fix mbarrier invalidate (NVIDIA#1494)

Daydream0929 · Jul 10, 2024 · 52fb43f · 52fb43f
1 parent 843adf0
commit 52fb43f
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/include/cutlass/arch/barrier.h b/include/cutlass/arch/barrier.h
@@ -360,7 +360,7 @@ struct ClusterBarrier {
     uint32_t smem_addr = cute::cast_smem_ptr_to_uint(smem_ptr);
     asm volatile(
         "{\n\t"
-        "mbarrier.ival.shared::cta.b64 [%0]; \n\t"
+        "mbarrier.inval.shared::cta.b64 [%0]; \n\t"
         "}"
         :
         : "r"(smem_addr));