From e4e70f8b0a05977f8f638ea3da9990b51987397f Mon Sep 17 00:00:00 2001
From: Qianfeng Zhang <Qianfeng.Zhang@amd.com>
Date: Sun, 18 May 2025 15:58:02 +0000
Subject: [PATCH] Set the block_per_cu to 3 for hdim-128

---
 .../ck_tile/18_hstu_attention/hstu_attention_fwd_pipeline.hpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/example/ck_tile/18_hstu_attention/hstu_attention_fwd_pipeline.hpp b/example/ck_tile/18_hstu_attention/hstu_attention_fwd_pipeline.hpp
index 6c3cb1e638..c605e1dfca 100644
--- a/example/ck_tile/18_hstu_attention/hstu_attention_fwd_pipeline.hpp
+++ b/example/ck_tile/18_hstu_attention/hstu_attention_fwd_pipeline.hpp
@@ -84,9 +84,9 @@ struct HstuAttentionFwdPipelineQRKSVS
             else if constexpr(kQKHeaddim == 96 || kQKHeaddim == 128)
             {
                 if constexpr(kHasBias)
-                    return 1;
-                else
                     return 2;
+                else
+                    return 3;
             }
             else if constexpr(kQKHeaddim == 256)
             {