1 /* Test of worker-private variables declared on loop directive, broadcasting
2 to vector-partitioned mode. Array worker variable. */
3
4 /* { dg-additional-options "--param=openacc-kernels=decompose" } */
5
6 /* { dg-additional-options "-fopt-info-omp-all" }
7 { dg-additional-options "-foffload=-fopt-info-omp-all" } */
8
9 /* { dg-additional-options "--param=openacc-privatization=noisy" }
10 { dg-additional-options "-foffload=--param=openacc-privatization=noisy" }
11 Prune a few: uninteresting:
12 { dg-prune-output {note: variable 'D\.[0-9]+' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} } */
13
14 /* It's only with Tcl 8.5 (released in 2007) that "the variable 'varName'
15 passed to 'incr' may be unset, and in that case, it will be set to [...]",
16 so to maintain compatibility with earlier Tcl releases, we manually
17 initialize counter variables:
18 { dg-line l_dummy[variable c_compute 0 c_loop_i 0 c_loop_j 0 c_loop_k 0] }
19 { dg-message "dummy" "" { target iN-VAl-Id } l_dummy } to avoid
20 "WARNING: dg-line var l_dummy defined, but not used". */
21
22 #include <assert.h>
23
24 int
25 main (int argc, char* argv[])
26 {
27 int i, arr[32 * 32 * 32];
28 int pt[2];
29
30 for (i = 0; i < 32 * 32 * 32; i++)
31 arr[i] = i;
32
33 /* "pt" is treated as "present_or_copy" on the kernels directive because it
34 is an array variable. */
35 #pragma acc kernels copy(arr) /* { dg-line l_compute[incr c_compute] } */
36 /* [PR104784] For some reason, for C++, the OpenACC 'kernels' decomposition
37 decides that a data region is needed for 'j', and subsequently requests it
38 to be made addressable.
39 { dg-note {OpenACC 'kernels' decomposition: variable 'j' declared in block requested to be made addressable} {} { target c++ } l_compute$c_compute }
40 { dg-note {variable 'j' made addressable} {} { target c++ } l_compute$c_compute }
41 { dg-note {variable 'j' declared in block is candidate for adjusting OpenACC privatization level} {} { target c++ } l_compute$c_compute } */
42 {
43 int j;
44
45 /* { dg-note {forwarded loop nest in OpenACC 'kernels' region to 'parloops' for analysis} {} { target *-*-* } .+1 } */
46 #pragma acc loop gang(num:32) /* { dg-line l_loop_i[incr c_loop_i] } */
47 /* { dg-note {variable 'j' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target c } l_loop_i$c_loop_i } */
48 /* { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_i$c_loop_i } */
49 for (i = 0; i < 32; i++)
50 {
51 /* But here, it is made private per-worker. */
52 #pragma acc loop worker(num:32) private(pt) /* { dg-line l_loop_j[incr c_loop_j] } */
53 /* { dg-note {variable 'pt' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_j$c_loop_j } */
54 /* { dg-note {variable 'j' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_j$c_loop_j } */
55 /* { dg-note {variable 'k' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_j$c_loop_j } */
56 for (j = 0; j < 32; j++)
57 {
58 int k;
59
60 pt[0] = i ^ j * 3;
61
62 #pragma acc loop vector(length:32) /* { dg-line l_loop_k[incr c_loop_k] } */
63 /* { dg-note {variable 'k' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_k$c_loop_k } */
64 for (k = 0; k < 32; k++)
65 arr[i * 1024 + j * 32 + k] += pt[0] * k;
66
67 pt[1] = i | j * 5;
68
69 #pragma acc loop vector(length:32) /* { dg-line l_loop_k[incr c_loop_k] } */
70 /* { dg-note {variable 'k' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_k$c_loop_k } */
71 for (k = 0; k < 32; k++)
72 arr[i * 1024 + j * 32 + k] += pt[1] * k;
73 }
74 }
75 /* { dg-optimized {assigned OpenACC seq loop parallelism} {} { target *-*-* } l_loop_i$c_loop_i } */
76 }
77
78 for (i = 0; i < 32; i++)
79 for (int j = 0; j < 32; j++)
80 for (int k = 0; k < 32; k++)
81 {
82 int idx = i * 1024 + j * 32 + k;
83 assert (arr[idx] == idx + (i ^ j * 3) * k + (i | j * 5) * k);
84 }
85
86 return 0;
87 }