diff --git a/dsv4/kernels/attention/fmha.py b/dsv4/kernels/attention/fmha.py
index d356ac25..6f9b3c98 100644
--- a/dsv4/kernels/attention/fmha.py
+++ b/dsv4/kernels/attention/fmha.py
@@ -430,6 +430,7 @@ class FmhaKernel:
                         rP_bf16_frg[None, j].store(s_vec.to(self.q_dtype))
 
                     cute.copy(tiled_tmem_store, rP_words, tTMEM_STOREtP)
+                    cute.arch.fence_view_async_tmem_store()
                 else:
                     # SMEM-P: compute softmax, write P to SMEM (TODO)
                     for j in range(frg_cnt):