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
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
|
#include "Halide.h"
#include "gpu_object_lifetime_tracker.h"
#include <iostream>
using namespace Halide;
Internal::GpuObjectLifetimeTracker tracker;
void halide_print(JITUserContext *user_context, const char *str) {
printf("%s", str);
tracker.record_gpu_debug(str);
}
int main(int argc, char *argv[]) {
Var x, xi;
Target target = get_jit_target_from_environment();
// We need to hook the default handler too, to catch the frees done by release_all
JITHandlers handlers;
handlers.custom_print = halide_print;
Internal::JITSharedRuntime::set_default_handlers(handlers);
// We need debug output to record object creation.
target.set_feature(Target::Debug);
{
// Verify that internal buffers in a sequence of device stages are
// released. This should generate some early frees of buffers with
// device allocations.
const int stage_count = 10;
Func f[stage_count];
f[0](x) = x;
for (int i = 1; i < stage_count; i++) {
f[i](x) = f[i - 1](x);
}
for (int i = 0; i < stage_count; i++) {
f[i].compute_root();
if (i % 3 != 0) {
if (target.has_gpu_feature()) {
f[i].gpu_tile(x, xi, 32);
} else if (target.has_feature(Target::HVX)) {
f[i].hexagon();
}
}
}
Func output = f[stage_count - 1];
output.realize({256}, target);
}
Internal::JITSharedRuntime::release_all();
int ret = tracker.validate_gpu_object_lifetime(true /* allow_globals */, true /* allow_none */, 1 /* max_globals */);
if (ret != 0) {
fprintf(stderr, "validate_gpu_object_lifetime() failed\n");
return 1;
}
printf("Success!\n");
return 0;
}
|
