize the arrays
init(A);
#pragma hmpp conv allocate
#pragma hmpp conv advancedload, args[A;B]
// Run GPU code
t_start = rtclock();
#pragma hmpp conv callsite, args[A;B].advancedload=true, asynchronous
conv3D(A, B_outputFromGpu);
#pragma hmpp conv synchronize
t_end = rtclock();
fprintf(stderr, "GPU Runtime: %0.6lfs\n", t_end - t_start);
#pragma hmpp conv delegatedstore, args[B]
#pragma hmpp conv release
t_start = rtclock();
conv3D(A, B);
t_end = rtclock();
fprintf(stderr, "CPU Runtime: %0.6lfs\n", t_end - t_start);
compareResults(B, B_outputFromGpu);
return 0;
}
int main(int argc, char *argv[])
{
int ni = NI;
int nj = NJ;
int nk = NK;
POLYBENCH_3D_ARRAY_DECL(A,DATA_TYPE,NI,NJ,NK,ni,nj,nk);
POLYBENCH_3D_ARRAY_DECL(B,DATA_TYPE,NI,NJ,NK,ni,nj,nk);
POLYBENCH_3D_ARRAY_DECL(B_outputFromGpu,DATA_TYPE,NI,NJ,NK,ni,nj,nk);
init(ni, nj, nk, POLYBENCH_ARRAY(A));
read_cl_file();
cl_initialization();
cl_mem_init(POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(B));
cl_load_prog();
cl_launch_kernel(ni, nj, nk);
errcode = clEnqueueReadBuffer(clCommandQue, b_mem_obj, CL_TRUE, 0, NI * NJ * NK * sizeof(DATA_TYPE), POLYBENCH_ARRAY(B_outputFromGpu), 0, NULL, NULL);
if(errcode != CL_SUCCESS) printf("Error in reading GPU mem\n");
#ifdef RUN_ON_CPU
/* Start timer. */
polybench_start_instruments;
conv3D(ni, nj, nk, POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(B));
/* Stop and print timer. */
printf("CPU Time in seconds:\n");
polybench_stop_instruments;
polybench_print_instruments;
compareResults(ni, nj, nk, POLYBENCH_ARRAY(B), POLYBENCH_ARRAY(B_outputFromGpu));
#else //prevent dead code elimination
polybench_prevent_dce(print_array(ni, nj, nk, POLYBENCH_ARRAY(B_outputFromGpu)));
#endif //RUN_ON_CPU
cl_clean_up();
POLYBENCH_FREE_ARRAY(A);
POLYBENCH_FREE_ARRAY(B);
POLYBENCH_FREE_ARRAY(B_outputFromGpu);
return 0;
}
