diff --git a/example/65_gemm_multiply_multiply/CMakeLists.txt b/example/65_gemm_multiply_multiply/CMakeLists.txt
index 61954720b6..6803782df3 100644
--- a/example/65_gemm_multiply_multiply/CMakeLists.txt
+++ b/example/65_gemm_multiply_multiply/CMakeLists.txt
@@ -1,9 +1,16 @@
+list(APPEND TILE_EXAPMLE_BLOCKSCALE_COMPILE_OPTIONS -mllvm -greedy-reverse-local-assignment=1)
+list(APPEND TILE_EXAPMLE_BLOCKSCALE_COMPILE_OPTIONS -v --save-temps -Wno-gnu-line-marker)
+ 
+add_example_executable(example_moe_gemm1 moe_gemm1.cpp)
+add_example_executable(example_moe_gemm2 moe_gemm2.cpp)
+target_compile_options(example_moe_gemm1 PRIVATE ${TILE_EXAPMLE_BLOCKSCALE_COMPILE_OPTIONS})
+target_compile_options(example_moe_gemm2 PRIVATE ${TILE_EXAPMLE_BLOCKSCALE_COMPILE_OPTIONS})
+
+
 add_example_executable(example_gemm_multiply_multiply_xdl_fp8 gemm_multiply_multiply_xdl_fp8.cpp)
 add_example_executable(example_gemm_multiply_multiply_xdl_fp8_ab_scale gemm_multiply_multiply_xdl_fp8_ab_scale.cpp)
 add_example_executable(example_gemm_multiply_multiply_xdl_fp8_bpreshuffle gemm_multiply_multiply_xdl_fp8_bpreshuffle.cpp)
 add_example_executable(example_gemm_add_add_xdl_fp16 gemm_add_add_xdl_fp16.cpp)
 add_example_executable(example_gemm_multiply_multiply_xdl_int8 gemm_multiply_multiply_xdl_int8.cpp)
-add_example_executable(example_moe_gemm1 moe_gemm1.cpp)
-add_example_executable(example_moe_gemm2 moe_gemm2.cpp)
 add_example_executable(example_moe_pk_i4_gemm1 moe_pk_i4_gemm1.cpp)
 add_example_executable(example_moe_pk_i4_gemm2 moe_pk_i4_gemm2.cpp)
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_moe_gemm.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_moe_gemm.hpp
index edd9d0a8df..85e3d13a9a 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_moe_gemm.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_moe_gemm.hpp
@@ -747,40 +747,19 @@ struct GridwiseMoeGemm
         // in some cases.
         else if constexpr(is_same<tensor_layout::gemm::RowMajor, ALayout>::value)
         {
-            constexpr auto MLdsLayer        = 32 * 4 / KPerBlock / sizeof(LDSTypeA) /APackedSize < 1
-                                                  ? 1
-                                                  : 32 * 4 / KPerBlock / sizeof(LDSTypeA);
-            constexpr auto a_lds_block_desc = make_naive_tensor_descriptor(
-                make_tuple(
-                    AK0Number * Number<MLdsLayer>{}, Number<MPerBlock / MLdsLayer>{}, AK1Number),
-                make_tuple(AK1Number, Number<KPerBlock * MLdsLayer>{}, I1));
+            constexpr auto a_lds_block_desc =
+                make_naive_tensor_descriptor(make_tuple(AK0Number, Number<MPerBlock>{}, AK1Number),
+                                             make_tuple(AK1Number, Number<KPerBlock>{}, I1));
 
             constexpr auto a_lds_block_desc_permuted = transform_tensor_descriptor(
                 a_lds_block_desc,
-                make_tuple(make_xor_with_modulo_transform(make_tuple(
-                               Number<MPerBlock / MLdsLayer>{}, Number<AK0Number * MLdsLayer>{})),
+                make_tuple(make_xor_with_modulo_transform(
+                               make_tuple(Number<MPerBlock>{}, Number<AK0Number>{})),
                            make_pass_through_transform(AK1Number)),
                 make_tuple(Sequence<1, 0>{}, Sequence<2>{}),
                 make_tuple(Sequence<1, 0>{}, Sequence<2>{}));
 
-            constexpr auto a_lds_block_desc_ak0_mldslayer_m_ak1 = transform_tensor_descriptor(
-                a_lds_block_desc_permuted,
-                make_tuple(make_unmerge_transform(make_tuple(AK0Number, Number<MLdsLayer>{})),
-                           make_pass_through_transform(Number<MPerBlock / MLdsLayer>{}),
-                           make_pass_through_transform(AK1Number)),
-                make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
-                make_tuple(Sequence<0, 2>{}, Sequence<1>{}, Sequence<3>{}));
-
-            constexpr auto a_lds_block_desc_ak0_m_ak1 = transform_tensor_descriptor(
-                a_lds_block_desc_ak0_mldslayer_m_ak1,
-                make_tuple(make_pass_through_transform(AK0Number),
-                           make_merge_transform_v3_division_mod(
-                               make_tuple(Number<MPerBlock / MLdsLayer>{}, Number<MLdsLayer>{})),
-                           make_pass_through_transform(AK1Number)),
-                make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3>{}),
-                make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
-
-            return a_lds_block_desc_ak0_m_ak1;
+            return a_lds_block_desc_permuted;
         }
         else // ColumnMajor A
         {