Merge branch 'develop' into gfx950

include/ck_tile/ops/gemm/warp/warp_gemm_impl.hpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -14,6 +14,11 @@ struct WarpGemmImpl
     static constexpr index_t kM = WarpGemmAttribute::kM;
     static constexpr index_t kN = WarpGemmAttribute::kN;
     static constexpr index_t kK = WarpGemmAttribute::kK;
+    /// @brief The number of elements in K dimension processed by single thread in wavefront.
+    ///
+    /// @note  Note that WarpGemm may run MFMA instruction multiple times (on different K).
+    ///        In such situation this value reflects this fact.
+    static constexpr index_t kKPerThread = WarpGemmAttribute::kKPerThread;
 
     using ADataType = typename WarpGemmAttribute::ADataType;
     using BDataType = typename WarpGemmAttribute::BDataType;
@@ -31,11 +36,21 @@ struct WarpGemmImpl
     using BWarpTensor = static_distributed_tensor<BDataType, BWarpDstr>;
     using CWarpTensor = static_distributed_tensor<CDataType, CWarpDstr>;
 
-    CK_TILE_DEVICE void operator()(CWarpTensor& c, const AWarpTensor& a, const BWarpTensor& b) const
+    CK_TILE_HOST_DEVICE static constexpr auto get_num_of_access()
     {
-        using AVec = ext_vector_t<ADataType, AWarpTensor::get_thread_buffer_size()>;
-        using BVec = ext_vector_t<BDataType, BWarpTensor::get_thread_buffer_size()>;
-        using CVec = ext_vector_t<CDataType, CWarpTensor::get_thread_buffer_size()>;
+        return WarpGemmAttribute_::get_num_of_access();
+    }
+
+    template <typename CTensor, typename ATensor, typename BTensor, bool post_nop_ = false>
+    CK_TILE_DEVICE void
+    operator()(CTensor& c, const ATensor& a, const BTensor& b, bool_constant<post_nop_> = {}) const
+    {
+        static_assert(detail::is_similiar_distributed_tensor_v<CTensor, CWarpTensor> &&
+                      detail::is_similiar_distributed_tensor_v<ATensor, AWarpTensor> &&
+                      detail::is_similiar_distributed_tensor_v<BTensor, BWarpTensor>);
+        using AVec = ext_vector_t<ADataType, ATensor::get_thread_buffer_size()>;
+        using BVec = ext_vector_t<BDataType, BTensor::get_thread_buffer_size()>;
+        using CVec = ext_vector_t<CDataType, CTensor::get_thread_buffer_size()>;
 
         constexpr auto I0 = number<0>{};
 
@@ -44,18 +59,49 @@ struct WarpGemmImpl
         auto c_vec       = c.get_thread_buffer().template get_as<CVec>()[I0];
 
         // c_vec += a_vec * b_vec
-        WarpGemmAttribute{}(c_vec, a_vec, b_vec);
+        WarpGemmAttribute{}(c_vec, a_vec, b_vec, bool_constant<post_nop_>{});
 
         c.get_thread_buffer().template set_as<CVec>(I0, c_vec);
     }
 
-    CK_TILE_DEVICE auto operator()(const AWarpTensor& a, const BWarpTensor& b) const
+    template <typename CTensor,
+              typename ATensor,
+              typename BTensor,
+              index_t i_subk,
+              bool post_nop_ = false>
+    CK_TILE_DEVICE void operator()(CTensor& c,
+                                   const ATensor& a,
+                                   const BTensor& b,
+                                   number<i_subk>,
+                                   bool_constant<post_nop_> = {}) const
     {
-        CWarpTensor c;
+        using AVec = ext_vector_t<ADataType, ATensor::get_thread_buffer_size()>;
+        using BVec = ext_vector_t<BDataType, BTensor::get_thread_buffer_size()>;
+        using CVec = ext_vector_t<CDataType, CTensor::get_thread_buffer_size()>;
+
+        constexpr auto I0 = number<0>{};
 
-        using AVec = ext_vector_t<ADataType, AWarpTensor::get_thread_buffer_size()>;
-        using BVec = ext_vector_t<BDataType, BWarpTensor::get_thread_buffer_size()>;
-        using CVec = ext_vector_t<CDataType, CWarpTensor::get_thread_buffer_size()>;
+        const auto a_vec = a.get_thread_buffer().template get_as<AVec>()[I0];
+        const auto b_vec = b.get_thread_buffer().template get_as<BVec>()[I0];
+        auto c_vec       = c.get_thread_buffer().template get_as<CVec>()[I0];
+
+        // c_vec += a_vec * b_vec
+        WarpGemmAttribute{}(c_vec, a_vec, b_vec, number<i_subk>{}, bool_constant<post_nop_>{});
+
+        c.get_thread_buffer().template set_as<CVec>(I0, c_vec);
+    }
+
+    template <typename ATensor, typename BTensor>
+    CK_TILE_DEVICE auto operator()(const ATensor& a, const BTensor& b) const
+    {
+        using CTensor = CWarpTensor;
+        static_assert(detail::is_similiar_distributed_tensor_v<ATensor, AWarpTensor> &&
+                      detail::is_similiar_distributed_tensor_v<BTensor, BWarpTensor>);
+        CTensor c;
+
+        using AVec = ext_vector_t<ADataType, ATensor::get_thread_buffer_size()>;
+        using BVec = ext_vector_t<BDataType, BTensor::get_thread_buffer_size()>;
+        using CVec = ext_vector_t<CDataType, CTensor::get_thread_buffer_size()>;
 
         constexpr auto I0 = number<0>{};
 

include/ck_tile/ops/smoothquant.hpp

@@ -3,6 +3,7 @@
 
 #pragma once
 
+#include "ck_tile/ops/smoothquant/kernel/moe_smoothquant_kernel.hpp"
 #include "ck_tile/ops/smoothquant/kernel/smoothquant_kernel.hpp"
 #include "ck_tile/ops/smoothquant/pipeline/smoothquant_pipeline_default_policy.hpp"
 #include "ck_tile/ops/smoothquant/pipeline/smoothquant_pipeline_one_pass.hpp"

include/ck_tile/ops/smoothquant/kernel/moe_smoothquant_kernel.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include "ck_tile/core.hpp"
+#include "ck_tile/ops/common.hpp"
+
+namespace ck_tile {
+
+// host side args
+struct MoeSmoothquantHostArgs
+{
+    const void* p_x;        // [tokens ,hidden_size], input, fp16/bf16
+    const void* p_xscale;   // [experts, hidden_size], input, columnwise scale, fp32
+    const void* p_topk_ids; // [tokens, topk]
+
+    void* p_yscale; // [topk * tokens,  1], output, rowwise quant scale
+    void* p_qy;     // [topk * tokens, hidden_size], output
+
+    index_t tokens;
+    index_t hidden_size;
+    index_t experts;
+    index_t topk;
+    index_t x_stride; // input x row stride
+    index_t y_stride; // output y stride(stride for topk)
+};
+
+// TODO: Extract some type to wrapper class
+template <typename Pipeline_>
+struct MoeSmoothquant
+{
+    using Pipeline = remove_cvref_t<Pipeline_>;
+    using Problem  = typename Pipeline::Problem;
+
+    using XDataType       = remove_cvref_t<typename Problem::XDataType>;
+    using XScaleDataType  = remove_cvref_t<typename Problem::XScaleDataType>;
+    using ComputeDataType = remove_cvref_t<typename Problem::ComputeDataType>;
+    using YScaleDataType  = remove_cvref_t<typename Problem::YScaleDataType>;
+    using QYDataType      = remove_cvref_t<typename Problem::QYDataType>;
+
+    static constexpr index_t Block_M = Problem::BlockShape::Block_M;
+    static constexpr index_t Block_N = Problem::BlockShape::Block_N;
+    static constexpr bool kPadM      = false; // always no need to pad along M
+    static constexpr bool kPadN      = Problem::kPadN;
+    static constexpr bool kTwoPass   = Problem::kTwoPass;
+
+    static constexpr index_t ThreadPerWarp_N = Problem::BlockShape::ThreadPerWarp_N;
+    static constexpr index_t Vector_N        = Problem::BlockShape::Vector_N;
+    static constexpr index_t Repeat_N        = Problem::BlockShape::Repeat_N;
+
+    static constexpr auto I0 = number<0>{};
+    static constexpr auto I1 = number<1>{};
+
+    static_assert(Problem::BlockShape::Repeat_M == 1);
+
+    struct Kargs
+    {
+        const void* p_x;        // [tokens ,hidden_size], input, fp16/bf16
+        const void* p_xscale;   // [experts, hidden_size], input, columnwise scale, fp32
+        const void* p_topk_ids; // [tokens, topk]
+
+        void* p_yscale; // [topk, tokens, 1], output, rowwise quant scale
+        void* p_qy;     // [topk, tokens, hidden_size], output
+
+        index_t tokens;
+        index_t hidden_size;
+        index_t experts;
+        index_t topk;
+        index_t x_stride; // input x row stride
+        index_t y_stride; // output y stride(stride for topk)
+    };
+    using Hargs = MoeSmoothquantHostArgs;
+
+    CK_TILE_HOST static constexpr Kargs MakeKargs(const Hargs& hargs)
+    {
+        return Kargs{hargs.p_x,
+                     hargs.p_xscale,
+                     hargs.p_topk_ids,
+                     hargs.p_yscale,
+                     hargs.p_qy,
+                     hargs.tokens,
+                     hargs.hidden_size,
+                     hargs.experts,
+                     hargs.topk,
+                     hargs.x_stride,
+                     hargs.y_stride};
+    }
+
+    CK_TILE_HOST static constexpr auto GridSize(const Hargs& hargs)
+    {
+        return dim3(hargs.topk, integer_divide_ceil(hargs.tokens, Block_M), 1);
+    }
+
+    CK_TILE_HOST static constexpr auto BlockSize() { return Problem::BlockShape::BlockSize; }
+
+    // clang-format off
+    template <typename T> struct t2s;
+    template <> struct t2s<float> { static constexpr const char * name = "fp32"; };
+    template <> struct t2s<ck_tile::fp16_t> { static constexpr const char * name = "fp16"; };
+    template <> struct t2s<ck_tile::bf16_t> { static constexpr const char * name = "bf16"; };
+    template <> struct t2s<ck_tile::fp8_t> { static constexpr const char * name = "fp8"; };
+    template <> struct t2s<ck_tile::bf8_t> { static constexpr const char * name = "bf8"; };
+    // clang-format on
+
+    // in byte
+    CK_TILE_HOST_DEVICE static constexpr index_t GetSmemSize() { return Pipeline::GetSmemSize(); }
+
+    CK_TILE_HOST static std::string GetName()
+    {
+        // clang-format off
+        using S_ = typename Problem::BlockShape;
+        auto surfix = [&] () {
+            std::string n;
+            if (kPadN) n += "_pn";
+            if (kTwoPass) n += "_2p";
+            return n; }();
+
+        #define _SS_  std::string
+        #define _TS_  std::to_string
+        return _SS_("moe_smoothquant_") + _SS_(t2s<XDataType>::name) + "_" +
+             _TS_(S_::Block_M) + "x" + _TS_(S_::Block_N) + "_" + _TS_(S_::WarpPerBlock_M) + "x" + _TS_(S_::WarpPerBlock_N) + "_" +
+             _TS_(S_::Warp_M) + "x" + _TS_(S_::Warp_N) + "_" + _TS_(S_::Vector_M) + "x" + _TS_(S_::Vector_N) + "_" +
+             _SS_(Pipeline::name) + surfix;
+        #undef _SS_
+        #undef _TS_
+        // clang-format on
+    }
+
+    CK_TILE_DEVICE void operator()(Kargs kargs) const
+    {
+        const index_t i_topk  = blockIdx.x;
+        const index_t i_token = blockIdx.y * Block_M;
+        const index_t i_token_in_thrd =
+            __builtin_amdgcn_readfirstlane(threadIdx.x / Problem::BlockShape::ThreadPerBlock_N);
+
+        const index_t i_expert = reinterpret_cast<const index_t*>(
+            kargs.p_topk_ids)[(i_token + i_token_in_thrd) * kargs.topk + i_topk];
+
+        // [tokens ,hidden_size]
+        const auto x_window = [&]() {
+            const auto tmp_ = make_naive_tensor_view<address_space_enum::global>(
+                static_cast<const XDataType*>(kargs.p_x),
+                make_tuple(kargs.tokens, kargs.hidden_size),
+                make_tuple(kargs.x_stride, 1),
+                number<Vector_N>{},
+                number<1>{});
+
+            const auto tmp2_ = pad_tensor_view(
+                tmp_, make_tuple(number<Block_M>{}, number<Block_N>{}), sequence<kPadM, kPadN>{});
+            return make_tile_window(
+                tmp2_, make_tuple(number<Block_M>{}, number<Block_N>{}), {i_token, 0});
+        }();
+
+        // [experts, hidden_size],
+        const auto xscale_window = [&]() {
+            const auto tmp_ = make_naive_tensor_view<address_space_enum::global>(
+                static_cast<const XScaleDataType*>(kargs.p_xscale) + i_expert * kargs.hidden_size,
+                make_tuple(kargs.hidden_size),
+                make_tuple(1),
+                number<Vector_N>{},
+                number<1>{});
+
+            const auto tmp2_ =
+                pad_tensor_view(tmp_, make_tuple(number<Block_N>{}), sequence<kPadN>{});
+
+            return make_tile_window(tmp2_, make_tuple(number<Block_N>{}), {0});
+        }();
+
+        // [topk, tokens]
+        auto yscale_window = [&]() {
+            const auto tmp_ = make_naive_tensor_view<address_space_enum::global>(
+                static_cast<YScaleDataType*>(kargs.p_yscale) + i_topk * kargs.tokens,
+                make_tuple(kargs.tokens),
+                make_tuple(1),
+                number<1>{});
+
+            const auto tmp2_ =
+                pad_tensor_view(tmp_, make_tuple(number<Block_M>{}), sequence<kPadM>{});
+
+            return make_tile_window(tmp2_, make_tuple(number<Block_M>{}), {i_token});
+        }();
+
+        // [topk, tokens, hidden_size]
+        auto qy_window = [&]() {
+            auto tmp_ = make_naive_tensor_view<address_space_enum::global>(
+                static_cast<QYDataType*>(kargs.p_qy) + i_topk * kargs.tokens * kargs.y_stride,
+                make_tuple(kargs.tokens, kargs.hidden_size),
+                make_tuple(kargs.y_stride, 1),
+                number<Vector_N>{},
+                number<1>{});
+
+            auto tmp2_ = pad_tensor_view(
+                tmp_, make_tuple(number<Block_M>{}, number<Block_N>{}), sequence<kPadM, kPadN>{});
+            return make_tile_window(
+                tmp2_, make_tuple(number<Block_M>{}, number<Block_N>{}), {i_token, 0});
+        }();
+
+        __shared__ char smem[GetSmemSize()];
+
+        Pipeline{}(x_window, xscale_window, yscale_window, qy_window, kargs.hidden_size, smem);
+    }
+};
+
+} // namespace ck_tile

include/ck_tile/ops/smoothquant/pipeline/smoothquant_pipeline_one_pass.hpp

@@ -25,6 +25,7 @@ struct SmoothquantPipelineOnePass
     static constexpr bool kNeedCrossWarpSync = Problem::kNeedCrossWarpSync;
     static constexpr bool kPadM              = false; // TODO - BlockSmoothquantProblem::kPadM
     static constexpr bool kPadN              = Problem::kPadN;
+    static constexpr bool UseMax3            = true; // TODO - Move to trait
 
     static constexpr const char* name = []() {
         if constexpr(kNeedCrossWarpSync)
@@ -52,7 +53,15 @@ struct SmoothquantPipelineOnePass
             xscale_window_, Policy::template MakeXScaleBlockTileDistribution<Problem>());
 
         auto reduce_absmax_func  = ReduceOp::AbsMax{};
-        auto reduce_max_func     = ReduceOp::Max{};
+        auto reduce_absmax3_func = [](auto acc_, auto v_0_, auto v_1_) {
+            float rtn;
+            asm volatile("v_max3_f32 %0, %1, abs(%2), abs(%3)"
+                         : "=v"(rtn)
+                         : "v"(acc_), "v"(v_0_), "v"(v_1_));
+            return rtn;
+        };
+        auto reduce_max_func = ReduceOp::Max{};
+
         auto block_reduce2d      = Policy::template GetBlockReduce2d<Problem>();
         auto block_reduce2d_sync = Policy::template GetBlockReduce2dSync<Problem>();
         auto block_reduce2d_cross_warp_sync =
@@ -68,8 +77,23 @@ struct SmoothquantPipelineOnePass
             xscale);
 
         // compute absmax, cross-lane->cross-warp
-        auto absmax = block_reduce2d(
-            y, reduce_absmax_func.GetIdentityValue<ComputeDataType>(), reduce_absmax_func);
+        auto absmax = [&]() {
+            constexpr auto x_size_per_row =
+                x.get_tile_distribution().get_ys_to_d_descriptor().get_lengths().at(number<1>{});
+            if constexpr(UseMax3 && std::is_same_v<ComputeDataType, float> &&
+                         x_size_per_row % 2 == 0)
+            {
+                return block_reduce2d(y,
+                                      reduce_absmax_func.GetIdentityValue<ComputeDataType>(),
+                                      reduce_absmax3_func,
+                                      sequence<1, 2>{});
+            }
+            else
+            {
+                return block_reduce2d(
+                    y, reduce_absmax_func.GetIdentityValue<ComputeDataType>(), reduce_absmax_func);
+            }
+        }();
         block_reduce2d_sync(absmax, reduce_max_func);
         block_reduce2d_cross_warp_sync(absmax, smem, reduce_max_func);
 

include/ck_tile/ops/smoothquant/pipeline/smoothquant_pipeline_two_pass.hpp

@@ -25,6 +25,7 @@ struct SmoothquantPipelineTwoPass
     static constexpr bool kNeedCrossWarpSync = Problem::kNeedCrossWarpSync;
     static constexpr bool kPadM              = false; // TODO - BlockSmoothquantProblem::kPadM
     static constexpr bool kPadN              = Problem::kPadN;
+    static constexpr bool UseMax3            = true; // TODO - Move to trait
 
     static constexpr const char* name = []() {
         if constexpr(kNeedCrossWarpSync)
@@ -56,6 +57,13 @@ struct SmoothquantPipelineTwoPass
             __builtin_amdgcn_readfirstlane(integer_divide_ceil(row_size, Block_N));
 
         auto reduce_absmax_func  = ReduceOp::AbsMax{};
+        auto reduce_absmax3_func = [](auto acc_, auto v_0_, auto v_1_) {
+            float rtn;
+            asm volatile("v_max3_f32 %0, %1, abs(%2), abs(%3)"
+                         : "=v"(rtn)
+                         : "v"(acc_), "v"(v_0_), "v"(v_1_));
+            return rtn;
+        };
         auto reduce_max_func     = ReduceOp::Max{};
         auto block_reduce2d      = Policy::template GetBlockReduce2d<Problem>();
         auto block_reduce2d_sync = Policy::template GetBlockReduce2dSync<Problem>();
@@ -77,7 +85,13 @@ struct SmoothquantPipelineTwoPass
                 x,
                 xscale);
 
-            block_reduce2d(y, absmax, reduce_absmax_func);
+            constexpr auto x_size_per_row =
+                x.get_tile_distribution().get_ys_to_d_descriptor().get_lengths().at(number<1>{});
+            if constexpr(UseMax3 && std::is_same_v<ComputeDataType, float> &&
+                         x_size_per_row % 2 == 0)
+                block_reduce2d(y, absmax, reduce_absmax3_func, sequence<1, 2>{});
+            else
+                block_reduce2d(y, absmax, reduce_absmax_func);
 
             move_tile_window(x_window, {0, Block_N});
             move_tile_window(xscale_window, {Block_N});

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

@@ -180,6 +180,58 @@ struct DeviceOperationInstanceFactory<
             }
         }
 #endif
+#ifdef CK_ENABLE_BF16
+        if constexpr(is_same_v<ADataType, ck::bhalf_t> && is_same_v<BDataType, ck::bhalf_t> &&
+                     is_same_v<CDataType, ck::bhalf_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_bf16_bf16_bf16_mk_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_bf16_bf16_bf16_mk_nk_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_bf16_bf16_bf16_km_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_bf16_bf16_bf16_km_nk_mn_instances(op_ptrs);
+            }
+        }
+#endif
+#ifdef CK_ENABLE_INT8
+        if constexpr(is_same_v<ADataType, int8_t> && is_same_v<BDataType, int8_t> &&
+                     is_same_v<CDataType, int8_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_int8_int8_int8_mk_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_int8_int8_int8_mk_nk_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_int8_int8_int8_km_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_int8_int8_int8_km_nk_mn_instances(op_ptrs);
+            }
+        }
+#endif
 #endif
 
 #ifdef CK_USE_XDL

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

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <vector>
+#include <memory>
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_batched_gemm_multiple_d_xdl_cshuffle_v3.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+#ifdef CK_ENABLE_BF16
+void add_device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instances(
+    std::vector<std::unique_ptr<DeviceBatchedGemmV2MultiD<Row,
+                                                          Col,
+                                                          Empty_Tuple,
+                                                          Row,
+                                                          BF16,
+                                                          BF16,
+                                                          Empty_Tuple,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          PassThrough>>>& instances);
+
+void add_device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<DeviceBatchedGemmV2MultiD<Row,
+                                                          Col,
+                                                          Empty_Tuple,
+                                                          Row,
+                                                          BF16,
+                                                          BF16,
+                                                          Empty_Tuple,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          PassThrough>>>& instances);
+
+void add_device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<DeviceBatchedGemmV2MultiD<Row,
+                                                          Col,
+                                                          Empty_Tuple,
+                                                          Row,
+                                                          BF16,
+                                                          BF16,
+                                                          Empty_Tuple,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          PassThrough>>>& instances);
+
+#endif
+
+#ifdef CK_ENABLE_FP8
+void add_device_batched_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_default_instances(
+    std::vector<std::unique_ptr<DeviceBatchedGemmV2MultiD<Row,
+                                                          Col,
+                                                          Empty_Tuple,
+                                                          Row,
+                                                          F8,
+                                                          F8,
+                                                          Empty_Tuple,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          PassThrough>>>& instances);
+
+void add_device_batched_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<DeviceBatchedGemmV2MultiD<Row,
+                                                          Col,
+                                                          Empty_Tuple,
+                                                          Row,
+                                                          F8,
+                                                          F8,
+                                                          Empty_Tuple,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          PassThrough>>>& instances);
+
+void add_device_batched_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<DeviceBatchedGemmV2MultiD<Row,
+                                                          Col,
+                                                          Empty_Tuple,
+                                                          Row,
+                                                          F8,
+                                                          F8,
+                                                          Empty_Tuple,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          PassThrough>>>& instances);
+
+#endif
+
+template <typename ADataType,
+          typename BDataType,
+          typename DsDataType,
+          typename CDataType,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename CLayout>
+struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceBatchedGemmV2MultiD<
+    ALayout,
+    BLayout,
+    DsLayout,
+    CLayout,
+    ADataType,
+    BDataType,
+    DsDataType,
+    CDataType,
+    ck::tensor_operation::element_wise::PassThrough,
+    ck::tensor_operation::element_wise::PassThrough,
+    ck::tensor_operation::element_wise::PassThrough>>
+{
+    using DeviceOp = DeviceBatchedGemmV2MultiD<ALayout,
+                                               BLayout,
+                                               DsLayout,
+                                               CLayout,
+                                               ADataType,
+                                               BDataType,
+                                               DsDataType,
+                                               CDataType,
+                                               ck::tensor_operation::element_wise::PassThrough,
+                                               ck::tensor_operation::element_wise::PassThrough,
+                                               ck::tensor_operation::element_wise::PassThrough>;
+
+    static auto GetInstances()
+    {
+        std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
+
+#ifdef CK_ENABLE_BF16
+        if constexpr(is_same_v<ADataType, bhalf_t> && is_same_v<BDataType, bhalf_t> &&
+                     is_same_v<CDataType, bhalf_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instances(
+                    op_ptrs);
+
+                add_device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instances(
+                    op_ptrs);
+
+                add_device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instances(
+                    op_ptrs);
+            }
+        }
+#endif
+
+#ifdef CK_ENABLE_FP8
+        if constexpr(is_same_v<ADataType, f8_t> && is_same_v<BDataType, f8_t> &&
+                     is_same_v<CDataType, bhalf_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_batched_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_default_instances(
+                    op_ptrs);
+
+                add_device_batched_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_mem_v1_default_instances(
+                    op_ptrs);
+
+                add_device_batched_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_mem_v2_default_instances(
+                    op_ptrs);
+            }
+        }
+#endif
+        return op_ptrs;
+    }
+};
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

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

@@ -237,6 +237,206 @@ 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,
@@ -327,6 +527,121 @@ 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/include/ck/library/tensor_operation_instance/gpu/gemm_wmma.inc

@@ -28,6 +28,46 @@ void add_device_gemm_wmma_f16_f16_f16_mk_nk_mn_instances(
         DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
         instances);
 
+void add_device_gemm_wmma_bf16_bf16_bf16_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_bf16_bf16_bf16_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_bf16_bf16_bf16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_bf16_bf16_bf16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_int8_int8_int8_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_int8_int8_int8_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_int8_int8_int8_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_int8_int8_int8_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation

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

@@ -95,6 +95,45 @@ void add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instances(
                                                   PassThrough,
                                                   PassThrough>>>& instances);
 
+void add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_pv1_inter_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_pv1_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_pv2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
 void add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_irregular_instances(
     std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
                                                   Col,
@@ -189,6 +228,124 @@ void add_device_grouped_gemm_multiple_d_xdl_two_stage_bf16_bf16_bf16_mk_nk_mn_in
                                                   PassThrough,
                                                   PassThrough,
                                                   PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_kn_mn_irregular_pv1_inter_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_kn_mn_irregular_pv1_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_kn_mn_irregular_pv2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_nk_mn_irregular_pv1_inter_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Col,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_nk_mn_irregular_pv1_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Col,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_nk_mn_irregular_pv2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Col,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_km_kn_mn_irregular_pv1_inter_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Col,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_km_kn_mn_irregular_pv1_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Col,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_km_kn_mn_irregular_pv2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Col,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  BF16,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
 #endif
 
 #if defined(CK_ENABLE_BF16) && defined(CK_ENABLE_INT8)
@@ -262,7 +419,11 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
             {
                 add_device_grouped_gemm_xdl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
                 add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
-                add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_instances(
+                add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_pv1_inter_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_pv1_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_pv2_instances(
                     op_ptrs);
                 add_device_grouped_gemm_multiple_d_xdl_two_stage_f16_f16_f16_mk_kn_mn_instances(
                     op_ptrs);
@@ -334,12 +495,34 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
             {
                 add_device_grouped_gemm_multiple_d_xdl_two_stage_bf16_bf16_bf16_mk_kn_mn_instances(
                     op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_kn_mn_irregular_pv1_inter_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_kn_mn_irregular_pv1_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_kn_mn_irregular_pv2_instances(
+                    op_ptrs);
             }
             else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
                               is_same_v<ELayout, Row>)
             {
                 add_device_grouped_gemm_multiple_d_xdl_two_stage_bf16_bf16_bf16_mk_nk_mn_instances(
                     op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_nk_mn_irregular_pv1_inter_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_nk_mn_irregular_pv1_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_mk_nk_mn_irregular_pv2_instances(
+                    op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
+                              is_same_v<ELayout, Row>)
+            {
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_km_kn_mn_irregular_pv1_inter_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_km_kn_mn_irregular_pv1_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_xdl_splitk_bf16_bf16_bf16_km_kn_mn_irregular_pv2_instances(
+                    op_ptrs);
             }
         }
 #endif

library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_instance.cpp → library/include/ck/library/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_instance.hpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
 
 #include <cstdlib>
 
@@ -7,16 +9,17 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_splitk_cshuffle.hpp"
-
-#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_selector.hpp"
+#include "ck/utility/loop_scheduler.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
 
-using F16 = ck::half_t;
-using F32 = float;
+using F16  = ck::half_t;
+using BF16 = ck::bhalf_t;
+using F32  = float;
 
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -25,97 +28,109 @@ template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
 
 using Empty_Tuple = ck::Tuple<>;
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
-using PassThrough                    = ck::tensor_operation::element_wise::PassThrough;
-static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+static constexpr auto PipelineV1         = ck::PipelineVersion::v1;
+static constexpr auto PipelineV2         = ck::PipelineVersion::v2;
+static constexpr auto DefaultScheduler   = ck::LoopScheduler::Default;
+static constexpr auto InterwaveScheduler = ck::LoopScheduler::Interwave;
+static constexpr auto GemmMNKPadding     = device::GemmSpecialization::MNKPadding;
+static constexpr auto GemmDefault        = device::GemmSpecialization::Default;
 
-using device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_tile_instances = std::tuple<
+template <typename T,
+          device::GemmSpecialization GemmSpec = GemmMNKPadding,
+          PipelineVersion Pipeline            = PipelineV1,
+          LoopScheduler Scheduler             = DefaultScheduler,
+          enable_if_t<sizeof(T) == 2, bool>   = false>
+using device_grouped_gemm_xdl_splitk_2Bt_rrr_instances = std::tuple<
     // clang-format off
-        //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
-        //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
-        //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
-        //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1>,
-
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v1,  LoopScheduler::Interwave>,
+        //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| Pipeline | Loop      |
+        //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector| Version  | Scheduler |
+        //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|          |           |
+        //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |          |           |
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,   Pipeline,  Scheduler>
+    // clang-format on
+    >;
 
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>,
-        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8, PipelineVersion::v2>
+template <typename T,
+          device::GemmSpecialization GemmSpec = GemmMNKPadding,
+          PipelineVersion Pipeline            = PipelineV1,
+          LoopScheduler Scheduler             = DefaultScheduler,
+          enable_if_t<sizeof(T) == 2, bool>   = false>
+using device_grouped_gemm_xdl_splitk_2Bt_rcr_instances = std::tuple<
+    // clang-format off
+        //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| Pipeline | Loop      |
+        //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector| Version  | Scheduler |
+        //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|          |           |
+        //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |          |           |
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8, Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Col, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8, Pipeline,  Scheduler>
     // clang-format on
     >;
 
-void add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_instances(
-    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
-                                                  Row,
-                                                  Empty_Tuple,
-                                                  Row,
-                                                  F16,
-                                                  F16,
-                                                  Empty_Tuple,
-                                                  F16,
-                                                  PassThrough,
-                                                  PassThrough,
-                                                  PassThrough>>>& instances)
-{
-    add_device_operation_instances(
-        instances, device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_tile_instances{});
-}
+template <typename T,
+          device::GemmSpecialization GemmSpec = GemmMNKPadding,
+          PipelineVersion Pipeline            = PipelineV1,
+          LoopScheduler Scheduler             = DefaultScheduler,
+          enable_if_t<sizeof(T) == 2, bool>   = false>
+using device_grouped_gemm_xdl_splitk_2Bt_crr_instances = std::tuple<
+    // clang-format off
+        //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| Pipeline | Loop      |
+        //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector| Version  | Scheduler |
+        //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|          |           |
+        //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |          |           |
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   192,    32,   8,   8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   192,    32,    32,   8,   8,   32,   32,    3,    1,  S<1, 2, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,    64,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,    32,    32,   8,   8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,    64,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,   Pipeline,  Scheduler>,
+        DeviceGroupedGemmXdlSplitKCShuffle<    Col,    Row, Empty_Tuple,    Row,     T,     T,     F32,        T, Empty_Tuple,     T, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,  S<1, 4, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,   Pipeline,  Scheduler>
+    // clang-format on
+    >;
 
 } // namespace instance
 } // namespace device

library/src/tensor_operation_instance/gpu/CMakeLists.txt

@@ -62,7 +62,7 @@ function(add_instance_library INSTANCE_NAME)
     endforeach()
     # Do not build mha instances if gfx94 or gfx90a targets are not on the target list
     foreach(source IN LISTS ARGN)
-	    if(NOT INST_TARGETS MATCHES "gfx94" AND NOT INST_TARGETS MATCHES "gfx95" AND NOT INST_TARGETS MATCHES "gfx90a" AND source MATCHES "mha")
+	    if(NOT INST_TARGETS MATCHES "gfx94" AND NOT INST_TARGETS MATCHES "gfx90a" AND source MATCHES "mha")
          message("removing mha instance ${source} ")
          list(REMOVE_ITEM ARGN "${source}")
     endif()
@@ -70,17 +70,29 @@ function(add_instance_library INSTANCE_NAME)
     # Do not build gemm_universal_f8 or gemm_multiply_multiply_f8 for any targets except gfx94
     if(NOT CK_USE_FP8_ON_UNSUPPORTED_ARCH)
     foreach(source IN LISTS ARGN)
-    if(NOT INST_TARGETS MATCHES "gfx94" AND NOT INST_TARGETS MATCHES "gfx95" AND source MATCHES "gemm_multiply_multiply_xdl_f8")
+    if(NOT INST_TARGETS MATCHES "gfx94" AND source MATCHES "gemm_multiply_multiply_xdl_f8")
          message("removing gemm_multiply_multiply_f8 instance ${source} ")
          list(REMOVE_ITEM ARGN "${source}")
     endif()
     endforeach()
     foreach(source IN LISTS ARGN)
-    if(NOT INST_TARGETS MATCHES "gfx94" AND NOT INST_TARGETS MATCHES "gfx95" AND source MATCHES "gemm_xdl_universal" AND source MATCHES "_f8_")
+    if(NOT INST_TARGETS MATCHES "gfx94" AND source MATCHES "gemm_xdl_universal" AND source MATCHES "_f8_")
          message("removing gemm_universal_f8 instance ${source} ")
          list(REMOVE_ITEM ARGN "${source}")
     endif()
     endforeach()
+    foreach(source IN LISTS ARGN)
+    if(NOT INST_TARGETS MATCHES "gfx94" AND source MATCHES "batched_gemm_xdl_universal" AND source MATCHES "_f8_")
+         message("removing batched_gemm_universal_f8 instance ${source} ")
+         list(REMOVE_ITEM ARGN "${source}")
+    endif()
+    endforeach()
+    foreach(source IN LISTS ARGN)
+    if(NOT INST_TARGETS MATCHES "gfx94" AND source MATCHES "gemm_xdl_universal_streamk" AND source MATCHES "_f8_")
+         message("removing gemm_universal_streamk_f8 instance ${source} ")
+         list(REMOVE_ITEM ARGN "${source}")
+    endif()
+    endforeach()
     endif()
     #only continue if there are some source files left on the list
     if(ARGN)
@@ -90,7 +102,7 @@ function(add_instance_library INSTANCE_NAME)
             if(source MATCHES "_xdl")
                 list(REMOVE_ITEM INST_TARGETS gfx900 gfx906 gfx906:xnack- gfx1030 gfx1100 gfx1101 gfx1102 gfx1103 gfx1200 gfx1201 gfx10.3-generic gfx11-generic gfx12-generic)
             elseif(source MATCHES "_wmma")
-                list(REMOVE_ITEM INST_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack+ gfx908:xnack- gfx90a:xnack+ gfx90a:xnack- gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030 gfx950)
+                list(REMOVE_ITEM INST_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack+ gfx908:xnack- gfx90a:xnack+ gfx90a:xnack- gfx908 gfx90a gfx940 gfx941 gfx942 gfx1030)
             elseif(source MATCHES "mha")
                 list(REMOVE_ITEM INST_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack- gfx908:xnack+ gfx908 gfx1030 gfx1100 gfx1101 gfx1102 gfx1103 gfx1200 gfx1201 gfx10.3-generic gfx11-generic gfx12-generic)
             endif()
@@ -102,6 +114,9 @@ function(add_instance_library INSTANCE_NAME)
                 if(source MATCHES "gemm_multiply_multiply_f8")
                     list(REMOVE_ITEM INST_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack- gfx908:xnack+ gfx90a:xnack+ gfx90a:xnack- gfx908 gfx90a gfx1030 gfx1100 gfx1101 gfx1102 gfx1103 gfx1200 gfx1201 gfx10.3-generic gfx11-generic gfx12-generic)
                 endif()
+                if(source MATCHES "bached_gemm_multiply_multiply_f8")
+                    list(REMOVE_ITEM INST_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack- gfx908:xnack+ gfx90a:xnack+ gfx90a:xnack- gfx908 gfx90a gfx1030 gfx1100 gfx1101 gfx1102 gfx1103 gfx1200 gfx1201 gfx10.3-generic gfx11-generic gfx12-generic)
+                endif()
             endif()
             set(offload_targets)
             foreach(target IN LISTS INST_TARGETS)
@@ -331,7 +346,7 @@ if(CK_DEVICE_CONV_INSTANCES)
 endif()
 if(CK_DEVICE_MHA_INSTANCES)
         set(gpu_list ${INST_TARGETS})
-	if(gpu_list MATCHES "gfx94" OR gpu_list MATCHES "gfx90a" OR gpu_list MATCHES "gfx95")
+	if(gpu_list MATCHES "gfx94" OR gpu_list MATCHES "gfx90a")
             add_library(device_mha_operations STATIC ${CK_DEVICE_MHA_INSTANCES})
             add_library(composablekernels::device_mha_operations ALIAS device_mha_operations)
             target_compile_features(device_mha_operations PUBLIC)

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

@@ -2,9 +2,7 @@ set(GEMM_INSTANCES)
 list(APPEND GEMM_INSTANCES device_gemm_xdl_f64_f64_f64_mk_kn_mn_instance.cpp
     device_gemm_xdl_f64_f64_f64_mk_nk_mn_instance.cpp
     device_gemm_xdl_f64_f64_f64_km_kn_mn_instance.cpp
-    device_gemm_xdl_f64_f64_f64_km_nk_mn_instance.cpp)
-
-list(APPEND GEMM_INSTANCES
+    device_gemm_xdl_f64_f64_f64_km_nk_mn_instance.cpp
     device_gemm_xdl_f32_f32_f32_mk_kn_mn_instance.cpp
     device_gemm_xdl_f32_f32_f32_mk_nk_mn_instance.cpp
     device_gemm_xdl_f32_f32_f32_km_kn_mn_instance.cpp
@@ -21,9 +19,6 @@ list(APPEND GEMM_INSTANCES
     device_gemm_dl_f32_f32_f32_mk_nk_mn_instance.cpp
     device_gemm_dl_f32_f32_f32_km_kn_mn_instance.cpp
     device_gemm_dl_f32_f32_f32_km_nk_mn_instance.cpp
-    )
-
-list(APPEND GEMM_INSTANCES
     device_gemm_dl_f16_f16_f16_mk_kn_mn_instance.cpp
     device_gemm_dl_f16_f16_f16_mk_kn_mn_irregular_instance.cpp
     device_gemm_dl_f16_f16_f16_mk_nk_mn_instance.cpp
@@ -78,9 +73,6 @@ list(APPEND GEMM_INSTANCES
     device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_default_pipeline_v1_instance.cpp
     device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_default_pipeline_v2_instance.cpp
     device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
-    )
-
-list(APPEND GEMM_INSTANCES
     device_gemm_dl_i8_i8_i8_mk_kn_mn_instance.cpp
     device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instance.cpp
     device_gemm_dl_i8_i8_i8_mk_nk_mn_instance.cpp
@@ -92,15 +84,11 @@ list(APPEND GEMM_INSTANCES
     device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp
     device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
     device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp
-    device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp)
-
-list(APPEND GEMM_INSTANCES
+    device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp
     device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instance.cpp
     device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
     device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instance.cpp
-    device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instance.cpp)
-
-list(APPEND GEMM_INSTANCES
+    device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instance.cpp
     device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_kn_mn_v1_default_instance.cpp
     device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_kn_mn_v1_interwave_default_instance.cpp
     device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_kn_mn_v2_default_instance.cpp
@@ -109,14 +97,19 @@ list(APPEND GEMM_INSTANCES
     device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_kn_mn_v2_padded_instance.cpp
     device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_nk_mn_instance.cpp
     device_gemm_xdl_c_shuffle_fp8_fp8_fp8_km_kn_mn_instance.cpp
-    device_gemm_xdl_c_shuffle_fp8_fp8_fp8_km_nk_mn_instance.cpp)
-
-
-list(APPEND GEMM_INSTANCES
+    device_gemm_xdl_c_shuffle_fp8_fp8_fp8_km_nk_mn_instance.cpp
     device_gemm_wmma_f16_f16_f16_mk_kn_mn_instance.cpp
     device_gemm_wmma_f16_f16_f16_mk_nk_mn_instance.cpp
     device_gemm_wmma_f16_f16_f16_km_kn_mn_instance.cpp
-    device_gemm_wmma_f16_f16_f16_km_nk_mn_instance.cpp)
+    device_gemm_wmma_f16_f16_f16_km_nk_mn_instance.cpp
+    device_gemm_wmma_bf16_bf16_bf16_mk_kn_mn_instance.cpp
+    device_gemm_wmma_bf16_bf16_bf16_mk_nk_mn_instance.cpp
+    device_gemm_wmma_bf16_bf16_bf16_km_kn_mn_instance.cpp
+    device_gemm_wmma_bf16_bf16_bf16_km_nk_mn_instance.cpp
+    device_gemm_wmma_int8_int8_int8_mk_kn_mn_instance.cpp
+    device_gemm_wmma_int8_int8_int8_mk_nk_mn_instance.cpp
+    device_gemm_wmma_int8_int8_int8_km_kn_mn_instance.cpp
+    device_gemm_wmma_int8_int8_int8_km_nk_mn_instance.cpp)
 
 add_instance_library(device_gemm_instance ${GEMM_INSTANCES})
 

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f32_f32_f32_km_kn_mn_instance.cpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <cstdlib>
 
@@ -14,15 +14,12 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-using F16 = ck::half_t;
 using F32 = float;
-
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
 
 template <ck::index_t... Is>
-using S = ck::Sequence<Is...>;
-
+using S           = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
@@ -34,7 +31,39 @@ using device_gemm_dl_f32_f32_f32_km_kn_mn_instances = std::tuple<
         //  ########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   | ThreadM111| ThreadN111| Thread| ClusterM110Xs| ClusterN110Xs| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|     DstVectorTensor|        SrcDstAccess|  SrcDstVectorDim| DstScalarPerVector|
         //  ########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |           |           |       |              |              |        K0_M0_M1_K1|          K0_M0_M1_K1|   ArrangeOrder|          Order| Lengths_K0_M0_M1_K1| ContiguousDimOrder|  Lengths_K0_M0_M1_K1|        K0_N0_N1_K1|          K0_N0_N1_K1|   ArrangeOrder|          Order| Lengths_K0_N0_N1_K1| ContiguousDimOrder| Lengths_K0_N0_N1_K1|               Order|                 |                   |
         //  ########|      |      |      |        |        |        |        |            |            |            |               |      |      |      |      |   |           |           |       |              |              |                   |                     |               |               |                    |                   |                     |                   |                     |               |               |                    |                   |                    |                    |                 |                   |
-        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  1,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<2, 1, 4, 1>,      S<8, 1,  32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 1>,      S<2, 1, 4, 1>,      S<8, 1,  32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,       S<1, 1, 4, 1>, S<0, 1, 2, 3, 4, 5>,                5,                  4>
+        // MPerBlock=128, NPerBlock=128
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<4, 4>,       S<4, 4>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<2, 8>,       S<2, 8>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=128, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,   128,    64,    16,  2,          4,          4,      1,       S<8, 2>,       S<4, 2>,      S<2, 1, 8, 2>,       S<8, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 8, 2>,        S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,   128,    64,    16,  2,          4,          4,      1,       S<2, 8>,       S<2, 4>,      S<2, 1, 8, 2>,       S<8, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 8, 2>,        S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=128
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,    64,   128,    16,  2,          4,          4,      1,       S<4, 2>,       S<8, 2>,      S<2, 1, 8, 2>,        S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 8, 2>,       S<8, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,    64,   128,    16,  2,          4,          4,      1,       S<2, 4>,       S<2, 8>,      S<2, 1, 8, 2>,        S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 8, 2>,       S<8, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<4, 2>,       S<4, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<2, 4>,       S<2, 4>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<8, 1>,       S<4, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<4, 2>,       S<8, 1>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=16, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    16,    64,    16,  2,          1,          4,      1,       S<4, 2>,       S<4, 2>,      S<1, 1, 4, 2>,       S<16, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<4, 1, 4, 2>,       S<4, 1, 16, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=16
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    16,    16,  2,          4,          1,      1,       S<4, 2>,       S<4, 2>,      S<4, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<1, 1, 4, 2>,       S<16, 1, 4, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        // MPerBlock=16, NPerBlock=16
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    16,    16,    16,    16,  2,          2,          2,      1,       S<2, 2>,       S<2, 2>,      S<4, 1, 4, 2>,        S<4, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<4, 1, 4, 2>,        S<4, 1, 4, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    16,    16,    16,    16,  2,          2,          2,      1,       S<1, 4>,       S<1, 4>,      S<4, 1, 4, 2>,        S<4, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<4, 1, 4, 2>,        S<4, 1, 4, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        // MPerBlock=8, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,     8,    64,    32,  2,          1,          2,      1,       S<4, 1>,       S<8, 2>,      S<1, 1, 4, 2>,       S<32, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<8, 1, 4, 2>,       S<4, 1, 16, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,     8,    64,    32,  2,          1,          2,      1,       S<2, 2>,       S<8, 2>,      S<1, 1, 4, 2>,       S<32, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<8, 1, 4, 2>,       S<4, 1, 16, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        // MPerBlock=64, NPerBlock=8
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,     8,    32,  2,          2,          1,      1,       S<8, 2>,       S<4, 1>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<1, 1, 4, 2>,       S<32, 1, 2, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,     8,    32,  2,          2,          1,      1,       S<8, 2>,       S<2, 2>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<1, 1, 4, 2>,       S<32, 1, 2, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        // MPerBlock=8, NPerBlock=8
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          1,          2,      1,       S<4, 1>,       S<2, 1>,      S<1, 1, 4, 2>,        S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<1, 1, 4, 2>,        S<4, 1, 2, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          1,          2,      1,       S<1, 4>,       S<1, 2>,      S<1, 1, 4, 2>,        S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<1, 1, 4, 2>,        S<4, 1, 2, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          2,          1,      1,       S<2, 1>,       S<4, 1>,      S<1, 1, 4, 2>,        S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<1, 1, 4, 2>,        S<4, 1, 2, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          2,          1,      1,       S<1, 2>,       S<1, 4>,      S<1, 1, 4, 2>,        S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,     S<1, 1, 4, 2>,        S<4, 1, 2, 1>,   S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>
     // clang-format on
     >;
 

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f32_f32_f32_km_nk_mn_instance.cpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <cstdlib>
 
@@ -14,15 +14,12 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-using F16 = ck::half_t;
 using F32 = float;
-
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
 
 template <ck::index_t... Is>
-using S = ck::Sequence<Is...>;
-
+using S           = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
@@ -35,7 +32,39 @@ using device_gemm_dl_f32_f32_f32_km_nk_mn_instances =
         //  ########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   | ThreadM111| ThreadN111| Thread| ClusterM110Xs| ClusterN110Xs| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor|        SrcDstAccess|  SrcDstVectorDim| DstScalarPerVector|
         //  ########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |           |           |       |              |              |        K0_M0_M1_K1|          K0_M0_M1_K1|   ArrangeOrder|          Order| Lengths_K0_M0_M1_K1| ContiguousDimOrder|  Lengths_K0_M0_M1_K1|        K0_N0_N1_K1|          K0_N0_N1_K1|   ArrangeOrder|          Order| Lengths_K0_N0_N1_K1| ContiguousDimOrder|  Lengths_K0_N0_N1_K1|               Order|                 |                   |
         //  ########|      |      |      |        |        |        |        |            |            |            |               |      |      |      |      |   |           |           |       |              |              |                   |                     |               |               |                    |                   |                     |                   |                     |               |               |                    |                   |                     |                    |                 |                   |
-        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  1,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<2, 1, 4, 1>,      S<8, 1,  32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 1>,      S<8, 1, 1, 1>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 1>,      S<1, 2, 0, 3>,        S<1, 1, 1, 1>, S<0, 1, 2, 3, 4, 5>,                5,                  4>
+        // MPerBlock=128, NPerBlock=128
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<4, 4>,       S<4, 4>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<2, 8>,       S<2, 8>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=128, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,   128,    64,    16,  2,          4,          4,      1,       S<8, 2>,       S<4, 2>,      S<2, 1, 8, 2>,       S<8, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,   128,    64,    16,  2,          4,          4,      1,       S<2, 8>,       S<2, 4>,      S<2, 1, 8, 2>,       S<8, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=128
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,    64,   128,    16,  2,          4,          4,      1,       S<4, 2>,       S<8, 2>,      S<2, 1, 8, 2>,        S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 2, 2>,       S<2, 1, 64, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,    64,   128,    16,  2,          4,          4,      1,       S<2, 4>,       S<2, 8>,      S<2, 1, 8, 2>,        S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 2, 2>,       S<2, 1, 64, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<4, 2>,       S<4, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<2, 4>,       S<2, 4>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<8, 1>,       S<4, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<4, 2>,       S<8, 1>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=16, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    16,    64,    16,  2,          1,          4,      1,       S<4, 2>,       S<4, 2>,      S<1, 1, 4, 2>,       S<16, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 4, 2>,       S<4, 1, 16, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=16
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    16,    16,  2,          4,          1,      1,       S<4, 2>,       S<4, 2>,      S<4, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 1, 2>,       S<4, 1, 16, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        // MPerBlock=16, NPerBlock=16
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    16,    16,    16,    16,  2,          2,          2,      1,       S<2, 2>,       S<2, 2>,      S<4, 1, 4, 2>,       S<4, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,        S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 4, 2>,        S<4, 1, 4, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    16,    16,    16,    16,  2,          2,          2,      1,       S<1, 4>,       S<1, 4>,      S<4, 1, 4, 2>,       S<4, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,        S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 4, 2>,        S<4, 1, 4, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        // MPerBlock=8, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,     8,    64,    32,  2,          1,          2,      1,       S<4, 1>,       S<8, 2>,      S<1, 1, 4, 2>,       S<32, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,     8,    64,    32,  2,          1,          2,      1,       S<2, 2>,       S<8, 2>,      S<1, 1, 4, 2>,       S<32, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        // MPerBlock=64, NPerBlock=8
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,     8,    32,  2,          2,          1,      1,       S<8, 2>,       S<4, 1>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 1, 2>,        S<8, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,     8,    32,  2,          2,          1,      1,       S<8, 2>,       S<2, 2>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 1, 2>,        S<8, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        // MPerBlock=8, NPerBlock=8
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          1,          2,      1,       S<4, 1>,       S<2, 1>,      S<1, 1, 4, 2>,       S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,        S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 1, 2>,        S<1, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          1,          2,      1,       S<1, 4>,       S<1, 2>,      S<1, 1, 4, 2>,       S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,        S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 1, 2>,        S<1, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          2,          1,      1,       S<2, 1>,       S<4, 1>,      S<1, 1, 4, 2>,       S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,        S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 1, 2>,        S<1, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Col,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          2,          1,      1,       S<1, 2>,       S<1, 4>,      S<1, 1, 4, 2>,       S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,        S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>,      S<4, 1, 1, 2>,        S<1, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>
         // clang-format on
         >;
 

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f32_f32_f32_mk_kn_mn_instance.cpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <cstdlib>
 
@@ -14,15 +14,12 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-using F16 = ck::half_t;
 using F32 = float;
-
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
 
 template <ck::index_t... Is>
-using S = ck::Sequence<Is...>;
-
+using S           = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
@@ -35,7 +32,39 @@ using device_gemm_dl_f32_f32_f32_mk_kn_mn_instances =
         //  ########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   | ThreadM111| ThreadN111| Thread| ClusterM110Xs| ClusterN110Xs| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor|        SrcDstAccess|  SrcDstVectorDim| DstScalarPerVector|
         //  ########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |           |           |       |              |              |        K0_M0_M1_K1|          K0_M0_M1_K1|   ArrangeOrder|          Order| Lengths_K0_M0_M1_K1| ContiguousDimOrder|  Lengths_K0_M0_M1_K1|        K0_N0_N1_K1|          K0_N0_N1_K1|   ArrangeOrder|          Order| Lengths_K0_N0_N1_K1| ContiguousDimOrder|  Lengths_K0_N0_N1_K1|               Order|                 |                   |
         //  ########|      |      |      |        |        |        |        |            |            |            |               |      |      |      |      |   |           |           |       |              |              |                   |                     |               |               |                    |                   |                     |                   |                     |               |               |                    |                   |                     |                    |                 |                   |
-        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  1,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<8, 1, 1, 1>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 1>,      S<1, 2, 0, 3>,        S<1, 1, 1, 1>,      S<2, 1, 4, 1>,      S<8, 1,  32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 1>, S<0, 1, 2, 3, 4, 5>,                5,                  4>
+        // MPerBlock=128, NPerBlock=128
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<4, 4>,       S<4, 4>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<2, 8>,       S<2, 8>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 4, 2>,       S<8, 1, 32, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=128, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,   128,    64,    16,  2,          4,          4,      1,       S<8, 2>,       S<4, 2>,      S<8, 1, 2, 2>,       S<2, 1, 64, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 8, 2>,        S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,   128,    64,    16,  2,          4,          4,      1,       S<2, 8>,       S<2, 4>,      S<8, 1, 2, 2>,       S<2, 1, 64, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 8, 2>,        S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=128
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,    64,   128,    16,  2,          4,          4,      1,       S<4, 2>,       S<8, 2>,      S<8, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 8, 2>,       S<8, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,    64,   128,    16,  2,          4,          4,      1,       S<2, 4>,       S<2, 8>,      S<8, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 8, 2>,       S<8, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<4, 2>,       S<4, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<2, 4>,       S<2, 4>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<8, 1>,       S<4, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<4, 2>,       S<8, 1>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<2, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=16, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    16,    64,    16,  2,          1,          4,      1,       S<4, 2>,       S<4, 2>,      S<4, 1, 1, 2>,      S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<4, 1, 4, 2>,        S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=16
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    16,    16,  2,          4,          1,      1,       S<4, 2>,       S<4, 2>,      S<4, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<1, 1, 4, 2>,       S<16, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        // MPerBlock=16, NPerBlock=16
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    16,    16,    16,    16,  2,          2,          2,      1,       S<2, 2>,       S<2, 2>,      S<4, 1, 4, 2>,        S<4, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<4, 1, 4, 2>,        S<4, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    16,    16,    16,    16,  2,          2,          2,      1,       S<1, 4>,       S<1, 4>,      S<4, 1, 4, 2>,        S<4, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<4, 1, 4, 2>,        S<4, 1, 4, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        // MPerBlock=8, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,     8,    64,    32,  2,          1,          2,      1,       S<4, 1>,       S<8, 2>,      S<4, 1, 1, 2>,       S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<8, 1, 4, 2>,        S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,     8,    64,    32,  2,          1,          2,      1,       S<2, 2>,       S<8, 2>,      S<4, 1, 1, 2>,       S<8, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<8, 1, 4, 2>,        S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        // MPerBlock=64, NPerBlock=8
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,     8,    32,  2,          2,          1,      1,       S<8, 2>,       S<4, 1>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<1, 1, 4, 2>,       S<32, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,     8,    32,  2,          2,          1,      1,       S<8, 2>,       S<2, 2>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<1, 1, 4, 2>,       S<32, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        // MPerBlock=8, NPerBlock=8
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          1,          2,      1,       S<4, 1>,       S<2, 1>,      S<4, 1, 1, 2>,        S<1, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<1, 1, 4, 2>,        S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          1,          2,      1,       S<1, 4>,       S<1, 2>,      S<4, 1, 1, 2>,        S<1, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<1, 1, 4, 2>,        S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          2,          1,      1,       S<2, 1>,       S<4, 1>,      S<4, 1, 1, 2>,        S<1, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<1, 1, 4, 2>,        S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          2,          1,      1,       S<1, 2>,       S<1, 4>,      S<4, 1, 1, 2>,        S<1, 1, 8, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<4, 1, 1, 2>,      S<0, 3, 1, 2>,        S<1, 1, 1, 2>,      S<1, 1, 4, 2>,        S<4, 1, 2, 1>,  S<0, 3, 1, 2>,  S<0, 3, 1, 2>,       S<1, 1, 4, 1>,      S<0, 3, 1, 2>,        S<1, 1, 4, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>
         // clang-format on
         >;
 

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f32_f32_f32_mk_nk_mn_instance.cpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <cstdlib>
 
@@ -14,15 +14,12 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-using F16 = ck::half_t;
 using F32 = float;
-
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
 
 template <ck::index_t... Is>
-using S = ck::Sequence<Is...>;
-
+using S           = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
@@ -35,7 +32,39 @@ using device_gemm_dl_f32_f32_f32_mk_nk_mn_instances =
         //  ########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   | ThreadM111| ThreadN111| Thread| ClusterM110Xs| ClusterN110Xs| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor| ThreadSliceLengths| ThreadClusterLengths|  ThreadCluster|      SrcAccess|     SrcVectorTensor|    SrcVectorTensor|      DstVectorTensor|        SrcDstAccess|  SrcDstVectorDim| DstScalarPerVector|
         //  ########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |           |           |       |              |              |        K0_M0_M1_K1|          K0_M0_M1_K1|   ArrangeOrder|          Order| Lengths_K0_M0_M1_K1| ContiguousDimOrder|  Lengths_K0_M0_M1_K1|        K0_N0_N1_K1|          K0_N0_N1_K1|   ArrangeOrder|          Order| Lengths_K0_N0_N1_K1| ContiguousDimOrder|  Lengths_K0_N0_N1_K1|               Order|                 |                   |
         //  ########|      |      |      |        |        |        |        |            |            |            |               |      |      |      |      |   |           |           |       |              |              |                   |                     |               |               |                    |                   |                     |                   |                     |               |               |                    |                   |                     |                    |                 |                   |
-        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  1,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<8, 1, 1, 1>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 1>,      S<1, 2, 0, 3>,        S<1, 1, 1, 1>,      S<8, 1, 1, 1>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 1>,      S<1, 2, 0, 3>,        S<1, 1, 1, 1>, S<0, 1, 2, 3, 4, 5>,                5,                  4>
+        // MPerBlock=128, NPerBlock=128
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<8, 2>,       S<8, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<4, 4>,       S<4, 4>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   256,   128,   128,    16,  2,          4,          4,      1,       S<2, 8>,       S<2, 8>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 1, 2>,      S<2, 1, 128, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // // MPerBlock=128, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,   128,    64,    16,  2,          4,          4,      1,       S<8, 2>,       S<4, 2>,      S<8, 1, 2, 2>,       S<2, 1, 64, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,   128,    64,    16,  2,          4,          4,      1,       S<2, 8>,       S<2, 4>,      S<8, 1, 2, 2>,       S<2, 1, 64, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // // MPerBlock=64, NPerBlock=128
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,    64,   128,    16,  2,          4,          4,      1,       S<4, 2>,       S<8, 2>,      S<8, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 2, 2>,       S<2, 1, 64, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,   128,    64,   128,    16,  2,          4,          4,      1,       S<2, 4>,       S<2, 8>,      S<8, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 2, 2>,       S<2, 1, 64, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<4, 2>,       S<4, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<2, 4>,       S<2, 4>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<8, 1>,       S<4, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    64,     8,  2,          4,          4,      1,       S<4, 2>,       S<8, 1>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 2, 2>,       S<2, 1, 32, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=16, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    16,    64,    16,  2,          1,          4,      1,       S<4, 2>,       S<4, 2>,      S<4, 1, 1, 2>,       S<4, 1, 16, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 4, 2>,      S<4, 1, 16, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  4>,
+        // MPerBlock=64, NPerBlock=16
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,    16,    16,  2,          4,          1,      1,       S<4, 2>,       S<4, 2>,      S<4, 1, 4, 2>,       S<4, 1, 16, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 1, 2>,      S<4, 1, 16, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        // MPerBlock=16, NPerBlock=16
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    16,    16,    16,    16,  2,          2,          2,      1,       S<2, 2>,       S<2, 2>,      S<4, 1, 4, 2>,       S<4, 1, 4, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,        S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 4, 2>,       S<4, 1, 4, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    16,    16,    16,    16,  2,          2,          2,      1,       S<1, 4>,       S<1, 4>,      S<4, 1, 4, 2>,       S<4, 1, 4, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,        S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 4, 2>,       S<4, 1, 4, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        // MPerBlock=8, NPerBlock=64
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,     8,    64,    32,  2,          1,          2,      1,       S<4, 1>,       S<8, 2>,      S<4, 1, 1, 2>,        S<8, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 4, 2>,      S<4, 1, 16, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,     8,    64,    32,  2,          1,          2,      1,       S<2, 2>,       S<8, 2>,      S<4, 1, 1, 2>,        S<8, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<8, 1, 4, 2>,      S<4, 1, 16, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        // MPerBlock=64, NPerBlock=8
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,     8,    32,  2,          2,          1,      1,       S<8, 2>,       S<4, 1>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 1, 2>,       S<8, 1, 8, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,    64,    64,     8,    32,  2,          2,          1,      1,       S<8, 2>,       S<2, 2>,      S<8, 1, 4, 2>,       S<4, 1, 16, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 1, 2>,       S<8, 1, 8, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        // MPerBlock=8, NPerBlock=8
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          1,          2,      1,       S<4, 1>,       S<2, 1>,      S<4, 1, 1, 2>,       S<1, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,        S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 1, 2>,       S<1, 1, 8, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          1,          2,      1,       S<1, 4>,       S<1, 2>,      S<4, 1, 1, 2>,       S<1, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,        S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 1, 2>,       S<1, 1, 8, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  2>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          2,          1,      1,       S<2, 1>,       S<4, 1>,      S<4, 1, 1, 2>,       S<1, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,        S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 1, 2>,       S<1, 1, 8, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>,
+        DeviceGemmDl<   F32,   F32,   F32,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,     8,     8,     8,     4,  2,          2,          1,      1,       S<1, 2>,       S<1, 4>,      S<4, 1, 1, 2>,       S<1, 1, 8, 1>,  S<1, 2, 0, 3>,  S<1, 2, 0, 3>,        S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>,      S<4, 1, 1, 2>,       S<1, 1, 8, 1>,   S<1, 2, 0, 3>,  S<1, 2, 0, 3>,       S<4, 1, 1, 2>,      S<1, 2, 0, 3>,        S<1, 1, 1, 2>, S<0, 1, 2, 3, 4, 5>,                5,                  1>
         // clang-format on
         >;
 

library/src/tensor_operation_instance/gpu/gemm/device_gemm_wmma_bf16_bf16_bf16_km_kn_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_wmma.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using BF16 = ck::bhalf_t;
+using F32  = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+// Compilation parameters for a[k, m] * b[k, n] = c[m, n]
+using device_gemm_wmma_bf16_bf16_bf16_km_kn_mn_instances = std::tuple<
+    // clang-format off
+        //######################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|          GEMM| NumPrefetch| Block|  MPer|  NPer|  KPer| K1| MPer| NPer|      M|       N|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CShuffleBlockTransfer| CShuffleBlockTransfer|
+        //######################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise|Specialization|            |  Size| Block| Block| Block|   | WMMA| WMMA| Repeat|  Repeat|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN|  MRepeat|  MRepeat|        ClusterLengths|       ScalarPerVector|
+        //######################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|              |            |      |      |      |      |   |     |     |       |        | Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          | PerStore| PerStore|      MBlock_MPerBlock|                      |
+        //######################|        |        |        |      |      |      |        |         |            |            |            |              |            |      |      |      |      |   |     |     |       |        |                |               |               |               |               |               |          |                |               |               |              |               |               |          |         |         |      NBlock_NPerBlock|                      |
+        /* Prefetch 2, consume enormous vgpr resource*/
+        // 8 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,   256,   128,   128,    32,  8,   16,   16,      4,       2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        // 4 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,   128,   128,    64,    64,  8,   16,   16,      4,       2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        // 2 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,    64,    64,    32,    32,  8,   16,   16,      4,       1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        // 1 Wave
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,    32,    16,    16,    32,  8,   16,   16,      1,       1,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>,
+        /* Prefetch 1, prefer larger KPerBlock value for better latency hiding*/
+        // 8 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   256,    64,  8,   16,   16,      4,       4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   128,    64,  8,   16,   16,      4,       2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   160,    64,  8,   16,   16,      2,       5,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 64, 1,  4>,                      8>,
+        // 4 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,   128,   128,    32,  8,   16,   16,      4,       4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,   256,    64,    64,  8,   16,   16,      8,       2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,    64,   256,    64,  8,   16,   16,      2,       8,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,    64,    80,    64,  8,   16,   16,      1,       5,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 64, 1,  2>,                      8>,
+        // 2 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    16,    64,    64,  8,   16,   16,      1,       2,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    64,    32,    64,  8,   16,   16,      4,       1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    32,    64,    64,  8,   16,   16,      2,       2,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        // 1 Wave
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    32,    16,    32,    64,  8,   16,   16,      1,       2,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    32,    16,    16,    64,  8,   16,   16,      1,       1,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>
+    // clang-format on
+    >;
+
+void add_device_gemm_wmma_bf16_bf16_bf16_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(instances, device_gemm_wmma_bf16_bf16_bf16_km_kn_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_wmma_bf16_bf16_bf16_km_nk_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_wmma.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using BF16 = ck::bhalf_t;
+using F32  = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+// Compilation parameters for a[k, m] * b[n, k] = c[m, n]
+using device_gemm_wmma_bf16_bf16_bf16_km_nk_mn_instances = std::tuple<
+    // clang-format off
+        //######################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|          GEMM| NumPrefetch| Block|  MPer|  NPer|  KPer| K1| MPer| NPer|      M|       N|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CShuffleBlockTransfer| CShuffleBlockTransfer|
+        //######################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise|Specialization|            |  Size| Block| Block| Block|   | WMMA| WMMA| Repeat|  Repeat|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN|  MRepeat|  MRepeat|        ClusterLengths|       ScalarPerVector|
+        //######################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|              |            |      |      |      |      |   |     |     |       |        | Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          | PerStore| PerStore|      MBlock_MPerBlock|                      |
+        //######################|        |        |        |      |      |      |        |         |            |            |            |              |            |      |      |      |      |   |     |     |       |        |                |               |               |               |               |               |          |                |               |               |              |               |               |          |         |         |      NBlock_NPerBlock|                      |
+        /* Prefetch 2, consume enormous vgpr resource*/
+        // 8 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,   256,   128,   128,    32,  8,   16,   16,      4,       2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        // 4 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,   128,   128,    64,    64,  8,   16,   16,      4,       2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        // 2 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,    64,    64,    32,    32,  8,   16,   16,      4,       1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        // 1 Wave
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,    32,    16,    16,    32,  8,   16,   16,      1,       1,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<2, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>,
+        /* Prefetch 1, prefer larger KPerBlock value for better latency hiding*/
+        // 8 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   256,    64,  8,   16,   16,      4,       4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   128,    64,  8,   16,   16,      4,       2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   160,    64,  8,   16,   16,      2,       5,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 64, 1,  4>,                      8>,
+        // 4 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,   128,   128,    32,  8,   16,   16,      4,       4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,   256,    64,    64,  8,   16,   16,      8,       2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,    64,   256,    64,  8,   16,   16,      2,       8,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,    64,    80,    64,  8,   16,   16,      1,       5,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 64, 1,  2>,                      8>,
+        // 2 Waves
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    16,    64,    64,  8,   16,   16,      1,       2,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    64,    32,    64,  8,   16,   16,      4,       1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    32,    64,    64,  8,   16,   16,      2,       2,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        // 1 Wave
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    32,    16,    32,    64,  8,   16,   16,      1,       2,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<2, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>,
+        DeviceGemmWmma_CShuffle<      Col,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    32,    16,    16,    64,  8,   16,   16,      1,       1,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<2, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>
+    // clang-format on
+    >;
+
+void add_device_gemm_wmma_bf16_bf16_bf16_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(instances, device_gemm_wmma_bf16_bf16_bf16_km_nk_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_wmma_bf16_bf16_bf16_mk_kn_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_gemm_wmma.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using BF16 = ck::bhalf_t;
+using F32  = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+using device_gemm_wmma_bf16_bf16_bf16_mk_kn_mn_instances = std::tuple<
+    // clang-format off
+        //######################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|          GEMM| NumPrefetch| Block|  MPer|  NPer|  KPer| K1| MPer| NPer|      M|       N|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CShuffleBlockTransfer| CShuffleBlockTransfer|
+        //######################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise|Specialization|            |  Size| Block| Block| Block|   | WMMA| WMMA| Repeat|  Repeat|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN|  MRepeat|  MRepeat|        ClusterLengths|       ScalarPerVector|
+        //######################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|              |            |      |      |      |      |   |     |     |       |        | Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          | PerStore| PerStore|      MBlock_MPerBlock|                      |
+        //######################|        |        |        |      |      |      |        |         |            |            |            |              |            |      |      |      |      |   |     |     |       |        |                |               |               |               |               |               |          |                |               |               |              |               |               |          |         |         |      NBlock_NPerBlock|                      |
+        /* Prefetch 2, consume enormous vgpr resource*/
+        // 8 Waves
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,   256,   128,   128,    32,  8,   16,   16,      4,       2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        // 4 Waves
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,   128,   128,    64,    64,  8,   16,   16,      4,       2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        // 2 Waves
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,    64,    64,    32,    32,  8,   16,   16,      4,       1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        // 1 Wave
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           2,    32,    16,    16,    32,  8,   16,   16,      1,       1,     S<2, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>,
+        /* Prefetch 1, prefer larger KPerBlock value for better latency hiding*/
+        // 8 Waves
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   256,    64,  8,   16,   16,      4,       4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   128,    64,  8,   16,   16,      4,       2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 32, 1,  8>,                      8>,
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   256,   128,   160,    64,  8,   16,   16,      2,       5,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 64, 1,  4>,                      8>,
+        // 4 Waves
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,   128,   128,    32,  8,   16,   16,      4,       4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,   256,    64,    64,  8,   16,   16,      8,       2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,    64,   256,    64,  8,   16,   16,      2,       8,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 32, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,   128,    64,    80,    64,  8,   16,   16,      1,       5,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 64, 1,  2>,                      8>,
+        // 2 Waves
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    16,    64,    64,  8,   16,   16,      1,       2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    64,    32,    64,  8,   16,   16,      4,       1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    64,    32,    64,    64,  8,   16,   16,      2,       2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 16, 1,  4>,                      8>,
+        // 1 Wave
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    32,    16,    32,    64,  8,   16,   16,      1,       2,     S<2, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>,
+        DeviceGemmWmma_CShuffle<      Row,     Row,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,           1,    32,    16,    16,    64,  8,   16,   16,      1,       1,     S<2, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<2, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,        1,        1,       S<1, 16, 1,  2>,                      8>
+    // clang-format on
+    >;
+
+void add_device_gemm_wmma_bf16_bf16_bf16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(instances, device_gemm_wmma_bf16_bf16_bf16_mk_kn_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck