Bug found, intra-row permute off caused

example/01_gemm/gemm_wmma_fp16.cpp

@@ -60,7 +60,7 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmWmma_CShuffle
            true,           
            4,           // C shuffle (M Repeat) Per store
            1,           // C shuffle (N Repeat) Per store
-           S<1, 32, 1,  8>,               
+           S<1, 16, 1,  16>,               
            8>;
 // clang-format on
 

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

@@ -87,11 +87,15 @@ struct DeviceGemmWmma_CShuffle : public DeviceGemm<ALayout,
     static constexpr auto NWaves = NPerBlock / (NRepeat * NPerWmma);
     static constexpr auto WmmaK  = 16;
 
-    static constexpr auto AEnableLds = NWaves == 1 ? false : true;
-    static constexpr auto BEnableLds = MWaves == 1 ? false : true;
-    // Unconditional enable double side LDS if uncommented following
-    // AEnableLds = true;
-    // BEnableLds = true;
+    static constexpr auto AEnableLds_auto = NWaves == 1 ? false : true;
+    static constexpr auto BEnableLds_auto = MWaves == 1 ? false : true;
+
+    // If true, LDS is used unconditionally
+    static constexpr auto AEnableLds_manu = false;
+    static constexpr auto BEnableLds_manu = false;
+
+    static constexpr auto AEnableLds = AEnableLds_auto || AEnableLds_manu;
+    static constexpr auto BEnableLds = BEnableLds_auto || BEnableLds_manu;
 
     static constexpr auto matrix_padder =
         MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};

include/ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_v1.hpp

@@ -360,7 +360,6 @@ struct GridwiseGemmPipeline_v1<1, false, true>
     }
 };
 
-// placeholder
 template <>
 struct GridwiseGemmPipeline_v1<1, true, false>
 {

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp

@@ -1401,6 +1401,7 @@ struct ThreadwiseTensorSliceTransfer_StaticToStatic_InterRow
                 // apply element-wise operation
                 element_op_(v_this_row, src_buf[Number<src_offset>{}]);
 
+                // apply intra-row permute.
                 if constexpr(IntraRowSwizzlePerm)
                 {
                     temp = __builtin_amdgcn_permlane16(