AbstractLinAlgPack::value_type
AbstractLinAlgPack::sum( const Vector& v_rhs )
{
sum_op.reduct_obj_reinit(&*sum_targ);
const Vector* vecs[1] = { &v_rhs };
apply_op(sum_op,1,vecs,0,NULL,&*sum_targ);
return RTOp_ROp_sum_val(sum_op(*sum_targ));
}
int *
arr_int_mkrand(size_t nints, int *sum_ptr)
{
int sum = 0;
int *ret;
ret = xmalloc(sizeof(int)*nints);
for (size_t i=0; i<nints; i++) {
ret[i] = rand() % 100;
sum += sum_op(ret[i]);
}
*sum_ptr = sum;
return ret;
}
int
main(int argc, const char *argv[])
{
unsigned nthreads;
size_t nints;
int sum1, sum2;
int *arr;
nints = 0;
if (argc > 1)
nints = atol(argv[1]);
if (nints == 0)
nints = 100000;
#pragma omp parallel
#pragma omp master
nthreads = omp_get_num_threads();
printf("Number of threads: %u\n", nthreads);
printf("number of ints: %lu\n", nints);
arr = arr_int_mkrand(nints, &sum1);
sum2 = 0;
tsc_t t; tsc_init(&t); tsc_start(&t);
#pragma omp parallel for reduction(+:sum2)
for (size_t i=0; i<nints; i++) {
sum2 += sum_op(arr[i]);
}
tsc_pause(&t);
tsc_report("sum_OMP", &t);
if (sum1 != sum2) {
fprintf(stderr, "Error in sum: %d vs %d\n", sum1, sum2);
abort();
}
printf("DONE\n");
return 0;
}
