| 1 |
/* { dg-additional-options "-fopt-info-note-omp" } |
| 2 |
{ dg-additional-options "--param=openacc-privatization=noisy" } |
| 3 |
{ dg-additional-options "-foffload=-fopt-info-note-omp" } |
| 4 |
{ dg-additional-options "-foffload=--param=openacc-privatization=noisy" } |
| 5 |
for testing/documenting aspects of that functionality. */ |
| 6 |
|
| 7 |
#include <stdio.h> |
| 8 |
#include <openacc.h> |
| 9 |
#include <gomp-constants.h> |
| 10 |
|
| 11 |
#define N (32*32*32+17) |
| 12 |
int main () |
| 13 |
{ |
| 14 |
int ix; |
| 15 |
int ondev = 0; |
| 16 |
int t = 0, h = 0; |
| 17 |
int workersize, vectorsize; |
| 18 |
|
| 19 |
#define NW 32 |
| 20 |
#define VL 32 |
| 21 |
#pragma acc parallel num_workers(NW) vector_length(VL) \ |
| 22 |
copy(ondev) |
| 23 |
/* { dg-note {variable 'ix' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } .-2 } */ |
| 24 |
{ |
| 25 |
#pragma acc loop worker vector reduction (+:t) |
| 26 |
/* { dg-note {variable 'ix' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } .-1 } */ |
| 27 |
/* { dg-note {variable 'val' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } .-2 } */ |
| 28 |
/* { dg-note {variable 'g' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } .-3 } */ |
| 29 |
/* { dg-note {variable 'w' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } .-4 } */ |
| 30 |
/* { dg-note {variable 'v' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } .-5 } */ |
| 31 |
for (unsigned ix = 0; ix < N; ix++) |
| 32 |
{ |
| 33 |
int val = ix; |
| 34 |
|
| 35 |
if (acc_on_device (acc_device_not_host)) |
| 36 |
{ |
| 37 |
int g, w, v; |
| 38 |
|
| 39 |
g = __builtin_goacc_parlevel_id (GOMP_DIM_GANG); |
| 40 |
w = __builtin_goacc_parlevel_id (GOMP_DIM_WORKER); |
| 41 |
v = __builtin_goacc_parlevel_id (GOMP_DIM_VECTOR); |
| 42 |
val = (g << 16) | (w << 8) | v; |
| 43 |
ondev = 1; |
| 44 |
} |
| 45 |
t += val; |
| 46 |
} |
| 47 |
} |
| 48 |
workersize = NW; |
| 49 |
vectorsize = VL; |
| 50 |
#ifdef ACC_DEVICE_TYPE_radeon |
| 51 |
/* AMD GCN has an upper limit of 'num_workers(16)'. */ |
| 52 |
if (workersize > 16) |
| 53 |
workersize = 16; |
| 54 |
/* AMD GCN uses the autovectorizer for the vector dimension: the use |
| 55 |
of a function call in vector-partitioned code in this test is not |
| 56 |
currently supported. */ |
| 57 |
vectorsize = 1; |
| 58 |
#endif |
| 59 |
|
| 60 |
for (ix = 0; ix < N; ix++) |
| 61 |
{ |
| 62 |
int val = ix; |
| 63 |
if(ondev) |
| 64 |
{ |
| 65 |
int g = 0; |
| 66 |
int w = (ix / vectorsize) % workersize; |
| 67 |
int v = ix % vectorsize; |
| 68 |
|
| 69 |
val = (g << 16) | (w << 8) | v; |
| 70 |
} |
| 71 |
h += val; |
| 72 |
} |
| 73 |
if (t != h) |
| 74 |
{ |
| 75 |
printf ("t=%x expected %x\n", t, h); |
| 76 |
return 1; |
| 77 |
} |
| 78 |
|
| 79 |
return 0; |
| 80 |
} |