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
|
#include "Halide.h"
#include <iostream>
#include <stdio.h>
using namespace Halide;
int main(int argc, char **argv) {
Var x("x"), y("y");
Func f("f"), g("g"), h("h");
printf("Defining function...\n");
f(x, y) = max(x, y);
g(x, y) = min(x, y);
h(x, y) = clamp(x + y, 20, 100);
Var xo("xo"), yo("yo"), xi("xi"), yi("yi");
Target target = get_jit_target_from_environment();
if (target.has_gpu_feature()) {
// Resolve why OpenCL used 32,1 tiling
f.gpu_tile(x, y, xo, yo, xi, yi, 8, 8);
g.gpu_tile(x, y, xo, yo, xi, yi, 8, 8);
h.gpu_tile(x, y, xo, yo, xi, yi, 8, 8);
} else if (target.has_feature(Target::HVX)) {
f.hexagon().vectorize(x, 32);
g.hexagon().vectorize(x, 32);
h.hexagon().vectorize(x, 32);
}
printf("Realizing function...\n");
Buffer<int> imf = f.realize({32, 32}, target);
Buffer<int> img = g.realize({32, 32}, target);
Buffer<int> imh = h.realize({32, 32}, target);
// Check the result was what we expected
for (int i = 0; i < 32; i++) {
for (int j = 0; j < 32; j++) {
if (imf(i, j) != (i > j ? i : j)) {
printf("imf[%d, %d] = %d\n", i, j, imf(i, j));
return 1;
}
if (img(i, j) != (i < j ? i : j)) {
printf("img[%d, %d] = %d\n", i, j, img(i, j));
return 1;
}
int href = i + j;
if (href < 20) href = 20;
if (href > 100) href = 100;
if (imh(i, j) != href) {
printf("imh[%d, %d] = %d (not %d)\n", i, j, imh(i, j), href);
return 1;
}
}
}
printf("Success!\n");
return 0;
}
|
