From d1749b3aaee9b082cf9723bae9f1f76d14729a10 Mon Sep 17 00:00:00 2001
From: Qianfeng Zhang <Qianfeng.Zhang@amd.com>
Date: Tue, 15 Apr 2025 15:08:46 +0000
Subject: [PATCH] Use kM0=128 kN0=64 to completely remove the vgprs spilling

---
 .../ck_tile/18_hstu_attention/hstu_attention_fwd_setting.hpp    | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/example/ck_tile/18_hstu_attention/hstu_attention_fwd_setting.hpp b/example/ck_tile/18_hstu_attention/hstu_attention_fwd_setting.hpp
index c842df341b..5d82a3680c 100644
--- a/example/ck_tile/18_hstu_attention/hstu_attention_fwd_setting.hpp
+++ b/example/ck_tile/18_hstu_attention/hstu_attention_fwd_setting.hpp
@@ -35,7 +35,7 @@ struct HstuAttentionFwdBlockTile<64>
 template <>
 struct HstuAttentionFwdBlockTile<128>
 {
-    using type        = ck_tile::sequence<128, 128, 32, 128, 32, 128>;
+    using type        = ck_tile::sequence<128, 64, 32, 128, 32, 128>;
     using gemm0_warps = ck_tile::sequence<4, 1, 1>;
     using gemm1_warps = ck_tile::sequence<4, 1, 1>;
 };