fp8 streamk

example/01_gemm/CMakeLists.txt

@@ -23,7 +23,6 @@ add_example_executable(example_gemm_xdl_fp16_v2 gemm_xdl_fp16_v2.cpp)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_v2)
 
 add_example_executable(example_gemm_xdl_fp16_streamk_v3 gemm_xdl_fp16_streamk_v3.cpp)
-target_compile_options(example_gemm_xdl_fp16_streamk_v3 PRIVATE -ggdb -O1 -march=native)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_streamk_v3)
 add_example_executable(example_gemm_xdl_fp16_v3 gemm_xdl_fp16_v3.cpp)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_v3)
@@ -78,6 +77,9 @@ add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp8)
 add_example_executable(example_gemm_xdl_fp8_bf8 gemm_xdl_fp8_bf8.cpp)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp8_bf8)
 
+add_example_executable(example_gemm_xdl_fp8_streamk_v3 gemm_xdl_fp8_streamk_v3.cpp)
+add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp8_streamk_v3)
+
 add_example_executable(example_gemm_xdl_fp16_fp8 gemm_xdl_fp16_fp8.cpp)
 add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_fp8)
 

example/01_gemm/gemm_xdl_fp16_streamk_v3.cpp

@@ -43,6 +43,16 @@ using DeviceGemmV2_Streamk_Instance =
 
 using ReferenceGemmInstance = ck::tensor_operation::host::
     ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;
+using ReferenceGemmInstanceGPU = ck::tensor_operation::device::ReferenceGemm<ALayout,
+                                                                             BLayout,
+                                                                             CLayout,
+                                                                             ADataType,
+                                                                             BDataType,
+                                                                             CDataType,
+                                                                             AccDataType,
+                                                                             AElementOp,
+                                                                             BElementOp,
+                                                                             CElementOp>;
 
 #include "run_gemm_example_streamk_v2.inc"
 

example/01_gemm/gemm_xdl_fp8_streamk_v3.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "common.hpp"
+
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_streamk_v3.hpp"
+
+using ADataType   = ck::f8_t;
+using BDataType   = ck::f8_t;
+using AccDataType = float;
+// using CShuffleDataType = ck::f8_t;
+using CShuffleDataType = ck::half_t;
+using CDataType        = ck::half_t;
+
+using ALayout = Row;
+using BLayout = Col;
+using CLayout = Row;
+
+using AElementOp = PassThrough;
+using BElementOp = PassThrough;
+using CElementOp = PassThrough;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+
+// clang-format off
+using DeviceGemmV2_Streamk_Instance = 
+    ck::tensor_operation::device::DeviceGemm_Xdl_CShuffle_Streamk_V3<
+        ALayout,   BLayout,  CLayout,   
+        ADataType,   BDataType,  CDataType,  AccDataType,  CShuffleDataType, 
+        PassThrough, PassThrough, PassThrough, GemmDefault, 
+          256,
+        128, 256, 
+        128, 16, 16,
+        16,   16,
+        4,    8, 
+        S<8, 32, 1>,  S<1, 0, 2>,  S<1, 0, 2>, 
+        2, 16, 16, 1,
+        S<8, 32, 1>,  S<1, 0, 2>,  S<1, 0, 2>, 
+        2, 16, 16, 1,
+        1, 2, S<1, 32, 1, 8>, 8,
+        ck::BlockGemmPipelineScheduler::Intrawave,ck::BlockGemmPipelineVersion::v3, ck::f8_t>;
+// clang-format on
+
+using ReferenceGemmInstance = ck::tensor_operation::host::
+    ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;
+using ReferenceGemmInstanceGPU = ck::tensor_operation::device::ReferenceGemm<ALayout,
+                                                                             BLayout,
+                                                                             CLayout,
+                                                                             ADataType,
+                                                                             BDataType,
+                                                                             CDataType,
+                                                                             AccDataType,
+                                                                             AElementOp,
+                                                                             BElementOp,
+                                                                             CElementOp>;
+
+#include "run_gemm_example_streamk_v2.inc"
+
+int main(int argc, char* argv[]) { return !run_gemm_universal_streamk_example(argc, argv); }

example/01_gemm/run_gemm_example_streamk_v2.inc

@@ -176,6 +176,7 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
 
     Tensor<CDataType> c_m_n_host_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
     Tensor<CDataType> c_m_n_device_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
+    Tensor<CDataType> c_m_n_device_ref_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
 
     std::cout << "a_m_k: " << a_m_k.mDesc << std::endl;
     std::cout << "b_k_n: " << b_k_n.mDesc << std::endl;
@@ -196,6 +197,8 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
     DeviceMem a_m_k_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpaceSize());
     DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n.mDesc.GetElementSpaceSize());
     DeviceMem c_m_n_device_buf(sizeof(CDataType) * c_m_n_device_result.mDesc.GetElementSpaceSize());
+    DeviceMem c_m_n_device_ref_buf(sizeof(CDataType) *
+                                   c_m_n_device_ref_result.mDesc.GetElementSpaceSize());
 
     a_m_k_device_buf.ToDevice(a_m_k.mData.data());
     b_k_n_device_buf.ToDevice(b_k_n.mData.data());
@@ -260,14 +263,13 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
     // }
 
     bool pass = true;
-    if(config.do_verification)
+    if((config.do_verification == 1) || (config.do_verification == 3))
     {
         auto ref_gemm    = ReferenceGemmInstance{};
         auto ref_invoker = ref_gemm.MakeInvoker();
 
         auto ref_argument = ref_gemm.MakeArgument(
             a_m_k, b_k_n, c_m_n_host_result, PassThrough{}, PassThrough{}, PassThrough{});
-        printf("inside do verification\n");
         ref_invoker.Run(ref_argument);
 
         ave_time = invoker.Run(argument, StreamConfig{nullptr, false, 1});
@@ -280,28 +282,49 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
 
         return ck::utils::check_err(c_m_n_device_result_converted, c_m_n_host_result);
 #else
-        printf("device copy initiated\n"); // HS
-        // if((workspace_size != 0) && (Streamk_sel > 0))
-        // {
-        //     printf("entered if\n");
-        //     workspace.FromDevice(c_m_n_device_result.mData.data());
-        // }
-        // else
-        //     c_m_n_device_buf.FromDevice(c_m_n_device_result.mData.data());
-        // printf("device copy finished\n"); // HS
-        // pass &= ck::utils::check_err(c_m_n_device_result,
-        //                              c_m_n_host_result,
-        //                              "Error: Incorrect results!",
-        //                              get_rtol<CDataType>(),
-        //                              get_atol<CDataType>());
+        c_m_n_device_buf.FromDevice(c_m_n_device_result.mData.data());
+
+        pass &= ck::utils::check_err(c_m_n_device_result,
+                                     c_m_n_host_result,
+                                     "Error: Incorrect results!",
+                                     get_rtol<CDataType>(),
+                                     get_atol<CDataType>());
 #endif
     }
 
+    if((config.do_verification == 2) || (config.do_verification == 3))
+    {
+        // GPU verification
+        auto ref_gemm_gpu    = ReferenceGemmInstanceGPU{};
+        auto ref_invoker_gpu = ref_gemm_gpu.MakeInvoker();
+
+        auto ref_argument_gpu = ref_gemm_gpu.MakeArgument(
+            static_cast<ADataType*>(a_m_k_device_buf.GetDeviceBuffer()),
+            static_cast<BDataType*>(b_k_n_device_buf.GetDeviceBuffer()),
+            static_cast<CDataType*>(c_m_n_device_ref_buf.GetDeviceBuffer()),
+            M,
+            N,
+            K,
+            a_element_op,
+            b_element_op,
+            c_element_op);
+
+        std::cout << "Running verification on GPU." << std::endl;
+        ref_invoker_gpu.Run(ref_argument_gpu, StreamConfig{});
+
+        c_m_n_device_ref_buf.FromDevice(c_m_n_device_ref_result.mData.data());
+        c_m_n_device_buf.FromDevice(c_m_n_device_result.mData.data());
+
+        pass &= ck::utils::check_err(c_m_n_device_result,
+                                     c_m_n_device_ref_result,
+                                     "Error: Incorrect results!",
+                                     get_rtol<CDataType>(),
+                                     get_atol<CDataType>());
+    }
+
     if(config.time_kernel)
     {
-        printf("before running timing\n");
         ave_time         = invoker.Run(argument, StreamConfig{nullptr, config.time_kernel});
-        printf("after running timing\n");
         std::size_t flop = 2_uz * M * N * K;
         std::size_t num_btype =
             sizeof(ADataType) * M * K + sizeof(BDataType) * K * N + sizeof(CDataType) * M * N;

include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_streamk_v3.hpp

@@ -131,22 +131,22 @@ struct DeviceGemm_Xdl_CShuffle_Streamk_V3 : public DeviceGemm_Streamk_V2<ALayout
     {
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-            printf("inside run\n");
+
             if(stream_config.log_level_ > 0)
             {
                 arg.Print();
             }
-            printf("done printing arg\n");
+
             if(!GridwiseGemm::CheckValidity(arg))
             {
                 throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
             }
-            printf("done checking arg validity\n");
+
             float ave_time = 0;
 
             index_t k_grain = KPerBlock;
             index_t K_split = (arg.K + k_grain - 1) / k_grain * KPerBlock;
-            printf("done finding k_split\n");
+
             const bool has_main_k_block_loop = GridwiseGemm::CalculateHasMainKBlockLoop(K_split);
             if constexpr(GridwiseGemm::Block2CTileMap_streamk::ReductionStrategy ==
                          StreamKReductionStrategy::Atomic)
@@ -218,14 +218,13 @@ struct DeviceGemm_Xdl_CShuffle_Streamk_V3 : public DeviceGemm_Streamk_V2<ALayout
                             hipGetErrorString(hipMemsetAsync(
                                 workspace_semaphore,
                                 0,
-                                sizeof(uint32_t),
-                                //arg.block_2_ctile_map_streamk.get_workspace_size_for_semaphore(),
+                                // sizeof(uint32_t),
+                                arg.block_2_ctile_map_streamk.get_workspace_size_for_semaphore(),
                                 stream_config.stream_id_));
                         };
-                        printf("before ave_time\n");
+
                         ave_time = launch_and_time_kernel_with_preprocess(
                             stream_config, preprocess, kernel, grid_dim, dim3(BlockSize), 0, arg);
-                        printf("after ave_time\n");
                     }
                 }
             };
@@ -246,9 +245,8 @@ struct DeviceGemm_Xdl_CShuffle_Streamk_V3 : public DeviceGemm_Streamk_V2<ALayout
                                                         true,
                                                         InMemoryDataOperationEnum::Set,
                                                         minimum_occupancy>;
-                        printf("before running lambda\n");
+
                         Run(kernel);
-                        printf("after running lambda\n");
                     }
                 }
                 // Tail number could be One to Seven
@@ -516,9 +514,224 @@ struct DeviceGemm_Xdl_CShuffle_Streamk_V3 : public DeviceGemm_Streamk_V2<ALayout
                              BElementwiseOperation,
                              CElementwiseOperation)
     {
+        constexpr index_t minimum_occupancy =
+            BlkGemmPipeSched == BlockGemmPipelineScheduler::Intrawave ? 1 : 2;
+        index_t K_split                  = (K + KPerBlock - 1) / KPerBlock * KPerBlock;
+        const bool has_main_k_block_loop = GridwiseGemm::CalculateHasMainKBlockLoop(K_split);
+        int occupancy, num_cu;
+        const auto calculate_grid_size = [&](const auto& kernel) {
+            hipError_t rtn;
+            rtn = hipOccupancyMaxActiveBlocksPerMultiprocessor(&occupancy, kernel, BlockSize, 0);
+            hip_check_error(rtn);
+
+            hipDeviceProp_t dev_prop;
+            hipDevice_t dev;
+            rtn = hipGetDevice(&dev);
+            hip_check_error(rtn);
+            rtn = hipGetDeviceProperties(&dev_prop, dev);
+            hip_check_error(rtn);
+            num_cu = dev_prop.multiProcessorCount;
+
+            Grid_size = num_cu * occupancy;
+        };
+
+        if(has_main_k_block_loop)
+        {
+            // Tail number always full
+            if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v1 ||
+                         BlkGemmPipelineVer == BlockGemmPipelineVersion::v3)
+            {
+
+                {
+                    const auto kernel = kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                                    true,
+                                                                    InMemoryDataOperationEnum::Set,
+                                                                    minimum_occupancy>;
+                    calculate_grid_size(kernel);
+                }
+            }
+            // Tail number could be One to Seven
+            else if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v2)
+            {
+
+                {
+                    if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::One)
+                    {
+                        const auto kernel =
+                            kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                        true,
+                                                        InMemoryDataOperationEnum::Set,
+                                                        minimum_occupancy,
+                                                        TailNumber::One>;
+                        calculate_grid_size(kernel);
+                    }
+                    else if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Full)
+                    {
+                        const auto kernel =
+                            kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                        true,
+                                                        InMemoryDataOperationEnum::Set,
+                                                        minimum_occupancy,
+                                                        TailNumber::Full>;
+                        calculate_grid_size(kernel);
+                    }
+
+                    if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 2)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Two)
+                        {
+                            const auto kernel =
+                                kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                            true,
+                                                            InMemoryDataOperationEnum::Set,
+                                                            minimum_occupancy,
+                                                            TailNumber::Two>;
+                            calculate_grid_size(kernel);
+                        }
+                    }
+
+                    if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 3)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Three)
+                        {
+                            const auto kernel =
+                                kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                            true,
+                                                            InMemoryDataOperationEnum::Set,
+                                                            minimum_occupancy,
+                                                            TailNumber::Three>;
+                            calculate_grid_size(kernel);
+                        }
+                    }
+
+                    if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 4)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Four)
+                        {
+                            const auto kernel =
+                                kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                            true,
+                                                            InMemoryDataOperationEnum::Set,
+                                                            minimum_occupancy,
+                                                            TailNumber::Four>;
+                            calculate_grid_size(kernel);
+                        }
+                    }
+
+                    if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 5)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Five)
+                        {
+                            const auto kernel =
+                                kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                            true,
+                                                            InMemoryDataOperationEnum::Set,
+                                                            minimum_occupancy,
+                                                            TailNumber::Five>;
+                            calculate_grid_size(kernel);
+                        }
+                    }
+
+                    if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 6)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Six)
+                        {
+                            const auto kernel =
+                                kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                            true,
+                                                            InMemoryDataOperationEnum::Set,
+                                                            minimum_occupancy,
+                                                            TailNumber::Six>;
+                            calculate_grid_size(kernel);
+                        }
+                    }
+
+                    if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 7)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Seven)
+                        {
+                            const auto kernel =
+                                kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                            true,
+                                                            InMemoryDataOperationEnum::Set,
+                                                            minimum_occupancy,
+                                                            TailNumber::Seven>;
+                            calculate_grid_size(kernel);
+                        }
+                    }
+                }
+            }
+            // Tail number could be Odd or Even
+            else if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v4)
+            {
+
+                {
+                    if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
+                    {
+                        const auto kernel =
+                            kernel_gemm_xdl_cshuffle_v3_2lds<GridwiseGemm,
+                                                             true,
+                                                             InMemoryDataOperationEnum::Set,
+                                                             minimum_occupancy,
+                                                             TailNumber::Odd>;
+                        calculate_grid_size(kernel);
+                    }
+                    else
+                    {
+                        const auto kernel =
+                            kernel_gemm_xdl_cshuffle_v3_2lds<GridwiseGemm,
+                                                             true,
+                                                             InMemoryDataOperationEnum::Set,
+                                                             minimum_occupancy,
+                                                             TailNumber::Even>;
+                        calculate_grid_size(kernel);
+                    }
+                }
+            }
+            else
+            {
+
+                {
+                    if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
+                    {
+                        const auto kernel =
+                            kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                        true,
+                                                        InMemoryDataOperationEnum::Set,
+                                                        minimum_occupancy,
+                                                        TailNumber::Odd>;
+                        calculate_grid_size(kernel);
+                    }
+                    else
+                    {
+                        const auto kernel =
+                            kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                        true,
+                                                        InMemoryDataOperationEnum::Set,
+                                                        minimum_occupancy,
+                                                        TailNumber::Even>;
+                        calculate_grid_size(kernel);
+                    }
+                }
+            }
+        }
+        else
+        {
+            // Tail number always 1
+            if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v1)
+            {
+
+                {
+                    const auto kernel = kernel_gemm_xdl_cshuffle_v3<GridwiseGemm,
+                                                                    false,
+                                                                    InMemoryDataOperationEnum::Set,
+                                                                    minimum_occupancy>;
+                    calculate_grid_size(kernel);
+                }
+            }
+        }
 
-        return Argument{
-            p_a, p_b, p_c, M, N, K, StrideA, StrideB, StrideC, streamk_sel, Grid_size}; // HS
+        return Argument{p_a, p_b, p_c, M, N, K, StrideA, StrideB, StrideC, streamk_sel, Grid_size};
     }
 
     static auto MakeInvoker() { return Invoker{}; }

library/include/ck/library/tensor_operation_instance/gpu/gemm_universal_streamk.hpp

@@ -237,6 +237,204 @@ void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_mnkpaddin
                                                       PassThrough,
                                                       PassThrough>>>& instances);
 #endif
+#if(defined(CK_ENABLE_FP8))
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_mnpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_mnpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_mnpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Col, Row, F8, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
 template <typename ADataType,
           typename BDataType,
           typename CDataType,
@@ -326,6 +524,120 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGemm_S
             }
         }
 #endif
+#if(defined(CK_ENABLE_FP8))
+        if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, f8_t> &&
+                     is_same_v<CDataType, half_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_mnpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_mnkpadding_instances(
+                    op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_mnkpadding_instances(
+                    op_ptrs);
+            }
+        }
+        else if constexpr(is_same_v<ADataType, f8_t> && is_same_v<BDataType, half_t> &&
+                          is_same_v<CDataType, half_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_mnpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instances(
+                    op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_mnpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_mnkpadding_instances(
+                    op_ptrs);
+            }
+        }
+#endif
 
         return op_ptrs;
     }

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/CMakeLists.txt

@@ -21,6 +21,47 @@ list(APPEND GEMM_UNIVERSAL_STREAMK_INSTANCES
         device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v1_mnkpadding_instance.cpp
         device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp
         device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp
-        device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instance.cpp)
+        device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instance.cpp
 
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_mnpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v1_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_v2_mnkpadding_instance.cpp
+
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_mnpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_comp_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v1_mnkpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp
+        device_gemm_xdl_universal_streamk_f8_f16_f16/device_gemm_xdl_universal_streamk_f8_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instance.cpp)
 add_instance_library(device_gemm_universal_streamk_instance ${GEMM_UNIVERSAL_STREAMK_INSTANCES})

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_streamk_v3.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = f8_t;
+using F16 = half_t;
+using F32 = float;
+
+using Row = tensor_layout::gemm::RowMajor;
+using Col = tensor_layout::gemm::ColumnMajor;
+
+template <index_t... Is>
+using S = Sequence<Is...>;
+
+using PassThrough = element_wise::PassThrough;
+
+static constexpr auto GemmDefault    = GemmSpecialization::Default;
+static constexpr auto GemmKPadding   = GemmSpecialization::KPadding;
+static constexpr auto GemmMNPadding  = GemmSpecialization::MNPadding;
+static constexpr auto GemmMNKPadding = GemmSpecialization::MNKPadding;
+
+static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
+static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
+
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_instances = std::tuple<
+    // clang-format off
+        //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| 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|                         Block-wiseGemm|               Block-wiseGemm|
+        //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
+        //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
+        //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
+        
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   4,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   256,    32,   8,   4,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
+    // clang-format on
+    >;
+
+template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_instances = std::tuple<
+    // clang-format off
+        //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| 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|    Block-wiseGemm|               Block-wiseGemm|
+        //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
+        //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
+        //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
+
+        // Latency friendly
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   256,   8,   4,  16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,   256,   8,   4,  16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        // Memory friendly
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   256,   8,   4,  16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,   256,   8,   4,  16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,   128,   8,   4,  16,   16,    1,    2,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   64,   128,   8,   4,  32,   32,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,  128,    64,   8,   4,  32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    32,  256,    64,   8,   4,  32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+    // clang-format on
+    >;
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_default_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_instances<GemmDefault>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_kpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_instances<GemmKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_instances<GemmMNKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_mnpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_instances<GemmMNPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_default_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_instances<Intrawave,
+                                                                            GemmDefault>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_kpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_instances<Intrawave,
+                                                                            GemmKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v1_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_instances<Intrawave,
+                                                                            GemmMNKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_default_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_instances<Interwave,
+                                                                            GemmDefault>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_kpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_instances<Interwave,
+                                                                            GemmKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F16 = ck::half_t;
+using F32 = float;
+
+void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_v2_mnkpadding_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm_Streamk_V2<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_mem_instances<Interwave,
+                                                                            GemmMNKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f8_f16/device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_streamk_v3.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = f8_t;
+using F16 = half_t;
+using F32 = float;
+
+using Row = tensor_layout::gemm::RowMajor;
+using Col = tensor_layout::gemm::ColumnMajor;
+
+template <index_t... Is>
+using S = Sequence<Is...>;
+
+using PassThrough = element_wise::PassThrough;
+
+static constexpr auto GemmDefault    = GemmSpecialization::Default;
+static constexpr auto GemmKPadding   = GemmSpecialization::KPadding;
+static constexpr auto GemmMNPadding  = GemmSpecialization::MNPadding;
+static constexpr auto GemmMNKPadding = GemmSpecialization::MNKPadding;
+
+static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
+static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
+
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_comp_instances = std::tuple<
+    // clang-format off
+        //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| 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|                         Block-wiseGemm|               Block-wiseGemm|
+        //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
+        //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
+        //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
+        
+        // Compute friendly
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,  16,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
+    // clang-format on
+    >;
+
+template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_streamk_f16_f8_f16_mk_nk_mn_mem_instances = std::tuple<
+    // clang-format off
+        //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| 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|    Block-wiseGemm|               Block-wiseGemm|
+        //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
+        //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
+        //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
+
+        // Latency friendly 
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,   128,   8,   16,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   128,   8,   16,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 8, 1>,      S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   128,   8,   16,  16,   16,    1,    1,     S<8,  8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 8, 1>,      S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,   128,   8,   16,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        // Memory friendly
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,   128,   32,   128,   8,   16,  32,   32,    2,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,   128,   16,   128,   8,   16,  16,   16,    4,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    64,   32,   128,   8,   16,  32,   32,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    64,   16,   128,   8,   16,  16,   16,    2,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,   128,   8,   16,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   128,   8,   16,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 8, 1>,      S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   128,   8,   16,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 8, 1>,      S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,   128,   8,   16,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,   128,   8,   16,  16,   16,    1,    2,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   64,   128,   8,   16,  32,   32,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,   128,   8,   16,  16,   16,    1,    4,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,  128,   128,   8,   16,  32,   32,    1,    2,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,   128,   8,   16,  16,   16,    1,    4,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    32,  256,   128,   8,   16,  32,   32,    1,    2,     S<16, 16, 1>,    S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 16>,              8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+    // clang-format on
+    >;
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck