We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 6b9fd1f commit c24a686Copy full SHA for c24a686
src/tl_templates/cuda/common.h
@@ -55,14 +55,10 @@ using int4_t = int4;
55
} while (0)
56
57
// using cutlass abs function for half_t
58
-TL_PATCH TL_DEVICE half_t __habs(const half_t x) {
59
- return abs(x);
60
-}
+TL_PATCH TL_DEVICE half_t __habs(const half_t x) { return abs(x); }
61
62
// using cutlass abs function for bfloat_t
63
-TL_PATCH TL_DEVICE bfloat16_t __habs(const bfloat16_t x) {
64
65
+TL_PATCH TL_DEVICE bfloat16_t __habs(const bfloat16_t x) { return abs(x); }
66
67
// hrsqrt function for half_t
68
TL_PATCH TL_DEVICE half_t hrsqrt(const half_t x) {
0 commit comments