ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-02 04:31:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_TILE] Fix compilation errors introduced in #2320, #2219 and #2214 (#2388)

Browse Source 
* Fix compilation errors

* Fix more ck_tile example compilation errors
This commit is contained in:
Po Yen Chen
2025-06-23 12:29:15 +08:00
committed by GitHub
parent 0366fb2abc
commit 7d669440a6
10 changed files with 112 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
include/ck_tile/ops/norm_reduce/block/block_norm_reduce.hpp
View File

@@ -250,7 +250,7 @@ struct BlockNormReduceCrossWarpSync
// | w0 | w1 | w2 | w3 | -----> | w0123 |
//
// -> also store data from every wave into LDS
constexpr index_t num_warps = BlockShape::BlockSize / WarpSize;
constexpr index_t num_warps = BlockShape::BlockSize / get_warp_size();
return num_warps * 4 * thread_buf_size * sizeof(float);
}
@@ -276,7 +276,7 @@ struct BlockNormReduceCrossWarpSync
const index_t lane_id = get_lane_id();
const index_t warp_id = get_warp_id();
constexpr auto num_reduce_warps = GetReduceWarps<MeanDistributedTensor_>();
constexpr index_t num_warps = BlockShape::BlockSize / WarpSize;
constexpr index_t num_warps = BlockShape::BlockSize / get_warp_size();
const index_t smem_offset = warp_id;
// skip if nonthing to do