use larger vector for B

example/01_gemm/gemm_xdl_fp16.cpp

@@ -44,7 +44,7 @@ using CElementOp = ck::tensor_operation::element_wise::PassThrough;
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
 
 // clang-format off
-#if 0
+#if 1
 using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemm_Xdl_CShuffle
 //######| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
 //######|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|

example/01_gemm/gemm_xdl_fp16_splitk.cpp

@@ -50,7 +50,7 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdlSplitKCShu
 //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Spacialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
 //#########################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
 //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-                           <  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,    16,   128,     4,  8,   16,   16,    1,    2,  S<1, 4, 16, 4>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              2,              2,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 16, 1, 16>,               4>;
+                           <  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,    16,   128,     4,  8,   16,   16,    1,    2,  S<1, 4, 16, 4>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              2,              2,      true,  S<1, 4, 32, 2>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              4,      true,           1,           1,                   S<1, 16, 1, 16>,               4>;
                            //<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,    16,   256,     4,  8,   16,   16,    1,    4,  S<1, 4, 16, 4>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              2,              2,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 16>,               2>;
 
 

include/ck/tensor_operation/gpu/device/device_gemm_xdl_splitk_c_shuffle.hpp

@@ -245,7 +245,8 @@ struct DeviceGemmXdlSplitKCShuffle
         CShuffleMRepeatPerShuffle,
         CShuffleNRepeatPerShuffle,
         CBlockTransferScalarPerVector_NWaveNPerXDL,
-        CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock>;
+        CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
+        true>;
 
     // GridwiseGemm
     using GridwiseGemmAtomicAdd = GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2<
@@ -287,7 +288,8 @@ struct DeviceGemmXdlSplitKCShuffle
         CShuffleMRepeatPerShuffle,
         CShuffleNRepeatPerShuffle,
         CBlockTransferScalarPerVector_NWaveNPerXDL,
-        CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock>;
+        CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
+        true>;
 
     using CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock =
         decltype(GridwiseGemm::MakeCGridDesc_MBlock_MPerBlock_NBlock_NPerBlock(CGridDesc_M_N{}));