Merge branch 'amd-develop' into amd-master

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

@@ -98,30 +98,31 @@ struct DeviceOperationInstanceFactory<
         if constexpr(NumDimSpatial == 2 && is_same_v<InLayout, NHWC> &&
                      is_same_v<WeiLayout, KYXC> && is_same_v<OutLayout, NHWK>)
         {
+#ifdef CK_ENABLE_FP32
             if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
                          is_same_v<OutDataType, float>)
             {
                 add_device_conv2d_fwd_xdl_nhwc_kyxc_nhwk_f32_instances(op_ptrs);
             }
+#endif
 #ifdef CK_ENABLE_FP16
-            else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                              is_same_v<OutDataType, half_t>)
+            if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                         is_same_v<OutDataType, half_t>)
             {
                 add_device_conv2d_fwd_xdl_nhwc_kyxc_nhwk_f16_instances(op_ptrs);
                 add_device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_BF16
-            else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
-                              is_same_v<WeiDataType, ck::bhalf_t> &&
-                              is_same_v<OutDataType, ck::bhalf_t>)
+            if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
+                         is_same_v<WeiDataType, ck::bhalf_t> && is_same_v<OutDataType, ck::bhalf_t>)
             {
                 add_device_conv2d_fwd_xdl_nhwc_kyxc_nhwk_bf16_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_INT8
-            else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                              is_same_v<OutDataType, int8_t>)
+            if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                         is_same_v<OutDataType, int8_t>)
             {
                 add_device_conv2d_fwd_xdl_nhwc_kyxc_nhwk_int8_instances(op_ptrs);
             }

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

@@ -98,6 +98,26 @@ void add_device_gemm_xdl_splitk_f16_f8_f16_mk_nk_mn_instances(
     std::vector<std::unique_ptr<
         DeviceGemmSplitK<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
         instances);
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances);
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances);
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances);
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances);
 #endif
 
 template <typename ADataType,
@@ -105,7 +125,8 @@ template <typename ADataType,
           typename CDataType,
           typename ALayout,
           typename BLayout,
-          typename CLayout>
+          typename CLayout,
+          typename ComputeType>
 struct DeviceOperationInstanceFactory<
     ck::tensor_operation::device::DeviceGemmSplitK<ALayout,
                                                    BLayout,
@@ -115,7 +136,8 @@ struct DeviceOperationInstanceFactory<
                                                    CDataType,
                                                    ck::tensor_operation::element_wise::PassThrough,
                                                    ck::tensor_operation::element_wise::PassThrough,
-                                                   ck::tensor_operation::element_wise::PassThrough>>
+                                                   ck::tensor_operation::element_wise::PassThrough,
+                                                   ComputeType>>
 {
     using DeviceOp = DeviceGemmSplitK<ALayout,
                                       BLayout,
@@ -125,14 +147,15 @@ struct DeviceOperationInstanceFactory<
                                       CDataType,
                                       ck::tensor_operation::element_wise::PassThrough,
                                       ck::tensor_operation::element_wise::PassThrough,
-                                      ck::tensor_operation::element_wise::PassThrough>;
+                                      ck::tensor_operation::element_wise::PassThrough,
+                                      ComputeType>;
 
     static auto GetInstances()
     {
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
 #ifdef CK_ENABLE_FP32
         if constexpr(is_same_v<ADataType, float> && is_same_v<BDataType, float> &&
-                     is_same_v<CDataType, float>)
+                     is_same_v<CDataType, float> && is_same_v<ComputeType, float>)
         {
             if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
                          is_same_v<CLayout, Row>)
@@ -157,8 +180,8 @@ struct DeviceOperationInstanceFactory<
         }
 #endif
 #ifdef CK_ENABLE_FP16
-        else if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
-                          is_same_v<CDataType, half_t>)
+        if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
+                     is_same_v<CDataType, half_t> && is_same_v<ComputeType, half_t>)
         {
             if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
                          is_same_v<CLayout, Row>)
@@ -183,8 +206,8 @@ struct DeviceOperationInstanceFactory<
         }
 #endif
 #if(defined(CK_ENABLE_FP16) || defined(CK_ENABLE_FP8))
-        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<ADataType, f8_t> && is_same_v<BDataType, half_t> &&
+                     is_same_v<CDataType, half_t> && is_same_v<ComputeType, half_t>)
         {
             if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
                          is_same_v<CLayout, Row>)
@@ -207,8 +230,8 @@ struct DeviceOperationInstanceFactory<
                 add_device_gemm_xdl_splitk_f8_f16_f16_km_nk_mn_instances(op_ptrs);
             }
         }
-        else 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<ADataType, half_t> && is_same_v<BDataType, f8_t> &&
+                     is_same_v<CDataType, half_t> && is_same_v<ComputeType, half_t>)
         {
             if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
                          is_same_v<CLayout, Row>)
@@ -231,6 +254,31 @@ struct DeviceOperationInstanceFactory<
                 add_device_gemm_xdl_splitk_f16_f8_f16_km_nk_mn_instances(op_ptrs);
             }
         }
+        else if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
+                          is_same_v<CDataType, half_t> && is_same_v<ComputeType, f8_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_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_xdl_splitk_f16_f16_f16_comp_f8_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_xdl_splitk_f16_f16_f16_comp_f8_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_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances(op_ptrs);
+            }
+        }
+
 #endif
         return op_ptrs;
     }

library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_dl_instance.hpp

@@ -6,8 +6,6 @@
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_dl.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
-#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
 namespace ck {
 namespace tensor_operation {
 namespace device {

library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_wmma_instance.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_wmma_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/convolution_backward_weight_specialization.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F16 = ck::half_t;
+using F32 = float;
+using I8  = int8_t;
+using I32 = int32_t;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using namespace ck::tensor_layout::convolution;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto ConvBwdWeightDefault =
+    ck::tensor_operation::device::ConvolutionBackwardWeightSpecialization::Default;
+
+static constexpr auto ConvBwdWeightFilter1x1Stride1Pad0 =
+    ck::tensor_operation::device::ConvolutionBackwardWeightSpecialization::Filter1x1Stride1Pad0;
+
+template <index_t NDSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename CLayout,
+          ConvolutionBackwardWeightSpecialization ConvSpec>
+using device_grouped_conv_bwd_weight_wmma_f16_instances =
+    std::tuple<
+        // clang-format off
+        //#####################################|    NumDim|       A|       B|       C| AData| BData|  CData| AccData|            A|           B|            C|    ConvForward| Block|  MPer|  NPer|  KPer| K1|  MPer| NPer| MRepeat| NRepeat|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|       CShuffle|       CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //#####################################|   Spatial|  Layout|  Layout|  Layout|  Type|  Type|   Type|    Type|  Elementwise| Elementwise|  Elementwise| Specialization|  Size| Block| Block| Block|   |  WMMA| WMMA|        |        |   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MRepeatPerWave| NRepeatPerWave|            _MBlock_MPerBlock| ScalarPerVector|
+        //#####################################|          |        |        |        |      |      |       |        |    Operation|   Operation|    Operation|               |      |      |      |      |   |      |     |        |        | Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |     PerShuffle|     PerShuffle|            _NBlock_NPerBlock|      _NPerBlock|
+        //#####################################|          |        |        |        |      |      |       |        |             |            |             |               |      |      |      |      |   |      |     |        |        |                |               |               |               |               |               |          |                |               |               |              |               |               |          |               |               |                             |                |
+        // generic instance
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          128,    64,    64,     4,  8,    16,   16,       2,       2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 4>,               1>,
+        // blocksize=256
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          256,   128,   256,     8,  8,    16,   16,       2,       8,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               4>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          256,   256,   128,     8,  8,    16,   16,       8,       2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,               2>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          256,   256,    64,     8,  8,    16,   16,       4,       2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               4>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          256,    64,   256,     8,  8,    16,   16,       2,       4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 16, 1, 16>,              4>,
+        // blocksize=128
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          128,    64,   128,     8,  8,    16,   16,       2,       4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          128,   128,    64,     8,  8,    16,   16,       4,       2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          128,   128,   128,     8,  8,    16,   16,       4,       4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          128,    32,   256,     8,  8,    16,   16,       1,       8,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,          128,   256,    32,     8,  8,    16,   16,       8,       1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        // blocksize=64
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           64,    64,    32,     8,  8,    16,   16,       4,       1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           64,    32,    64,     8,  8,    16,   16,       1,       4,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 32, 1, 2>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           64,    64,    64,     8,  8,    16,   16,       2,       4,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 32, 1, 2>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           64,   128,    32,     8,  8,    16,   16,       8,       1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           64,    32,   128,     8,  8,    16,   16,       1,       8,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 32, 1, 2>,               8>,
+        // blocksize=32
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    16,    32,     8,  8,    16,   16,       1,       2,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    16,    64,     8,  8,    16,   16,       1,       4,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,  
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    32,    64,     8,  8,    16,   16,       2,       4,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,  
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    32,    32,     8,  8,    16,   16,       2,       2,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,  
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    64,    32,     8,  8,    16,   16,       4,       2,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,  
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    64,    16,     8,  8,    16,   16,       4,       1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,  
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  F16,   F16,  F16,  F32,  PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    32,    16,     8,  8,    16,   16,       2,       1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>
+        // clang-format on
+        >;
+
+template <index_t NDSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename CLayout,
+          ConvolutionBackwardWeightSpecialization ConvSpec>
+using device_grouped_conv_bwd_weight_wmma_i8_instances =
+    std::tuple<
+        // clang-format off
+        //#####################################|    NumDim|       A|       B|       C| AData| BData|  CData| AccData|            A|           B|            C|    ConvForward| Block|  MPer|  NPer|  KPer| K1|  MPer| NPer| MRepeat| NRepeat|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|       CShuffle|       CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //#####################################|   Spatial|  Layout|  Layout|  Layout|  Type|  Type|   Type|    Type|  Elementwise| Elementwise|  Elementwise| Specialization|  Size| Block| Block| Block|   |  WMMA| WMMA|        |        |   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MRepeatPerWave| NRepeatPerWave|            _MBlock_MPerBlock| ScalarPerVector|
+        //#####################################|          |        |        |        |      |      |       |        |    Operation|   Operation|    Operation|               |      |      |      |      |   |      |     |        |        | Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |     PerShuffle|     PerShuffle|            _NBlock_NPerBlock|      _NPerBlock|
+        //#####################################|          |        |        |        |      |      |       |        |             |            |             |               |      |      |      |      |   |      |     |        |        |                |               |               |               |               |               |          |                |               |               |              |               |               |          |               |               |                             |                |
+        // generic instance
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout,  I8,    I8,  I8,  I32,   PassThrough, PassThrough, PassThrough,       ConvSpec,          128,    64,    64,     4,  8,     16,   16,       2,       2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               1,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             1,             8,         1,           1,           1,               S<1, 32, 1, 4>,               1>,
+        // blocksize=256
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          256,    64,   256,     8,   8,    16,   16,       2,       4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               2,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          256,   256,    64,     8,   8,    16,   16,       4,       2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            2,              8,         1,           1,           1,               S<1, 64, 1, 4>,               8>,
+        // blocksize=128
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          128,   128,   256,     8,   8,    16,   16,       4,       8,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,           16,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          128,    64,   256,     8,   8,    16,   16,       2,       8,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               4,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,           16,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          128,    32,   256,     8,   8,    16,   16,       1,       8,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               2,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,           16,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          128,    64,   128,     8,   8,    16,   16,       2,       4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               4,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          128,   128,    64,     8,   8,    16,   16,       4,       2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            4,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          128,   256,    32,     8,   8,    16,   16,       8,       1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            2,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          128,   256,    64,     8,   8,    16,   16,       8,       2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            4,              8,         1,           1,           1,               S<1, 16, 1, 8>,               2>,      
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,          128,   256,   128,     8,   8,    16,   16,       8,       4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8>, 
+        // blocksize=64
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,           64,    32,   128,     8,   8,    16,   16,       1,       8,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               4,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,           16,              8,         1,           1,           1,               S<1, 32, 1, 2>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,           64,    64,   128,     8,   8,    16,   16,       2,       8,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,           16,              8,         1,           1,           1,               S<1, 32, 1, 2>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,           64,   128,    64,     8,   8,    16,   16,       8,       2,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8>,
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,           64,   128,    32,     8,   8,    16,   16,       8,       1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              8,         1,      S<8, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            4,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8>,
+        // blocksize=32
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    16,    64,     8,   8,    16,   16,       1,       4,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               4,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,           16,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,  
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    64,    64,     8,   8,    16,   16,       4,       4,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,           16,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,  
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    32,    32,     8,   8,    16,   16,       2,       2,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            8,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>,  
+        DeviceGroupedConvBwdWeight_Wmma_CShuffle<NDSpatial, ALayout, BLayout, CLayout, I8,    I8,  I8,  I32,    PassThrough, PassThrough, PassThrough,       ConvSpec,           32,    64,    16,     8,   8,    16,   16,       4,       1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              8,         1,      S<8, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,            4,              8,         1,           1,           1,               S<1, 16, 1, 2>,               8>
+        // clang-format on
+        >;
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

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

@@ -163,6 +163,30 @@ void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP32
 void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
@@ -177,6 +201,31 @@ void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances
                                                            PassThrough,
                                                            PassThrough>>>& instances);
 #endif
+#ifdef CK_ENABLE_INT8
+void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           int8_t,
+                                                           int8_t,
+                                                           int8_t,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           int8_t,
+                                                           int8_t,
+                                                           int8_t,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
 #ifdef CK_ENABLE_BF16
 void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
@@ -202,6 +251,30 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP32
 void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
@@ -231,6 +304,31 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_
                                                            BF8,
                                                            F8>>>& instances);
 #endif
+#ifdef CK_ENABLE_INT8
+void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           int8_t,
+                                                           int8_t,
+                                                           int8_t,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           int8_t,
+                                                           int8_t,
+                                                           int8_t,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
 
 #ifdef DL_KERNELS
 // dl
@@ -529,8 +627,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_FP16
-                else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                                  is_same_v<OutDataType, half_t>)
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_f16_instances(op_ptrs);
@@ -539,9 +637,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
-                                  is_same_v<WeiDataType, float> &&
-                                  is_same_v<OutDataType, ck::bhalf_t>)
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
@@ -552,8 +649,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
             }
-            else if constexpr(is_same_v<InLayout, NWGC> && is_same_v<WeiLayout, GKXC> &&
-                              is_same_v<OutLayout, NWGK>)
+            if constexpr(is_same_v<InLayout, NWGC> && is_same_v<WeiLayout, GKXC> &&
+                         is_same_v<OutLayout, NWGK>)
             {
 #ifdef DL_KERNELS
 #ifdef CK_ENABLE_FP32
@@ -564,16 +661,15 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_FP16
-                else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                                  is_same_v<OutDataType, half_t>)
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t>)
                 {
                     add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_f16_instances(op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
-                                  is_same_v<WeiDataType, float> &&
-                                  is_same_v<OutDataType, ck::bhalf_t>)
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t>)
                 {
                     add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_bf16_f32_bf16_instances(
                         op_ptrs);
@@ -582,7 +678,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
 #endif
             }
         }
-        else if constexpr(NumDimSpatial == 2)
+        if constexpr(NumDimSpatial == 2)
         {
             if constexpr(is_same_v<InLayout, GNHWC> && is_same_v<WeiLayout, GKYXC> &&
                          is_same_v<OutLayout, GNHWK>)
@@ -600,8 +696,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_FP16
-                else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                                  is_same_v<OutDataType, half_t>)
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_f16_instances(
@@ -612,9 +708,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
-                                  is_same_v<WeiDataType, float> &&
-                                  is_same_v<OutDataType, ck::bhalf_t>)
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
@@ -625,8 +720,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
             }
-            else if constexpr(is_same_v<InLayout, NHWGC> && is_same_v<WeiLayout, GKYXC> &&
-                              is_same_v<OutLayout, NHWGK>)
+            if constexpr(is_same_v<InLayout, NHWGC> && is_same_v<WeiLayout, GKYXC> &&
+                         is_same_v<OutLayout, NHWGK>)
             {
 #ifdef CK_ENABLE_FP32
                 if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
@@ -641,8 +736,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_FP16
-                else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                                  is_same_v<OutDataType, half_t>)
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_f16_instances(
@@ -653,9 +748,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
-                                  is_same_v<WeiDataType, float> &&
-                                  is_same_v<OutDataType, ck::bhalf_t>)
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
@@ -667,7 +761,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
 #endif
             }
         }
-        else if constexpr(NumDimSpatial == 3)
+        if constexpr(NumDimSpatial == 3)
         {
             if constexpr(is_same_v<InLayout, GNDHWC> && is_same_v<WeiLayout, GKZYXC> &&
                          is_same_v<OutLayout, GNDHWK>)
@@ -685,8 +779,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_FP16
-                else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                                  is_same_v<OutDataType, half_t>)
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_f16_instances(
@@ -694,12 +788,15 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
 #endif
                     add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
                         op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
+                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
-                                  is_same_v<WeiDataType, float> &&
-                                  is_same_v<OutDataType, ck::bhalf_t>)
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
@@ -708,10 +805,20 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                     add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
                         op_ptrs);
                 }
+#endif
+#ifdef CK_ENABLE_INT8
+                else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                                  is_same_v<OutDataType, int8_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
+                        op_ptrs);
+                }
 #endif
             }
-            else if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
-                              is_same_v<OutLayout, NDHWGK>)
+            if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
+                         is_same_v<OutLayout, NDHWGK>)
             {
 #ifdef CK_ENABLE_FP32
                 if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
@@ -726,10 +833,9 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 }
 #endif
 #ifdef CK_ENABLE_FP16
-                else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                                  is_same_v<OutDataType, half_t> &&
-                                  is_same_v<ComputeTypeA, half_t> &&
-                                  is_same_v<ComputeTypeB, half_t>)
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
+                             is_same_v<ComputeTypeB, half_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
@@ -737,12 +843,15 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
 #endif
                     add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
                         op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
+                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
-                                  is_same_v<WeiDataType, float> &&
-                                  is_same_v<OutDataType, ck::bhalf_t>)
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t>)
                 {
 #ifdef DL_KERNELS
                     add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
@@ -752,10 +861,20 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                         op_ptrs);
                 }
 #endif
+#ifdef CK_ENABLE_INT8
+                else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                                  is_same_v<OutDataType, int8_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
+                        op_ptrs);
+                }
+#endif
 #if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-                else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                                  is_same_v<OutDataType, half_t> &&
-                                  is_same_v<ComputeTypeA, bf8_t> && is_same_v<ComputeTypeB, f8_t>)
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, bf8_t> &&
+                             is_same_v<ComputeTypeB, f8_t>)
                 {
                     add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_instances(
                         op_ptrs);

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

@@ -19,13 +19,13 @@ namespace instance {
 #ifdef CK_ENABLE_FP16
 // FP16
 void add_device_normalization_rank_2_1_f16_instances(
-    std::vector<std::unique_ptr<DeviceNormalization<F16, F16, F16, F32, F16, PassThrough, 2, 1>>>&);
+    std::vector<std::unique_ptr<DeviceNormalization<F16, F16, F16, F16, F32, PassThrough, 2, 1>>>&);
 
 void add_device_normalization_rank_4_3_f16_instances(
-    std::vector<std::unique_ptr<DeviceNormalization<F16, F16, F16, F32, F16, PassThrough, 4, 3>>>&);
+    std::vector<std::unique_ptr<DeviceNormalization<F16, F16, F16, F16, F32, PassThrough, 4, 3>>>&);
 
 void add_device_normalization_rank_5_3_f16_instances(
-    std::vector<std::unique_ptr<DeviceNormalization<F16, F16, F16, F32, F16, PassThrough, 5, 3>>>&);
+    std::vector<std::unique_ptr<DeviceNormalization<F16, F16, F16, F16, F32, PassThrough, 5, 3>>>&);
 #endif
 #ifdef CK_ENABLE_FP32
 // FP32
@@ -42,14 +42,15 @@ template <typename XDataType,
           typename GammaDataType,
           typename BetaDataType,
           typename YDataType,
+          typename SaveMeanInvStdDataType,
           index_t Rank,
           index_t NumReduceDim>
 struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceNormalization<
     XDataType,
     GammaDataType,
     BetaDataType,
-    F32,
     YDataType,
+    SaveMeanInvStdDataType,
     ck::tensor_operation::element_wise::PassThrough,
     Rank,
     NumReduceDim>>
@@ -57,8 +58,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceNormal
     using DeviceOp = DeviceNormalization<XDataType,
                                          GammaDataType,
                                          BetaDataType,
-                                         F32,
                                          YDataType,
+                                         SaveMeanInvStdDataType,
                                          ck::tensor_operation::element_wise::PassThrough,
                                          Rank,
                                          NumReduceDim>;
@@ -68,7 +69,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceNormal
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
 #ifdef CK_ENABLE_FP16
         if constexpr(is_same_v<XDataType, F16> && is_same_v<GammaDataType, F16> &&
-                     is_same_v<BetaDataType, F16> && is_same_v<YDataType, F16>)
+                     is_same_v<BetaDataType, F16> && is_same_v<YDataType, F16> &&
+                     is_same_v<SaveMeanInvStdDataType, F32>)
         {
             if constexpr(Rank == 2 && NumReduceDim == 1)
             {
@@ -86,7 +88,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceNormal
 #endif
 #ifdef CK_ENABLE_FP32
         if constexpr(is_same_v<XDataType, F32> && is_same_v<GammaDataType, F32> &&
-                     is_same_v<BetaDataType, F32> && is_same_v<YDataType, F32>)
+                     is_same_v<BetaDataType, F32> && is_same_v<YDataType, F32> &&
+                     is_same_v<SaveMeanInvStdDataType, F32>)
         {
             if constexpr(Rank == 2 && NumReduceDim == 1)
             {

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

@@ -19,7 +19,7 @@ namespace instance {
 
 // FP16
 void add_device_normalization_rank_5_3_swish_f16_instances(
-    std::vector<std::unique_ptr<DeviceNormalization<F16, F16, F16, F32, F16, Swish, 5, 3>>>&);
+    std::vector<std::unique_ptr<DeviceNormalization<F16, F16, F16, F16, F32, Swish, 5, 3>>>&);
 
 // FP32
 void add_device_normalization_rank_5_3_swish_f32_instances(
@@ -27,20 +27,21 @@ void add_device_normalization_rank_5_3_swish_f32_instances(
 
 // [x, gamma, beta, y] = [f16, f32, f32, f16]
 void add_device_normalization_rank_5_3_swish_f16_f32_f32_f16_instances(
-    std::vector<std::unique_ptr<DeviceNormalization<F16, F32, F32, F32, F16, Swish, 5, 3>>>&);
+    std::vector<std::unique_ptr<DeviceNormalization<F16, F32, F32, F16, F32, Swish, 5, 3>>>&);
 
 template <typename XDataType,
           typename GammaDataType,
           typename BetaDataType,
           typename YDataType,
+          typename SaveMeanInvStdDataType,
           index_t Rank,
           index_t NumReduceDim>
 struct DeviceOperationInstanceFactory<
     ck::tensor_operation::device::DeviceNormalization<XDataType,
                                                       GammaDataType,
                                                       BetaDataType,
-                                                      F32,
                                                       YDataType,
+                                                      SaveMeanInvStdDataType,
                                                       ck::tensor_operation::element_wise::Swish,
                                                       Rank,
                                                       NumReduceDim>>
@@ -48,8 +49,8 @@ struct DeviceOperationInstanceFactory<
     using DeviceOp = DeviceNormalization<XDataType,
                                          GammaDataType,
                                          BetaDataType,
-                                         F32,
                                          YDataType,
+                                         SaveMeanInvStdDataType,
                                          ck::tensor_operation::element_wise::Swish,
                                          Rank,
                                          NumReduceDim>;
@@ -59,7 +60,8 @@ struct DeviceOperationInstanceFactory<
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
 
         if constexpr(is_same_v<XDataType, F16> && is_same_v<GammaDataType, F16> &&
-                     is_same_v<BetaDataType, F16> && is_same_v<YDataType, F16>)
+                     is_same_v<BetaDataType, F16> && is_same_v<YDataType, F16> &&
+                     is_same_v<SaveMeanInvStdDataType, F32>)
         {
             if constexpr(Rank == 5 && NumReduceDim == 3)
             {
@@ -67,7 +69,8 @@ struct DeviceOperationInstanceFactory<
             }
         }
         else if constexpr(is_same_v<XDataType, F32> && is_same_v<GammaDataType, F32> &&
-                          is_same_v<BetaDataType, F32> && is_same_v<YDataType, F32>)
+                          is_same_v<BetaDataType, F32> && is_same_v<YDataType, F32> &&
+                          is_same_v<SaveMeanInvStdDataType, F32>)
         {
             if constexpr(Rank == 5 && NumReduceDim == 3)
             {
@@ -75,7 +78,8 @@ struct DeviceOperationInstanceFactory<
             }
         }
         else if constexpr(is_same_v<XDataType, F16> && is_same_v<GammaDataType, F32> &&
-                          is_same_v<BetaDataType, F32> && is_same_v<YDataType, F16>)
+                          is_same_v<BetaDataType, F32> && is_same_v<YDataType, F16> &&
+                          is_same_v<SaveMeanInvStdDataType, F32>)
         {
             if constexpr(Rank == 5 && NumReduceDim == 3)
             {

library/include/ck/library/utility/check_err.hpp

@@ -230,7 +230,6 @@ check_err(const Range& out,
     return res;
 }
 
-#if defined CK_ENABLE_FP8
 template <typename Range, typename RefRange>
 std::enable_if_t<(std::is_same_v<ranges::range_value_t<Range>, ranges::range_value_t<RefRange>> &&
                   std::is_same_v<ranges::range_value_t<Range>, f8_t>),
@@ -275,9 +274,7 @@ check_err(const Range& out,
     }
     return res;
 }
-#endif
 
-#if defined CK_ENABLE_BF8
 template <typename Range, typename RefRange>
 std::enable_if_t<(std::is_same_v<ranges::range_value_t<Range>, ranges::range_value_t<RefRange>> &&
                   std::is_same_v<ranges::range_value_t<Range>, bf8_t>),
@@ -322,7 +319,6 @@ check_err(const Range& out,
     }
     return res;
 }
-#endif
 
 } // namespace utils
 } // namespace ck

library/include/ck/library/utility/host_common_util.hpp

@@ -22,7 +22,7 @@ static inline void dumpBufferToFile(const char* fileName, T* data, size_t dataNu
     std::ofstream outFile(fileName, std::ios::binary);
     if(outFile)
     {
-        outFile.write(reinterpret_cast<char*>(data), dataNumItems * sizeof(T));
+        outFile.write(reinterpret_cast<const char*>(data), dataNumItems * sizeof(T));
         outFile.close();
         std::cout << "Write output to file " << fileName << std::endl;
     }

library/include/ck/library/utility/host_tensor_generator.hpp

@@ -200,10 +200,11 @@ struct GeneratorTensor_3<ck::bf8_t>
 template <typename T>
 struct GeneratorTensor_4
 {
-    std::default_random_engine generator;
+    std::mt19937 generator;
     std::normal_distribution<float> distribution;
 
-    GeneratorTensor_4(float mean, float stddev) : generator(1), distribution(mean, stddev){};
+    GeneratorTensor_4(float mean, float stddev, unsigned int seed = 1)
+        : generator(seed), distribution(mean, stddev){};
 
     template <typename... Is>
     T operator()(Is...)

library/src/tensor_operation_instance/gpu/CMakeLists.txt

@@ -2,44 +2,44 @@ function(add_instance_library INSTANCE_NAME)
     message("adding instance ${INSTANCE_NAME}")
     set(result 1)
     if(DEFINED DTYPES)
-      foreach(source IN LISTS ARGN)
-        set(test 0)
-        foreach(type IN LISTS DTYPES)
+        foreach(source IN LISTS ARGN)
+            set(test 0)
+            foreach(type IN LISTS DTYPES)
                 if(type MATCHES "fp16")
-                        set(type1 "_f16")
+                    set(type1 "_f16")
                 elseif(type MATCHES "fp32")
-                        set(type1 "_f32")
+                    set(type1 "_f32")
                 elseif(type MATCHES "fp8")
-                        set(type1 "_f8")
+                    set(type1 "_f8")
                 elseif(type MATCHES "bf16")
-                        set(type1 "_b16")
+                    set(type1 "_b16")
                 elseif(type MATCHES "fp64")
-                        set(type1 "_f64")
+                    set(type1 "_f64")
                 elseif(type MATCHES "int8")
-                        set(type1 "_i8")
+                    set(type1 "_i8")
                 endif()
                 #make an exception for reduction kernels
-                if("${source}" MATCHES "${type}" OR "${source}" MATCHES "${type1}" OR "${source}" MATCHES "device_reduce_instance")
-                        #if filename matches any selected type, exit type loop and do no exclude the file from the list
-                        set(test 0)
-                        break()
+                if("${source}" MATCHES "${type}" OR "${source}" MATCHES "${type1}" OR "${source}" MATCHES "device_reduce_instance" OR ${source} MATCHES "device_image_to_column")
+                    #if filename matches any selected type, exit type loop and do no exclude the file from the list
+                    set(test 0)
+                    break()
                 elseif((source MATCHES "fp8" OR source MATCHES "fp32" OR source MATCHES "fp64" OR source MATCHES "bf16" OR source MATCHES "int8" OR source MATCHES "fp16" OR
-                  source MATCHES "_f8" OR source MATCHES "_f32" OR source MATCHES "_f64" OR source MATCHES "_i8" OR source MATCHES "_f16" OR source MATCHES "_b16") AND
-                  NOT(source MATCHES type OR source MATCHES type1))
-                        #if filename contains a type which doesn't match any selected type, mark it for removal
-                        set(test 1)
+                    source MATCHES "_f8" OR source MATCHES "_f32" OR source MATCHES "_f64" OR source MATCHES "_i8" OR source MATCHES "_f16" OR source MATCHES "_b16") AND
+                    NOT(source MATCHES type OR source MATCHES type1))
+                    #if filename contains a type which doesn't match any selected type, mark it for removal
+                    set(test 1)
                 endif()
-        endforeach()
-        if(test EQUAL 1)
+            endforeach()
+            if(test EQUAL 1)
                 message("removing instance ${source} ")
                 list(REMOVE_ITEM ARGN "${source}")
-        endif()
-      endforeach()
+            endif()
+        endforeach()
     endif()
     foreach(source IN LISTS ARGN)
         if(NOT DEFINED DL_KERNELS AND source MATCHES "_dl")
-                message("removing dl instance ${source} ")
-                list(REMOVE_ITEM ARGN "${source}")
+            message("removing dl instance ${source} ")
+            list(REMOVE_ITEM ARGN "${source}")
         endif()
     endforeach()
     #only continue if there are some source files left on the list
@@ -49,8 +49,10 @@ function(add_instance_library INSTANCE_NAME)
         set_target_properties(${INSTANCE_NAME} PROPERTIES POSITION_INDEPENDENT_CODE ON)
         clang_tidy_check(${INSTANCE_NAME})
         set(result 0)
+        message("add_instance_library ${INSTANCE_NAME}")
+    else()
+        message("skip_instance_libary ${INSTANCE_NAME}")
     endif()
-    #message("add_instance_library returns ${result}")
     set(result ${result} PARENT_SCOPE)
 endfunction(add_instance_library INSTANCE_NAME)
 
@@ -58,65 +60,74 @@ endfunction(add_instance_library INSTANCE_NAME)
 file(GLOB dir_list LIST_DIRECTORIES true *)
 set(CK_DEVICE_INSTANCES)
 FOREACH(subdir_path ${dir_list})
-set(target_dir)
-IF(IS_DIRECTORY "${subdir_path}")
-    set(cmake_instance)
-    file(READ "${subdir_path}/CMakeLists.txt" cmake_instance)
-    set(add_inst 0)
-    if(("${cmake_instance}" MATCHES "_fp8" OR "${cmake_instance}" MATCHES "_f8") AND DTYPES MATCHES "fp8")
+    set(target_dir)
+    IF(IS_DIRECTORY "${subdir_path}")
+        set(cmake_instance)
+        file(READ "${subdir_path}/CMakeLists.txt" cmake_instance)
+        set(add_inst 0)
+        if(("${cmake_instance}" MATCHES "_fp8" OR "${cmake_instance}" MATCHES "_f8") AND DTYPES MATCHES "fp8")
             message("fp8 instance found!")
             set(add_inst 1)
-    endif()
-    if(("${cmake_instance}" MATCHES "_fp16" OR "${cmake_instance}" MATCHES "_f16") AND DTYPES MATCHES "fp16")
+        endif()
+        if(("${cmake_instance}" MATCHES "_bf8" OR "${cmake_instance}" MATCHES "_b8") AND DTYPES MATCHES "bf8")
+            message("bf8 instance found!")
+            set(add_inst 1)
+        endif()
+        if(("${cmake_instance}" MATCHES "_fp16" OR "${cmake_instance}" MATCHES "_f16") AND DTYPES MATCHES "fp16")
             message("fp16 instance found!")
             set(add_inst 1)
-    endif()
-    if(("${cmake_instance}" MATCHES "_fp32" OR "${cmake_instance}" MATCHES "_f32") AND DTYPES MATCHES "fp32")
+        endif()
+        if(("${cmake_instance}" MATCHES "_fp32" OR "${cmake_instance}" MATCHES "_f32") AND DTYPES MATCHES "fp32")
             message("fp32 instance found!")
             set(add_inst 1)
-    endif()
-    if(("${cmake_instance}" MATCHES "_fp64" OR "${cmake_instance}" MATCHES "_f64") AND DTYPES MATCHES "fp64")
+        endif()
+        if(("${cmake_instance}" MATCHES "_fp64" OR "${cmake_instance}" MATCHES "_f64") AND DTYPES MATCHES "fp64")
             message("fp64 instance found!")
             set(add_inst 1)
-    endif()
-    if("${cmake_instance}" MATCHES "_bf16" AND DTYPES MATCHES "bf16")
+        endif()
+        if("${cmake_instance}" MATCHES "_bf16" AND DTYPES MATCHES "bf16")
             message("bf16 instance found!")
             set(add_inst 1)
-    endif()
-    if(("${cmake_instance}" MATCHES "_int8" OR "${cmake_instance}" MATCHES "_i8") AND DTYPES MATCHES "int8")
+        endif()
+        if(("${cmake_instance}" MATCHES "_int8" OR "${cmake_instance}" MATCHES "_i8") AND DTYPES MATCHES "int8")
             message("int8 instance found!")
             set(add_inst 1)
-    endif()
-    if(NOT "${cmake_instance}" MATCHES "_fp8" OR
-                   NOT "${cmake_instance}" MATCHES "_f8" OR
-                   NOT "${cmake_instance}" MATCHES "_fp16" OR
-                   NOT "${cmake_instance}" MATCHES "_f16" OR
-                   NOT "${cmake_instance}" MATCHES "_fp32" OR
-                   NOT "${cmake_instance}" MATCHES "_f32" OR
-                   NOT "${cmake_instance}" MATCHES "_fp64" OR
-                   NOT "${cmake_instance}" MATCHES "_f64" OR
-                   NOT "${cmake_instance}" MATCHES "_bf16" OR
-                   NOT "${cmake_instance}" MATCHES "_int8" OR
-                   NOT "${cmake_instance}" MATCHES "_i8" OR
-                   NOT "${cmake_instance}" MATCHES "_int4" OR
-                   NOT DEFINED DTYPES)
-             message("instance should be built for all types!")
-             set(add_inst 1)
-     endif()
-    if("${cmake_instance}" MATCHES "quantization" AND DEFINED DTYPES AND NOT DTYPES MATCHES "int8")
-           message("quantization instances will not be built!")
-           set(add_inst 0)
-    endif()
-    if("${cmake_instance}" MATCHES "ONLY DL_KERNELS" AND NOT DEFINED DL_KERNELS)
-	    message("Found only dl instances, but DL_KERNELS is not set. Skipping.")
+        endif()
+        if(NOT ("${cmake_instance}" MATCHES "_fp8" OR
+                "${cmake_instance}" MATCHES "_f8" OR
+                "${cmake_instance}" MATCHES "_fp16" OR
+                "${cmake_instance}" MATCHES "_f16" OR
+                "${cmake_instance}" MATCHES "_fp32" OR
+                "${cmake_instance}" MATCHES "_f32" OR
+                "${cmake_instance}" MATCHES "_fp64" OR
+                "${cmake_instance}" MATCHES "_f64" OR
+                "${cmake_instance}" MATCHES "_bf16" OR
+                "${cmake_instance}" MATCHES "_int8" OR
+                "${cmake_instance}" MATCHES "_i8" OR
+                "${cmake_instance}" MATCHES "_int4"))
+            message("instance should be built for all types!")
+            set(add_inst 1)
+        endif()
+        if(NOT DEFINED DTYPES)
+            set(add_inst 1)
+        endif()
+        if(("${cmake_instance}" MATCHES "quantization") AND (DEFINED DTYPES) AND (NOT DTYPES MATCHES "int8"))
+            message("quantization instances will not be built!")
             set(add_inst 0)
-    endif()
-    if(add_inst EQUAL 1)
-      get_filename_component(target_dir ${subdir_path} NAME)
-      add_subdirectory(${target_dir})
-      list(APPEND CK_DEVICE_INSTANCES $<TARGET_OBJECTS:device_${target_dir}_instance>)
-    endif()
-ENDIF()
+        endif()
+        if(("${cmake_instance}" MATCHES "ONLY DL_KERNELS") AND (NOT DEFINED DL_KERNELS))
+            message("Found only dl instances, but DL_KERNELS is not set. Skipping.")
+            set(add_inst 0)
+        endif()
+        if((add_inst EQUAL 1))
+            get_filename_component(target_dir ${subdir_path} NAME)
+            add_subdirectory(${target_dir})
+            list(APPEND CK_DEVICE_INSTANCES $<TARGET_OBJECTS:device_${target_dir}_instance>)
+            message("add_instance_directory ${subdir_path}")
+        else()
+            message("skip_instance_directory ${subdir_path}")
+        endif()
+    ENDIF()
 ENDFOREACH()
 
 add_library(device_operations STATIC ${CK_DEVICE_INSTANCES})
@@ -158,11 +169,11 @@ target_compile_options(device_operations PRIVATE
 
 # install(TARGETS device_operations LIBRARY DESTINATION lib)
 rocm_install(TARGETS device_operations
-        EXPORT device_operationsTargets)
+    EXPORT device_operationsTargets)
 
 rocm_install(DIRECTORY ${DEV_OPS_INC_DIRS} DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/ck)
 rocm_install(EXPORT device_operationsTargets
-        FILE composable_kerneldevice_operationsTargets.cmake
-        NAMESPACE composable_kernel::
-        DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/composable_kernel
+    FILE composable_kerneldevice_operationsTargets.cmake
+    NAMESPACE composable_kernel::
+    DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/composable_kernel
 )

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

@@ -96,13 +96,9 @@ list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_kn_mn_instan
 
 add_instance_library(device_gemm_instance ${GEMM_INSTANCES})
 
-set(ENABLE_PIPELINE_V2_OPT OFF)
+set(ENABLE_PIPELINE_V2_OPT)
 
 if (ENABLE_PIPELINE_V2_OPT)
-    set(MAX_ILP_OPTS
-       -mllvm
-       -amdgpu-enable-max-ilp-scheduling-strategy
-    )
     set(WAVES_PER_EU_DEFS
        CK_USE_WAVES_PER_EU=1
        CK_MIN_WAVES_PER_EU=1
@@ -118,7 +114,7 @@ if (ENABLE_PIPELINE_V2_OPT)
        COMPILE_DEFINITIONS "${WAVES_PER_EU_DEFS};${IGLP_OPT_DEFS}")
     # layout=NN
     set_source_files_properties(device_gemm_xdl_f16_f16_f16/km_nk_mn_default_pipeline_v2_opt_instance.cpp PROPERTIES
-       COMPILE_OPTIONS "${MAX_ILP_OPTS}"
+       COMPILE_OPTIONS ";;"
        COMPILE_DEFINITIONS "${WAVES_PER_EU_DEFS};${IGLP_OPT_DEFS}")
     # layout=TT
     set_source_files_properties(device_gemm_xdl_f16_f16_f16/mk_kn_mn_default_pipeline_v2_opt_instance.cpp PROPERTIES
@@ -126,7 +122,7 @@ if (ENABLE_PIPELINE_V2_OPT)
        COMPILE_DEFINITIONS "${WAVES_PER_EU_DEFS};${IGLP_OPT_DEFS}")
     # layout=TN
     set_source_files_properties(device_gemm_xdl_f16_f16_f16/mk_nk_mn_default_pipeline_v2_opt_instance.cpp PROPERTIES
-       COMPILE_OPTIONS "${MAX_ILP_OPTS}"
+       COMPILE_OPTIONS ";;"
        COMPILE_DEFINITIONS "${WAVES_PER_EU_DEFS};${IGLP_OPT_DEFS}")
 endif(ENABLE_PIPELINE_V2_OPT)
 

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp

@@ -29,7 +29,10 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
 
+static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
 // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
 using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances = std::tuple<
     // clang-format off
         //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           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|          LoopScheduler|                    Pipeline|
@@ -37,61 +40,61 @@ using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances = std::tuple<
         //#####################|        |        |        |      |      |      |        |         |   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|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
         ,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>
 #endif
 #if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES        
         // pipeline v2, 1 wave
         ,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>
 #endif
     // clang-format on
     >;
@@ -101,8 +104,11 @@ void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(
         DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
         instances)
 {
-    add_device_operation_instances(instances,
-                                   device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances{});
+    add_device_operation_instances(
+        instances, device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances<GemmDefault>{});
+
+    add_device_operation_instances(
+        instances, device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances<MNKPadding>{});
 }
 
 } // namespace instance

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp

@@ -29,6 +29,9 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
 
+static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
 // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
 using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances = std::tuple<
     // clang-format off
@@ -37,52 +40,52 @@ using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances = std::tuple<
         //#####################|        |        |        |      |      |      |        |         |   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|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
         ,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Interwave,        PipelineVersion::v1>
 #endif
 #if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES        
         // pipeline v2, 1 wave
         ,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v2>
 #endif
     // clang-format on
     >;
@@ -92,8 +95,11 @@ void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(
         DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
         instances)
 {
-    add_device_operation_instances(instances,
-                                   device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances{});
+    add_device_operation_instances(
+        instances, device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances<GemmDefault>{});
+
+    add_device_operation_instances(
+        instances, device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances<MNKPadding>{});
 }
 
 } // namespace instance

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_kn_mn_instance.cpp

@@ -27,7 +27,10 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
 
+static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
 // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
 using device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances =
     std::tuple<
         // clang-format off
@@ -35,22 +38,22 @@ using device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances =
         //#####################|        |        |        |   Type|   Type|   Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //#####################|        |        |        |       |       |       |        |         |   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|
         //#####################|        |        |        |       |       |       |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    64,  16,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    64,  16,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    64,  16,  16,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    64,  16,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,             16,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    64,  16,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    64,  16,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 2>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    64,  16,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,             16,         1,           1,           1,               S<1, 64, 1, 2>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    64,  16,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,             16,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    64,  16,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    64,  16,  16,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    64,  16,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    64,  16,  16,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    64,  16,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    64,  16,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    64,  16,  16,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    64,  16,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,             16,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    64,  16,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    64,  16,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 2>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    64,  16,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,             16,         1,           1,           1,               S<1, 64, 1, 2>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    64,  16,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,             16,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    64,  16,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    64,  16,  16,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    64,  16,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         0,           1,           1,               S<1, 64, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Row,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    64,  16,  16,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,             16,         1,           1,           1,               S<1, 64, 1, 4>,             16>
         // clang-format on
         >;
 
@@ -58,8 +61,11 @@ void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances(
     std::vector<std::unique_ptr<
         DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances)
 {
-    add_device_operation_instances(instances,
-                                   device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances{});
+    add_device_operation_instances(
+        instances, device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances<GemmDefault>{});
+
+    add_device_operation_instances(
+        instances, device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances<MNKPadding>{});
 }
 
 } // namespace instance

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_fp8_fp8_fp8_mk_nk_mn_instance.cpp

@@ -27,7 +27,10 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
 
+static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
 // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
+template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
 using device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances =
     std::tuple<
         // clang-format off
@@ -35,19 +38,19 @@ using device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances =
         //#####################|        |        |        |   Type|   Type|   Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //#####################|        |        |        |       |       |       |        |         |   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|
         //#####################|        |        |        |       |       |       |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    64,  16,  16,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    64,    64,  16,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 2>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    64,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    64,    64,  16,  16,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,    64,    64,  16,  16,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,   128,    64,  16,  16,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,   128,    32,    64,  16,  16,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 2>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    64,  16,  16,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    32,    64,  16,  16,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,             16>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    64,  16,  16,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,             16>
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    64,  16,  16,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    64,    64,  16,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 2>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    64,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    64,    64,  16,  16,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,    64,    64,  16,  16,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,   128,    64,  16,  16,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,   128,    32,    64,  16,  16,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 2>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   128,    32,   128,    64,  16,  16,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    32,    64,  16,  16,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,             16>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,     F8,     F8,     F8,     F32,       F8, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    32,    64,    64,  16,  16,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,             16>
         // clang-format on
         >;
 
@@ -55,8 +58,11 @@ void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances(
     std::vector<std::unique_ptr<
         DeviceGemm<Row, Col, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances)
 {
-    add_device_operation_instances(instances,
-                                   device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances{});
+    add_device_operation_instances(
+        instances, device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances<GemmDefault>{});
+
+    add_device_operation_instances(
+        instances, device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances<MNKPadding>{});
 }
 
 } // namespace instance

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

@@ -15,6 +15,10 @@ list(APPEND GEMM_SPLITK_INSTANCES device_gemm_xdl_splitk_f32_f32_f32_mk_kn_mn_in
                                   device_gemm_xdl_splitk_f16_fp8_f16_mk_kn_mn_instance.cpp
                                   device_gemm_xdl_splitk_f16_fp8_f16_mk_nk_mn_instance.cpp
                                   device_gemm_xdl_splitk_f16_fp8_f16_km_kn_mn_instance.cpp
-                                  device_gemm_xdl_splitk_f16_fp8_f16_km_nk_mn_instance.cpp)
+                                  device_gemm_xdl_splitk_f16_fp8_f16_km_nk_mn_instance.cpp
+                                  device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_mk_kn_mn_instance.cpp
+                                  device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_mk_nk_mn_instance.cpp
+                                  device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_km_kn_mn_instance.cpp
+                                  device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_km_nk_mn_instance.cpp)
 
 add_instance_library(device_gemm_splitk_instance ${GEMM_SPLITK_INSTANCES})

library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_km_kn_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, 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_xdl_splitk_c_shuffle.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+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 PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+
+// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances = std::tuple<
+    // clang-format off
+        //#########################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|    Type|
+        //#########################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,     F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,    GemmDefault,  256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              1,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,     F8>
+    // clang-format on
+    >;
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances)
+{
+    add_device_operation_instances(instances,
+                                   device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_kn_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_km_nk_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, 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_xdl_splitk_c_shuffle.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+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 PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+
+// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances = std::tuple<
+    // clang-format off
+        //#########################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|    Type|
+        //#########################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              4,              8,      true,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              2,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,              2,              1,              8,      true,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>
+    // clang-format on
+    >;
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances)
+{
+    add_device_operation_instances(instances,
+                                   device_gemm_xdl_splitk_f16_f16_f16_comp_f8_km_nk_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_comp_fp8_mk_kn_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, 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_xdl_splitk_c_shuffle.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+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 PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+// static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
+static constexpr auto GemmMNPadding = ck::tensor_operation::device::GemmSpecialization::MNPadding;
+
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_generic_instances = std::tuple<
+    // clang-format off
+        //#########################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|    Type|
+        //#########################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    32,    64,     4,  8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              1,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               2,    F8>
+    // clang-format on
+    >;
+
+// Compilation parameters for a[m, k] * b[k, n] = c[m, n]
+using device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances = std::tuple<
+    // clang-format off
+        //#########################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer| Compute|
+        //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|    Type|
+        //#########################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|        |
+        //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |        |
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,    64,   192,     4,  8,   32,   32,    1,    3,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 48, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   192,    64,     4,  8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 16, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    32,   192,     4,  8,   32,   32,    1,    3,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 24, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,   192,    32,     4,  8,   32,   32,    3,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    32,    64,     4,  8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    64,    32,     4,  8,   32,   32,    1,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,    32,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,    F8>,
+        DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,      true,           1,           1,                   S<1, 32, 1, 4>,               8,    F8>
+    // clang-format on
+    >;
+
+void add_device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmSplitK<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough, F8>>>&
+        instances)
+{
+    add_device_operation_instances(
+        instances, device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_generic_instances{});
+    add_device_operation_instances(instances,
+                                   device_gemm_xdl_splitk_f16_f16_f16_comp_f8_mk_kn_mn_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck