Revision 799236949867b6a2be0d492137b36a0b67011622 authored by Andrew Adams on 07 December 2021, 16:16:50 UTC, committed by GitHub on 07 December 2021, 16:16:50 UTC
* Add a version of fast_integer_divide that rounds towards zero * clang-format * Fix test condition * Clean up debugging code * Add explanatory comment to performance test * Pacify clang tidy
1 parent fb305fd
gpu_arg_types.cpp
#include "Halide.h"
using namespace Halide;
int main(int argc, char *argv[]) {
if (!get_jit_target_from_environment().has_gpu_feature()) {
printf("[SKIP] No GPU target enabled.\n");
return 0;
}
Func f, g;
Var x, y, tx, ty;
Param<int16_t> foo;
Expr e = select(foo > x, cast<int16_t>(255), foo + cast<int16_t>(x));
f(x) = e;
g(x) = e;
foo.set(-1);
f.gpu_tile(x, tx, 8);
Buffer<int16_t> out = f.realize({256});
Buffer<int16_t> out2 = g.realize({256});
out.copy_to_host();
for (int i = 0; i < 256; i++) {
if (out(i) != out2(i)) {
printf("Incorrect result at %d: %d != %d\n", i, out(i), out2(i));
printf("Failed\n");
return -1;
}
}
printf("Success!\n");
return 0;
}
Computing file changes ...
