Fix RCR gemm bug

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_wmma_cshuffle.hpp

@@ -121,12 +121,12 @@ struct DeviceGemmMultipleD_Wmma_CShuffle : public DeviceGemmMultipleD<ALayout,
     static auto MakeBGridDescriptor_K0_N_K1(index_t KRaw, index_t NRaw, index_t StrideB)
     {
         const auto b_grid_desc_nraw_kraw = [&]() {
-            if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
+            if constexpr(is_same_v<tensor_layout::gemm::RowMajor, BLayout>)
             {
                 return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
                                                     make_tuple(I1, StrideB));
             }
-            else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
+            else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, BLayout>)
             {
                 return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
                                                     make_tuple(StrideB, I1));

include/ck/tensor_operation/gpu/device/impl/device_gemm_wmma.hpp

@@ -114,12 +114,12 @@ struct DeviceGemmWmma_CShuffle : public DeviceGemm<ALayout,
     static auto MakeBGridDescriptor_K0_N_K1(index_t KRaw, index_t NRaw, index_t StrideB)
     {
         const auto b_grid_desc_nraw_kraw = [&]() {
-            if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
+            if constexpr(is_same_v<tensor_layout::gemm::RowMajor, BLayout>)
             {
                 return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
                                                     make_tuple(I1, StrideB));
             }
-            else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
+            else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, BLayout>)
             {
                 return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
                                                     make_tuple(StrideB, I1));

include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_wmma_cshuffle.hpp

@@ -716,7 +716,6 @@ struct GridwiseGemmMultipleD_k0mk1_k0nk1_mn_wmma_cshuffle
                                                           c_thread_buf,
                                                           K0BlockMainLoop);
 /*******************************************************************************/
-        //printf("safe 1");
         // write out to C, implement shuffle
         {
             constexpr auto c_thread_desc_mrepeat_mwave_msubgroup_nrepeat_nwave_nthreadpersubgroup_maccvgprs =