Skip to content

Commit c0e5ca6

Browse files
committed
[SYCL][CUDA] Only align if necessary
1 parent 9a718f1 commit c0e5ca6

File tree

1 file changed

+4
-2
lines changed

1 file changed

+4
-2
lines changed

sycl/plugins/cuda/pi_cuda.hpp

+4-2
Original file line numberDiff line numberDiff line change
@@ -647,8 +647,10 @@ struct _pi_kernel {
647647
const size_t alignment = std::min(max_alignment, size);
648648

649649
// align the argument
650-
size_t alignedLocalOffset =
651-
localOffset + alignment - (localOffset % alignment);
650+
size_t alignedLocalOffset = localOffset;
651+
if (localOffset % alignment != 0) {
652+
alignedLocalOffset += alignment - (localOffset % alignment);
653+
}
652654

653655
add_arg(index, sizeof(size_t), (const void *)&(alignedLocalOffset),
654656
size + (alignedLocalOffset - localOffset));

0 commit comments

Comments
 (0)