Merge remote-tracking branch 'origin/develop' into aosewski/drop_cshuffle

example/13_pool2d_fwd/pool2d_fwd_fp16.cpp

@@ -2,7 +2,6 @@
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <iostream>
-#include <cstdlib>
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
@@ -10,9 +9,9 @@
 
 #include "pool2d_fwd_common.hpp"
 
-using InDataType  = ck::half_t;
-using OutDataType = ck::half_t;
-using AccDataType = float;
+using InDataType      = ck::half_t;
+using OutDataType     = ck::half_t;
+using ComputeDataType = float;
 
 using IndexDataType = int32_t;
 
@@ -91,7 +90,7 @@ int main(int argc, char* argv[])
 
     bool pass = pool_test<InDataType,
                           OutDataType,
-                          AccDataType,
+                          ComputeDataType,
                           IndexDataType,
                           InLayout,
                           OutLayout,

example/13_pool2d_fwd/pool2d_fwd_fp32.cpp

@@ -2,7 +2,6 @@
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <iostream>
-#include <cstdlib>
 
 #include "ck/ck.hpp"
 #include "ck/utility/reduction_enums.hpp"
@@ -10,9 +9,9 @@
 
 #include "pool2d_fwd_common.hpp"
 
-using InDataType  = float;
-using OutDataType = float;
-using AccDataType = float;
+using InDataType      = float;
+using OutDataType     = float;
+using ComputeDataType = float;
 
 using IndexDataType = int32_t;
 
@@ -91,7 +90,7 @@ int main(int argc, char* argv[])
 
     bool pass = pool_test<InDataType,
                           OutDataType,
-                          AccDataType,
+                          ComputeDataType,
                           IndexDataType,
                           InLayout,
                           OutLayout,

example/14_gemm_quantization/CMakeLists.txt

-add_example_executable(example_gemm_xdl_bias_relu_quantization_int8 gemm_xdl_bias_relu_quantization_int8.cpp)
-add_example_executable(example_gemm_xdl_quantization_int8 gemm_xdl_quantization_int8.cpp)
\ No newline at end of file
+# dlops
+add_example_executable(example_gemm_dl_quantization_int8 gemm_dl_quantization_int8.cpp)
+
+# xdlops
+if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR GPU_TARGETS MATCHES "gfx940")
+   add_example_executable(example_gemm_xdl_bias_relu_quantization_int8 gemm_xdl_bias_relu_quantization_int8.cpp)
+   add_example_executable(example_gemm_xdl_quantization_int8 gemm_xdl_quantization_int8.cpp)
+endif()
\ No newline at end of file

example/14_gemm_quantization/gemm_dl_quantization_int8.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <iostream>
+#include <numeric>
+#include <initializer_list>
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_dl.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/utility/device_memory.hpp"
+#include "ck/library/utility/host_tensor.hpp"
+#include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
+#include "ck/library/utility/check_err.hpp"
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using I8  = int8_t;
+using I32 = int32_t;
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+using PassThrough  = ck::tensor_operation::element_wise::PassThrough;
+using AElementOp   = PassThrough;
+using BElementOp   = PassThrough;
+using ActivationOp = PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Activation_Mul_Clamp<ActivationOp>;
+
+using ADataType        = I8;
+using BDataType        = I8;
+using AccDataType      = I32;
+using CShuffleDataType = I32;
+using DsDataType       = ck::Tuple<>;
+using EDataType        = I8;
+
+using ALayout  = Row;
+using BLayout  = Col;
+using DsLayout = ck::Tuple<>;
+using ELayout  = Row;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+
+using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmMultipleD_Dl<
+    ALayout,
+    BLayout,
+    DsLayout,
+    ELayout,
+    ADataType,
+    BDataType,
+    AccDataType,
+    DsDataType,
+    EDataType,
+    AElementOp,
+    BElementOp,
+    CDEElementOp,
+    GemmDefault,
+    256,                 // BlockSize
+    128,                 // MPerBlock
+    128,                 // NPerBlock
+    16,                  // K0PerBlock
+    4,                   // K1
+    4,                   // M1PerThread
+    4,                   // N1PerThread
+    1,                   // KPerThread
+    S<8, 2>,             // M1N1ThreadClusterM1Xs
+    S<8, 2>,             // M1N1ThreadClusterN1Xs
+    S<8, 1, 1, 4>,       // ABlockTransferThreadSliceLengths_K0_M0_M1_K1
+    S<2, 1, 128, 1>,     // ABlockTransferThreadClusterLengths_K0_M0_M1_K1
+    S<1, 2, 0, 3>,       // ABlockTransferThreadClusterArrangeOrder
+    S<1, 2, 0, 3>,       // ABlockTransferSrcAccessOrder
+    S<4, 1, 1, 4>,       // ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1
+    S<1, 2, 0, 3>,       // ABlockTransferSrcVectorTensorContiguousDimOrder
+    S<1, 1, 1, 4>,       // ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1
+    S<8, 1, 1, 4>,       // BBlockTransferThreadSliceLengths_K0_N0_N1_K1
+    S<2, 1, 128, 1>,     // BBlockTransferThreadClusterLengths_K0_N0_N1_K1
+    S<1, 2, 0, 3>,       // BBlockTransferThreadClusterArrangeOrder
+    S<1, 2, 0, 3>,       // BBlockTransferSrcAccessOrder
+    S<4, 1, 1, 4>,       // BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1
+    S<1, 2, 0, 3>,       // BBlockTransferSrcVectorTensorContiguousDimOrder
+    S<1, 1, 1, 4>,       // BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1
+    S<0, 1, 2, 3, 4, 5>, // CThreadTransferSrcDstAccessOrder
+    5,                   // CThreadTransferSrcDstVectorDim
+    4>;                  // CThreadTransferDstScalarPerVector
+
+using ReferenceGemmInstance = ck::tensor_operation::host::
+    ReferenceGemm<ADataType, BDataType, EDataType, float, PassThrough, PassThrough, CDEElementOp>;
+
+int main()
+{
+    bool do_verification = true;
+    bool time_kernel     = false;
+
+    // GEMM shape
+    ck::index_t M = 1024;
+    ck::index_t N = 1024;
+    ck::index_t K = 1024;
+
+    ck::index_t StrideA = 1024;
+    ck::index_t StrideB = 1024;
+    ck::index_t StrideE = 1024;
+
+    float requant_scale = 0.03;
+
+    auto f_host_tensor_descriptor =
+        [](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
+            using namespace ck::literals;
+
+            if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::value)
+            {
+                return HostTensorDescriptor(std::vector<std::size_t>({row, col}),
+                                            std::vector<std::size_t>({stride, 1_uz}));
+            }
+            else
+            {
+                return HostTensorDescriptor(std::vector<std::size_t>({row, col}),
+                                            std::vector<std::size_t>({1_uz, stride}));
+            }
+        };
+
+    Tensor<ADataType> a_m_k(f_host_tensor_descriptor(M, K, StrideA, ALayout{}));
+    Tensor<BDataType> b_k_n(f_host_tensor_descriptor(K, N, StrideB, BLayout{}));
+    Tensor<EDataType> e_m_n_host_result(f_host_tensor_descriptor(M, N, StrideE, ELayout{}));
+    Tensor<EDataType> e_m_n_device_result(f_host_tensor_descriptor(M, N, StrideE, ELayout{}));
+
+    std::cout << "a_m_k: " << a_m_k.mDesc << std::endl;
+    std::cout << "b_k_n: " << b_k_n.mDesc << std::endl;
+    std::cout << "e_m_n: " << e_m_n_host_result.mDesc << std::endl;
+
+    a_m_k.GenerateTensorValue(GeneratorTensor_2<ADataType>{-5, 5});
+    b_k_n.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
+
+    DeviceMem a_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpaceSize());
+    DeviceMem b_device_buf(sizeof(BDataType) * b_k_n.mDesc.GetElementSpaceSize());
+    DeviceMem e_device_buf(sizeof(EDataType) * e_m_n_device_result.mDesc.GetElementSpaceSize());
+
+    a_device_buf.ToDevice(a_m_k.mData.data());
+    b_device_buf.ToDevice(b_k_n.mData.data());
+
+    auto a_element_op   = AElementOp{};
+    auto b_element_op   = BElementOp{};
+    auto cde_element_op = CDEElementOp{requant_scale, ActivationOp{}};
+
+    // do GEMM
+    auto gemm     = DeviceGemmInstance{};
+    auto invoker  = gemm.MakeInvoker();
+    auto argument = gemm.MakeArgument(a_device_buf.GetDeviceBuffer(),
+                                      b_device_buf.GetDeviceBuffer(),
+                                      {},
+                                      e_device_buf.GetDeviceBuffer(),
+                                      M,
+                                      N,
+                                      K,
+                                      StrideA,
+                                      StrideB,
+                                      {},
+                                      StrideE,
+                                      a_element_op,
+                                      b_element_op,
+                                      cde_element_op);
+
+    if(!gemm.IsSupportedArgument(argument))
+    {
+        throw std::runtime_error(
+            "wrong! device_gemm with the specified compilation parameters does "
+            "not support this GEMM problem");
+    }
+
+    float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
+
+    std::size_t flop = std::size_t(2) * M * N * K;
+    std::size_t num_btype =
+        sizeof(ADataType) * M * K + sizeof(BDataType) * K * N + sizeof(EDataType) * M * N;
+
+    float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
+
+    float gb_per_sec = num_btype / 1.E6 / ave_time;
+
+    std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
+              << gemm.GetTypeString() << std::endl;
+
+    e_device_buf.FromDevice(e_m_n_device_result.mData.data());
+
+    if(do_verification)
+    {
+        auto ref_gemm    = ReferenceGemmInstance{};
+        auto ref_invoker = ref_gemm.MakeInvoker();
+
+        auto ref_argument = ref_gemm.MakeArgument(
+            a_m_k, b_k_n, e_m_n_host_result, a_element_op, b_element_op, cde_element_op);
+
+        ref_invoker.Run(ref_argument);
+
+        return ck::utils::check_err(e_m_n_device_result, e_m_n_host_result) ? 0 : 1;
+    }
+
+    return 0;
+}

example/15_grouped_gemm/CMakeLists.txt

@@ -4,12 +4,17 @@ add_example_executable(example_grouped_gemm_xdl_fp32 grouped_gemm_xdl_fp32.cpp)
 add_example_executable(example_grouped_gemm_xdl_fp16 grouped_gemm_xdl_fp16.cpp)
 add_example_executable(example_grouped_gemm_xdl_bfp16 grouped_gemm_xdl_bfp16.cpp)
 add_example_executable(example_grouped_gemm_xdl_int8 grouped_gemm_xdl_int8.cpp)
+add_example_executable(example_grouped_gemm_multiple_d_dl_fp16 grouped_gemm_multiple_d_dl_fp16.cpp)
+add_example_executable(example_grouped_gemm_xdl_splitk_fp16 grouped_gemm_xdl_splitk_fp16.cpp)
+
 
 add_dependencies(example_grouped_gemm_xdl
                  example_grouped_gemm_xdl_fp32
                  example_grouped_gemm_xdl_fp16
                  example_grouped_gemm_xdl_bfp16
-                 example_grouped_gemm_xdl_int8)
+                 example_grouped_gemm_xdl_int8
+                 example_grouped_gemm_multiple_d_dl_fp16
+                 example_grouped_gemm_xdl_splitk_fp16)
 
 if(USE_BITINT_EXTENSION_INT4)
   add_example_executable(example_grouped_gemm_xdl_int4 grouped_gemm_xdl_int4.cpp)

example/15_grouped_gemm/grouped_gemm_multiple_d_dl_fp16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <algorithm>
+#include <cstddef>
+#include <initializer_list>
+#include <iostream>
+#include <numeric>
+#include <stdexcept>
+#include <string>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_dl.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/utility/check_err.hpp"
+#include "ck/library/utility/device_memory.hpp"
+#include "ck/library/utility/host_tensor.hpp"
+#include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
+#include "ck/utility/data_type.hpp"
+#include "ck/utility/tuple.hpp"
+#include "ck/utility/sequence.hpp"
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+using ADataType   = F16;
+using BDataType   = F16;
+using AccDataType = F32;
+using DsDataType  = ck::Tuple<>;
+using EDataType   = F16;
+
+using ALayout  = Row;
+using BLayout  = Row;
+using DsLayout = ck::Tuple<>;
+using ELayout  = Row;
+
+using AElementOp   = PassThrough;
+using BElementOp   = PassThrough;
+using CDEElementOp = PassThrough;
+
+static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNPadding;
+
+// clang-format off
+using DeviceGemmInstance = ck::tensor_operation::device::
+        //   ##################| ALayout| BLayout| DsLayout| ELayout|     AData|     BData|     AccData|     DsData|     EData|           A|           B|          CDE|           GEMM| Block|  MPer|  NPer| K0Per| K1|      M1Per|      N1Per|   KPer|  M11N11Thread|  M11N11Thread|     ABlockTransfer|       ABlockTransfer| ABlockTransfer| ABlockTransfer|      ABlockTransfer|     ABlockTransfer|       ABlockTransfer|     BBlockTransfer|       BBlockTransfer| BBlockTransfer| BBlockTransfer|      BBlockTransfer|     BBlockTransfer|       BBlockTransfer|     CThreadTransfer|  CThreadTransfer|    CThreadTransfer|
+        //   ##################|        |        |         |        |      Type|      Type|        Type|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization|  Size| Block| Block| Block|   | ThreadM111| ThreadN111| Thread| ClusterM110Xs| ClusterN110Xs| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor|        SrcDstAccess|  SrcDstVectorDim| DstScalarPerVector|
+        //   ##################|        |        |         |        |          |          |            |           |          |   Operation|   Operation|    Operation|               |      |      |      |      |   |           |           |       |              |              |        K0_M0_M1_K1|          K0_M0_M1_K1|   ArrangeOrder|          Order| Lengths_K0_M0_M1_K1| ContiguousDimOrder|  Lengths_K0_M0_M1_K1|        K0_N0_N1_K1|          K0_N0_N1_K1|   ArrangeOrder|          Order| Lengths_K0_N0_N1_K1| ContiguousDimOrder|  Lengths_K0_N0_N1_K1|               Order|                 |                   |
+        //   ##################|        |        |         |        |          |          |            |           |          |            |            |             |               |      |      |      |      |   |           |           |       |              |              |                   |                     |               |               |                    |                   |                     |                   |                     |               |               |                    |                   |                     |                    |                 |                   |
+  DeviceGroupedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,   256,   128,   128,    16,  2,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 4, 2>,      S<8, 1,  32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>;
+// clang-format on
+
+#include "run_grouped_gemm_example.inc"
+
+int main(int argc, char* argv[]) { return !run_grouped_gemm_example(argc, argv); }

example/15_grouped_gemm/grouped_gemm_xdl_splitk_fp16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <iostream>
+#include <numeric>
+#include <initializer_list>
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_splitk_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/utility/check_err.hpp"
+#include "ck/library/utility/device_memory.hpp"
+#include "ck/library/utility/host_tensor.hpp"
+#include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/literals.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+using ADataType        = F16;
+using BDataType        = F16;
+using AccDataType      = F32;
+using CShuffleDataType = F16;
+using DsDataType       = ck::Tuple<>;
+using EDataType        = F16;
+
+using ALayout  = Row;
+using BLayout  = Col;
+using DsLayout = ck::Tuple<>;
+using ELayout  = Row;
+
+using AElementOp   = PassThrough;
+using BElementOp   = PassThrough;
+using CDEElementOp = PassThrough;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+using DeviceGemmInstance = ck::tensor_operation::device::DeviceGroupedGemmXdlSplitKCShuffle
+    // clang-format off
+//######| ALayout| BLayout| DsLayout| ELayout|     AData|     BData|     AccData|         CShuffle|     DsData|     EData|           A|           B|          CDE|           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|         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|
+//######|        |        |         |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        < ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>;
+// clang-format on
+
+#include "run_grouped_gemm_example.inc"
+
+int main(int argc, char* argv[])
+{
+    ProblemSize problem_size;
+    ExecutionConfig config;
+
+    problem_size.group_count = 16;
+
+    problem_size.Ms = {
+        167, 183, 177, 181, 153, 139, 156, 173, 163, 150, 204, 184, 168, 156, 168, 148};
+
+    for(int i = 0; i < problem_size.group_count; i++)
+    {
+        problem_size.Ns.push_back(768);
+        problem_size.Ks.push_back(4608);
+
+        problem_size.stride_As.push_back(problem_size.Ks[i]);
+        problem_size.stride_Bs.push_back(problem_size.Ks[i]);
+        problem_size.stride_Cs.push_back(problem_size.Ns[i]);
+    }
+
+    if(argc == 4)
+    {
+        config.do_verification = std::stoi(argv[1]);
+        config.init_method     = std::stoi(argv[2]);
+        config.time_kernel     = std::stoi(argv[3]);
+    }
+    else
+    {
+        printf("arg1: verification (0=no, 1=yes)\n");
+        printf("arg2: initialization (0=no init, 1=integer value, 2=decimal value)\n");
+        printf("arg3: time kernel (0=n0, 1=yes)\n");
+        exit(0);
+    }
+
+    return !run_grouped_gemm(problem_size, config);
+}

example/15_grouped_gemm/run_grouped_gemm_example.inc

@@ -147,6 +147,7 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
 #else
         a_tensors_device[i]->ToDevice(a_tensors[i].mData.data());
         b_tensors_device[i]->ToDevice(b_tensors[i].mData.data());
+        c_tensors_device[i]->SetZero();
 #endif
 
         p_a.push_back(a_tensors_device[i]->GetDeviceBuffer());

example/16_gemm_multi_d_multi_reduces/CMakeLists.txt

-add_custom_target(example_gemm_reduce_xdl)
-add_custom_target(example_gemm_reduce_xdl_max)
-add_custom_target(example_gemm_reduce_xdl_mean_meansquare)
-add_custom_target(example_gemm_add_add_mean_meansquare_xdl)
+if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR GPU_TARGETS MATCHES "gfx940")
+   add_custom_target(example_gemm_reduce_xdl)
+   add_custom_target(example_gemm_reduce_xdl_max)
+   add_custom_target(example_gemm_reduce_xdl_mean_meansquare)
+   add_custom_target(example_gemm_add_add_mean_meansquare_xdl)
 
-add_example_executable(example_gemm_max_xdl_fp16 gemm_max_xdl_fp16.cpp)
-add_example_executable(example_gemm_max_xdl_int8 gemm_max_xdl_int8.cpp)
-add_example_executable(example_gemm_max_xdl_fp32 gemm_max_xdl_fp32.cpp)
-add_example_executable(example_gemm_max_xdl_bf16 gemm_max_xdl_bf16.cpp)
+   add_example_executable(example_gemm_max_xdl_fp16 gemm_max_xdl_fp16.cpp)
+   add_example_executable(example_gemm_max_xdl_int8 gemm_max_xdl_int8.cpp)
+   add_example_executable(example_gemm_max_xdl_fp32 gemm_max_xdl_fp32.cpp)
+   add_example_executable(example_gemm_max_xdl_bf16 gemm_max_xdl_bf16.cpp)
 
-add_example_executable(example_gemm_add_add_mean_meansquare_xdl_fp16 gemm_add_add_mean_meansquare_xdl_fp16.cpp)
+   add_example_executable(example_gemm_add_add_mean_meansquare_xdl_fp16 gemm_add_add_mean_meansquare_xdl_fp16.cpp)
 
-add_example_executable(example_gemm_mean_meansquare_xdl_fp16 gemm_mean_meansquare_xdl_fp16.cpp)
-add_example_executable(example_gemm_mean_meansquare_xdl_fp32 gemm_mean_meansquare_xdl_fp32.cpp)
-add_example_executable(example_gemm_mean_meansquare_xdl_bf16 gemm_mean_meansquare_xdl_bf16.cpp)
-add_example_executable(example_gemm_add_addsquare_xdl_int8 gemm_add_addsquare_xdl_int8.cpp)
+   add_example_executable(example_gemm_mean_meansquare_xdl_fp16 gemm_mean_meansquare_xdl_fp16.cpp)
+   add_example_executable(example_gemm_mean_meansquare_xdl_fp32 gemm_mean_meansquare_xdl_fp32.cpp)
+   add_example_executable(example_gemm_mean_meansquare_xdl_bf16 gemm_mean_meansquare_xdl_bf16.cpp)
+   add_example_executable(example_gemm_add_addsquare_xdl_int8 gemm_add_addsquare_xdl_int8.cpp)
 
-add_dependencies(example_gemm_reduce_xdl_max
+   add_dependencies(example_gemm_reduce_xdl_max
                  example_gemm_max_xdl_bf16
                  example_gemm_max_xdl_fp16
                  example_gemm_max_xdl_fp32
                  example_gemm_max_xdl_int8)
 
-add_dependencies(example_gemm_reduce_xdl_mean_meansquare
+   add_dependencies(example_gemm_reduce_xdl_mean_meansquare
                  example_gemm_mean_meansquare_xdl_fp16
                  example_gemm_mean_meansquare_xdl_fp32
                  example_gemm_mean_meansquare_xdl_bf16
                  example_gemm_add_addsquare_xdl_int8)
 
-add_dependencies(example_gemm_add_add_mean_meansquare_xdl example_gemm_add_add_mean_meansquare_xdl_fp16)
+   add_dependencies(example_gemm_add_add_mean_meansquare_xdl example_gemm_add_add_mean_meansquare_xdl_fp16)
 
-add_dependencies(example_gemm_reduce_xdl
+   add_dependencies(example_gemm_reduce_xdl
                  example_gemm_reduce_xdl_mean_meansquare
                  example_gemm_reduce_xdl_max
                  example_gemm_add_add_mean_meansquare_xdl)
 
-if(USE_BITINT_EXTENSION_INT4)
-  add_example_executable(example_gemm_max_xdl_int4 gemm_max_xdl_int4.cpp)
-  add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_int4)
+   if(USE_BITINT_EXTENSION_INT4)
+      add_example_executable(example_gemm_max_xdl_int4 gemm_max_xdl_int4.cpp)
+      add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_int4)
+   endif()
 endif()

example/17_convnd_bwd_data/CMakeLists.txt

-add_example_executable(example_convnd_bwd_data_xdl_fp16 convnd_bwd_data_xdl_fp16.cpp)
-target_link_libraries(example_convnd_bwd_data_xdl_fp16 PRIVATE utility)
-
+if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR GPU_TARGETS MATCHES "gfx940")
+   add_example_executable(example_convnd_bwd_data_xdl_fp16 convnd_bwd_data_xdl_fp16.cpp)
+   target_link_libraries(example_convnd_bwd_data_xdl_fp16 PRIVATE utility)
+endif()
 add_example_executable(example_convnd_bwd_data_dl_fp16 convnd_bwd_data_dl_fp16.cpp)
 target_link_libraries(example_convnd_bwd_data_dl_fp16 PRIVATE utility)

example/18_batched_gemm_reduce/CMakeLists.txt

-add_example_executable(example_batched_gemm_reduce_xdl_fp16 batched_gemm_reduce_xdl_fp16.cpp)
+if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR GPU_TARGETS MATCHES "gfx940")
+   add_example_executable(example_batched_gemm_reduce_xdl_fp16 batched_gemm_reduce_xdl_fp16.cpp)
+endif()
 

example/19_binary_elementwise/broadcast_add_2d_amn_bn.cpp

@@ -6,7 +6,7 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/element/binary_element_wise_operation.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_elementwise.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_elementwise_impl.hpp"
 
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
@@ -23,13 +23,13 @@ using CDataType  = F16;
 using Add = ck::tensor_operation::element_wise::Add;
 
 using DeviceElementwiseAddInstance =
-    ck::tensor_operation::device::DeviceElementwise<ck::Tuple<ABDataType, ABDataType>,
-                                                    ck::Tuple<CDataType>,
-                                                    Add,
-                                                    2,
-                                                    8,
-                                                    ck::Sequence<8, 8>,
-                                                    ck::Sequence<8>>;
+    ck::tensor_operation::device::DeviceElementwiseImpl<ck::Tuple<ABDataType, ABDataType>,
+                                                        ck::Tuple<CDataType>,
+                                                        Add,
+                                                        2,
+                                                        8,
+                                                        ck::Sequence<8, 8>,
+                                                        ck::Sequence<8>>;
 
 template <typename HostTensorA,
           typename HostTensorB,

example/19_binary_elementwise/broadcast_add_3d_am_bmnk.cpp

@@ -6,7 +6,7 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/element/binary_element_wise_operation.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_elementwise.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_elementwise_impl.hpp"
 
 #include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
@@ -23,13 +23,13 @@ using CDataType  = F16;
 using Add = ck::tensor_operation::element_wise::Add;
 
 using DeviceElementwiseAddInstance =
-    ck::tensor_operation::device::DeviceElementwise<ck::Tuple<ABDataType, ABDataType>,
-                                                    ck::Tuple<CDataType>,
-                                                    Add,
-                                                    3,
-                                                    8,
-                                                    ck::Sequence<1, 8>,
-                                                    ck::Sequence<8>>;
+    ck::tensor_operation::device::DeviceElementwiseImpl<ck::Tuple<ABDataType, ABDataType>,
+                                                        ck::Tuple<CDataType>,
+                                                        Add,
+                                                        3,
+                                                        8,
+                                                        ck::Sequence<1, 8>,
+                                                        ck::Sequence<8>>;
 
 template <typename HostTensorA, typename HostTensorB, typename HostTensorC, typename Functor>
 void host_broadcast3D_am_bmnk(HostTensorC& C,

example/19_binary_elementwise/elementwise_add_1d.cpp

@@ -5,7 +5,7 @@
 #include <cstdlib>
 
 #include "ck/ck.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_elementwise.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_elementwise_impl.hpp"
 #include "ck/tensor_operation/gpu/element/binary_element_wise_operation.hpp"
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
@@ -21,13 +21,13 @@ using CDataType  = F16;
 using Add = ck::tensor_operation::element_wise::Add;
 
 using DeviceElementwiseAddInstance =
-    ck::tensor_operation::device::DeviceElementwise<ck::Tuple<ABDataType, ABDataType>,
-                                                    ck::Tuple<CDataType>,
-                                                    Add,
-                                                    1,
-                                                    8,
-                                                    ck::Sequence<8, 8>,
-                                                    ck::Sequence<8>>;
+    ck::tensor_operation::device::DeviceElementwiseImpl<ck::Tuple<ABDataType, ABDataType>,
+                                                        ck::Tuple<CDataType>,
+                                                        Add,
+                                                        1,
+                                                        8,
+                                                        ck::Sequence<8, 8>,
+                                                        ck::Sequence<8>>;
 
 template <typename HostTensorA, typename HostTensorB, typename HostTensorC, typename Functor>
 void host_elementwise1D(

example/19_binary_elementwise/elementwise_add_4d.cpp

@@ -6,7 +6,7 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/element/binary_element_wise_operation.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_elementwise.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_elementwise_impl.hpp"
 
 #include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
@@ -23,13 +23,13 @@ using CDataType  = F16;
 using Add = ck::tensor_operation::element_wise::Add;
 
 using DeviceElementwiseAddInstance =
-    ck::tensor_operation::device::DeviceElementwise<ck::Tuple<ABDataType, ABDataType>,
-                                                    ck::Tuple<CDataType>,
-                                                    Add,
-                                                    4,
-                                                    8,
-                                                    ck::Sequence<8, 8>,
-                                                    ck::Sequence<8>>;
+    ck::tensor_operation::device::DeviceElementwiseImpl<ck::Tuple<ABDataType, ABDataType>,
+                                                        ck::Tuple<CDataType>,
+                                                        Add,
+                                                        4,
+                                                        8,
+                                                        ck::Sequence<8, 8>,
+                                                        ck::Sequence<8>>;
 
 template <typename HostTensorA, typename HostTensorB, typename HostTensorC, typename Functor>
 void host_elementwise4D(HostTensorC& C,

example/20_grouped_conv_bwd_weight/CMakeLists.txt

-add_custom_target(example_grouped_conv_bwd_weight)
+if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR GPU_TARGETS MATCHES "gfx940")
+   add_custom_target(example_grouped_conv_bwd_weight)
 
-add_example_executable(example_grouped_conv_bwd_weight_xdl_fp16 grouped_conv_bwd_weight_xdl_fp16.cpp)
-add_example_executable(example_grouped_conv_bwd_weight_xdl_bf16 grouped_conv_bwd_weight_xdl_bf16.cpp)
+   add_example_executable(example_grouped_conv_bwd_weight_xdl_fp16 grouped_conv_bwd_weight_xdl_fp16.cpp)
+   add_example_executable(example_grouped_conv_bwd_weight_xdl_bf16 grouped_conv_bwd_weight_xdl_bf16.cpp)
 
-
-add_dependencies(example_grouped_conv_bwd_weight example_grouped_conv_bwd_weight_xdl_fp16
+   add_dependencies(example_grouped_conv_bwd_weight example_grouped_conv_bwd_weight_xdl_fp16
                                                  example_grouped_conv_bwd_weight_xdl_bf16)
+endif()
+
+add_custom_target(example_grouped_conv_bwd_weight_dl)
+
+add_example_executable(example_grouped_conv_bwd_weight_dl_fp16 grouped_conv_bwd_weight_dl_fp16.cpp)
+
+add_dependencies(example_grouped_conv_bwd_weight_dl example_grouped_conv_bwd_weight_dl_fp16)

example/20_grouped_conv_bwd_weight/common.hpp

@@ -9,7 +9,6 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/convolution_backward_weight_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_gnwc_gkxc_gnwk_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 

example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_dl_fp16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "common.hpp"
+
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_gnwc_gkxc_gnwk_dl.hpp"
+
+using InDataType  = F16;
+using WeiDataType = F16;
+using OutDataType = F16;
+using AccDataType = F32;
+
+using InElementOp  = PassThrough;
+using WeiElementOp = PassThrough;
+using OutElementOp = PassThrough;
+
+template <ck::index_t NDimSpatial>
+using DeviceConvBwdWeightInstance =
+    ck::tensor_operation::device::DeviceGroupedConvBwdWeightGnwcGkxcGnwk_Dl<
+        NDimSpatial,          // NDimSpatial
+        InDataType,           // InDataType
+        WeiDataType,          // WeiDataType
+        OutDataType,          // OutDataType
+        AccDataType,          // AccDataType
+        InElementOp,          // InElementwiseOperation
+        WeiElementOp,         // WeiElementwiseOperation
+        OutElementOp,         // OutElementwiseOperation
+        ConvBwdWeightDefault, // ConvBackwardWeightSpecialization
+        256,                  // BlockSize
+        128,                  // MPerBlock
+        128,                  // NPerBlock
+        16,                   // K0PerBlock
+        2,                    // K1
+        4,                    // M1PerThread
+        4,                    // N1PerThread
+        1,                    // KPerThread
+        S<8, 2>,              // M1N1ThreadClusterM1Xs
+        S<8, 2>,              // M1N1ThreadClusterN1Xs
+        S<1, 8, 1, 1, 2>,     // ABlockTransferThreadSliceLengths_K0_M0_M1_K1
+        S<1, 2, 1, 128, 1>,   // ABlockTransferThreadClusterLengths_K0_M0_M1_K1
+        S<0, 2, 3, 1, 4>,     // ABlockTransferThreadClusterArrangeOrder
+        S<0, 2, 3, 1, 4>,     // ABlockTransferSrcAccessOrder
+        S<1, 1, 1, 1, 1>,     // ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1
+        S<0, 2, 3, 1, 4>,     // ABlockTransferSrcVectorTensorContiguousDimOrder
+        S<1, 1, 1, 1, 1>,     // ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1
+        S<1, 1, 1, 8, 2>,     // BBlockTransferThreadSliceLengths_K0_N0_N1_K1
+        S<1, 16, 1, 16, 1>,   // BBlockTransferThreadClusterLengths_K0_N0_N1_K1
+        S<0, 1, 4, 2, 3>,     // BBlockTransferThreadClusterArrangeOrder
+        S<0, 1, 4, 2, 3>,     // BBlockTransferSrcAccessOrder
+        S<1, 1, 1, 8, 1>,     // BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1
+        S<0, 1, 4, 2, 3>,     // BBlockTransferSrcVectorTensorContiguousDimOrder
+        S<1, 1, 1, 1, 2>,     // BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1
+        S<0, 1, 2, 3, 4, 5>,  // CThreadTransferSrcDstAccessOrder
+        5,                    // CThreadTransferSrcDstVectorDim
+        4>;                   // CThreadTransferDstScalarPerVector
+
+#include "run_grouped_conv_bwd_weight_example.inc"
+
+int main(int argc, char* argv[]) { return !run_grouped_conv_bwd_weight_example(argc, argv); }

example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_bf16.cpp

@@ -3,6 +3,8 @@
 
 #include "common.hpp"
 
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_gnwc_gkxc_gnwk_xdl_cshuffle.hpp"
+
 using InDataType = BF16;
 // bf16 kernel use fp32 atomic add to accumulate Weight tensor into global memory
 using WeiDataType = F32;
@@ -13,6 +15,46 @@ using InElementOp  = PassThrough;
 using WeiElementOp = PassThrough;
 using OutElementOp = PassThrough;
 
+template <ck::index_t NDimSpatial>
+using DeviceConvBwdWeightInstance =
+    ck::tensor_operation::device::DeviceGroupedConvBwdWeightGnwcGkxcGnwk_Xdl_CShuffle<
+        NDimSpatial,          // NDimSpatial
+        InDataType,           // InDataType
+        WeiDataType,          // WeiDataType
+        OutDataType,          // OutDataType
+        AccDataType,          // AccDataType
+        InElementOp,          // InElementwiseOperation
+        WeiElementOp,         // WeiElementwiseOperation
+        OutElementOp,         // OutElementwiseOperation
+        ConvBwdWeightDefault, // ConvolutionBackwardWeightSpecialization
+        256,                  // BlockSize
+        128,                  // MPerBlock
+        128,                  // NPerBlock
+        4,                    // K0PerBlock
+        8,                    // K1
+        32,                   // MPerXdl
+        32,                   // NPerXdl
+        2,                    // MXdlPerWave
+        2,                    // NXdlPerWave
+        S<1, 4, 16, 4>,       // ABlockTransferThreadClusterLengths_K0_M_K1
+        S<0, 3, 1, 2>,        // ABlockTransferThreadClusterArrangeOrder
+        S<0, 2, 1, 3>,        // ABlockTransferSrcAccessOrder
+        2,                    // ABlockTransferSrcVectorDim
+        8,                    // ABlockTransferSrcScalarPerVector
+        2,                    // ABlockTransferDstScalarPerVector_K1
+        true,                 // ABlockLdsAddExtraM
+        S<1, 4, 16, 4>,       // BBlockTransferThreadClusterLengths_K0_N_K1
+        S<0, 3, 1, 2>,        // BBlockTransferThreadClusterArrangeOrder
+        S<0, 2, 1, 3>,        // BBlockTransferSrcAccessOrder
+        2,                    // BBlockTransferSrcVectorDim
+        8,                    // BBlockTransferSrcScalarPerVector
+        2,                    // BBlockTransferDstScalarPerVector_K1
+        true,                 // BBlockLdsAddExtraN
+        1,                    // CShuffleMXdlPerWavePerShuffle
+        1,                    // CShuffleNXdlPerWavePerShuffle
+        S<1, 32, 1, 4>,       // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
+        128 / (sizeof(WeiDataType) * CHAR_BIT)>; // CBlockTransferScalarPerVector_NWaveNPerXdl
+
 #include "run_grouped_conv_bwd_weight_example.inc"
 
 int main(int argc, char* argv[]) { return !run_grouped_conv_bwd_weight_example(argc, argv); }

example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16.cpp

@@ -3,6 +3,8 @@
 
 #include "common.hpp"
 
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_gnwc_gkxc_gnwk_xdl_cshuffle.hpp"
+
 using InDataType  = F16;
 using WeiDataType = F16;
 using OutDataType = F16;
@@ -12,6 +14,46 @@ using InElementOp  = PassThrough;
 using WeiElementOp = PassThrough;
 using OutElementOp = PassThrough;
 
+template <ck::index_t NDimSpatial>
+using DeviceConvBwdWeightInstance =
+    ck::tensor_operation::device::DeviceGroupedConvBwdWeightGnwcGkxcGnwk_Xdl_CShuffle<
+        NDimSpatial,          // NDimSpatial
+        InDataType,           // InDataType
+        WeiDataType,          // WeiDataType
+        OutDataType,          // OutDataType
+        AccDataType,          // AccDataType
+        InElementOp,          // InElementwiseOperation
+        WeiElementOp,         // WeiElementwiseOperation
+        OutElementOp,         // OutElementwiseOperation
+        ConvBwdWeightDefault, // ConvolutionBackwardWeightSpecialization
+        256,                  // BlockSize
+        128,                  // MPerBlock
+        128,                  // NPerBlock
+        4,                    // K0PerBlock
+        8,                    // K1
+        32,                   // MPerXdl
+        32,                   // NPerXdl
+        2,                    // MXdlPerWave
+        2,                    // NXdlPerWave
+        S<1, 4, 16, 4>,       // ABlockTransferThreadClusterLengths_K0_M_K1
+        S<0, 3, 1, 2>,        // ABlockTransferThreadClusterArrangeOrder
+        S<0, 2, 1, 3>,        // ABlockTransferSrcAccessOrder
+        2,                    // ABlockTransferSrcVectorDim
+        8,                    // ABlockTransferSrcScalarPerVector
+        2,                    // ABlockTransferDstScalarPerVector_K1
+        true,                 // ABlockLdsAddExtraM
+        S<1, 4, 16, 4>,       // BBlockTransferThreadClusterLengths_K0_N_K1
+        S<0, 3, 1, 2>,        // BBlockTransferThreadClusterArrangeOrder
+        S<0, 2, 1, 3>,        // BBlockTransferSrcAccessOrder
+        2,                    // BBlockTransferSrcVectorDim
+        8,                    // BBlockTransferSrcScalarPerVector
+        2,                    // BBlockTransferDstScalarPerVector_K1
+        true,                 // BBlockLdsAddExtraN
+        1,                    // CShuffleMXdlPerWavePerShuffle
+        1,                    // CShuffleNXdlPerWavePerShuffle
+        S<1, 32, 1, 4>,       // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
+        128 / (sizeof(WeiDataType) * CHAR_BIT)>; // CBlockTransferScalarPerVector_NWaveNPerXdl
+
 #include "run_grouped_conv_bwd_weight_example.inc"
 
 int main(int argc, char* argv[]) { return !run_grouped_conv_bwd_weight_example(argc, argv); }