int main(int argc, char *argv[])
{
int n;
REAL *y_omp, *y_ompacc, *x;
REAL a = 123.456;
#pragma omp target device(mpi:all) begin
n = VEC_LEN;
y_omp = (REAL *) malloc(n * sizeof(REAL));
y_ompacc = (REAL *) malloc(n * sizeof(REAL));
x = (REAL *) malloc(n * sizeof(REAL));
#pragma omp target device(mpi:all) end
#pragma omp target device(mpi:master) begin
srand48(1<<12);
init(x, n);
init(y_ompacc, n);
memcpy(y_ompacc, y_omp, n*sizeof(REAL));
#pragma omp target device(mpi:master) end
int num_threads;
// #pragma omp parallel shared (num_threads)
{
if (omp_get_thread_num() == 0)
num_threads = omp_get_num_threads();
}
/* CPU threading version*/
double omp_time = read_timer();
axpy_omp(x, y_omp, n, a);
omp_time = read_timer() - omp_time;
/* openmp acc version */
double ompacc_time = read_timer();
axpy_ompacc(x, y_ompacc, n, a);
ompacc_time = read_timer() - ompacc_time;
printf("axpy(%d): checksum: %g; time(s):\tOMP(%d threads)\tOMPACC\n", n, check(y_omp, y_ompacc, n),num_threads);
printf("\t\t\t\t\t\t%4f\t%4f\n", omp_time, ompacc_time);
free(y_omp);
free(y_ompacc);
free(x);
return 0;
}
int main(int argc, char *argv[])
{
int n;
REAL *y_ompacc, *x;
REAL a = 123.456;
n = VEC_LEN;
y_ompacc = (REAL *) malloc(n * sizeof(REAL));
x = (REAL *) malloc(n * sizeof(REAL));
srand48(1<<12);
init(x, n);
init(y_ompacc, n);
/* openmp acc version */
axpy_ompacc(x, y_ompacc, n, a);
free(y_ompacc);
free(x);
return 0;
}
