sgemm_impl: barrier dumb dumb
This commit is contained in:
@@ -1042,6 +1042,9 @@ inline void thread_block_gemm(const T *A, const T *B, float *C,
|
|||||||
load_tile_to_smem<T, MemLayout::MN_major, MemLayout::MN_major, BN, BK,
|
load_tile_to_smem<T, MemLayout::MN_major, MemLayout::MN_major, BN, BK,
|
||||||
threads_per_threadblock>(dim_n, block_n, block_k, B,
|
threads_per_threadblock>(dim_n, block_n, block_k, B,
|
||||||
local_b, tid_in_threadblock);
|
local_b, tid_in_threadblock);
|
||||||
|
|
||||||
|
threadblock_barrier(threadblock_id_in_cluster,
|
||||||
|
warps_per_threadblock_per_core);
|
||||||
#endif
|
#endif
|
||||||
|
|
||||||
// consumer code: SMEM->RF and compute
|
// consumer code: SMEM->RF and compute
|
||||||
|
|||||||
Reference in New Issue
Block a user