Merge commit '4c57157d508e4c102626730aa372c8111670a878' into develop

This commit is contained in:
assistant-librarian[bot]
2025-06-17 19:12:43 +00:00
parent dedfb2e7e6
commit 0afe9102b4
31 changed files with 213 additions and 206 deletions

View File

@@ -35,7 +35,7 @@ CK_TILE_DEVICE T warp_shuffle_up(const T& v_local, uint32_t lane_delta)
#elif 1
static_assert(sizeof(T) == sizeof(int32_t), "wrong!");
const uint32_t wrap_around_lane_delta = warpSize - lane_delta;
const uint32_t wrap_around_lane_delta = get_warp_size() - lane_delta;
const int32_t v_remote_tmp = __builtin_amdgcn_ds_bpermute(
(__lane_id() << 2) + (wrap_around_lane_delta << 2), bit_cast<int32_t>(v_local));