void a9_compute(int tid, struct buffer* buffers, int start_indx, int end_indx)
{
int i ;
for(i=0 ; i<NUM_ITER ; i++)
{
if(tid == 0) {
call_barrier(0) ;
pthread_barrier_wait(&barrier0) ;
doitgen(buffers[0].bo[0]->map, buffers[1].bo[0]->map, C41,
start_indx, end_indx) ;
} else {
pthread_barrier_wait(&barrier0) ;
doitgen(buffers[0].bo[0]->map, buffers[1].bo[0]->map, C42,
start_indx, end_indx) ;
}
}
}
void common_wrapper(UArg arg0, UArg arg1)
{
taskArgs* t = (taskArgs*)arg0 ;
int tid = (int)arg1 ;
int i ;
for(i=0 ; i<NUM_ITER ; i++) {
if(tid == 0) {
callBarrier(0, /*lock_id=*/4) ;
callLocalBarrier() ;
doitgen(t->buffer1, t->buffer2, t->buffer3, t->start_indx, t->end_indx) ;
} else {
callLocalBarrier() ;
doitgen(t->buffer1, t->buffer2, t->buffer3, t->start_indx, t->end_indx) ;
}
}
if(tid == 0)
Event_post(edgeDetectEvent, Event_Id_00) ;
else
Event_post(edgeDetectEvent, Event_Id_01) ;
}
int main(void) {
double t_start, t_end;
int i;
DATA_TYPE* A, *A_2;
DATA_TYPE* C4, *C4_2;
DATA_TYPE* sum, *sum_2;
A = (DATA_TYPE*)malloc(NR * NQ * NP * sizeof(DATA_TYPE));
C4 = (DATA_TYPE*)malloc(NP * NP * sizeof(DATA_TYPE));
sum = (DATA_TYPE*)malloc(NR * NQ * NP * sizeof(DATA_TYPE));
A_2 = (DATA_TYPE*)malloc(NR * NQ * NP * sizeof(DATA_TYPE));
C4_2 = (DATA_TYPE*)malloc(NP * NP * sizeof(DATA_TYPE));
sum_2 = (DATA_TYPE*)malloc(NR * NQ * NP * sizeof(DATA_TYPE));
init_array(A, C4);
init_array(A_2, C4_2);
read_cl_file();
cl_initialization();
cl_mem_init(A, C4, sum);
cl_load_prog();
t_start = rtclock();
int r;
for (r = 0; r < NR; r++)
{
cl_launch_kernel1(r);
cl_launch_kernel2(r);
}
t_end = rtclock();
errcode = clEnqueueReadBuffer(clCommandQue, c_mem_obj, CL_TRUE, 0, NR * NQ * NP * sizeof(DATA_TYPE), sum, 0, NULL, NULL);
if(errcode != CL_SUCCESS) printf("Error in reading GPU mem\n");
fprintf(stdout, "GPU Runtime: %0.6lfs\n", t_end - t_start);
t_start = rtclock();
doitgen(sum_2, A_2, C4_2);
t_end = rtclock();
fprintf(stdout, "CPU Runtime: %0.6lfs\n", t_end - t_start);
compareResults(sum, sum_2);
cl_clean_up();
return 0;
}
Int32 fxnTest1(UInt32 size, UInt32 *data)
{
UInt32 start_indx, end_indx ;
FxnArgs *args = (FxnArgs *)((UInt32)data + sizeof(map_info_type));
int* buffer1 = (int*)args->a ;
int* buffer2 = (int*)args->b ;
int* buffer3 = (int*)args->e ;
start_indx = args->start_indx;
end_indx = args->end_indx;
int i ;
for(i=0 ; i<NUM_ITER ; i++) {
callBarrier(0, /*lock_id=*/4) ;
doitgen(buffer1, buffer2, buffer3, start_indx, end_indx) ;
}
return 1 ;
}
