Merge remote-tracking branch 'origin/develop' into migraphx-update

example/24_batched_gemm/batched_gemm_xdl_fp8_rowwise_v3.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, 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_batched_gemm_multiple_d_xdl_cshuffle_v3.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/reference_tensor_operation/cpu/reference_batched_gemm.hpp"
+#include "ck/library/utility/literals.hpp"
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using F8   = ck::f8_t;
+using BF16 = ck::bhalf_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 MultiplyMultiply = ck::tensor_operation::element_wise::MultiplyMultiply;
+
+using ADataType        = F8;
+using BDataType        = F8;
+using AccDataType      = F32;
+using CShuffleDataType = F32;
+using D0DataType       = F32;
+using D1DataType       = F32;
+using DsDataType       = ck::Tuple<D0DataType, D1DataType>;
+using EDataType        = BF16;
+
+using ALayout  = Row;
+using BLayout  = Col;
+using D0Layout = Row;
+using D1Layout = Col;
+using DsLayout = ck::Tuple<D0Layout, D1Layout>;
+using ELayout  = Row;
+
+using AElementOp   = PassThrough;
+using BElementOp   = PassThrough;
+using CDEElementOp = MultiplyMultiply;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+
+using DeviceGemmInstance = ck::tensor_operation::device::DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<
+    ALayout,
+    BLayout,
+    DsLayout,
+    ELayout,
+    ADataType,
+    BDataType,
+    DsDataType,
+    EDataType,
+    AccDataType,
+    CShuffleDataType,
+    AElementOp,
+    BElementOp,
+    CDEElementOp,
+    GemmDefault,
+    256,            // BlockSize
+    256,            // MPerBlock
+    128,            // NPerBlock
+    32,             // KPerBlock
+    8,              // AK1
+    8,              // BK1
+    32,             // MPerXDL
+    32,             // NPerXDL
+    4,              // MXdlPerWave
+    2,              // NXdlPerWave
+    S<4, 64, 1>,    // ABlockTransferThreadClusterLengths_AK0_M_AK1
+    S<1, 0, 2>,     // ABlockTransferThreadClusterArrangeOrder
+    S<1, 0, 2>,     // ABlockTransferSrcAccessOrder
+    2,              // ABlockTransferSrcVectorDim
+    8,              // ABlockTransferSrcScalarPerVector
+    8,              // ABlockTransferDstScalarPerVector_AK1
+    1,              // ABlockLdsExtraM
+    S<4, 64, 1>,    // BBlockTransferThreadClusterLengths_BK0_N_BK1
+    S<1, 0, 2>,     // BBlockTransferThreadClusterArrangeOrder
+    S<1, 0, 2>,     // BBlockTransferSrcAccessOrder
+    2,              // BBlockTransferSrcVectorDim
+    8,              // BBlockTransferSrcScalarPerVector
+    8,              // BBlockTransferDstScalarPerVector_BK1
+    1,              // BBlockLdsExtraN
+    1,              // CShuffleMXdlPerWavePerShuffle
+    1,              // CShuffleNXdlPerWavePerShuffle
+    S<1, 32, 1, 8>, // CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
+    S<8, 8, 1>,     // CDEShuffleBlockTransferScalarPerVectors
+    ck::BlockGemmPipelineScheduler::Interwave, // BlockGemmPipelineScheduler
+    ck::BlockGemmPipelineVersion::v1,          // BlockGemmPipelineVersion
+    F8                                         // ComputeTypeA
+    >;
+
+#include "run_batched_gemm_example_rowwise.inc"
+
+int main(int argc, char* argv[]) { return !run_batched_gemm_rowwise_example(argc, argv); }

example/24_batched_gemm/run_batched_gemm_example.inc

@@ -210,17 +210,9 @@ bool run_batched_gemm_example(int argc, char* argv[])
 
     problem_size.M = 256 * (dis(gen) + 1);
     problem_size.N = 128 * (dis(gen) + 1);
-    problem_size.K = 64 * (dis(gen) + 2);
+    problem_size.K = 128 * (dis(gen) + 2);
 
-    problem_size.stride_A = problem_size.K;
-    problem_size.stride_B = problem_size.K;
-    problem_size.stride_C = problem_size.N;
-
-    problem_size.batch_stride_A = problem_size.M * problem_size.K;
-    problem_size.batch_stride_B = problem_size.K * problem_size.N;
-    problem_size.batch_stride_C = problem_size.M * problem_size.N;
-
-    problem_size.batch_count = 16;
+    problem_size.batch_count = 2;
 
     if(argc == 4)
     {
@@ -228,13 +220,37 @@ bool run_batched_gemm_example(int argc, char* argv[])
         config.init_method     = std::stoi(argv[2]);
         config.time_kernel     = std::stoi(argv[3]);
     }
+    else if(argc == 8)
+    {
+        config.do_verification   = std::stoi(argv[1]);
+        config.init_method       = std::stoi(argv[2]);
+        config.time_kernel       = std::stoi(argv[3]);
+        problem_size.M           = std::stoi(argv[4]);
+        problem_size.N           = std::stoi(argv[5]);
+        problem_size.K           = std::stoi(argv[6]);
+        problem_size.batch_count = std::stoi(argv[7]);
+    }
     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");
+        printf("optinal\n");
+        printf("arg4-7: M = %d N = %d K = %d Batch = %d\n",
+               problem_size.M,
+               problem_size.N,
+               problem_size.K,
+               problem_size.batch_count);
         exit(0);
     }
 
+    problem_size.stride_A = problem_size.K;
+    problem_size.stride_B = problem_size.K;
+    problem_size.stride_C = problem_size.N;
+
+    problem_size.batch_stride_A = problem_size.M * problem_size.K;
+    problem_size.batch_stride_B = problem_size.K * problem_size.N;
+    problem_size.batch_stride_C = problem_size.M * problem_size.N;
+
     return run_batched_gemm(problem_size, config);
 }

example/24_batched_gemm/run_batched_gemm_example_rowwise.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+#include <random>
+
+#pragma once
+
+struct ProblemSize final
+{
+    ck::index_t M = 3840;
+    ck::index_t N = 4096;
+    ck::index_t K = 4096;
+
+    ck::index_t stride_A = K;
+    ck::index_t stride_B = K;
+    ck::index_t stride_C = N;
+
+    ck::index_t stride_D0 = 0;
+    ck::index_t stride_D1 = 0;
+
+    ck::index_t batch_stride_A = M * K;
+    ck::index_t batch_stride_B = K * N;
+    ck::index_t batch_stride_C = M * N;
+
+    ck::index_t batch_stride_D0 = N;
+    ck::index_t batch_stride_D1 = M;
+
+    ck::index_t batch_count = 16;
+};
+
+struct ExecutionConfig final
+{
+    bool do_verification = true;
+    int init_method      = 1;
+    bool time_kernel     = false;
+};
+
+bool run_batched_gemm_rowwise(const ProblemSize& problem_size, const ExecutionConfig& config)
+{
+    using namespace ck::literals;
+
+    auto& [M,
+           N,
+           K,
+           stride_A,
+           stride_B,
+           stride_C,
+           stride_D0,
+           stride_D1,
+           batch_stride_A,
+           batch_stride_B,
+           batch_stride_C,
+           batch_stride_D0,
+           batch_stride_D1,
+           batch_count] = problem_size;
+
+    // GEMM shape
+    auto f_host_tensor_descriptor = [](std::size_t batch_count_,
+                                       std::size_t row,
+                                       std::size_t col,
+                                       std::size_t stride,
+                                       std::size_t batch_stride,
+                                       auto layout) {
+        using namespace ck::literals;
+
+        if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::value)
+        {
+            return HostTensorDescriptor({batch_count_, row, col}, {batch_stride, stride, 1_uz});
+        }
+        else
+        {
+            return HostTensorDescriptor({batch_count_, row, col}, {batch_stride, 1_uz, stride});
+        }
+    };
+
+    Tensor<ADataType> a_g_m_k(
+        f_host_tensor_descriptor(batch_count, M, K, stride_A, batch_stride_A, ALayout{}));
+    Tensor<BDataType> b_g_k_n(
+        f_host_tensor_descriptor(batch_count, K, N, stride_B, batch_stride_B, BLayout{}));
+    Tensor<D0DataType> d0_g_m_n(
+        f_host_tensor_descriptor(batch_count, M, N, stride_D0, batch_stride_D0, D0Layout{}));
+    Tensor<D1DataType> d1_g_m_n(
+        f_host_tensor_descriptor(batch_count, M, N, stride_D1, batch_stride_D1, D1Layout{}));
+    Tensor<EDataType> e_g_m_n_device_result(
+        f_host_tensor_descriptor(batch_count, M, N, stride_C, batch_stride_C, ELayout{}));
+
+    std::cout << "a_g_m_k: " << a_g_m_k.mDesc << std::endl;
+    std::cout << "b_g_k_n: " << b_g_k_n.mDesc << std::endl;
+    std::cout << "d0_g_m_n: " << d0_g_m_n.mDesc << std::endl;
+    std::cout << "d1_g_m_n: " << d1_g_m_n.mDesc << std::endl;
+    std::cout << "e_g_m_n: " << e_g_m_n_device_result.mDesc << std::endl;
+
+    switch(config.init_method)
+    {
+    case 0: break;
+    case 1:
+        a_g_m_k.GenerateTensorValue(GeneratorTensor_2<ADataType>{-5, 5});
+        b_g_k_n.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
+        break;
+    default:
+        a_g_m_k.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
+        b_g_k_n.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
+        break;
+    }
+
+    d0_g_m_n.GenerateTensorValue(GeneratorTensor_3<D0DataType>{0.0, 1.0});
+    d1_g_m_n.GenerateTensorValue(GeneratorTensor_3<D1DataType>{0.0, 1.0});
+
+    DeviceMem a_device_buf(sizeof(ADataType) * a_g_m_k.mDesc.GetElementSpaceSize());
+    DeviceMem b_device_buf(sizeof(BDataType) * b_g_k_n.mDesc.GetElementSpaceSize());
+    DeviceMem d0_device_buf(sizeof(D0DataType) * d0_g_m_n.mDesc.GetElementSpaceSize());
+    DeviceMem d1_device_buf(sizeof(D1DataType) * d1_g_m_n.mDesc.GetElementSpaceSize());
+    DeviceMem c_device_buf(sizeof(EDataType) * e_g_m_n_device_result.mDesc.GetElementSpaceSize());
+
+    a_device_buf.ToDevice(a_g_m_k.mData.data());
+    b_device_buf.ToDevice(b_g_k_n.mData.data());
+
+    d0_device_buf.ToDevice(d0_g_m_n.mData.data());
+    d1_device_buf.ToDevice(d1_g_m_n.mData.data());
+
+    auto a_element_op   = AElementOp{};
+    auto b_element_op   = BElementOp{};
+    auto cde_element_op = CDEElementOp{};
+
+    auto gemm    = DeviceGemmInstance{};
+    auto invoker = gemm.MakeInvoker();
+
+    // do GEMM
+    auto argument =
+        gemm.MakeArgument(a_device_buf.GetDeviceBuffer(),
+                          b_device_buf.GetDeviceBuffer(),
+                          {d0_device_buf.GetDeviceBuffer(), d1_device_buf.GetDeviceBuffer()},
+                          c_device_buf.GetDeviceBuffer(),
+                          M,
+                          N,
+                          K,
+                          batch_count,
+                          stride_A,
+                          stride_B,
+                          {stride_D0, stride_D1},
+                          stride_C,
+                          batch_stride_A,
+                          batch_stride_B,
+                          {batch_stride_D0, batch_stride_D1},
+                          batch_stride_C,
+                          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");
+    }
+
+    invoker.Run(argument, StreamConfig{nullptr, false});
+    bool pass = true;
+
+    if(config.do_verification)
+    {
+        c_device_buf.FromDevice(e_g_m_n_device_result.mData.data());
+
+        Tensor<CShuffleDataType> c_g_m_n({batch_count, M, N});
+
+        using ReferenceBatchedGemmInstance =
+            ck::tensor_operation::host::ReferenceBatchedGemm<ADataType,
+                                                             BDataType,
+                                                             CShuffleDataType,
+                                                             AccDataType,
+                                                             AElementOp,
+                                                             BElementOp,
+                                                             PassThrough>;
+
+        auto ref_batched_gemm = ReferenceBatchedGemmInstance{};
+        auto ref_invoker      = ref_batched_gemm.MakeInvoker();
+
+        Tensor<EDataType> e_g_m_n_host_result(
+            f_host_tensor_descriptor(batch_count, M, N, stride_C, batch_stride_C, ELayout{}));
+
+        auto ref_argument = ref_batched_gemm.MakeArgument(
+            a_g_m_k, b_g_k_n, c_g_m_n, a_element_op, b_element_op, PassThrough{});
+
+        ref_invoker.Run(ref_argument);
+
+        for(int b = 0; b < batch_count; ++b)
+        {
+            for(int m = 0; m < M; ++m)
+            {
+                for(int n = 0; n < N; ++n)
+                {
+                    cde_element_op(e_g_m_n_host_result(b, m, n),
+                                   c_g_m_n(b, m, n),
+                                   d0_g_m_n(b, m, n),
+                                   d1_g_m_n(b, m, n));
+                }
+            }
+        }
+
+        pass = ck::utils::check_err(
+            e_g_m_n_device_result, e_g_m_n_host_result, "Error: Incorrect results c");
+    }
+
+    if(config.time_kernel)
+    {
+        float ave_time = invoker.Run(argument, StreamConfig{nullptr, config.time_kernel});
+
+        std::size_t flop      = std::size_t(2) * batch_count * M * N * K;
+        std::size_t num_btype = sizeof(ADataType) * batch_count * M * K +
+                                sizeof(BDataType) * batch_count * K * N +
+                                sizeof(EDataType) * batch_count * 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;
+    }
+
+    return pass ? 0 : 1;
+}
+
+bool run_batched_gemm_rowwise_example(int argc, char* argv[])
+{
+    ProblemSize problem_size;
+    ExecutionConfig config;
+
+    std::mt19937 gen(11939);
+    std::uniform_int_distribution<int> dis(0, 15);
+
+    problem_size.M = 256 * (dis(gen) + 1);
+    problem_size.N = 128 * (dis(gen) + 1);
+    problem_size.K = 128 * (dis(gen) + 2);
+
+    problem_size.batch_count = 2;
+
+    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 if(argc == 8)
+    {
+        config.do_verification   = std::stoi(argv[1]);
+        config.init_method       = std::stoi(argv[2]);
+        config.time_kernel       = std::stoi(argv[3]);
+        problem_size.M           = std::stoi(argv[4]);
+        problem_size.N           = std::stoi(argv[5]);
+        problem_size.K           = std::stoi(argv[6]);
+        problem_size.batch_count = std::stoi(argv[7]);
+    }
+    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");
+        printf("optinal\n");
+        printf("arg4-7: M = %d N = %d K = %d Batch = %d\n",
+               problem_size.M,
+               problem_size.N,
+               problem_size.K,
+               problem_size.batch_count);
+        exit(0);
+    }
+
+    problem_size.stride_A = problem_size.K;
+    problem_size.stride_B = problem_size.K;
+    problem_size.stride_C = problem_size.N;
+
+    problem_size.stride_D0 = 0;
+    problem_size.stride_D1 = 0;
+
+    problem_size.batch_stride_A = problem_size.M * problem_size.K;
+    problem_size.batch_stride_B = problem_size.K * problem_size.N;
+    problem_size.batch_stride_C = problem_size.M * problem_size.N;
+
+    problem_size.batch_stride_D0 = problem_size.N;
+    problem_size.batch_stride_D1 = problem_size.M;
+
+    return run_batched_gemm_rowwise(problem_size, config);
+}

example/25_gemm_bias_e_permute/CMakeLists.txt

-if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
-    add_example_executable(example_gemm_bias_e_permute_g1m3n2k1_xdl_fp16 gemm_bias_e_permute_g1m3n2k1_xdl_fp16.cpp)
-    add_example_executable(example_gemm_bias_e_permute_g1m2n3k1_xdl_fp16 gemm_bias_e_permute_g1m2n3k1_xdl_fp16.cpp)
-endif()
+add_example_executable(example_gemm_bias_e_permute_g1m3n2k1_xdl_fp16 gemm_bias_e_permute_g1m3n2k1_xdl_fp16.cpp)
+add_example_executable(example_gemm_bias_e_permute_g1m2n3k1_xdl_fp16 gemm_bias_e_permute_g1m2n3k1_xdl_fp16.cpp)

example/26_contraction/CMakeLists.txt

-if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
-    add_example_executable(example_contraction_bilinear_xdl_fp32 contraction_bilinear_xdl_fp32.cpp)
-    add_example_executable(example_contraction_scale_xdl_fp32 contraction_scale_xdl_fp32.cpp)
-endif()
-if(DTYPES MATCHES "fp64" OR NOT DEFINED DTYPES)
-    add_example_executable(example_contraction_bilinear_xdl_fp64 contraction_bilinear_xdl_fp64.cpp)
-    add_example_executable(example_contraction_scale_xdl_fp64 contraction_scale_xdl_fp64.cpp)
-endif()
+add_custom_target(example_contraction)
+add_custom_target(example_contraction_scale)
+add_custom_target(example_contraction_bilinear)
+
+# FP32
+add_example_executable(example_contraction_bilinear_xdl_fp32 contraction_bilinear_xdl_fp32.cpp)
+add_example_dependencies(example_contraction_bilinear example_contraction_bilinear_xdl_fp32)
+
+add_example_executable(example_contraction_scale_xdl_fp32 contraction_scale_xdl_fp32.cpp)
+add_example_dependencies(example_contraction_scale example_contraction_scale_xdl_fp32)
+
+add_example_executable(example_contraction_bilinear_xdl_fp32_compute_bf16 contraction_bilinear_xdl_fp32_compute_bf16.cpp)
+add_example_dependencies(example_contraction_bilinear example_contraction_bilinear_xdl_fp32_compute_bf16)
+
+add_example_executable(example_contraction_scale_xdl_fp32_compute_bf16 contraction_scale_xdl_fp32_compute_bf16.cpp)
+add_example_dependencies(example_contraction_scale example_contraction_scale_xdl_fp32_compute_bf16)
+
+add_example_executable(example_contraction_bilinear_xdl_fp32_compute_fp16 contraction_bilinear_xdl_fp32_compute_fp16.cpp)
+add_example_dependencies(example_contraction_bilinear example_contraction_bilinear_xdl_fp32_compute_fp16)
+
+add_example_executable(example_contraction_scale_xdl_fp32_compute_fp16 contraction_scale_xdl_fp32_compute_fp16.cpp)
+add_example_dependencies(example_contraction_scale example_contraction_scale_xdl_fp32_compute_fp16)
+
+# FP64
+add_example_executable(example_contraction_bilinear_xdl_fp64 contraction_bilinear_xdl_fp64.cpp)
+add_example_dependencies(example_contraction_bilinear example_contraction_bilinear_xdl_fp64)
+
+add_example_executable(example_contraction_scale_xdl_fp64 contraction_scale_xdl_fp64.cpp)
+add_example_dependencies(example_contraction_scale example_contraction_scale_xdl_fp64)
+
+add_example_executable(example_contraction_bilinear_xdl_fp64_compute_fp32 contraction_bilinear_xdl_fp64_compute_fp32.cpp)
+add_example_dependencies(example_contraction_bilinear example_contraction_bilinear_xdl_fp64_compute_fp32)
+
+add_example_executable(example_contraction_scale_xdl_fp64_compute_fp32 contraction_scale_xdl_fp64_compute_fp32.cpp)
+add_example_dependencies(example_contraction_scale example_contraction_scale_xdl_fp64_compute_fp32)
+
+# FP16
+add_example_executable(example_contraction_bilinear_xdl_fp16_compute_fp32 contraction_bilinear_xdl_fp16_compute_fp32.cpp)
+add_example_dependencies(example_contraction_bilinear example_contraction_bilinear_xdl_fp16_compute_fp32)
+
+add_example_executable(example_contraction_scale_xdl_fp16_compute_fp32 contraction_scale_xdl_fp16_compute_fp32.cpp)
+add_example_dependencies(example_contraction_scale example_contraction_scale_xdl_fp16_compute_fp32)
+
+# BF16
+add_example_executable(example_contraction_bilinear_xdl_bf16_compute_fp32 contraction_bilinear_xdl_bf16_compute_fp32.cpp)
+add_example_dependencies(example_contraction_bilinear example_contraction_bilinear_xdl_bf16_compute_fp32)
+
+add_example_executable(example_contraction_scale_xdl_bf16_compute_fp32 contraction_scale_xdl_bf16_compute_fp32.cpp)
+add_example_dependencies(example_contraction_scale example_contraction_scale_xdl_bf16_compute_fp32)
+
+add_example_dependencies(example_contraction example_contraction_scale)
+add_example_dependencies(example_contraction example_contraction_bilinear)

example/26_contraction/README.md

@@ -7,14 +7,3 @@
 #arg3: time kernel (0=no, 1=yes)
 ./bin/example_contraction_bilinear_xdl_fp32 1 1 1
 ```
-
-Result (MI100 @ dynammic freq, 46TFlops peak FP32)
-```
-a_ms_ks: dim 4, lengths {30, 128, 32, 64}, strides {524288, 4096, 128, 1}
-b_ks_ns: dim 4, lengths {32, 64, 32, 64}, strides {128, 1, 524288, 4096}
-c_ms_ns: dim 4, lengths {30, 128, 32, 64}, strides {524288, 4096, 128, 1}
-launch_and_time_kernel: grid_dim {240, 1, 1}, block_dim {256, 1, 1}
-Warm up 1 time
-Start running 10 times...
-Perf: 0.843286 ms, 38.1985 TFlops, 94.5014 GB/s, DeviceContractionMultipleD_Xdl_CShuffle<256, 256, 128, 16, 4, 4>
-```

example/26_contraction/common_instances.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_cshuffle.hpp"
+
+using F16  = ck::half_t;
+using BF16 = ck::bhalf_t;
+using F32  = float;
+using F64  = double;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+// Generic instances for fp32, fp16 and bf16 data types.
+template <ck::index_t NumDimM,
+          ck::index_t NumDimN,
+          ck::index_t NumDimK,
+          typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CShuffleDataType,
+          typename DsDataType,
+          typename EDataType,
+          typename ComputeDataType,
+          typename AElementOp,
+          typename BElementOp,
+          typename CDEElementOp>
+// clang-format off
+using DeviceOpInstanceKK_Generic = ck::tensor_operation::device::
+        //#####################################| NumDimM| NumDimN| NumDimK|     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|         Compute|
+        //#####################################|        |        |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization| 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|            Data|
+        //#####################################|        |        |        |          |          |            |                 |           |          |   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|            Type|
+        //#####################################|        |        |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
+        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>;
+// clang-format on
+
+template <ck::index_t NumDimM,
+          ck::index_t NumDimN,
+          ck::index_t NumDimK,
+          typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CShuffleDataType,
+          typename DsDataType,
+          typename EDataType,
+          typename ComputeDataType,
+          typename AElementOp,
+          typename BElementOp,
+          typename CDEElementOp>
+// clang-format off
+using DeviceOpInstanceKN_Generic = ck::tensor_operation::device::
+        //#####################################| NumDimM| NumDimN| NumDimK|     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|         Compute|
+        //#####################################|        |        |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization| 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|            Data|
+        //#####################################|        |        |        |          |          |            |                 |           |          |   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|            Type|
+        //#####################################|        |        |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
+        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   4,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>;
+// clang-format on
+
+template <ck::index_t NumDimM,
+          ck::index_t NumDimN,
+          ck::index_t NumDimK,
+          typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CShuffleDataType,
+          typename DsDataType,
+          typename EDataType,
+          typename ComputeDataType,
+          typename AElementOp,
+          typename BElementOp,
+          typename CDEElementOp>
+// clang-format off
+using DeviceOpInstanceMK_Generic = ck::tensor_operation::device::
+        //#####################################| NumDimM| NumDimN| NumDimK|     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|         Compute|
+        //#####################################|        |        |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization| 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|            Data|
+        //#####################################|        |        |        |          |          |            |                 |           |          |   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|            Type|
+        //#####################################|        |        |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
+        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   1,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>;
+// clang-format on
+
+template <ck::index_t NumDimM,
+          ck::index_t NumDimN,
+          ck::index_t NumDimK,
+          typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CShuffleDataType,
+          typename DsDataType,
+          typename EDataType,
+          typename ComputeDataType,
+          typename AElementOp,
+          typename BElementOp,
+          typename CDEElementOp>
+// clang-format off
+using DeviceOpInstanceMN_Generic = ck::tensor_operation::device::
+        //#####################################| NumDimM| NumDimN| NumDimK|     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|         Compute|
+        //#####################################|        |        |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization| 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|            Data|
+        //#####################################|        |        |        |          |          |            |                 |           |          |   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|            Type|
+        //#####################################|        |        |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
+        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   1,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>;
+// clang-format on
+
+// Fp64 instances.
+template <ck::index_t NumDimM,
+          ck::index_t NumDimN,
+          ck::index_t NumDimK,
+          typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CShuffleDataType,
+          typename DsDataType,
+          typename EDataType,
+          typename ComputeDataType,
+          typename AElementOp,
+          typename BElementOp,
+          typename CDEElementOp>
+// clang-format off
+using DeviceOpInstanceKK_FP64 = ck::tensor_operation::device::
+        //#####################################| NumDimM| NumDimN| NumDimK|     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|         Compute|
+        //#####################################|        |        |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization| 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|            Data|
+        //#####################################|        |        |        |          |          |            |                 |           |          |   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|            Type|
+        //#####################################|        |        |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
+        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              2,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>;
+// clang-format on
+
+template <ck::index_t NumDimM,
+          ck::index_t NumDimN,
+          ck::index_t NumDimK,
+          typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CShuffleDataType,
+          typename DsDataType,
+          typename EDataType,
+          typename ComputeDataType,
+          typename AElementOp,
+          typename BElementOp,
+          typename CDEElementOp>
+// clang-format off
+using DeviceOpInstanceKN_FP64 = ck::tensor_operation::device::
+        //#####################################| NumDimM| NumDimN| NumDimK|     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|         Compute|
+        //#####################################|        |        |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization| 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|            Data|
+        //#####################################|        |        |        |          |          |            |                 |           |          |   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|            Type|
+        //#####################################|        |        |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
+        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   2,   1,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              2,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>;
+// clang-format on
+
+template <ck::index_t NumDimM,
+          ck::index_t NumDimN,
+          ck::index_t NumDimK,
+          typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CShuffleDataType,
+          typename DsDataType,
+          typename EDataType,
+          typename ComputeDataType,
+          typename AElementOp,
+          typename BElementOp,
+          typename CDEElementOp>
+// clang-format off
+using DeviceOpInstanceMK_FP64 = ck::tensor_operation::device::
+        //#####################################| NumDimM| NumDimN| NumDimK|     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|         Compute|
+        //#####################################|        |        |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization| 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|            Data|
+        //#####################################|        |        |        |          |          |            |                 |           |          |   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|            Type|
+        //#####################################|        |        |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
+        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   1,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>;
+// clang-format on
+
+template <ck::index_t NumDimM,
+          ck::index_t NumDimN,
+          ck::index_t NumDimK,
+          typename ADataType,
+          typename BDataType,
+          typename AccDataType,
+          typename CShuffleDataType,
+          typename DsDataType,
+          typename EDataType,
+          typename ComputeDataType,
+          typename AElementOp,
+          typename BElementOp,
+          typename CDEElementOp>
+// clang-format off
+using DeviceOpInstanceMN_FP64 = ck::tensor_operation::device::
+        //#####################################| NumDimM| NumDimN| NumDimK|     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|         Compute|
+        //#####################################|        |        |        |      Type|      Type|        Type|         DataType|       Type|      Type| Elementwise| Elementwise|  Elementwise| Specialization| 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|            Data|
+        //#####################################|        |        |        |          |          |            |                 |           |          |   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|            Type|
+        //#####################################|        |        |        |          |          |            |                 |           |          |            |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
+        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType,  AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   1,   1,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>;
+// clang-format on

example/26_contraction/contraction_bilinear_xdl_bf16_compute_fp32.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = BF16;
+using BDataType        = BF16;
+using AccDataType      = F32;
+using CShuffleDataType = BF16;
+using DDataType        = BF16;
+using DsDataType       = ck::Tuple<DDataType>;
+using EDataType        = BF16;
+using ComputeDataType  = F32;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Bilinear;
+
+using DeviceOpInstanceKKNN = DeviceOpInstanceKK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceKNNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceMKNN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceMNNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKNN;
+
+#include "run_contraction_bilinear_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_bilinear_example(argc, argv); }

example/26_contraction/contraction_bilinear_xdl_fp16_compute_fp32.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = F16;
+using BDataType        = F16;
+using AccDataType      = F32;
+using CShuffleDataType = F16;
+using DDataType        = F16;
+using DsDataType       = ck::Tuple<DDataType>;
+using EDataType        = F16;
+using ComputeDataType  = F32;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Bilinear;
+
+using DeviceOpInstanceKKNN = DeviceOpInstanceKK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceKNNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceMKNN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceMNNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKNN;
+
+#include "run_contraction_bilinear_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_bilinear_example(argc, argv); }

example/26_contraction/contraction_bilinear_xdl_fp32.cpp

 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2023, 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/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_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/numeric.hpp"
-#include "ck/library/reference_tensor_operation/cpu/reference_contraction.hpp"
-
-template <ck::index_t... Is>
-using S = ck::Sequence<Is...>;
-
-using F32 = float;
-
-using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+#include "common_instances.hpp"
 
 using ADataType        = F32;
 using BDataType        = F32;
@@ -32,6 +13,7 @@ using CShuffleDataType = F32;
 using DDataType        = F32;
 using DsDataType       = ck::Tuple<DDataType>;
 using EDataType        = F32;
+using ComputeDataType  = F32;
 
 static constexpr ck::index_t NumDimM = 2;
 static constexpr ck::index_t NumDimN = 2;
@@ -41,253 +23,64 @@ using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
 using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
 using CDEElementOp = ck::tensor_operation::element_wise::Bilinear;
 
-static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
-
-// clang-format off
-using DeviceOpInstanceKKNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F32,   F32,     F32,      F32, DsDataType,   F32,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4>;
-
-using DeviceOpInstanceKNNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F32,   F32,     F32,      F32, DsDataType,   F32,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   4,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4>;
-
-using DeviceOpInstanceMKNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F32,   F32,     F32,      F32, DsDataType,   F32,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   1,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4>;
-
-using DeviceOpInstanceMNNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F32,   F32,     F32,      F32, DsDataType,   F32,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   1,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4>;
-// clang-format on
-
-using DeviceOpInstance = DeviceOpInstanceKKNN;
-
-int main(int argc, char* argv[])
-{
-    bool do_verification = true;
-    int init_method      = 1;
-    bool time_kernel     = false;
-
-    // A[M0, M1, K0, K1]
-    std::vector<ck::index_t> a_ms_ks_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> a_ms_ks_strides{524288, 4096, 128, 1};
-    // B[N0, N1, K0, K1]
-    std::vector<ck::index_t> b_ns_ks_lengths{32, 64, 32, 64};
-    std::vector<ck::index_t> b_ns_ks_strides{524288, 4096, 128, 1};
-    // D[M0, M1, N0, N1]
-    std::vector<ck::index_t> d_ms_ns_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> d_ms_ns_strides{524288, 4096, 128, 1};
-    // E[M0, M1, N0, N1]
-    std::vector<ck::index_t> e_ms_ns_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> e_ms_ns_strides{524288, 4096, 128, 1};
-
-    float alpha = 1.f;
-    float beta  = 1.f;
-
-    if(argc == 1)
-    {
-        // use default case
-    }
-    else if(argc == 4)
-    {
-        do_verification = std::stoi(argv[1]);
-        init_method     = std::stoi(argv[2]);
-        time_kernel     = std::stoi(argv[3]);
-    }
-    else if(argc == 28)
-    {
-        do_verification = std::stoi(argv[1]);
-        init_method     = std::stoi(argv[2]);
-        time_kernel     = std::stoi(argv[3]);
-
-        const ck::index_t M0 = std::stoi(argv[4]);
-        const ck::index_t M1 = std::stoi(argv[5]);
-
-        const ck::index_t N0 = std::stoi(argv[6]);
-        const ck::index_t N1 = std::stoi(argv[7]);
-
-        const ck::index_t K0 = std::stoi(argv[8]);
-        const ck::index_t K1 = std::stoi(argv[9]);
-
-        a_ms_ks_lengths = {M0, M1, K0, K1};
-        a_ms_ks_strides = {
-            std::stoi(argv[10]), std::stoi(argv[11]), std::stoi(argv[12]), std::stoi(argv[13])};
-
-        b_ns_ks_lengths = {N0, N1, K0, K1};
-        b_ns_ks_strides = {
-            std::stoi(argv[14]), std::stoi(argv[15]), std::stoi(argv[16]), std::stoi(argv[17])};
-
-        d_ms_ns_lengths = {M0, M1, N0, N1};
-        d_ms_ns_strides = {
-            std::stoi(argv[18]), std::stoi(argv[19]), std::stoi(argv[20]), std::stoi(argv[21])};
-
-        e_ms_ns_lengths = {M0, M1, N0, N1};
-        e_ms_ns_strides = {
-            std::stoi(argv[22]), std::stoi(argv[23]), std::stoi(argv[24]), std::stoi(argv[25])};
-
-        alpha = std::stof(argv[26]);
-        beta  = std::stof(argv[27]);
-    }
-    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=no, 1=yes)\n");
-        printf("arg4 to 7: M0, M1, N0, N1, K0, K1\n");
-        printf("arg10 to 13: Stride_A_M0, Stride_A_M1, Stride_A_K0, Stride_A_K1\n");
-        printf("arg14 to 17: Stride_B_N0, Stride_B_N1, Stride_B_K0, Stride_B_K1\n");
-        printf("arg18 to 21: Stride_D_M0, Stride_D_M1, Stride_D_N0, Stride_D_N1\n");
-        printf("arg22 to 25: Stride_E_M0, Stride_E_M1, Stride_E_N0, Stride_E_N1\n");
-        printf("arg26 to 27: alpha, beta\n");
-        exit(0);
-    }
-
-    Tensor<ADataType> a_ms_ks(a_ms_ks_lengths, a_ms_ks_strides);
-    Tensor<BDataType> b_ns_ks(b_ns_ks_lengths, b_ns_ks_strides);
-    Tensor<EDataType> d_ms_ns(d_ms_ns_lengths, d_ms_ns_strides);
-    Tensor<EDataType> e_ms_ns_host_result(e_ms_ns_lengths, e_ms_ns_strides);
-    Tensor<EDataType> e_ms_ns_device_result(e_ms_ns_lengths, e_ms_ns_strides);
-
-    std::cout << "a_ms_ks: " << a_ms_ks.mDesc << std::endl;
-    std::cout << "b_ns_ks: " << b_ns_ks.mDesc << std::endl;
-    std::cout << "d_ms_ns: " << d_ms_ns.mDesc << std::endl;
-    std::cout << "e_ms_ns: " << e_ms_ns_host_result.mDesc << std::endl;
-
-    switch(init_method)
-    {
-    case 0: break;
-    case 1:
-        a_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-5, 5});
-        b_ns_ks.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
-        d_ms_ns.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
-        break;
-    default:
-        a_ms_ks.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
-        b_ns_ks.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
-        d_ms_ns.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
-        break;
-    }
-
-    DeviceMem a_device_buf(sizeof(ADataType) * a_ms_ks.mDesc.GetElementSpaceSize());
-    DeviceMem b_device_buf(sizeof(BDataType) * b_ns_ks.mDesc.GetElementSpaceSize());
-    DeviceMem d_device_buf(sizeof(DDataType) * d_ms_ns.mDesc.GetElementSpaceSize());
-    DeviceMem e_device_buf(sizeof(EDataType) * e_ms_ns_device_result.mDesc.GetElementSpaceSize());
-
-    a_device_buf.ToDevice(a_ms_ks.mData.data());
-    b_device_buf.ToDevice(b_ns_ks.mData.data());
-    d_device_buf.ToDevice(d_ms_ns.mData.data());
-
-    // set zero
-    e_device_buf.SetZero();
-
-    auto a_element_op   = AElementOp{};
-    auto b_element_op   = BElementOp{};
-    auto cde_element_op = CDEElementOp{alpha, beta};
-
-    // device operation
-    auto op       = DeviceOpInstance{};
-    auto invoker  = op.MakeInvoker();
-    auto argument = op.MakeArgument(a_device_buf.GetDeviceBuffer(),
-                                    b_device_buf.GetDeviceBuffer(),
-                                    std::array<const void*, 1>{d_device_buf.GetDeviceBuffer()},
-                                    e_device_buf.GetDeviceBuffer(),
-                                    a_ms_ks_lengths,
-                                    a_ms_ks_strides,
-                                    b_ns_ks_lengths,
-                                    b_ns_ks_strides,
-                                    std::array<std::vector<ck::index_t>, 1>{d_ms_ns_lengths},
-                                    std::array<std::vector<ck::index_t>, 1>{d_ms_ns_strides},
-                                    e_ms_ns_lengths,
-                                    e_ms_ns_strides,
-                                    a_element_op,
-                                    b_element_op,
-                                    cde_element_op);
-
-    if(!op.IsSupportedArgument(argument))
-    {
-        std::cout << op.GetTypeString() << " does not support this problem" << std::endl;
-
-        return 0;
-    }
-
-    float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
-
-    ck::index_t M =
-        ck::accumulate_n<ck::index_t>(e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
-
-    ck::index_t N = ck::accumulate_n<ck::index_t>(
-        e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
-
-    ck::index_t K = ck::accumulate_n<ck::index_t>(
-        a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
-
-    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(DDataType) * M * 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, "
-              << op.GetTypeString() << std::endl;
-
-    e_device_buf.FromDevice(e_ms_ns_device_result.mData.data());
-
-    if(do_verification)
-    {
-        Tensor<CShuffleDataType> c_ms_ns_host_result(e_ms_ns_lengths, e_ms_ns_strides);
-
-        using ReferenceOpInstance =
-            ck::tensor_operation::host::ReferenceContraction_M2_N2_K2<NumDimM,
+using DeviceOpInstanceKKNN = DeviceOpInstanceKK_Generic<NumDimM,
                                                         NumDimN,
                                                         NumDimK,
                                                         ADataType,
                                                         BDataType,
-                                                                      CShuffleDataType,
                                                         AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
                                                         AElementOp,
-                                                                      BElementOp>;
+                                                        BElementOp,
+                                                        CDEElementOp>;
 
-        auto ref_op      = ReferenceOpInstance{};
-        auto ref_invoker = ref_op.MakeInvoker();
+using DeviceOpInstanceKNNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
 
-        auto ref_argument =
-            ref_op.MakeArgument(a_ms_ks, b_ns_ks, c_ms_ns_host_result, a_element_op, b_element_op);
+using DeviceOpInstanceMKNN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
 
-        ref_invoker.Run(ref_argument);
+using DeviceOpInstanceMNNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
 
-        for(size_t m0 = 0; m0 < e_ms_ns_host_result.mDesc.GetLengths()[0]; ++m0)
-        {
-            for(size_t m1 = 0; m1 < e_ms_ns_host_result.mDesc.GetLengths()[1]; ++m1)
-            {
-                for(size_t n0 = 0; n0 < e_ms_ns_host_result.mDesc.GetLengths()[2]; ++n0)
-                {
-                    for(size_t n1 = 0; n1 < e_ms_ns_host_result.mDesc.GetLengths()[3]; ++n1)
-                    {
-                        cde_element_op(e_ms_ns_host_result(m0, m1, n0, n1),
-                                       c_ms_ns_host_result(m0, m1, n0, n1),
-                                       d_ms_ns(m0, m1, n0, n1));
-                    }
-                }
-            }
-        }
+using DeviceOpInstance = DeviceOpInstanceKKNN;
 
-        return ck::utils::check_err(e_ms_ns_device_result, e_ms_ns_host_result) ? 0 : 1;
-    }
+#include "run_contraction_bilinear_example.inc"
 
-    return 0;
-}
+int main(int argc, char* argv[]) { return run_contraction_bilinear_example(argc, argv); }

example/26_contraction/contraction_bilinear_xdl_fp32_compute_bf16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = F32;
+using BDataType        = F32;
+using AccDataType      = F32;
+using CShuffleDataType = F32;
+using DDataType        = F32;
+using DsDataType       = ck::Tuple<DDataType>;
+using EDataType        = F32;
+using ComputeDataType  = BF16;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Bilinear;
+
+using DeviceOpInstanceKKNN = DeviceOpInstanceKK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceKNNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceMKNN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceMNNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKNN;
+
+#include "run_contraction_bilinear_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_bilinear_example(argc, argv); }

example/26_contraction/contraction_bilinear_xdl_fp32_compute_fp16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = F32;
+using BDataType        = F32;
+using AccDataType      = F32;
+using CShuffleDataType = F32;
+using DDataType        = F32;
+using DsDataType       = ck::Tuple<DDataType>;
+using EDataType        = F32;
+using ComputeDataType  = F16;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Bilinear;
+
+using DeviceOpInstanceKKNN = DeviceOpInstanceKK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceKNNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceMKNN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstanceMNNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                        NumDimN,
+                                                        NumDimK,
+                                                        ADataType,
+                                                        BDataType,
+                                                        AccDataType,
+                                                        CShuffleDataType,
+                                                        DsDataType,
+                                                        EDataType,
+                                                        ComputeDataType,
+                                                        AElementOp,
+                                                        BElementOp,
+                                                        CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKNN;
+
+#include "run_contraction_bilinear_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_bilinear_example(argc, argv); }

example/26_contraction/contraction_bilinear_xdl_fp64.cpp

 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2023, 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/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_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/numeric.hpp"
-#include "ck/library/reference_tensor_operation/cpu/reference_contraction.hpp"
-
-template <ck::index_t... Is>
-using S = ck::Sequence<Is...>;
-
-using F64 = double;
-
-using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+#include "common_instances.hpp"
 
 using ADataType        = F64;
 using BDataType        = F64;
@@ -32,6 +13,7 @@ using CShuffleDataType = F64;
 using DDataType        = F64;
 using DsDataType       = ck::Tuple<DDataType>;
 using EDataType        = F64;
+using ComputeDataType  = F64;
 
 static constexpr ck::index_t NumDimM = 2;
 static constexpr ck::index_t NumDimN = 2;
@@ -41,253 +23,64 @@ using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
 using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
 using CDEElementOp = ck::tensor_operation::element_wise::Bilinear;
 
-static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
-
-// clang-format off
-using DeviceOpInstanceKKNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F64,   F64,     F64,      F64, DsDataType,   F64,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              2,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,         1,           1,           1,              S<1, 16, 1, 16>,               1>;
-
-using DeviceOpInstanceKNNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F64,   F64,     F64,      F64, DsDataType,   F64,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   2,   1,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              2,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1>;
-
-using DeviceOpInstanceMKNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F64,   F64,     F64,      F64, DsDataType,   F64,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   1,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,         1,           1,           1,              S<1, 16, 1, 16>,               1>;
-
-using DeviceOpInstanceMNNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F64,   F64,     F64,      F64, DsDataType,   F64,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   1,   1,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1>;
-// clang-format on
-
-using DeviceOpInstance = DeviceOpInstanceKKNN;
-
-int main(int argc, char* argv[])
-{
-    bool do_verification = true;
-    int init_method      = 1;
-    bool time_kernel     = false;
-
-    // A[M0, M1, K0, K1]
-    std::vector<ck::index_t> a_ms_ks_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> a_ms_ks_strides{524288, 4096, 128, 1};
-    // B[N0, N1, K0, K1]
-    std::vector<ck::index_t> b_ns_ks_lengths{32, 64, 32, 64};
-    std::vector<ck::index_t> b_ns_ks_strides{524288, 4096, 128, 1};
-    // D[M0, M1, N0, N1]
-    std::vector<ck::index_t> d_ms_ns_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> d_ms_ns_strides{524288, 4096, 128, 1};
-    // E[M0, M1, N0, N1]
-    std::vector<ck::index_t> e_ms_ns_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> e_ms_ns_strides{524288, 4096, 128, 1};
-
-    float alpha = 1.f;
-    float beta  = 1.f;
-
-    if(argc == 1)
-    {
-        // use default case
-    }
-    else if(argc == 4)
-    {
-        do_verification = std::stoi(argv[1]);
-        init_method     = std::stoi(argv[2]);
-        time_kernel     = std::stoi(argv[3]);
-    }
-    else if(argc == 28)
-    {
-        do_verification = std::stoi(argv[1]);
-        init_method     = std::stoi(argv[2]);
-        time_kernel     = std::stoi(argv[3]);
-
-        const ck::index_t M0 = std::stoi(argv[4]);
-        const ck::index_t M1 = std::stoi(argv[5]);
-
-        const ck::index_t N0 = std::stoi(argv[6]);
-        const ck::index_t N1 = std::stoi(argv[7]);
-
-        const ck::index_t K0 = std::stoi(argv[8]);
-        const ck::index_t K1 = std::stoi(argv[9]);
-
-        a_ms_ks_lengths = {M0, M1, K0, K1};
-        a_ms_ks_strides = {
-            std::stoi(argv[10]), std::stoi(argv[11]), std::stoi(argv[12]), std::stoi(argv[13])};
-
-        b_ns_ks_lengths = {N0, N1, K0, K1};
-        b_ns_ks_strides = {
-            std::stoi(argv[14]), std::stoi(argv[15]), std::stoi(argv[16]), std::stoi(argv[17])};
-
-        d_ms_ns_lengths = {M0, M1, N0, N1};
-        d_ms_ns_strides = {
-            std::stoi(argv[18]), std::stoi(argv[19]), std::stoi(argv[20]), std::stoi(argv[21])};
-
-        e_ms_ns_lengths = {M0, M1, N0, N1};
-        e_ms_ns_strides = {
-            std::stoi(argv[22]), std::stoi(argv[23]), std::stoi(argv[24]), std::stoi(argv[25])};
-
-        alpha = std::stof(argv[26]);
-        beta  = std::stof(argv[27]);
-    }
-    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=no, 1=yes)\n");
-        printf("arg4 to 7: M0, M1, N0, N1, K0, K1\n");
-        printf("arg10 to 13: Stride_A_M0, Stride_A_M1, Stride_A_K0, Stride_A_K1\n");
-        printf("arg14 to 17: Stride_B_N0, Stride_B_N1, Stride_B_K0, Stride_B_K1\n");
-        printf("arg18 to 21: Stride_D_M0, Stride_D_M1, Stride_D_N0, Stride_D_N1\n");
-        printf("arg22 to 25: Stride_E_M0, Stride_E_M1, Stride_E_N0, Stride_E_N1\n");
-        printf("arg26 to 27: alpha, beta\n");
-        exit(0);
-    }
-
-    Tensor<ADataType> a_ms_ks(a_ms_ks_lengths, a_ms_ks_strides);
-    Tensor<BDataType> b_ns_ks(b_ns_ks_lengths, b_ns_ks_strides);
-    Tensor<EDataType> d_ms_ns(d_ms_ns_lengths, d_ms_ns_strides);
-    Tensor<EDataType> e_ms_ns_host_result(e_ms_ns_lengths, e_ms_ns_strides);
-    Tensor<EDataType> e_ms_ns_device_result(e_ms_ns_lengths, e_ms_ns_strides);
-
-    std::cout << "a_ms_ks: " << a_ms_ks.mDesc << std::endl;
-    std::cout << "b_ns_ks: " << b_ns_ks.mDesc << std::endl;
-    std::cout << "d_ms_ns: " << d_ms_ns.mDesc << std::endl;
-    std::cout << "e_ms_ns: " << e_ms_ns_host_result.mDesc << std::endl;
-
-    switch(init_method)
-    {
-    case 0: break;
-    case 1:
-        a_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-5, 5});
-        b_ns_ks.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
-        d_ms_ns.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
-        break;
-    default:
-        a_ms_ks.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
-        b_ns_ks.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
-        d_ms_ns.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
-        break;
-    }
-
-    DeviceMem a_device_buf(sizeof(ADataType) * a_ms_ks.mDesc.GetElementSpaceSize());
-    DeviceMem b_device_buf(sizeof(BDataType) * b_ns_ks.mDesc.GetElementSpaceSize());
-    DeviceMem d_device_buf(sizeof(DDataType) * d_ms_ns.mDesc.GetElementSpaceSize());
-    DeviceMem e_device_buf(sizeof(EDataType) * e_ms_ns_device_result.mDesc.GetElementSpaceSize());
-
-    a_device_buf.ToDevice(a_ms_ks.mData.data());
-    b_device_buf.ToDevice(b_ns_ks.mData.data());
-    d_device_buf.ToDevice(d_ms_ns.mData.data());
-
-    // set zero
-    e_device_buf.SetZero();
-
-    auto a_element_op   = AElementOp{};
-    auto b_element_op   = BElementOp{};
-    auto cde_element_op = CDEElementOp{alpha, beta};
-
-    // device operation
-    auto op       = DeviceOpInstance{};
-    auto invoker  = op.MakeInvoker();
-    auto argument = op.MakeArgument(a_device_buf.GetDeviceBuffer(),
-                                    b_device_buf.GetDeviceBuffer(),
-                                    std::array<const void*, 1>{d_device_buf.GetDeviceBuffer()},
-                                    e_device_buf.GetDeviceBuffer(),
-                                    a_ms_ks_lengths,
-                                    a_ms_ks_strides,
-                                    b_ns_ks_lengths,
-                                    b_ns_ks_strides,
-                                    std::array<std::vector<ck::index_t>, 1>{d_ms_ns_lengths},
-                                    std::array<std::vector<ck::index_t>, 1>{d_ms_ns_strides},
-                                    e_ms_ns_lengths,
-                                    e_ms_ns_strides,
-                                    a_element_op,
-                                    b_element_op,
-                                    cde_element_op);
-
-    if(!op.IsSupportedArgument(argument))
-    {
-        std::cout << op.GetTypeString() << " does not support this problem" << std::endl;
-
-        return 0;
-    }
-
-    float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
-
-    ck::index_t M =
-        ck::accumulate_n<ck::index_t>(e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
-
-    ck::index_t N = ck::accumulate_n<ck::index_t>(
-        e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
-
-    ck::index_t K = ck::accumulate_n<ck::index_t>(
-        a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
-
-    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(DDataType) * M * 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, "
-              << op.GetTypeString() << std::endl;
-
-    e_device_buf.FromDevice(e_ms_ns_device_result.mData.data());
-
-    if(do_verification)
-    {
-        Tensor<CShuffleDataType> c_ms_ns_host_result(e_ms_ns_lengths, e_ms_ns_strides);
-
-        using ReferenceOpInstance =
-            ck::tensor_operation::host::ReferenceContraction_M2_N2_K2<NumDimM,
+using DeviceOpInstanceKKNN = DeviceOpInstanceKK_FP64<NumDimM,
                                                      NumDimN,
                                                      NumDimK,
                                                      ADataType,
                                                      BDataType,
-                                                                      CShuffleDataType,
                                                      AccDataType,
+                                                     CShuffleDataType,
+                                                     DsDataType,
+                                                     EDataType,
+                                                     ComputeDataType,
                                                      AElementOp,
-                                                                      BElementOp>;
+                                                     BElementOp,
+                                                     CDEElementOp>;
 
-        auto ref_op      = ReferenceOpInstance{};
-        auto ref_invoker = ref_op.MakeInvoker();
+using DeviceOpInstanceKNNN = DeviceOpInstanceKN_FP64<NumDimM,
+                                                     NumDimN,
+                                                     NumDimK,
+                                                     ADataType,
+                                                     BDataType,
+                                                     AccDataType,
+                                                     CShuffleDataType,
+                                                     DsDataType,
+                                                     EDataType,
+                                                     ComputeDataType,
+                                                     AElementOp,
+                                                     BElementOp,
+                                                     CDEElementOp>;
 
-        auto ref_argument =
-            ref_op.MakeArgument(a_ms_ks, b_ns_ks, c_ms_ns_host_result, a_element_op, b_element_op);
+using DeviceOpInstanceMKNN = DeviceOpInstanceMK_FP64<NumDimM,
+                                                     NumDimN,
+                                                     NumDimK,
+                                                     ADataType,
+                                                     BDataType,
+                                                     AccDataType,
+                                                     CShuffleDataType,
+                                                     DsDataType,
+                                                     EDataType,
+                                                     ComputeDataType,
+                                                     AElementOp,
+                                                     BElementOp,
+                                                     CDEElementOp>;
 
-        ref_invoker.Run(ref_argument);
+using DeviceOpInstanceMNNN = DeviceOpInstanceMN_FP64<NumDimM,
+                                                     NumDimN,
+                                                     NumDimK,
+                                                     ADataType,
+                                                     BDataType,
+                                                     AccDataType,
+                                                     CShuffleDataType,
+                                                     DsDataType,
+                                                     EDataType,
+                                                     ComputeDataType,
+                                                     AElementOp,
+                                                     BElementOp,
+                                                     CDEElementOp>;
 
-        for(size_t m0 = 0; m0 < e_ms_ns_host_result.mDesc.GetLengths()[0]; ++m0)
-        {
-            for(size_t m1 = 0; m1 < e_ms_ns_host_result.mDesc.GetLengths()[1]; ++m1)
-            {
-                for(size_t n0 = 0; n0 < e_ms_ns_host_result.mDesc.GetLengths()[2]; ++n0)
-                {
-                    for(size_t n1 = 0; n1 < e_ms_ns_host_result.mDesc.GetLengths()[3]; ++n1)
-                    {
-                        cde_element_op(e_ms_ns_host_result(m0, m1, n0, n1),
-                                       c_ms_ns_host_result(m0, m1, n0, n1),
-                                       d_ms_ns(m0, m1, n0, n1));
-                    }
-                }
-            }
-        }
+using DeviceOpInstance = DeviceOpInstanceKKNN;
 
-        return ck::utils::check_err(e_ms_ns_device_result, e_ms_ns_host_result) ? 0 : 1;
-    }
+#include "run_contraction_bilinear_example.inc"
 
-    return 0;
-}
+int main(int argc, char* argv[]) { return run_contraction_bilinear_example(argc, argv); }

example/26_contraction/contraction_bilinear_xdl_fp64_compute_fp32.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = F64;
+using BDataType        = F64;
+using AccDataType      = F32;
+using CShuffleDataType = F64;
+using DDataType        = F64;
+using DsDataType       = ck::Tuple<DDataType>;
+using EDataType        = F64;
+using ComputeDataType  = F32;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Bilinear;
+
+using DeviceOpInstanceKKNN = DeviceOpInstanceKK_FP64<NumDimM,
+                                                     NumDimN,
+                                                     NumDimK,
+                                                     ADataType,
+                                                     BDataType,
+                                                     AccDataType,
+                                                     CShuffleDataType,
+                                                     DsDataType,
+                                                     EDataType,
+                                                     ComputeDataType,
+                                                     AElementOp,
+                                                     BElementOp,
+                                                     CDEElementOp>;
+
+using DeviceOpInstanceKNNN = DeviceOpInstanceKN_FP64<NumDimM,
+                                                     NumDimN,
+                                                     NumDimK,
+                                                     ADataType,
+                                                     BDataType,
+                                                     AccDataType,
+                                                     CShuffleDataType,
+                                                     DsDataType,
+                                                     EDataType,
+                                                     ComputeDataType,
+                                                     AElementOp,
+                                                     BElementOp,
+                                                     CDEElementOp>;
+
+using DeviceOpInstanceMKNN = DeviceOpInstanceMK_FP64<NumDimM,
+                                                     NumDimN,
+                                                     NumDimK,
+                                                     ADataType,
+                                                     BDataType,
+                                                     AccDataType,
+                                                     CShuffleDataType,
+                                                     DsDataType,
+                                                     EDataType,
+                                                     ComputeDataType,
+                                                     AElementOp,
+                                                     BElementOp,
+                                                     CDEElementOp>;
+
+using DeviceOpInstanceMNNN = DeviceOpInstanceMN_FP64<NumDimM,
+                                                     NumDimN,
+                                                     NumDimK,
+                                                     ADataType,
+                                                     BDataType,
+                                                     AccDataType,
+                                                     CShuffleDataType,
+                                                     DsDataType,
+                                                     EDataType,
+                                                     ComputeDataType,
+                                                     AElementOp,
+                                                     BElementOp,
+                                                     CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKNN;
+
+#include "run_contraction_bilinear_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_bilinear_example(argc, argv); }

example/26_contraction/contraction_scale_xdl_bf16_compute_fp32.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = BF16;
+using BDataType        = BF16;
+using AccDataType      = F32;
+using CShuffleDataType = BF16;
+using DsDataType       = ck::Tuple<>;
+using EDataType        = BF16;
+using ComputeDataType  = F32;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Scale;
+
+using DeviceOpInstanceKKN = DeviceOpInstanceKK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceKNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceMKN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceMNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKN;
+
+#include "run_contraction_scale_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_scale_example(argc, argv); }

example/26_contraction/contraction_scale_xdl_fp16_compute_fp32.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = F16;
+using BDataType        = F16;
+using AccDataType      = F32;
+using CShuffleDataType = F16;
+using DsDataType       = ck::Tuple<>;
+using EDataType        = F16;
+using ComputeDataType  = F32;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Scale;
+
+using DeviceOpInstanceKKN = DeviceOpInstanceKK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceKNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceMKN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceMNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKN;
+
+#include "run_contraction_scale_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_scale_example(argc, argv); }

example/26_contraction/contraction_scale_xdl_fp32.cpp

 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2023, 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/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_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/numeric.hpp"
-#include "ck/library/reference_tensor_operation/cpu/reference_contraction.hpp"
-
-template <ck::index_t... Is>
-using S = ck::Sequence<Is...>;
-
-using F32 = float;
-
-using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+#include "common_instances.hpp"
 
 using ADataType        = F32;
 using BDataType        = F32;
@@ -31,6 +12,7 @@ using AccDataType      = F32;
 using CShuffleDataType = F32;
 using DsDataType       = ck::Tuple<>;
 using EDataType        = F32;
+using ComputeDataType  = F32;
 
 static constexpr ck::index_t NumDimM = 2;
 static constexpr ck::index_t NumDimN = 2;
@@ -40,237 +22,64 @@ using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
 using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
 using CDEElementOp = ck::tensor_operation::element_wise::Scale;
 
-static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
-
-// clang-format off
-using DeviceOpInstanceKKN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F32,   F32,     F32,      F32, DsDataType,   F32,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4>;
-
-using DeviceOpInstanceKNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F32,   F32,     F32,      F32, DsDataType,   F32,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   4,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4>;
-
-using DeviceOpInstanceMKN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F32,   F32,     F32,      F32, DsDataType,   F32,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   1,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4>;
-
-using DeviceOpInstanceMNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F32,   F32,     F32,      F32, DsDataType,   F32,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   256,   128,    16,   1,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4>;
-// clang-format on
-
-using DeviceOpInstance = DeviceOpInstanceKKN;
-
-int main(int argc, char* argv[])
-{
-    bool do_verification = true;
-    int init_method      = 1;
-    bool time_kernel     = false;
-
-    // A[M0, M1, K0, K1]
-    std::vector<ck::index_t> a_ms_ks_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> a_ms_ks_strides{524288, 4096, 128, 1};
-    // B[N0, N1, K0, K1]
-    std::vector<ck::index_t> b_ns_ks_lengths{32, 64, 32, 64};
-    std::vector<ck::index_t> b_ns_ks_strides{524288, 4096, 128, 1};
-    // E[M0, M1, N0, N1]
-    std::vector<ck::index_t> e_ms_ns_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> e_ms_ns_strides{524288, 4096, 128, 1};
-
-    float scale = 1.f;
-
-    if(argc == 1)
-    {
-        // use default case
-    }
-    else if(argc == 4)
-    {
-        do_verification = std::stoi(argv[1]);
-        init_method     = std::stoi(argv[2]);
-        time_kernel     = std::stoi(argv[3]);
-    }
-    else if(argc == 23)
-    {
-        do_verification = std::stoi(argv[1]);
-        init_method     = std::stoi(argv[2]);
-        time_kernel     = std::stoi(argv[3]);
-
-        const ck::index_t M0 = std::stoi(argv[4]);
-        const ck::index_t M1 = std::stoi(argv[5]);
-
-        const ck::index_t N0 = std::stoi(argv[6]);
-        const ck::index_t N1 = std::stoi(argv[7]);
-
-        const ck::index_t K0 = std::stoi(argv[8]);
-        const ck::index_t K1 = std::stoi(argv[9]);
-
-        a_ms_ks_lengths = {M0, M1, K0, K1};
-        a_ms_ks_strides = {
-            std::stoi(argv[10]), std::stoi(argv[11]), std::stoi(argv[12]), std::stoi(argv[13])};
-
-        b_ns_ks_lengths = {N0, N1, K0, K1};
-        b_ns_ks_strides = {
-            std::stoi(argv[14]), std::stoi(argv[15]), std::stoi(argv[16]), std::stoi(argv[17])};
-
-        e_ms_ns_lengths = {M0, M1, N0, N1};
-        e_ms_ns_strides = {
-            std::stoi(argv[18]), std::stoi(argv[19]), std::stoi(argv[20]), std::stoi(argv[21])};
-
-        scale = std::stof(argv[22]);
-    }
-    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=no, 1=yes)\n");
-        printf("arg4 to 9: M0, M1, N0, N1, K0, K1\n");
-        printf("arg10 to 13: Stride_A_M0, Stride_A_M1, Stride_A_K0, Stride_A_K1\n");
-        printf("arg14 to 17: Stride_B_N0, Stride_B_N1, Stride_B_K0, Stride_B_K1\n");
-        printf("arg18 to 21: Stride_E_M0, Stride_E_M1, Stride_E_N0, Stride_E_N1\n");
-        printf("arg22: scale\n");
-        exit(0);
-    }
-
-    Tensor<ADataType> a_ms_ks(a_ms_ks_lengths, a_ms_ks_strides);
-    Tensor<BDataType> b_ns_ks(b_ns_ks_lengths, b_ns_ks_strides);
-    Tensor<EDataType> e_ms_ns_host_result(e_ms_ns_lengths, e_ms_ns_strides);
-    Tensor<EDataType> e_ms_ns_device_result(e_ms_ns_lengths, e_ms_ns_strides);
-
-    std::cout << "a_ms_ks: " << a_ms_ks.mDesc << std::endl;
-    std::cout << "b_ns_ks: " << b_ns_ks.mDesc << std::endl;
-    std::cout << "e_ms_ns: " << e_ms_ns_host_result.mDesc << std::endl;
-
-    switch(init_method)
-    {
-    case 0: break;
-    case 1:
-        a_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-5, 5});
-        b_ns_ks.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
-        break;
-    default:
-        a_ms_ks.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
-        b_ns_ks.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
-        break;
-    }
-
-    DeviceMem a_device_buf(sizeof(ADataType) * a_ms_ks.mDesc.GetElementSpaceSize());
-    DeviceMem b_device_buf(sizeof(BDataType) * b_ns_ks.mDesc.GetElementSpaceSize());
-    DeviceMem e_device_buf(sizeof(EDataType) * e_ms_ns_device_result.mDesc.GetElementSpaceSize());
-
-    a_device_buf.ToDevice(a_ms_ks.mData.data());
-    b_device_buf.ToDevice(b_ns_ks.mData.data());
-
-    // set zero
-    e_device_buf.SetZero();
-
-    auto a_element_op   = AElementOp{};
-    auto b_element_op   = BElementOp{};
-    auto cde_element_op = CDEElementOp{scale};
-
-    // device operation
-    auto op       = DeviceOpInstance{};
-    auto invoker  = op.MakeInvoker();
-    auto argument = op.MakeArgument(a_device_buf.GetDeviceBuffer(),
-                                    b_device_buf.GetDeviceBuffer(),
-                                    std::array<const void*, 0>{},
-                                    e_device_buf.GetDeviceBuffer(),
-                                    a_ms_ks_lengths,
-                                    a_ms_ks_strides,
-                                    b_ns_ks_lengths,
-                                    b_ns_ks_strides,
-                                    std::array<std::vector<ck::index_t>, 0>{},
-                                    std::array<std::vector<ck::index_t>, 0>{},
-                                    e_ms_ns_lengths,
-                                    e_ms_ns_strides,
-                                    a_element_op,
-                                    b_element_op,
-                                    cde_element_op);
-
-    if(!op.IsSupportedArgument(argument))
-    {
-        std::cout << op.GetTypeString() << " does not support this problem" << std::endl;
-
-        return 0;
-    }
-
-    float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
-
-    ck::index_t M =
-        ck::accumulate_n<ck::index_t>(e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
-
-    ck::index_t N = ck::accumulate_n<ck::index_t>(
-        e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
-
-    ck::index_t K = ck::accumulate_n<ck::index_t>(
-        a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
-
-    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, "
-              << op.GetTypeString() << std::endl;
-
-    e_device_buf.FromDevice(e_ms_ns_device_result.mData.data());
-
-    if(do_verification)
-    {
-        Tensor<CShuffleDataType> c_ms_ns_host_result(e_ms_ns_lengths, e_ms_ns_strides);
-
-        using ReferenceOpInstance =
-            ck::tensor_operation::host::ReferenceContraction_M2_N2_K2<NumDimM,
+using DeviceOpInstanceKKN = DeviceOpInstanceKK_Generic<NumDimM,
                                                        NumDimN,
                                                        NumDimK,
                                                        ADataType,
                                                        BDataType,
-                                                                      CShuffleDataType,
                                                        AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
                                                        AElementOp,
-                                                                      BElementOp>;
+                                                       BElementOp,
+                                                       CDEElementOp>;
 
-        auto ref_op      = ReferenceOpInstance{};
-        auto ref_invoker = ref_op.MakeInvoker();
+using DeviceOpInstanceKNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
 
-        Tensor<float> empty_tensor(std::vector<ck::index_t>{}, std::vector<ck::index_t>{});
-        auto ref_argument =
-            ref_op.MakeArgument(a_ms_ks, b_ns_ks, c_ms_ns_host_result, a_element_op, b_element_op);
+using DeviceOpInstanceMKN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
 
-        ref_invoker.Run(ref_argument);
+using DeviceOpInstanceMNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
 
-        for(size_t m0 = 0; m0 < e_ms_ns_host_result.mDesc.GetLengths()[0]; ++m0)
-        {
-            for(size_t m1 = 0; m1 < e_ms_ns_host_result.mDesc.GetLengths()[1]; ++m1)
-            {
-                for(size_t n0 = 0; n0 < e_ms_ns_host_result.mDesc.GetLengths()[2]; ++n0)
-                {
-                    for(size_t n1 = 0; n1 < e_ms_ns_host_result.mDesc.GetLengths()[3]; ++n1)
-                    {
-                        cde_element_op(e_ms_ns_host_result(m0, m1, n0, n1),
-                                       c_ms_ns_host_result(m0, m1, n0, n1));
-                    }
-                }
-            }
-        }
+using DeviceOpInstance = DeviceOpInstanceKKN;
 
-        return ck::utils::check_err(e_ms_ns_device_result, e_ms_ns_host_result) ? 0 : 1;
-    }
+#include "run_contraction_scale_example.inc"
 
-    return 0;
-}
+int main(int argc, char* argv[]) { return run_contraction_scale_example(argc, argv); }

example/26_contraction/contraction_scale_xdl_fp32_compute_bf16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = F32;
+using BDataType        = F32;
+using AccDataType      = F32;
+using CShuffleDataType = F32;
+using DsDataType       = ck::Tuple<>;
+using EDataType        = F32;
+using ComputeDataType  = BF16;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Scale;
+
+using DeviceOpInstanceKKN = DeviceOpInstanceKK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceKNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceMKN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceMNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKN;
+
+#include "run_contraction_scale_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_scale_example(argc, argv); }

example/26_contraction/contraction_scale_xdl_fp32_compute_fp16.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "common_instances.hpp"
+
+using ADataType        = F32;
+using BDataType        = F32;
+using AccDataType      = F32;
+using CShuffleDataType = F32;
+using DsDataType       = ck::Tuple<>;
+using EDataType        = F32;
+using ComputeDataType  = F16;
+
+static constexpr ck::index_t NumDimM = 2;
+static constexpr ck::index_t NumDimN = 2;
+static constexpr ck::index_t NumDimK = 2;
+
+using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
+using CDEElementOp = ck::tensor_operation::element_wise::Scale;
+
+using DeviceOpInstanceKKN = DeviceOpInstanceKK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceKNN = DeviceOpInstanceKN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceMKN = DeviceOpInstanceMK_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstanceMNN = DeviceOpInstanceMN_Generic<NumDimM,
+                                                       NumDimN,
+                                                       NumDimK,
+                                                       ADataType,
+                                                       BDataType,
+                                                       AccDataType,
+                                                       CShuffleDataType,
+                                                       DsDataType,
+                                                       EDataType,
+                                                       ComputeDataType,
+                                                       AElementOp,
+                                                       BElementOp,
+                                                       CDEElementOp>;
+
+using DeviceOpInstance = DeviceOpInstanceKKN;
+
+#include "run_contraction_scale_example.inc"
+
+int main(int argc, char* argv[]) { return run_contraction_scale_example(argc, argv); }

example/26_contraction/contraction_scale_xdl_fp64.cpp

 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2023, 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/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_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/numeric.hpp"
-#include "ck/library/reference_tensor_operation/cpu/reference_contraction.hpp"
-
-template <ck::index_t... Is>
-using S = ck::Sequence<Is...>;
-
-using F64 = double;
-
-using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+#include "common_instances.hpp"
 
 using ADataType        = F64;
 using BDataType        = F64;
@@ -31,6 +12,7 @@ using AccDataType      = F64;
 using CShuffleDataType = F64;
 using DsDataType       = ck::Tuple<>;
 using EDataType        = F64;
+using ComputeDataType  = F64;
 
 static constexpr ck::index_t NumDimM = 2;
 static constexpr ck::index_t NumDimN = 2;
@@ -40,237 +22,64 @@ using AElementOp   = ck::tensor_operation::element_wise::PassThrough;
 using BElementOp   = ck::tensor_operation::element_wise::PassThrough;
 using CDEElementOp = ck::tensor_operation::element_wise::Scale;
 
-static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
-
-// clang-format off
-using DeviceOpInstanceKKN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F64,   F64,     F64,      F64, DsDataType,   F64,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              2,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,         1,           1,           1,              S<1, 16, 1, 16>,               1>;
-
-using DeviceOpInstanceKNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F64,   F64,     F64,      F64, DsDataType,   F64,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   2,   1,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              2,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1>;
-
-using DeviceOpInstanceMKN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F64,   F64,     F64,      F64, DsDataType,   F64,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   1,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,         1,           1,           1,              S<1, 16, 1, 16>,               1>;
-
-using DeviceOpInstanceMNN = ck::tensor_operation::device::
-        //#####################################| NumDimM| NumDimN| NumDimK| 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|
-        //#####################################|        |        |        |      |      |        |         |           |      |             |            |             |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceContractionMultipleD_Xdl_CShuffle< NumDimM, NumDimN, NumDimK,   F64,   F64,     F64,      F64, DsDataType,   F64,   AElementOp,  BElementOp, CDEElementOp,       GemmSpec,        1,   256,   128,   128,    16,   1,   1,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1>;
-// clang-format on
-
-using DeviceOpInstance = DeviceOpInstanceKKN;
-
-int main(int argc, char* argv[])
-{
-    bool do_verification = true;
-    int init_method      = 1;
-    bool time_kernel     = false;
-
-    // A[M0, M1, K0, K1]
-    std::vector<ck::index_t> a_ms_ks_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> a_ms_ks_strides{524288, 4096, 128, 1};
-    // B[N0, N1, K0, K1]
-    std::vector<ck::index_t> b_ns_ks_lengths{32, 64, 32, 64};
-    std::vector<ck::index_t> b_ns_ks_strides{524288, 4096, 128, 1};
-    // E[M0, M1, N0, N1]
-    std::vector<ck::index_t> e_ms_ns_lengths{30, 128, 32, 64};
-    std::vector<ck::index_t> e_ms_ns_strides{524288, 4096, 128, 1};
-
-    float scale = 1.f;
-
-    if(argc == 1)
-    {
-        // use default case
-    }
-    else if(argc == 4)
-    {
-        do_verification = std::stoi(argv[1]);
-        init_method     = std::stoi(argv[2]);
-        time_kernel     = std::stoi(argv[3]);
-    }
-    else if(argc == 23)
-    {
-        do_verification = std::stoi(argv[1]);
-        init_method     = std::stoi(argv[2]);
-        time_kernel     = std::stoi(argv[3]);
-
-        const ck::index_t M0 = std::stoi(argv[4]);
-        const ck::index_t M1 = std::stoi(argv[5]);
-
-        const ck::index_t N0 = std::stoi(argv[6]);
-        const ck::index_t N1 = std::stoi(argv[7]);
-
-        const ck::index_t K0 = std::stoi(argv[8]);
-        const ck::index_t K1 = std::stoi(argv[9]);
-
-        a_ms_ks_lengths = {M0, M1, K0, K1};
-        a_ms_ks_strides = {
-            std::stoi(argv[10]), std::stoi(argv[11]), std::stoi(argv[12]), std::stoi(argv[13])};
-
-        b_ns_ks_lengths = {N0, N1, K0, K1};
-        b_ns_ks_strides = {
-            std::stoi(argv[14]), std::stoi(argv[15]), std::stoi(argv[16]), std::stoi(argv[17])};
-
-        e_ms_ns_lengths = {M0, M1, N0, N1};
-        e_ms_ns_strides = {
-            std::stoi(argv[18]), std::stoi(argv[19]), std::stoi(argv[20]), std::stoi(argv[21])};
-
-        scale = std::stof(argv[22]);
-    }
-    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=no, 1=yes)\n");
-        printf("arg4 to 9: M0, M1, N0, N1, K0, K1\n");
-        printf("arg10 to 13: Stride_A_M0, Stride_A_M1, Stride_A_K0, Stride_A_K1\n");
-        printf("arg14 to 17: Stride_B_N0, Stride_B_N1, Stride_B_K0, Stride_B_K1\n");
-        printf("arg18 to 21: Stride_E_M0, Stride_E_M1, Stride_E_N0, Stride_E_N1\n");
-        printf("arg22: scale\n");
-        exit(0);
-    }
-
-    Tensor<ADataType> a_ms_ks(a_ms_ks_lengths, a_ms_ks_strides);
-    Tensor<BDataType> b_ns_ks(b_ns_ks_lengths, b_ns_ks_strides);
-    Tensor<EDataType> e_ms_ns_host_result(e_ms_ns_lengths, e_ms_ns_strides);
-    Tensor<EDataType> e_ms_ns_device_result(e_ms_ns_lengths, e_ms_ns_strides);
-
-    std::cout << "a_ms_ks: " << a_ms_ks.mDesc << std::endl;
-    std::cout << "b_ns_ks: " << b_ns_ks.mDesc << std::endl;
-    std::cout << "e_ms_ns: " << e_ms_ns_host_result.mDesc << std::endl;
-
-    switch(init_method)
-    {
-    case 0: break;
-    case 1:
-        a_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-5, 5});
-        b_ns_ks.GenerateTensorValue(GeneratorTensor_2<BDataType>{-5, 5});
-        break;
-    default:
-        a_ms_ks.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
-        b_ns_ks.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
-        break;
-    }
-
-    DeviceMem a_device_buf(sizeof(ADataType) * a_ms_ks.mDesc.GetElementSpaceSize());
-    DeviceMem b_device_buf(sizeof(BDataType) * b_ns_ks.mDesc.GetElementSpaceSize());
-    DeviceMem e_device_buf(sizeof(EDataType) * e_ms_ns_device_result.mDesc.GetElementSpaceSize());
-
-    a_device_buf.ToDevice(a_ms_ks.mData.data());
-    b_device_buf.ToDevice(b_ns_ks.mData.data());
-
-    // set zero
-    e_device_buf.SetZero();
-
-    auto a_element_op   = AElementOp{};
-    auto b_element_op   = BElementOp{};
-    auto cde_element_op = CDEElementOp{scale};
-
-    // device operation
-    auto op       = DeviceOpInstance{};
-    auto invoker  = op.MakeInvoker();
-    auto argument = op.MakeArgument(a_device_buf.GetDeviceBuffer(),
-                                    b_device_buf.GetDeviceBuffer(),
-                                    std::array<const void*, 0>{},
-                                    e_device_buf.GetDeviceBuffer(),
-                                    a_ms_ks_lengths,
-                                    a_ms_ks_strides,
-                                    b_ns_ks_lengths,
-                                    b_ns_ks_strides,
-                                    std::array<std::vector<ck::index_t>, 0>{},
-                                    std::array<std::vector<ck::index_t>, 0>{},
-                                    e_ms_ns_lengths,
-                                    e_ms_ns_strides,
-                                    a_element_op,
-                                    b_element_op,
-                                    cde_element_op);
-
-    if(!op.IsSupportedArgument(argument))
-    {
-        std::cout << op.GetTypeString() << " does not support this problem" << std::endl;
-
-        return 0;
-    }
-
-    float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
-
-    ck::index_t M =
-        ck::accumulate_n<ck::index_t>(e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
-
-    ck::index_t N = ck::accumulate_n<ck::index_t>(
-        e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
-
-    ck::index_t K = ck::accumulate_n<ck::index_t>(
-        a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
-
-    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, "
-              << op.GetTypeString() << std::endl;
-
-    e_device_buf.FromDevice(e_ms_ns_device_result.mData.data());
-
-    if(do_verification)
-    {
-        Tensor<CShuffleDataType> c_ms_ns_host_result(e_ms_ns_lengths, e_ms_ns_strides);
-
-        using ReferenceOpInstance =
-            ck::tensor_operation::host::ReferenceContraction_M2_N2_K2<NumDimM,
+using DeviceOpInstanceKKN = DeviceOpInstanceKK_FP64<NumDimM,
                                                     NumDimN,
                                                     NumDimK,
                                                     ADataType,
                                                     BDataType,
-                                                                      CShuffleDataType,
                                                     AccDataType,
+                                                    CShuffleDataType,
+                                                    DsDataType,
+                                                    EDataType,
+                                                    ComputeDataType,
                                                     AElementOp,
-                                                                      BElementOp>;
+                                                    BElementOp,
+                                                    CDEElementOp>;
 
-        auto ref_op      = ReferenceOpInstance{};
-        auto ref_invoker = ref_op.MakeInvoker();
+using DeviceOpInstanceKNN = DeviceOpInstanceKN_FP64<NumDimM,
+                                                    NumDimN,
+                                                    NumDimK,
+                                                    ADataType,
+                                                    BDataType,
+                                                    AccDataType,
+                                                    CShuffleDataType,
+                                                    DsDataType,
+                                                    EDataType,
+                                                    ComputeDataType,
+                                                    AElementOp,
+                                                    BElementOp,
+                                                    CDEElementOp>;
 
-        Tensor<float> empty_tensor(std::vector<ck::index_t>{}, std::vector<ck::index_t>{});
-        auto ref_argument =
-            ref_op.MakeArgument(a_ms_ks, b_ns_ks, c_ms_ns_host_result, a_element_op, b_element_op);
+using DeviceOpInstanceMKN = DeviceOpInstanceMK_FP64<NumDimM,
+                                                    NumDimN,
+                                                    NumDimK,
+                                                    ADataType,
+                                                    BDataType,
+                                                    AccDataType,
+                                                    CShuffleDataType,
+                                                    DsDataType,
+                                                    EDataType,
+                                                    ComputeDataType,
+                                                    AElementOp,
+                                                    BElementOp,
+                                                    CDEElementOp>;
 
-        ref_invoker.Run(ref_argument);
+using DeviceOpInstanceMNN = DeviceOpInstanceMN_FP64<NumDimM,
+                                                    NumDimN,
+                                                    NumDimK,
+                                                    ADataType,
+                                                    BDataType,
+                                                    AccDataType,
+                                                    CShuffleDataType,
+                                                    DsDataType,
+                                                    EDataType,
+                                                    ComputeDataType,
+                                                    AElementOp,
+                                                    BElementOp,
+                                                    CDEElementOp>;
 
-        for(size_t m0 = 0; m0 < e_ms_ns_host_result.mDesc.GetLengths()[0]; ++m0)
-        {
-            for(size_t m1 = 0; m1 < e_ms_ns_host_result.mDesc.GetLengths()[1]; ++m1)
-            {
-                for(size_t n0 = 0; n0 < e_ms_ns_host_result.mDesc.GetLengths()[2]; ++n0)
-                {
-                    for(size_t n1 = 0; n1 < e_ms_ns_host_result.mDesc.GetLengths()[3]; ++n1)
-                    {
-                        cde_element_op(e_ms_ns_host_result(m0, m1, n0, n1),
-                                       c_ms_ns_host_result(m0, m1, n0, n1));
-                    }
-                }
-            }
-        }
+using DeviceOpInstance = DeviceOpInstanceKKN;
 
-        return ck::utils::check_err(e_ms_ns_device_result, e_ms_ns_host_result) ? 0 : 1;
-    }
+#include "run_contraction_scale_example.inc"
 
-    return 0;
-}
+int main(int argc, char* argv[]) { return run_contraction_scale_example(argc, argv); }