1 /* Test of worker-private variables declared on a loop directive, broadcasting
2 to vector-partitioned mode. Aggregate 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 typedef struct
25 {
26 int x, y;
27 } vec2;
28
29 int
30 main (int argc, char* argv[])
31 {
32 int i, arr[32 * 32 * 32];
33 vec2 pt;
34
35 for (i = 0; i < 32 * 32 * 32; i++)
36 arr[i] = i;
37
38 #pragma acc kernels copy(arr) /* { dg-line l_compute[incr c_compute] } */
39 /* [PR104784] For some reason, for C++, the OpenACC 'kernels' decomposition
40 decides that a data region is needed for 'j', and subsequently requests it
41 to be made addressable.
42 { dg-note {OpenACC 'kernels' decomposition: variable 'j' declared in block requested to be made addressable} {} { target c++ } l_compute$c_compute }
43 { dg-note {variable 'j' made addressable} {} { target c++ } l_compute$c_compute }
44 { dg-note {variable 'j' declared in block is candidate for adjusting OpenACC privatization level} {} { target c++ } l_compute$c_compute } */
45 {
46 int j;
47
48 /* { dg-note {forwarded loop nest in OpenACC 'kernels' region to 'parloops' for analysis} {} { target *-*-* } .+1 } */
49 #pragma acc loop gang(num:32) /* { dg-line l_loop_i[incr c_loop_i] } */
50 /* { dg-note {variable 'j' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target c } l_loop_i$c_loop_i } */
51 /* { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_i$c_loop_i } */
52 for (i = 0; i < 32; i++)
53 {
54 #pragma acc loop worker(num:32) private(pt) /* { dg-line l_loop_j[incr c_loop_j] } */
55 /* { dg-note {variable 'pt' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_j$c_loop_j } */
56 /* { dg-note {variable 'j' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_j$c_loop_j } */
57 /* { dg-note {variable 'k' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_j$c_loop_j } */
58 for (j = 0; j < 32; j++)
59 {
60 int k;
61
62 pt.x = i ^ j * 3;
63 pt.y = i | j * 5;
64
65 #pragma acc loop vector(length:32) /* { dg-line l_loop_k[incr c_loop_k] } */
66 /* { dg-note {variable 'k' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_k$c_loop_k } */
67 for (k = 0; k < 32; k++)
68 arr[i * 1024 + j * 32 + k] += pt.x * k;
69
70 #pragma acc loop vector(length:32) /* { dg-line l_loop_k[incr c_loop_k] } */
71 /* { dg-note {variable 'k' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} {} { target *-*-* } l_loop_k$c_loop_k } */
72 for (k = 0; k < 32; k++)
73 arr[i * 1024 + j * 32 + k] += pt.y * k;
74 }
75 }
76 /* { dg-optimized {assigned OpenACC seq loop parallelism} {} { target *-*-* } l_loop_i$c_loop_i } */
77 }
78
79 for (i = 0; i < 32; i++)
80 for (int j = 0; j < 32; j++)
81 for (int k = 0; k < 32; k++)
82 {
83 int idx = i * 1024 + j * 32 + k;
84 assert (arr[idx] == idx + (i ^ j * 3) * k + (i | j * 5) * k);
85 }
86
87 return 0;
88 }