From 2aee73922ce159fcaf9492a0acb7a301ae135f37 Mon Sep 17 00:00:00 2001
From: questa-quan-wang <questaw@nvidia.com>
Date: Mon, 5 Jan 2026 16:36:03 +0800
Subject: [PATCH] Minor fix for testing of blockscaled dense GEMM with TMA
 prefetch (#2930)

* new example with TMA prefetch feature targeting for DRAM latency bound cases

* minor fix to resitrct as 100a arch

* typo

* apply arch for whole pytest

---------

Co-authored-by: Questa Wang <questaw@computelab-frontend-7.nvidia.com>
Co-authored-by: Questa Wang <questaw@umbriel-b200-145.ipp4a1.colossus.nvidia.com>
---
 .../sm_100a/test_dense_blockscaled_gemm_persistent_prefetch.py  | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/test/examples/CuTeDSL/sm_100a/test_dense_blockscaled_gemm_persistent_prefetch.py b/test/examples/CuTeDSL/sm_100a/test_dense_blockscaled_gemm_persistent_prefetch.py
index 623f9814..825c1697 100644
--- a/test/examples/CuTeDSL/sm_100a/test_dense_blockscaled_gemm_persistent_prefetch.py
+++ b/test/examples/CuTeDSL/sm_100a/test_dense_blockscaled_gemm_persistent_prefetch.py
@@ -47,7 +47,7 @@ from blackwell.dense_blockscaled_gemm_persistent_prefetch import (
 )
 
 import cutlass
-
+pytestmark = [pytest.mark.arch(["100a"])]
 
 @pytest.mark.invalid_case(
     lambda: not Sm100BlockScaledPersistentDenseGemmKernel.can_implement(