change save o to lds data type to float

include/ck_tile/ops/fused_moe/pipeline/fused_moegemm_pipeline_general.hpp

@@ -256,9 +256,9 @@ struct FusedMoeGemmPipeline_General
 
         constexpr auto w_dstr =
             make_static_tile_distribution(detail::make_reduce_tile_distribution_encoding(
-                s_acc.get_tile_distribution().get_static_tile_distribution_encoding(), sequence<1>{}));
-        auto w_global_to_dram_window = make_tile_window(
-            w_window_.get_bottom_tensor_view(),
+                s_acc.get_tile_distribution().get_static_tile_distribution_encoding(),
+                sequence<1>{}));
+        auto w_global_to_dram_window = make_tile_window(w_window_.get_bottom_tensor_view(),
                                                         make_tuple(number<BlockShape::Block_M0>{}),
                                                         w_window_.get_window_origin(),
                                                         w_dstr);
@@ -307,9 +307,10 @@ struct FusedMoeGemmPipeline_General
         PrintMem(d,"D",0);
 #endif
         // add to LDS
+        CK_TILE_LDS_ADDR float* smem_3 = reinterpret_cast<CK_TILE_LDS_ADDR float*>(smem);
         auto o_lds_view =
-            make_naive_tensor_view<address_space_enum::lds, memory_operation_enum::atomic_add>(
-                smem_0,
+            make_naive_tensor_view<address_space_enum::lds, memory_operation_enum::set>(
+                smem_3,
                 make_tuple(number<128>{}, number<32>{}),
                 make_tuple(32, 1),
                 number<8>{},
@@ -333,12 +334,16 @@ struct FusedMoeGemmPipeline_General
                         move_tile_window(o_olds_win, {32, 0});
                         auto o1 = load_tile(o_olds_win);
                         static_for<0, thread_buffer_size, 1>{}([&](auto i) {
-                            o0.get_thread_buffer()(i) = type_convert<ODataType>(
-                                type_convert<float>(o0.get_thread_buffer()[i]) +
+                            o0.get_thread_buffer()(i) =
+                                type_convert<float>(type_convert<float>(o0.get_thread_buffer()[i]) +
                                                     type_convert<float>(o1.get_thread_buffer()[i]));
                         });
                     });
-                    update_tile(o_window_, o0);
+                    // tile_elementwise_inout([&weight](auto& x) { x = x *
+                    // type_convert<float>(weight); },
+                    //                o0);
+                    auto o = cast_tile<ODataType>(o0);
+                    update_tile(o_window_, o);
                     // restore pos
                     move_tile_window(o_olds_win, {-32 * (BlockShape::Repeat_K1 - 1), 0});
                 }
@@ -359,8 +364,8 @@ struct FusedMoeGemmPipeline_General
             // move out window and save data
             tile_elementwise_inout([&weight](auto& x) { x = x * type_convert<float>(weight); },
                                    o_acc);
-            auto o = cast_tile<ODataType>(o_acc);
-            store_tile(o_alds_win, o);
+            // auto o = cast_tile<ODataType>(o_acc);
+            store_tile(o_alds_win, o_acc);
             block_sync_lds();
             save_o();
 
@@ -375,10 +380,10 @@ struct FusedMoeGemmPipeline_General
             gemm_1(o_acc, y, d);
 
             // block_sync_lds();
-            tile_elementwise_inout(
-                [&weight](auto& x) { x = x * type_convert<float>(weight); }, o_acc);
-            auto o = cast_tile<ODataType>(o_acc);
-            store_tile(o_alds_win, o);
+            tile_elementwise_inout([&weight](auto& x) { x = x * type_convert<float>(weight); },
+                                   o_acc);
+            // auto o = cast_tile<ODataType>(o_acc);
+            store_tile(o_alds_win, o_acc);
             block_sync_lds();
             save_o();
             // store_tile(o_window_, o);