clean code

composable_kernel/include/tensor_operation/gridwise_dynamic_gemm_xdlops.hpp

@@ -530,9 +530,9 @@ struct GridwiseDynamicGemm_km_kn_m0m1n0n1_xdlops_v1
                                 decltype(c_m0_m1_m2_n_thread_desc),
                                 decltype(c_m0_m1_m2_n_global_desc),
                                 Sequence<M0, 1, M2, 1>,
-                                Sequence<0, 1, 2, 3>, // CThreadTransferSrcDstAccessOrder,
-                                3,                    // CThreadTransferSrcDstVectorDim,
-                                1,                    // CThreadTransferDstScalarPerVector,
+                                CThreadTransferSrcDstAccessOrder,
+                                CThreadTransferSrcDstVectorDim,
+                                CThreadTransferDstScalarPerVector,
                                 CGlobalMemoryDataOperation,
                                 1,
                                 true>{c_m0_m1_m2_n_global_desc,

composable_kernel/include/tensor_operation/gridwise_dynamic_gemm_xdlops_v2.hpp

@@ -445,9 +445,9 @@ struct GridwiseDynamicGemm_km_kn_m0m1n0n1_xdlops_v2
                                 decltype(c_m0_m1_m2_n_thread_desc),
                                 decltype(c_m0_m1_m2_n_global_desc),
                                 Sequence<M0, 1, M2, 1>,
-                                Sequence<0, 1, 2, 3>, // CThreadTransferSrcDstAccessOrder,
-                                3,                    // CThreadTransferSrcDstVectorDim,
-                                1,                    // CThreadTransferDstScalarPerVector,
+                                CThreadTransferSrcDstAccessOrder,
+                                CThreadTransferSrcDstVectorDim,
+                                CThreadTransferDstScalarPerVector,
                                 CGlobalMemoryDataOperation,
                                 1,
                                 true>{c_m0_m1_m2_n_global_desc,

driver/include/device_dynamic_convolution_forward_implicit_gemm_v4r4_xdlops_nchw_kcyx_nkhw.hpp

@@ -152,8 +152,12 @@ void device_dynamic_convolution_forward_implicit_gemm_v4r4_xdlops_nchw_kcyx_nkhw
 
     for(index_t i = 0; i < 5; ++i)
     {
-        float ave_time = launch_kernel_dynamic_gemm_xdlops_v2<
-            BlockSize,
+#if 0
+        float ave_time = launch_kernel_dynamic_gemm_xdlops_v1
+#else
+        float ave_time = launch_kernel_dynamic_gemm_xdlops_v2
+#endif
+        <BlockSize,
          TInWei,
          TAcc,
          TOut,
@@ -185,8 +189,8 @@ void device_dynamic_convolution_forward_implicit_gemm_v4r4_xdlops_nchw_kcyx_nkhw
          1,
          GemmBBlockTransferSrcScalarPerVector_GemmN,
          GemmBBlockTransferDstScalarPerVector_KPack,
-            false, // don't move back src coordinate after threadwise copy, which will be fused with
-                   // MoveSrcSliceWindow() to save addr computation
+         false, // don't move back src coordinate after threadwise copy, which will be fused
+                // with MoveSrcSliceWindow() to save addr computation
          Sequence<2, 3, 0, 1>,
          3,
          GemmCThreadTransferDstScalarPerVector_GemmN1,