Fix timing issue in CK_TILE GEMM example (#2940)

Author: hyoon1

example/ck_tile/03_gemm/gemm_splitk_two_stage_invoker.hpp

@@ -231,7 +231,7 @@ struct SplitKTwoStageInvoker
                 preprocess = clear_gemm_output;
             }
 
-            return ck_tile::launch_kernel_time_mask(
+            ave_time = ck_tile::launch_kernel_time_mask(
                 s,
                 preprocess,
                 ck_tile::make_kernel<GemmConfig::kBlockPerCu>(
@@ -245,6 +245,8 @@ struct SplitKTwoStageInvoker
                                                   ck_tile::make_tuple(args.N, 1), // Output Stride
                                                   input_tensors,
                                                   static_cast<CDataType*>(c_ptr)));
+
+            return ave_time;
         };
 
         const auto RunSplitk = [&](const auto has_hot_loop_, const auto tail_number_) {

example/ck_tile/03_gemm/universal_gemm_invoker.hpp

@@ -174,10 +174,12 @@ struct UniversalInvoker
                 preprocess = clear_gemm_output;
             }
 
-            return ck_tile::launch_kernel_time_mask(
+            ave_time = ck_tile::launch_kernel_time_mask(
                 s,
                 preprocess,
                 ck_tile::make_kernel<GemmConfig::kBlockPerCu>(Kernel{}, grids, blocks, 0, kargs));
+
+            return ave_time;
         };
 
         const auto RunSplitk = [&](const auto has_hot_loop_, const auto tail_number_) {