NVIDIA/cutlass
1
0
Fork 0
mirror of https://github.com/NVIDIA/cutlass.git synced 2026-04-19 22:38:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

add missing condition for sync (#2889)

Browse Source
This commit is contained in:
dongxiao
2025-12-19 11:00:30 +08:00
committed by GitHub
parent ebf3165efb
commit 331e2f451c
1 changed files with 6 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
examples/python/CuTeDSL/blackwell/dense_gemm.py
View File

@@ -787,11 +787,13 @@ class DenseGemmKernel:
# Async arrive AB buffer empty
consumer_handle.release()
# Peek (try_wait) AB buffer empty for k_tile = prefetch_k_tile_cnt + k_tile + 1
peek_ab_empty_status = ab_producer.try_acquire()
if k_tile_idx + 1 < k_tile_cnt - prefetch_k_tile_cnt:
# Peek (try_wait) AB buffer empty for k_tile = prefetch_k_tile_cnt + k_tile + 1
peek_ab_empty_status = ab_producer.try_acquire()
# Peek (try_wait) AB buffer full for k_tile = k_tile + 1
peek_ab_full_status = ab_consumer.try_wait()
if k_tile_idx + 1 < k_tile_cnt and is_leader_cta:
# Peek (try_wait) AB buffer full for k_tile = k_tile + 1
peek_ab_full_status = ab_consumer.try_wait()
# Async arrive accumulator buffer full
if is_leader_cta: