From 5a0219b7c7cc272de1984bd77dad81079d5b376c Mon Sep 17 00:00:00 2001
From: valarLip <340077269@qq.com>
Date: Wed, 6 Aug 2025 10:09:33 +0000
Subject: [PATCH] reproduce

---
 example/65_gemm_multiply_multiply/CMakeLists.txt        | 2 ++
 example/65_gemm_multiply_multiply/moe_gemm1_xdl_fp8.cpp | 2 +-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/example/65_gemm_multiply_multiply/CMakeLists.txt b/example/65_gemm_multiply_multiply/CMakeLists.txt
index d1e1a51afd..719a506c7e 100644
--- a/example/65_gemm_multiply_multiply/CMakeLists.txt
+++ b/example/65_gemm_multiply_multiply/CMakeLists.txt
@@ -40,6 +40,8 @@ foreach(gpu IN LISTS GPU_TARGETS)
 endforeach()
 
 set(GEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1")
+
+list(APPEND GEMM_OPTIONS -Wno-error --save-temps)
 set(BLOCKSCALE_GEMM_OPTIONS )
 check_cxx_compiler_flag("-mllvm --misched-bottomup=1" HAS_MISCHED_BOTTOMUP)
 check_cxx_compiler_flag("-mllvm --misched-prera-direction=bottomup" HAS_MISCHED_PRERA_DIRECTION)
diff --git a/example/65_gemm_multiply_multiply/moe_gemm1_xdl_fp8.cpp b/example/65_gemm_multiply_multiply/moe_gemm1_xdl_fp8.cpp
index 9fe9fdde78..b6b87a0642 100644
--- a/example/65_gemm_multiply_multiply/moe_gemm1_xdl_fp8.cpp
+++ b/example/65_gemm_multiply_multiply/moe_gemm1_xdl_fp8.cpp
@@ -157,7 +157,7 @@ using AElementOp = PassThrough;
 using BElementOp = PassThrough;
 
 static constexpr auto GemmSpec         = ck::tensor_operation::device::GemmSpecialization::Default;
-static constexpr ck::index_t MPerBlock = 128;
+static constexpr ck::index_t MPerBlock = 256;
 static constexpr ck::index_t NPerBlock = 128;
 static constexpr ck::index_t MNPerXDL  = 16;
 static constexpr ck::index_t MXDLPerWave = MPerBlock / (MNPerXDL * 1);