blob: 425bc4617aac1d9c0592d79cdda4e434bdd44886 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
|
// RUN: %libomptarget-compilexx-run-and-check-generic
//
// REQUIRES: gpu
// XFAIL: intelgpu
#include <assert.h>
#include <ompx.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#pragma omp begin declare variant match(device = {arch(amdgcn)})
unsigned get_warp_size() { return __builtin_amdgcn_wavefrontsize(); }
#pragma omp end declare variant
#pragma omp begin declare variant match(device = {arch(nvptx64)})
unsigned get_warp_size() { return __nvvm_read_ptx_sreg_warpsize(); }
#pragma omp end declare variant
#pragma omp begin declare variant match(device = {kind(cpu)})
unsigned get_warp_size() { return 1; }
#pragma omp end declare variant
int main(int argc, char *argv[]) {
const int num_blocks = 1;
const int block_size = 256;
const int N = num_blocks * block_size;
int *res = (int *)malloc(N * sizeof(int));
#pragma omp target teams ompx_bare num_teams(num_blocks) \
thread_limit(block_size) map(from : res[0 : N])
{
int tid = ompx_thread_id_x();
uint64_t mask = ompx_ballot_sync(~0LU, tid & 0x1);
if (get_warp_size() == 64)
res[tid] = mask == 0xaaaaaaaaaaaaaaaa;
else
res[tid] = mask == 0xaaaaaaaa;
}
for (int i = 0; i < N; ++i)
assert(res[i]);
// CHECK: PASS
printf("PASS\n");
free(res);
return 0;
}
|
