1 | // RUN: %libomptarget-compile-generic -O3 && %libomptarget-run-generic | %fcheck-generic |
2 | |
3 | #include <omp.h> |
4 | #include <stdio.h> |
5 | #include <stdlib.h> |
6 | |
7 | #define MAX_N 25000 |
8 | |
9 | void reset_input(double *a, double *a_h, double *b, double *c) { |
10 | for(int i = 0 ; i < MAX_N ; i++) { |
11 | a[i] = a_h[i] = i; |
12 | b[i] = i*2; |
13 | c[i] = i-3; |
14 | } |
15 | } |
16 | |
17 | int main(int argc, char *argv[]) { |
18 | double *a = (double *)calloc(MAX_N, size: sizeof(double)); |
19 | double *a_h = (double *)calloc(MAX_N, size: sizeof(double)); |
20 | double *d = (double *)calloc(MAX_N, size: sizeof(double)); |
21 | double *d_h = (double *)calloc(MAX_N, size: sizeof(double)); |
22 | double *b = (double *)calloc(MAX_N, size: sizeof(double)); |
23 | double *c = (double *)calloc(MAX_N, size: sizeof(double)); |
24 | |
25 | #pragma omp target enter data map(to:a[:MAX_N],b[:MAX_N],c[:MAX_N],d[:MAX_N]) |
26 | |
27 | for (int n = 32 ; n < MAX_N ; n+=5000) { |
28 | reset_input(a, a_h, b, c); |
29 | |
30 | #pragma omp target update to(a[:n],b[:n],c[:n],d[:n]) |
31 | int t = 0; |
32 | for (int tms = 1 ; tms <= 256 ; tms *= 2) { // 8 times |
33 | for (int ths = 32 ; ths <= 1024 ; ths *= 2) { // 6 times |
34 | t++; |
35 | #pragma omp target |
36 | #pragma omp teams num_teams(tms) thread_limit(ths) |
37 | { |
38 | #pragma omp distribute parallel for |
39 | for (int i = 0; i < n; ++i) { |
40 | a[i] += b[i] + c[i]; |
41 | } |
42 | #pragma omp distribute parallel for |
43 | for (int i = 0; i < n; ++i) { |
44 | d[i] -= b[i] + c[i]; |
45 | } |
46 | } |
47 | } // loop over 'ths' |
48 | } // loop over 'tms' |
49 | |
50 | // check results for each 'n' |
51 | for (int times = 0 ; times < t ; times++) { |
52 | for (int i = 0; i < n; ++i) { |
53 | a_h[i] += b[i] + c[i]; |
54 | } |
55 | for (int i = 0; i < n; ++i) |
56 | d_h[i] -= b[i] + c[i]; |
57 | } |
58 | #pragma omp target update from(a[:n],d[:n]) |
59 | |
60 | for (int i = 0; i < n; ++i) { |
61 | if (a_h[i] != a[i]) { |
62 | printf(format: "A Error at n = %d, i = %d: host = %f, device = %f\n" , n, i, a_h[i], a[i]); |
63 | return 1; |
64 | } |
65 | if (d_h[i] != d[i]) { |
66 | printf(format: "D Error at n = %d, i = %d: host = %lf, device = %lf\n" , n, i, d_h[i], d[i]); |
67 | return 1; |
68 | } |
69 | } |
70 | } // loop over 'n' |
71 | |
72 | // CHECK: Succeeded |
73 | printf(format: "Succeeded\n" ); |
74 | return 0; |
75 | } |
76 | |