Fix interface tests

test/contraction/test_contraction_interface.cpp

@@ -38,7 +38,7 @@ class ContractionInstanceWrapper
         //#####################################|        |        |        |  Type|  Type|    Type| DataType|           Type|  Type|  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|
         //#####################################|        |        |        |      |      |        |         |               |      |    Operation|   Operation|    Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |                           |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |                           |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|                   _NWaveNPerXdl|
         //#####################################|        |        |        |      |      |        |         |               |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |                           |               |               |          |                |               |               |                           |               |               |          |            |            |                             |                                |
-        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,   F32,   F32,     F32,      F32, ck::Tuple<F32>,   F32,         Pass,        Pass,     Bilinear,       GemmSpec,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>, ABlockTransferSrcVectorDim,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>, BBlockTransferSrcVectorDim,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>, CDEBlockTransferScalarPerVector>;
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,   F32,   F32,     F32,      F32, ck::Tuple<F32>,   F32,         Pass,        Pass,     Bilinear,       GemmSpec,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>, ABlockTransferSrcVectorDim,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>, BBlockTransferSrcVectorDim,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>, CDEBlockTransferScalarPerVector>;
     // clang-format on
 
     bool isSupported(std::vector<ck::index_t>& ADims,

test/grouped_gemm/test_grouped_gemm_util.hpp

@@ -147,14 +147,14 @@ struct DeviceGroupedGemmSplitkInstanceWrapper
             32,
             4,
             2,
-            S<1, 4, 32, 1>,
+            S<1, 4, 16, 1>,
             ABlockTransferThreadClusterArrageOrder,
             ABlockTransferSrcAccessOrder,
             ABlockTransferSrcVectorDim::value,
             ABlockTransferSrcScalarPerVector,
             ABlockTransferDstScalarPerVector_K1::value,
             ABlockLdsAddExtraM::value,
-            S<1, 4, 32, 1>,
+            S<1, 4, 16, 1>,
             BBlockTransferThreadClusterArrageOrder,
             BBlockTransferSrcAccessOrder,
             BBlockTransferSrcVectorDim::value,