From 50cc0a13a69aa6bd59a35bc84759688c3552cc45 Mon Sep 17 00:00:00 2001
From: Rostyslav Geyyer <46627076+geyyer@users.noreply.github.com>
Date: Tue, 9 Apr 2024 13:56:54 -0500
Subject: [PATCH 1/7] Add an example (#1225)

---
 example/09_convnd_fwd/CMakeLists.txt          |  1 +
 .../convnd_fwd_xdl_fp16_comp_fp8.cpp          | 81 +++++++++++++++++++
 2 files changed, 82 insertions(+)
 create mode 100644 example/09_convnd_fwd/convnd_fwd_xdl_fp16_comp_fp8.cpp

diff --git a/example/09_convnd_fwd/CMakeLists.txt b/example/09_convnd_fwd/CMakeLists.txt
index afbe741212..778e81872d 100644
--- a/example/09_convnd_fwd/CMakeLists.txt
+++ b/example/09_convnd_fwd/CMakeLists.txt
@@ -5,6 +5,7 @@ add_example_executable(example_convnd_fwd_xdl_int8 convnd_fwd_xdl_int8.cpp)
 add_example_executable(example_convnd_fwd_xdl_fp8 convnd_fwd_xdl_fp8.cpp)
 add_example_executable(example_convnd_fwd_xdl_fp64 convnd_fwd_xdl_fp64.cpp)
 add_example_executable(example_convnd_fwd_xdl_bf8 convnd_fwd_xdl_bf8.cpp)
+add_example_executable(example_convnd_fwd_xdl_fp16_comp_fp8 convnd_fwd_xdl_fp16_comp_fp8.cpp)
 add_example_executable(example_convnd_fwd_xdl_fp8_bf8 convnd_fwd_xdl_fp8_bf8.cpp)
 add_example_executable(example_convnd_fwd_dl_fp16 convnd_fwd_dl_fp16.cpp)
 add_example_executable(example_convnd_fwd_dl_fp32 convnd_fwd_dl_fp32.cpp)
diff --git a/example/09_convnd_fwd/convnd_fwd_xdl_fp16_comp_fp8.cpp b/example/09_convnd_fwd/convnd_fwd_xdl_fp16_comp_fp8.cpp
new file mode 100644
index 0000000000..346ab8d953
--- /dev/null
+++ b/example/09_convnd_fwd/convnd_fwd_xdl_fp16_comp_fp8.cpp
@@ -0,0 +1,81 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "convnd_fwd_common.hpp"
+
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_abd_xdl_cshuffle.hpp"
+
+#include "ck/library/utility/convolution_host_tensor_descriptor_helper.hpp"
+
+using InDataType       = ck::half_t;
+using WeiDataType      = ck::half_t;
+using AccDataType      = float;
+using CShuffleDataType = ck::half_t;
+using OutDataType      = ck::half_t;
+using ComputeType      = ck::f8_t;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using InElementOp  = ck::tensor_operation::element_wise::PassThrough;
+using WeiElementOp = ck::tensor_operation::element_wise::PassThrough;
+using OutElementOp = ck::tensor_operation::element_wise::UnaryConvert;
+
+static constexpr auto ConvSpec =
+    ck::tensor_operation::device::ConvolutionForwardSpecialization::Default;
+
+static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+template <ck::index_t NDimSpatial, typename InLayout, typename WeiLayout, typename OutLayout>
+using DeviceGroupedConvNDFwdInstance =
+    ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<
+        NDimSpatial,
+        InLayout,
+        WeiLayout,
+        ck::Tuple<>,
+        OutLayout,
+        InDataType,
+        WeiDataType,
+        AccDataType,
+        CShuffleDataType,
+        ck::Tuple<>,
+        OutDataType,
+        InElementOp,
+        WeiElementOp,
+        OutElementOp,
+        ConvSpec,    // ConvForwardSpecialization
+        GemmSpec,    // GemmSpecialization
+        1,           //
+        256,         // BlockSize
+        128,         // MPerBlock
+        256,         // NPerBlock
+        32,          // KPerBlock
+        8,           // AK1
+        8,           // BK1
+        32,          // MPerXdl
+        32,          // NPerXdl
+        2,           // MXdlPerWave
+        4,           // NXdlPerWave
+        S<4, 64, 1>, // ABlockTransferThreadClusterLengths_AK0_M_AK1
+        S<1, 0, 2>,  // ABlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,  // ABlockTransferSrcAccessOrder
+        2,           // ABlockTransferSrcVectorDim
+        8,           // ABlockTransferSrcScalarPerVector
+        8,           // ABlockTransferDstScalarPerVector_AK1
+        1,           // ABlockLdsExtraM
+        S<4, 64, 1>, // BBlockTransferThreadClusterLengths_BK0_N_BK1
+        S<1, 0, 2>,  // BBlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,  // BBlockTransferSrcAccessOrder
+        2,           // BBlockTransferSrcVectorDim
+        8,           // BBlockTransferSrcScalarPerVector
+        8,           // BBlockTransferDstScalarPerVector_BK1
+        1,           // BBlockLdsExtraN
+        1,
+        1,
+        S<1, 32, 1, 8>,
+        8,
+        ComputeType>;
+
+#include "run_convnd_fwd_example.inc"
+
+int main(int argc, char* argv[]) { return run_convnd_fwd_example(argc, argv) ? 0 : 1; }

From 366592b0ff47f1fe986d08c4b78fa8c87f6a3751 Mon Sep 17 00:00:00 2001
From: Rostyslav Geyyer <46627076+geyyer@users.noreply.github.com>
Date: Tue, 9 Apr 2024 13:57:32 -0500
Subject: [PATCH 2/7] Add an example (#1227)

---
 .../CMakeLists.txt                            |  3 ++
 .../common.hpp                                |  2 +
 ...ed_conv_bwd_data_xdl_fp16_comp_bf8_fp8.cpp | 38 +++++++++++++++++++
 3 files changed, 43 insertions(+)
 create mode 100644 example/38_grouped_conv_bwd_data_multiple_d/grouped_conv_bwd_data_xdl_fp16_comp_bf8_fp8.cpp

diff --git a/example/38_grouped_conv_bwd_data_multiple_d/CMakeLists.txt b/example/38_grouped_conv_bwd_data_multiple_d/CMakeLists.txt
index 72e6959649..ce951f6353 100644
--- a/example/38_grouped_conv_bwd_data_multiple_d/CMakeLists.txt
+++ b/example/38_grouped_conv_bwd_data_multiple_d/CMakeLists.txt
@@ -3,6 +3,9 @@ add_custom_target(example_grouped_conv_bwd_data)
 add_example_executable(example_grouped_conv_bwd_data_xdl_fp16 grouped_conv_bwd_data_xdl_fp16.cpp)
 add_example_dependencies(example_grouped_conv_bwd_data example_grouped_conv_bwd_data_xdl_fp16)
 
+add_example_executable(example_grouped_conv_bwd_data_xdl_fp16_comp_bf8_fp8 grouped_conv_bwd_data_xdl_fp16_comp_bf8_fp8.cpp)
+add_example_dependencies(example_grouped_conv_bwd_data example_grouped_conv_bwd_data_xdl_fp16_comp_bf8_fp8)
+
 add_example_executable(example_grouped_conv_bwd_data_bias_relu_xdl_fp16 grouped_conv_bwd_data_bias_relu_xdl_fp16.cpp)
 add_example_dependencies(example_grouped_conv_bwd_data example_grouped_conv_bwd_data_bias_relu_xdl_fp16)
 
diff --git a/example/38_grouped_conv_bwd_data_multiple_d/common.hpp b/example/38_grouped_conv_bwd_data_multiple_d/common.hpp
index ebb1c606cb..8a0474156c 100644
--- a/example/38_grouped_conv_bwd_data_multiple_d/common.hpp
+++ b/example/38_grouped_conv_bwd_data_multiple_d/common.hpp
@@ -34,6 +34,8 @@ static constexpr auto ConvBwdDataDefault =
 
 using FP16 = ck::half_t;
 using FP32 = float;
+using FP8  = ck::f8_t;
+using BF8  = ck::bf8_t;
 
 struct ExecutionConfig final
 {
diff --git a/example/38_grouped_conv_bwd_data_multiple_d/grouped_conv_bwd_data_xdl_fp16_comp_bf8_fp8.cpp b/example/38_grouped_conv_bwd_data_multiple_d/grouped_conv_bwd_data_xdl_fp16_comp_bf8_fp8.cpp
new file mode 100644
index 0000000000..41023ef82a
--- /dev/null
+++ b/example/38_grouped_conv_bwd_data_multiple_d/grouped_conv_bwd_data_xdl_fp16_comp_bf8_fp8.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_data_multiple_d_xdl_cshuffle_v1.hpp"
+#include "common.hpp"
+
+using OutDataType      = FP16;
+using WeiDataType      = FP16;
+using AccDataType      = FP32;
+using CShuffleDataType = FP16;
+using DsDataType       = ck::Tuple<>;
+using InDataType       = FP16;
+using AComputeType     = BF8;
+using BComputeType     = FP8;
+
+using OutLayout = ck::tensor_layout::convolution::GNHWK;
+using WeiLayout = ck::tensor_layout::convolution::GKYXC;
+using DsLayout  = ck::Tuple<>;
+using InLayout  = ck::tensor_layout::convolution::GNHWC;
+
+using OutElementOp = PassThrough;
+using WeiElementOp = PassThrough;
+using InElementOp  = PassThrough;
+
+static constexpr auto LoopSched = ck::make_default_loop_scheduler();
+
+// clang-format off
+using DeviceConvInstance = ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
+// ######| NDimSpatial|   ALayout|   BLayout|   DsLayout|  ELayout|       AData|       BData|     AccData|         CShuffle|       DsData|      EData| AElementwise| BElementwise| CDEElementwise| ConvolutionBackward| DoPad| DoPad|      NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer|    MXdl|    NXdl|    ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|    BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffleMXdl| CShuffleNXdl|   CDEBlockTransfer| CDEBlockTransfer|      Loop|     ACompute|     BCompute|
+// ######|            |          |          |           |         |        Type|        Type|        Type|         DataType|         Type|       Type|    Operation|    Operation|      Operation|  DataSpecialization| GemmM| GemmN| PrefetchStage|  Size| Block| Block| Block|    |    |  XDL|  XDL| PerWave| PerWave|     ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar|    ExtraM|     ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar|    ExtraN|      PerWave|      PerWave|  _MBlock_MPerBlock|  ScalarPerVector| Scheduler|         Type|         Type|
+// ######|            |          |          |           |         |            |            |            |                 |             |           |             |             |               |                    |      |      |              |      |      |      |      |    |    |     |     |        |        | Lengths_AK0_M_AK1|   ArrangeOrder|               |               |      PerVector|  PerVector_AK1|          | Lengths_BK0_N_BK1|   ArrangeOrder|               |               |      PerVector|  PerVector_BK1|          |   PerShuffle|   PerShuffle|  _NBlock_NPerBlock|       _NPerBlock|          |             |             |
+// ######|            |          |          |           |         |            |            |            |                 |             |           |             |             |               |                    |      |      |              |      |      |      |      |    |    |     |     |        |        |                  |               |               |               |               |               |          |                  |               |               |               |               |               |          |             |             |                   |                 |          |             |             |
+         < NDimSpatial, OutLayout, WeiLayout,   DsLayout, InLayout, OutDataType, WeiDataType, AccDataType, CShuffleDataType,   DsDataType, InDataType, OutElementOp, WeiElementOp,    InElementOp,  ConvBwdDataDefault,  true,  true,             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, LoopSched, AComputeType, BComputeType>;
+// clang-format on
+
+#include "run_grouped_conv_bwd_data_example.inc"
+
+int main(int argc, char* argv[]) { return run_grouped_conv_bwd_data_example(argc, argv); }

From ced5af16f7ac072315d9fb270ac86b77cff6f8c5 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bart=C5=82omiej=20Kocot?= <barkocot@amd.com>
Date: Tue, 9 Apr 2024 23:46:21 +0200
Subject: [PATCH 3/7] Extend support for contraction 6D (#1207)

* Extend support for contraction up to 5D

* Extend contraction bilinear instances

* Fix interface test

* Add 6d support, remove 3d,4d,5d

* Fixes

* Fix readme

* Make defualt dim for contraction instances
---
 .../cpu/reference_contraction.hpp             | 132 ++++-
 .../device_contraction_instance.hpp           | 252 +++++-----
 .../gpu/contraction_bilinear.hpp              | 464 +++++++++++++++++-
 .../gpu/contraction_scale.hpp                 | 464 +++++++++++++++++-
 .../ck/library/utility/host_tensor.hpp        |  33 +-
 ...16_bf16_bf16_compute_f32_kknn_instance.cpp |   5 +-
 ...16_bf16_bf16_compute_f32_knnn_instance.cpp |   5 +-
 ...16_bf16_bf16_compute_f32_mknn_instance.cpp |   5 +-
 ...16_bf16_bf16_compute_f32_mnnn_instance.cpp |   5 +-
 ..._f16_f16_f16_compute_f32_kknn_instance.cpp |   5 +-
 ..._f16_f16_f16_compute_f32_knnn_instance.cpp |   5 +-
 ..._f16_f16_f16_compute_f32_mknn_instance.cpp |   5 +-
 ..._f16_f16_f16_compute_f32_mnnn_instance.cpp |   5 +-
 ...f32_f32_f32_compute_bf16_kknn_instance.cpp |   5 +-
 ...f32_f32_f32_compute_bf16_knnn_instance.cpp |   5 +-
 ...f32_f32_f32_compute_bf16_mknn_instance.cpp |   5 +-
 ...f32_f32_f32_compute_bf16_mnnn_instance.cpp |   5 +-
 ..._f32_f32_f32_compute_f16_kknn_instance.cpp |   5 +-
 ..._f32_f32_f32_compute_f16_knnn_instance.cpp |   5 +-
 ..._f32_f32_f32_compute_f16_mknn_instance.cpp |   5 +-
 ..._f32_f32_f32_compute_f16_mnnn_instance.cpp |   5 +-
 ..._shuffle_f32_f32_f32_f32_kknn_instance.cpp |   5 +-
 ..._shuffle_f32_f32_f32_f32_knnn_instance.cpp |   5 +-
 ..._shuffle_f32_f32_f32_f32_mknn_instance.cpp |   5 +-
 ..._shuffle_f32_f32_f32_f32_mnnn_instance.cpp |   5 +-
 ..._f64_f64_f64_compute_f32_kknn_instance.cpp |   5 +-
 ..._f64_f64_f64_compute_f32_knnn_instance.cpp |   5 +-
 ..._f64_f64_f64_compute_f32_mknn_instance.cpp |   5 +-
 ..._f64_f64_f64_compute_f32_mnnn_instance.cpp |   5 +-
 ..._shuffle_f64_f64_f64_f64_kknn_instance.cpp |   5 +-
 ..._shuffle_f64_f64_f64_f64_knnn_instance.cpp |   5 +-
 ..._shuffle_f64_f64_f64_f64_mknn_instance.cpp |   5 +-
 ..._shuffle_f64_f64_f64_f64_mnnn_instance.cpp |   5 +-
 ...16_bf16_bf16_compute_f32_kknn_instance.cpp |  58 +++
 ...16_bf16_bf16_compute_f32_knnn_instance.cpp |  58 +++
 ...16_bf16_bf16_compute_f32_mknn_instance.cpp |  58 +++
 ...16_bf16_bf16_compute_f32_mnnn_instance.cpp |  58 +++
 ..._f16_f16_f16_compute_f32_kknn_instance.cpp |  58 +++
 ..._f16_f16_f16_compute_f32_knnn_instance.cpp |  58 +++
 ..._f16_f16_f16_compute_f32_mknn_instance.cpp |  58 +++
 ..._f16_f16_f16_compute_f32_mnnn_instance.cpp |  58 +++
 ...f32_f32_f32_compute_bf16_kknn_instance.cpp |  58 +++
 ...f32_f32_f32_compute_bf16_knnn_instance.cpp |  58 +++
 ...f32_f32_f32_compute_bf16_mknn_instance.cpp |  58 +++
 ...f32_f32_f32_compute_bf16_mnnn_instance.cpp |  58 +++
 ..._f32_f32_f32_compute_f16_kknn_instance.cpp |  58 +++
 ..._f32_f32_f32_compute_f16_knnn_instance.cpp |  58 +++
 ..._f32_f32_f32_compute_f16_mknn_instance.cpp |  58 +++
 ..._f32_f32_f32_compute_f16_mnnn_instance.cpp |  58 +++
 ..._shuffle_f32_f32_f32_f32_kknn_instance.cpp |  58 +++
 ..._shuffle_f32_f32_f32_f32_knnn_instance.cpp |  58 +++
 ..._shuffle_f32_f32_f32_f32_mknn_instance.cpp |  58 +++
 ..._shuffle_f32_f32_f32_f32_mnnn_instance.cpp |  58 +++
 ..._f64_f64_f64_compute_f32_kknn_instance.cpp |  58 +++
 ..._f64_f64_f64_compute_f32_knnn_instance.cpp |  58 +++
 ..._f64_f64_f64_compute_f32_mknn_instance.cpp |  58 +++
 ..._f64_f64_f64_compute_f32_mnnn_instance.cpp |  58 +++
 ..._shuffle_f64_f64_f64_f64_kknn_instance.cpp |  58 +++
 ..._shuffle_f64_f64_f64_f64_knnn_instance.cpp |  58 +++
 ..._shuffle_f64_f64_f64_f64_mknn_instance.cpp |  58 +++
 ..._shuffle_f64_f64_f64_f64_mnnn_instance.cpp |  58 +++
 .../gpu/contraction_bilinear/CMakeLists.txt   |  70 +--
 ...f16_bf16_bf16_compute_f32_kkn_instance.cpp |   5 +-
 ...f16_bf16_bf16_compute_f32_knn_instance.cpp |   5 +-
 ...f16_bf16_bf16_compute_f32_mkn_instance.cpp |   5 +-
 ...f16_bf16_bf16_compute_f32_mnn_instance.cpp |   5 +-
 ...e_f16_f16_f16_compute_f32_kkn_instance.cpp |   5 +-
 ...e_f16_f16_f16_compute_f32_knn_instance.cpp |   5 +-
 ...e_f16_f16_f16_compute_f32_mkn_instance.cpp |   5 +-
 ...e_f16_f16_f16_compute_f32_mnn_instance.cpp |   5 +-
 ..._f32_f32_f32_compute_bf16_kkn_instance.cpp |   5 +-
 ..._f32_f32_f32_compute_bf16_knn_instance.cpp |   5 +-
 ..._f32_f32_f32_compute_bf16_mkn_instance.cpp |   5 +-
 ..._f32_f32_f32_compute_bf16_mnn_instance.cpp |   5 +-
 ...e_f32_f32_f32_compute_f16_kkn_instance.cpp |   5 +-
 ...e_f32_f32_f32_compute_f16_knn_instance.cpp |   5 +-
 ...e_f32_f32_f32_compute_f16_mkn_instance.cpp |   5 +-
 ...e_f32_f32_f32_compute_f16_mnn_instance.cpp |   5 +-
 ...xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp |   5 +-
 ...xdl_c_shuffle_f32_f32_f32_knn_instance.cpp |   5 +-
 ...xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp |   5 +-
 ...xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp |   5 +-
 ...e_f64_f64_f64_compute_f32_kkn_instance.cpp |   5 +-
 ...e_f64_f64_f64_compute_f32_knn_instance.cpp |   5 +-
 ...e_f64_f64_f64_compute_f32_mkn_instance.cpp |   5 +-
 ...e_f64_f64_f64_compute_f32_mnn_instance.cpp |   5 +-
 ...xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp |   5 +-
 ...xdl_c_shuffle_f64_f64_f64_knn_instance.cpp |   5 +-
 ...xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp |   5 +-
 ...xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp |   5 +-
 ...f16_bf16_bf16_compute_f32_kkn_instance.cpp |  58 +++
 ...f16_bf16_bf16_compute_f32_knn_instance.cpp |  58 +++
 ...f16_bf16_bf16_compute_f32_mkn_instance.cpp |  58 +++
 ...f16_bf16_bf16_compute_f32_mnn_instance.cpp |  58 +++
 ...e_f16_f16_f16_compute_f32_kkn_instance.cpp |  58 +++
 ...e_f16_f16_f16_compute_f32_knn_instance.cpp |  58 +++
 ...e_f16_f16_f16_compute_f32_mkn_instance.cpp |  58 +++
 ...e_f16_f16_f16_compute_f32_mnn_instance.cpp |  58 +++
 ..._f32_f32_f32_compute_bf16_kkn_instance.cpp |  58 +++
 ..._f32_f32_f32_compute_bf16_knn_instance.cpp |  58 +++
 ..._f32_f32_f32_compute_bf16_mkn_instance.cpp |  58 +++
 ..._f32_f32_f32_compute_bf16_mnn_instance.cpp |  58 +++
 ...e_f32_f32_f32_compute_f16_kkn_instance.cpp |  58 +++
 ...e_f32_f32_f32_compute_f16_knn_instance.cpp |  58 +++
 ...e_f32_f32_f32_compute_f16_mkn_instance.cpp |  58 +++
 ...e_f32_f32_f32_compute_f16_mnn_instance.cpp |  58 +++
 ...xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp |  57 +++
 ...xdl_c_shuffle_f32_f32_f32_knn_instance.cpp |  57 +++
 ...xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp |  57 +++
 ...xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp |  57 +++
 ...e_f64_f64_f64_compute_f32_kkn_instance.cpp |  58 +++
 ...e_f64_f64_f64_compute_f32_knn_instance.cpp |  58 +++
 ...e_f64_f64_f64_compute_f32_mkn_instance.cpp |  58 +++
 ...e_f64_f64_f64_compute_f32_mnn_instance.cpp |  58 +++
 ...xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp |  57 +++
 ...xdl_c_shuffle_f64_f64_f64_knn_instance.cpp |  57 +++
 ...xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp |  57 +++
 ...xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp |  57 +++
 .../gpu/contraction_scale/CMakeLists.txt      |  70 +--
 profiler/README.md                            |  24 +-
 .../profiler/profile_contraction_impl.hpp     |  80 ++-
 .../profiler/profile_contraction_utils.hpp    |  28 +-
 profiler/src/profile_contraction_bilinear.cpp | 152 +++---
 profiler/src/profile_contraction_scale.cpp    | 143 ++++--
 .../test_contraction_interface_xdl.cpp        |  14 +-
 test/contraction/test_contraction_xdl.cpp     | 138 ++++--
 126 files changed, 5033 insertions(+), 551 deletions(-)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_bilinear/{ => 2D}/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp (94%)
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp (94%)
 rename library/src/tensor_operation_instance/gpu/contraction_scale/{ => 2D}/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp (94%)
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp

diff --git a/library/include/ck/library/reference_tensor_operation/cpu/reference_contraction.hpp b/library/include/ck/library/reference_tensor_operation/cpu/reference_contraction.hpp
index 527dac6d3b..38557e349e 100644
--- a/library/include/ck/library/reference_tensor_operation/cpu/reference_contraction.hpp
+++ b/library/include/ck/library/reference_tensor_operation/cpu/reference_contraction.hpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -15,7 +15,6 @@ namespace ck {
 namespace tensor_operation {
 namespace host {
 
-// hardcoded for NumDimM == NumDimN == NumDimK == 2
 template <ck::index_t NumDimM,
           ck::index_t NumDimN,
           ck::index_t NumDimK,
@@ -26,7 +25,9 @@ template <ck::index_t NumDimM,
           typename ComputeDataType,
           typename AElementwiseOperation,
           typename BElementwiseOperation,
-          ck::enable_if_t<NumDimM == 2 && NumDimN == 2 && NumDimK == 2, bool> = false>
+          ck::enable_if_t<(NumDimM == 2 || NumDimM == 6) && (NumDimN == 2 || NumDimN == 6) &&
+                              (NumDimK == 2 || NumDimK == 6),
+                          bool> = false>
 struct ReferenceContraction_M2_N2_K2 : public ck::tensor_operation::device::BaseOperator
 {
     // Argument
@@ -60,9 +61,28 @@ struct ReferenceContraction_M2_N2_K2 : public ck::tensor_operation::device::Base
 
         float Run(const Argument& arg)
         {
-            auto f_ms_ns = [&](auto m0, auto m1, auto n0, auto n1) {
-                const ck::index_t K0 = arg.a_ms_ks_.mDesc.GetLengths()[2];
-                const ck::index_t K1 = arg.a_ms_ks_.mDesc.GetLengths()[3];
+            auto f_ms_ns = [&](auto m0,
+                               auto m1,
+                               auto m2,
+                               auto m3,
+                               auto m4,
+                               auto m5,
+                               auto n0,
+                               auto n1,
+                               auto n2,
+                               auto n3,
+                               auto n4,
+                               auto n5) {
+                const ck::index_t K0 = arg.a_ms_ks_.mDesc.GetLengths()[NumDimM];
+                const ck::index_t K1 = arg.a_ms_ks_.mDesc.GetLengths()[NumDimM + 1];
+                const ck::index_t K2 =
+                    NumDimK >= 3 ? arg.a_ms_ks_.mDesc.GetLengths()[NumDimM + 2] : 1;
+                const ck::index_t K3 =
+                    NumDimK >= 4 ? arg.a_ms_ks_.mDesc.GetLengths()[NumDimM + 3] : 1;
+                const ck::index_t K4 =
+                    NumDimK >= 5 ? arg.a_ms_ks_.mDesc.GetLengths()[NumDimM + 4] : 1;
+                const ck::index_t K5 =
+                    NumDimK >= 6 ? arg.a_ms_ks_.mDesc.GetLengths()[NumDimM + 5] : 1;
 
                 AccDataType v_acc = 0;
 
@@ -70,32 +90,96 @@ struct ReferenceContraction_M2_N2_K2 : public ck::tensor_operation::device::Base
                 {
                     for(ck::index_t k1 = 0; k1 < K1; ++k1)
                     {
-                        // Simulate the possible casting when ComputeDataType is different than the
-                        // A/B data types
-                        ComputeDataType v_a_compute_input =
-                            ck::type_convert<ComputeDataType>(arg.a_ms_ks_(m0, m1, k0, k1));
-                        ComputeDataType v_b_compute_input =
-                            ck::type_convert<ComputeDataType>(arg.b_ns_ks_(n0, n1, k0, k1));
+                        for(ck::index_t k2 = 0; k2 < K2; ++k2)
+                        {
+                            for(ck::index_t k3 = 0; k3 < K3; ++k3)
+                            {
+                                for(ck::index_t k4 = 0; k4 < K4; ++k4)
+                                {
+                                    for(ck::index_t k5 = 0; k5 < K5; ++k5)
+                                    {
+                                        ComputeDataType v_a_compute_input;
+                                        ComputeDataType v_b_compute_input;
 
-                        AccDataType v_a;
-                        AccDataType v_b;
+                                        // Simulate the possible casting when ComputeDataType is
+                                        // different than the A/B data types
+                                        if constexpr(NumDimK == 2)
+                                        {
+                                            v_a_compute_input = ck::type_convert<ComputeDataType>(
+                                                arg.a_ms_ks_(m0, m1, k0, k1));
+                                            v_b_compute_input = ck::type_convert<ComputeDataType>(
+                                                arg.b_ns_ks_(n0, n1, k0, k1));
+                                        }
+                                        else if constexpr(NumDimK == 6)
+                                        {
+                                            v_a_compute_input = ck::type_convert<
+                                                ComputeDataType>(arg.a_ms_ks_(
+                                                m0, m1, m2, m3, m4, m5, k0, k1, k2, k3, k4, k5));
+                                            v_b_compute_input = ck::type_convert<
+                                                ComputeDataType>(arg.b_ns_ks_(
+                                                n0, n1, n2, n3, n4, n5, k0, k1, k2, k3, k4, k5));
+                                        }
 
-                        arg.a_element_op_(v_a, ck::type_convert<AccDataType>(v_a_compute_input));
-                        arg.b_element_op_(v_b, ck::type_convert<AccDataType>(v_b_compute_input));
+                                        AccDataType v_a;
+                                        AccDataType v_b;
 
-                        v_acc += v_a * v_b;
+                                        arg.a_element_op_(
+                                            v_a, ck::type_convert<AccDataType>(v_a_compute_input));
+                                        arg.b_element_op_(
+                                            v_b, ck::type_convert<AccDataType>(v_b_compute_input));
+
+                                        v_acc += v_a * v_b;
+                                    }
+                                }
+                            }
+                        }
                     }
                 }
 
-                arg.c_ms_ns_(m0, m1, n0, n1) = ck::type_convert<CDataType>(v_acc);
+                if constexpr(NumDimK == 2)
+                {
+                    arg.c_ms_ns_(m0, m1, n0, n1) = ck::type_convert<CDataType>(v_acc);
+                }
+                else if constexpr(NumDimK == 6)
+                {
+                    arg.c_ms_ns_(m0, m1, m2, m3, m4, m5, n0, n1, n2, n3, n4, n5) =
+                        ck::type_convert<CDataType>(v_acc);
+                }
             };
 
-            make_ParallelTensorFunctor(f_ms_ns,
-                                       arg.c_ms_ns_.mDesc.GetLengths()[0],
-                                       arg.c_ms_ns_.mDesc.GetLengths()[1],
-                                       arg.c_ms_ns_.mDesc.GetLengths()[2],
-                                       arg.c_ms_ns_.mDesc.GetLengths()[3])(
-                std::thread::hardware_concurrency());
+            if constexpr(NumDimK == 2)
+            {
+                make_ParallelTensorFunctor(f_ms_ns,
+                                           arg.c_ms_ns_.mDesc.GetLengths()[0],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[1],
+                                           1,
+                                           1,
+                                           1,
+                                           1,
+                                           arg.c_ms_ns_.mDesc.GetLengths()[2],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[3],
+                                           1,
+                                           1,
+                                           1,
+                                           1)(std::thread::hardware_concurrency());
+            }
+            else if constexpr(NumDimK == 6)
+            {
+                make_ParallelTensorFunctor(f_ms_ns,
+                                           arg.c_ms_ns_.mDesc.GetLengths()[0],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[1],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[2],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[3],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[4],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[5],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[6],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[7],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[8],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[9],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[10],
+                                           arg.c_ms_ns_.mDesc.GetLengths()[11])(
+                    std::thread::hardware_concurrency());
+            }
 
             return 0;
         }
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp b/library/include/ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp
index b67119ad19..84b9764391 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 #pragma once
 
 #include <cstdlib>
@@ -42,30 +42,31 @@ template <typename ADataType,
           typename ComputeDataType,
           typename AElementwiseOp,
           typename BElementwiseOp,
-          typename CDEElementwiseOp>
+          typename CDEElementwiseOp,
+          index_t NumDim = 2>
 using device_contraction_kk_instance = std::tuple<
     // clang-format off
         //#####################################| NumDimM| NumDimN| NumDimK|      AData|     BData|     AccData|         CShuffle|     DsData|     EData|              A|              B|              CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|         Compute|
         //#####################################|        |        |        |       Type|      Type|        Type|         DataType|       Type|      Type|    Elementwise|    Elementwise|      Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            Data|
         //#####################################|        |        |        |           |          |            |                 |           |          |      Operation|      Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            Type|
         //#####################################|        |        |        |           |          |            |                 |           |          |               |               |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    32,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    32,    64,    16,   4,   4,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    32,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    32,    64,    16,   4,   4,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1,  8>,               4, ComputeDataType>,
         // Small scalar per vector
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              4,         1,           1,           1,              S<1, 16, 1,  8>,               2, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              4,         1,           1,           1,              S<1, 16, 1,  8>,               2, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
     // clang-format on
     >;
 
@@ -78,33 +79,34 @@ template <typename ADataType,
           typename ComputeDataType,
           typename AElementwiseOp,
           typename BElementwiseOp,
-          typename CDEElementwiseOp>
+          typename CDEElementwiseOp,
+          index_t NumDim = 2>
 using device_contraction_kn_instance = std::tuple<
     // clang-format off
         //#####################################| NumDimM| NumDimN| NumDimK|      AData|     BData|     AccData|         CShuffle|     DsData|     EData|              A|              B|              CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|         Compute|
         //#####################################|        |        |        |       Type|      Type|        Type|         DataType|       Type|      Type|    Elementwise|    Elementwise|      Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            Data|
         //#####################################|        |        |        |           |          |            |                 |           |          |      Operation|      Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            Type|
         //#####################################|        |        |        |           |          |            |                 |           |          |               |               |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   1,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   1,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   1,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   1,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   1,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   1,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   1,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   1,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   1,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   1,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   1,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   1,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   1,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   1,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
         // Small scalar per vector
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1,  8>,               2, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              4,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              2,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1,  8>,               2, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              4,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
     // clang-format on
     >;
 
@@ -117,33 +119,34 @@ template <typename ADataType,
           typename ComputeDataType,
           typename AElementwiseOp,
           typename BElementwiseOp,
-          typename CDEElementwiseOp>
+          typename CDEElementwiseOp,
+          index_t NumDim = 2>
 using device_contraction_mk_instance = std::tuple<
     // clang-format off
         //#####################################| NumDimM| NumDimN| NumDimK|      AData|     BData|     AccData|         CShuffle|     DsData|     EData|              A|              B|              CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|         Compute|
         //#####################################|        |        |        |       Type|      Type|        Type|         DataType|       Type|      Type|    Elementwise|    Elementwise|      Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            Data|
         //#####################################|        |        |        |           |          |            |                 |           |          |      Operation|      Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            Type|
         //#####################################|        |        |        |           |          |            |                 |           |          |               |               |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   1,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   1,   4,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   1,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   4,   32,   32,    2,    2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   4,   32,   32,    2,    1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   4,   32,   32,    1,    2,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   1,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   1,   4,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   1,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   4,   32,   32,    2,    2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   4,   32,   32,    2,    1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   4,   32,   32,    1,    2,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
         // Small scalar per vector
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              4,         1,           1,           1,              S<1, 16, 1,  8>,               2, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              4,         1,           1,           1,              S<1, 16, 1,  8>,               2, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
     // clang-format on
     >;
 
@@ -156,33 +159,34 @@ template <typename ADataType,
           typename ComputeDataType,
           typename AElementwiseOp,
           typename BElementwiseOp,
-          typename CDEElementwiseOp>
+          typename CDEElementwiseOp,
+          index_t NumDim = 2>
 using device_contraction_mn_instance = std::tuple<
     // clang-format off
         //#####################################| NumDimM| NumDimN| NumDimK|      AData|     BData|     AccData|         CShuffle|     DsData|     EData|              A|              B|              CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|         Compute|
         //#####################################|        |        |        |       Type|      Type|        Type|         DataType|       Type|      Type|    Elementwise|    Elementwise|      Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            Data|
         //#####################################|        |        |        |           |          |            |                 |           |          |      Operation|      Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            Type|
         //#####################################|        |        |        |           |          |            |                 |           |          |               |               |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   1,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   1,   1,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   1,   1,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   1,   32,   32,    2,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   1,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   1,   32,   32,    2,    2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   1,   32,   32,    2,    1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   1,   32,   32,    1,    2,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   1,   1,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   1,   1,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    16,   4,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   1,   1,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,   128,    16,   4,   4,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   1,   32,   32,    2,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   1,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1,  8>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   1,   32,   32,    2,    2,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,         1,           1,           1,              S<1,  8, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   1,   32,   32,    2,    1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   4,   4,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   1,   32,   32,    1,    2,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              1,         0,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   4,   4,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,         1,           1,           1,              S<1, 16, 1, 16>,               4, ComputeDataType>,
         // Small scalar per vector
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1,  8>,               2, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   4,   4,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1, 16, 1,  8>,               2, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   4,   4,   32,   32,    2,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              4,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              4,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
     // clang-format on
     >;
 
@@ -195,23 +199,24 @@ template <typename ADataType,
           typename ComputeDataType,
           typename AElementwiseOp,
           typename BElementwiseOp,
-          typename CDEElementwiseOp>
+          typename CDEElementwiseOp,
+          index_t NumDim = 2>
 using device_contraction_f64_kk_instance = std::tuple<
     // clang-format off
         //#####################################| NumDimM| NumDimN| NumDimK|      AData|     BData|     AccData|         CShuffle|     DsData|     EData|              A|              B|              CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|         Compute|
         //#####################################|        |        |        |       Type|      Type|        Type|         DataType|       Type|      Type|    Elementwise|    Elementwise|      Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            Data|
         //#####################################|        |        |        |           |          |            |                 |           |          |      Operation|      Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            Type|
         //#####################################|        |        |        |           |          |            |                 |           |          |               |               |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   2,   2,   16,   16,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    32,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   2,   2,   16,   16,    4,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    32,    64,    16,   2,   2,   16,   16,    2,    4,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    32,    16,   2,   2,   16,   16,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    32,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    64,    32,    16,   2,   2,   16,   16,    4,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,    64,    32,    64,    16,   2,   2,   16,   16,    2,    4,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,              S<1,  8, 1,  8>,               1, ComputeDataType>
     // clang-format on
     >;
 
@@ -224,23 +229,24 @@ template <typename ADataType,
           typename ComputeDataType,
           typename AElementwiseOp,
           typename BElementwiseOp,
-          typename CDEElementwiseOp>
+          typename CDEElementwiseOp,
+          index_t NumDim = 2>
 using device_contraction_f64_kn_instance = std::tuple<
     // clang-format off
         //#####################################| NumDimM| NumDimN| NumDimK|      AData|     BData|     AccData|         CShuffle|     DsData|     EData|              A|              B|              CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|         Compute|
         //#####################################|        |        |        |       Type|      Type|        Type|         DataType|       Type|      Type|    Elementwise|    Elementwise|      Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            Data|
         //#####################################|        |        |        |           |          |            |                 |           |          |      Operation|      Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            Type|
         //#####################################|        |        |        |           |          |            |                 |           |          |               |               |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   1,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   1,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   1,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   1,   16,   16,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   1,   16,   16,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   1,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   1,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   1,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   1,   16,   16,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   1,   16,   16,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>
     // clang-format on
     >;
 
@@ -253,23 +259,24 @@ template <typename ADataType,
           typename ComputeDataType,
           typename AElementwiseOp,
           typename BElementwiseOp,
-          typename CDEElementwiseOp>
+          typename CDEElementwiseOp,
+          index_t NumDim = 2>
 using device_contraction_f64_mk_instance = std::tuple<
     // clang-format off
         //#####################################| NumDimM| NumDimN| NumDimK|      AData|     BData|     AccData|         CShuffle|     DsData|     EData|              A|              B|              CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|         Compute|
         //#####################################|        |        |        |       Type|      Type|        Type|         DataType|       Type|      Type|    Elementwise|    Elementwise|      Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            Data|
         //#####################################|        |        |        |           |          |            |                 |           |          |      Operation|      Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            Type|
         //#####################################|        |        |        |           |          |            |                 |           |          |               |               |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   2,   16,   16,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1,  8>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1,  8>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   2,   16,   16,    4,    4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1,  8, 1, 16>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1,  8, 1, 16>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   2,   16,   16,    4,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   2,   16,   16,    2,    4,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   2,   16,   16,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1,  8>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1,  8>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   2,   16,   16,    4,    4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1,  8, 1, 16>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1,  8, 1, 16>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   2,   16,   16,    4,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   2,   16,   16,    2,    4,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,         1,           1,           1,               S<1, 16, 1, 16>,              1, ComputeDataType>
     // clang-format on
     >;
 
@@ -282,23 +289,24 @@ template <typename ADataType,
           typename ComputeDataType,
           typename AElementwiseOp,
           typename BElementwiseOp,
-          typename CDEElementwiseOp>
+          typename CDEElementwiseOp,
+          index_t NumDim = 2>
 using device_contraction_f64_mn_instance = std::tuple<
     // clang-format off
         //#####################################| NumDimM| NumDimN| NumDimK|      AData|     BData|     AccData|         CShuffle|     DsData|     EData|              A|              B|              CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|         Compute|
         //#####################################|        |        |        |       Type|      Type|        Type|         DataType|       Type|      Type|    Elementwise|    Elementwise|      Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            Data|
         //#####################################|        |        |        |           |          |            |                 |           |          |      Operation|      Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            Type|
         //#####################################|        |        |        |           |          |            |                 |           |          |               |               |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                |
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   1,   16,   16,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   1,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   1,   16,   16,    4,    4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   1,   16,   16,    4,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   1,   16,   16,    2,    4,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
-        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   1,   16,   16,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   1,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   1,   16,   16,    4,    4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   1,   16,   16,    4,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   1,   16,   16,    2,    4,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>,
+        DeviceContractionMultipleD_Xdl_CShuffle<  NumDim,  NumDim,  NumDim,  ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1, ComputeDataType>
     // clang-format on
     >;
 
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/contraction_bilinear.hpp b/library/include/ck/library/tensor_operation_instance/gpu/contraction_bilinear.hpp
index d06cab1196..948c7ff3aa 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/contraction_bilinear.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/contraction_bilinear.hpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -16,6 +16,7 @@ namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
+// 2D
 #ifdef CK_ENABLE_FP32
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
@@ -387,6 +388,378 @@ void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_
                                                            Bilinear,
                                                            F32>>>& instances);
 #endif // CK_ENABLE_FP16
+// 6D
+#ifdef CK_ENABLE_FP32
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F16>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F16>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F16>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F16>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           BF16>>>& instances);
+#endif // CK_ENABLE_FP32
+
+#ifdef CK_ENABLE_FP64
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F64>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F64>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F64>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F64>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP64
+
+#ifdef CK_ENABLE_FP16
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           F16_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           F16_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           F16_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           F16_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP16
+
+#ifdef CK_ENABLE_BF16
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP16
 
 // Contraction + Bilinear
 template <index_t NumDimM,
@@ -465,6 +838,42 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, ck::half_t>)
+                {
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, ck::bhalf_t>)
+                {
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP32
 #ifdef CK_ENABLE_FP64
@@ -496,6 +905,31 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, double>)
+                {
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP64
 #ifdef CK_ENABLE_FP16
@@ -516,6 +950,20 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP16
 #ifdef CK_ENABLE_BF16
@@ -536,6 +984,20 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance(
+                        op_ptrs);
+                    add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_BF16
         return op_ptrs;
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/contraction_scale.hpp b/library/include/ck/library/tensor_operation_instance/gpu/contraction_scale.hpp
index 8e994d61ae..86becdb313 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/contraction_scale.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/contraction_scale.hpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -16,6 +16,7 @@ namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
+// 2D
 #ifdef CK_ENABLE_FP32
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
@@ -387,6 +388,378 @@ void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_
                                                            Scale,
                                                            F32>>>& instances);
 #endif // CK_ENABLE_FP16
+// 6D
+#ifdef CK_ENABLE_FP32
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances);
+#endif // CK_ENABLE_FP32
+
+#ifdef CK_ENABLE_FP64
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP64
+
+#ifdef CK_ENABLE_FP16
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP16
+
+#ifdef CK_ENABLE_BF16
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP16
 
 // Contraction + Scale
 template <index_t NumDimM,
@@ -464,6 +837,42 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, ck::half_t>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, ck::bhalf_t>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP32
 #ifdef CK_ENABLE_FP64
@@ -495,6 +904,31 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, double>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP64
 #ifdef CK_ENABLE_FP16
@@ -515,6 +949,20 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP16
 #ifdef CK_ENABLE_BF16
@@ -535,6 +983,20 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_BF16
         return op_ptrs;
diff --git a/library/include/ck/library/utility/host_tensor.hpp b/library/include/ck/library/utility/host_tensor.hpp
index 816d834130..ddbd16ad9a 100644
--- a/library/include/ck/library/utility/host_tensor.hpp
+++ b/library/include/ck/library/utility/host_tensor.hpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -408,6 +408,37 @@ struct Tensor
                                        mDesc.GetLengths()[5])(num_thread);
             break;
         }
+        case 12: {
+            auto f = [&](auto i0,
+                         auto i1,
+                         auto i2,
+                         auto i3,
+                         auto i4,
+                         auto i5,
+                         auto i6,
+                         auto i7,
+                         auto i8,
+                         auto i9,
+                         auto i10,
+                         auto i11) {
+                (*this)(i0, i1, i2, i3, i4, i5, i6, i7, i8, i9, i10, i11) =
+                    g(i0, i1, i2, i3, i4, i5, i6, i7, i8, i9, i10, i11);
+            };
+            make_ParallelTensorFunctor(f,
+                                       mDesc.GetLengths()[0],
+                                       mDesc.GetLengths()[1],
+                                       mDesc.GetLengths()[2],
+                                       mDesc.GetLengths()[3],
+                                       mDesc.GetLengths()[4],
+                                       mDesc.GetLengths()[5],
+                                       mDesc.GetLengths()[6],
+                                       mDesc.GetLengths()[7],
+                                       mDesc.GetLengths()[8],
+                                       mDesc.GetLengths()[9],
+                                       mDesc.GetLengths()[10],
+                                       mDesc.GetLengths()[11])(num_thread);
+            break;
+        }
         default: throw std::runtime_error("unspported dimension");
         }
     }
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
index 56dc1d2c71..fbfaaa447d 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_com
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
index 7b8163814e..79a133e367 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_com
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
index ba202202f9..ed59114c4d 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_com
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp
index b6acab1c18..87c9b5a39f 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_com
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
index 1bd3495b36..73ddec4ca6 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
index 6817546d3d..7fbd5378fa 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
index cedace2cb2..a184af5d6e 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp
index b8383e5268..d11fe67345 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
index 2e73bdb804..fd1ef79278 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute
                                    BF16,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
index ae69db03ac..3ac06b8e0a 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute
                                    BF16,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
index fc6b6003f7..9335a491b3 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute
                                    BF16,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp
index b81984acc4..51142cf58f 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute
                                    BF16,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
index 8da6a79f0a..dd987efc7c 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute
                                    F16,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
index b05c443e85..d8d6c4be8d 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute
                                    F16,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
index c2cbb9f25f..52bdc5995b 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute
                                    F16,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp
index 887eaad7eb..83cf74764a 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute
                                    F16,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
index 9a3c71acb5..fa8e21b0ca 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_in
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
index 5659d477d4..fad88126ca 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_in
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
index 7cbe1e34be..4168956a4d 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_in
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp
index e26dcf394b..e90515652c 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_in
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Bilinear>;
+                                   Bilinear,
+                                   2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
index 74c5c9c653..3326dbf04a 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute
                                        F32,
                                        PassThrough,
                                        PassThrough,
-                                       Bilinear>;
+                                       Bilinear,
+                                       2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
index 8a2705f722..73151526ec 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute
                                        F32,
                                        PassThrough,
                                        PassThrough,
-                                       Bilinear>;
+                                       Bilinear,
+                                       2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
index 47924a843d..a14dec910b 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute
                                        F32,
                                        PassThrough,
                                        PassThrough,
-                                       Bilinear>;
+                                       Bilinear,
+                                       2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp
index d6c684e2b0..a682d92fdc 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute
                                        F32,
                                        PassThrough,
                                        PassThrough,
-                                       Bilinear>;
+                                       Bilinear,
+                                       2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
index 75bd8e3f4a..f1e1a8d8d6 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_in
                                        F64,
                                        PassThrough,
                                        PassThrough,
-                                       Bilinear>;
+                                       Bilinear,
+                                       2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
index c515f37916..0287f929d5 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_in
                                        F64,
                                        PassThrough,
                                        PassThrough,
-                                       Bilinear>;
+                                       Bilinear,
+                                       2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
index ff9fa61e0b..2be2171b2c 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_in
                                        F64,
                                        PassThrough,
                                        PassThrough,
-                                       Bilinear>;
+                                       Bilinear,
+                                       2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
index f274c45b58..537872663c 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/2D/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_in
                                        F64,
                                        PassThrough,
                                        PassThrough,
-                                       Bilinear>;
+                                       Bilinear,
+                                       2>;
 
 void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
new file mode 100644
index 0000000000..8403a51e5e
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance =
+    device_contraction_kk_instance<BF16,
+                                   BF16,
+                                   F32,
+                                   BF16,
+                                   BF16_Tuple,
+                                   BF16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
new file mode 100644
index 0000000000..a2da04f702
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance =
+    device_contraction_kn_instance<BF16,
+                                   BF16,
+                                   F32,
+                                   BF16,
+                                   BF16_Tuple,
+                                   BF16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
new file mode 100644
index 0000000000..7f076c9856
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance =
+    device_contraction_mk_instance<BF16,
+                                   BF16,
+                                   F32,
+                                   BF16,
+                                   BF16_Tuple,
+                                   BF16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp
new file mode 100644
index 0000000000..c9e003feaa
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance =
+    device_contraction_mn_instance<BF16,
+                                   BF16,
+                                   F32,
+                                   BF16,
+                                   BF16_Tuple,
+                                   BF16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
new file mode 100644
index 0000000000..6557151183
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance =
+    device_contraction_kk_instance<F16,
+                                   F16,
+                                   F32,
+                                   F16,
+                                   F16_Tuple,
+                                   F16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           F16_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
new file mode 100644
index 0000000000..4904360ea5
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance =
+    device_contraction_kn_instance<F16,
+                                   F16,
+                                   F32,
+                                   F16,
+                                   F16_Tuple,
+                                   F16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           F16_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
new file mode 100644
index 0000000000..069a8a744e
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance =
+    device_contraction_mk_instance<F16,
+                                   F16,
+                                   F32,
+                                   F16,
+                                   F16_Tuple,
+                                   F16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           F16_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp
new file mode 100644
index 0000000000..6052371389
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance =
+    device_contraction_mn_instance<F16,
+                                   F16,
+                                   F32,
+                                   F16,
+                                   F16_Tuple,
+                                   F16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           F16_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
new file mode 100644
index 0000000000..16540cf115
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance =
+    device_contraction_kk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   BF16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           BF16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
new file mode 100644
index 0000000000..2f1206bbf4
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance =
+    device_contraction_kn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   BF16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           BF16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
new file mode 100644
index 0000000000..9bd03b37a8
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance =
+    device_contraction_mk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   BF16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           BF16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp
new file mode 100644
index 0000000000..55a1fb301f
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance =
+    device_contraction_mn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   BF16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           BF16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
new file mode 100644
index 0000000000..5394dc80db
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance =
+    device_contraction_kk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   F16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
new file mode 100644
index 0000000000..f5967e75ec
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance =
+    device_contraction_kn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   F16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
new file mode 100644
index 0000000000..10713d7417
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance =
+    device_contraction_mk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   F16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp
new file mode 100644
index 0000000000..1dfcfaae16
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance =
+    device_contraction_mn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   F16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
new file mode 100644
index 0000000000..112b5e1a00
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance =
+    device_contraction_kk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
new file mode 100644
index 0000000000..f281a9fc74
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance =
+    device_contraction_kn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
new file mode 100644
index 0000000000..9c79c15a36
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance =
+    device_contraction_mk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp
new file mode 100644
index 0000000000..1a4425e45c
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance =
+    device_contraction_mn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   F32_Tuple,
+                                   F32,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Bilinear,
+                                   6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           F32_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
new file mode 100644
index 0000000000..52c8fb1c11
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance =
+    device_contraction_f64_kk_instance<F64,
+                                       F64,
+                                       F32,
+                                       F64,
+                                       F64_Tuple,
+                                       F64,
+                                       F32,
+                                       PassThrough,
+                                       PassThrough,
+                                       Bilinear,
+                                       6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
new file mode 100644
index 0000000000..d023e438df
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance =
+    device_contraction_f64_kn_instance<F64,
+                                       F64,
+                                       F32,
+                                       F64,
+                                       F64_Tuple,
+                                       F64,
+                                       F32,
+                                       PassThrough,
+                                       PassThrough,
+                                       Bilinear,
+                                       6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
new file mode 100644
index 0000000000..72bb9eda33
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance =
+    device_contraction_f64_mk_instance<F64,
+                                       F64,
+                                       F32,
+                                       F64,
+                                       F64_Tuple,
+                                       F64,
+                                       F32,
+                                       PassThrough,
+                                       PassThrough,
+                                       Bilinear,
+                                       6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp
new file mode 100644
index 0000000000..97ef213be6
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance =
+    device_contraction_f64_mn_instance<F64,
+                                       F64,
+                                       F32,
+                                       F64,
+                                       F64_Tuple,
+                                       F64,
+                                       F32,
+                                       PassThrough,
+                                       PassThrough,
+                                       Bilinear,
+                                       6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
new file mode 100644
index 0000000000..06b1ce8fc0
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance =
+    device_contraction_f64_kk_instance<F64,
+                                       F64,
+                                       F64,
+                                       F64,
+                                       F64_Tuple,
+                                       F64,
+                                       F64,
+                                       PassThrough,
+                                       PassThrough,
+                                       Bilinear,
+                                       6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F64>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
new file mode 100644
index 0000000000..078facb336
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance =
+    device_contraction_f64_kn_instance<F64,
+                                       F64,
+                                       F64,
+                                       F64,
+                                       F64_Tuple,
+                                       F64,
+                                       F64,
+                                       PassThrough,
+                                       PassThrough,
+                                       Bilinear,
+                                       6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F64>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
new file mode 100644
index 0000000000..14eae08fd1
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance =
+    device_contraction_f64_mk_instance<F64,
+                                       F64,
+                                       F64,
+                                       F64,
+                                       F64_Tuple,
+                                       F64,
+                                       F64,
+                                       PassThrough,
+                                       PassThrough,
+                                       Bilinear,
+                                       6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F64>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
new file mode 100644
index 0000000000..982e4dd5f8
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/6D/device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance =
+    device_contraction_f64_mn_instance<F64,
+                                       F64,
+                                       F64,
+                                       F64,
+                                       F64_Tuple,
+                                       F64,
+                                       F64,
+                                       PassThrough,
+                                       PassThrough,
+                                       Bilinear,
+                                       6>;
+
+void add_device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear,
+                                                           F64>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_bilinear/CMakeLists.txt b/library/src/tensor_operation_instance/gpu/contraction_bilinear/CMakeLists.txt
index a28c6717dd..70e4bbfe57 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/CMakeLists.txt
@@ -1,43 +1,49 @@
 # ONLY XDL_KERNELS
 set(DEVICE_CONTRACTION_BILINEAR_INSTANCES)
 
-# FP32
-list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp)
+list(APPEND DIMS 2 6)
 
-list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp)
+foreach(idx IN LISTS DIMS)
+    set(PREFIX ${idx}D/device_contraction_bilinear_m${idx}_n${idx}_k${idx})
 
-list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp)
+    # FP32
+    list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_kknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_knnn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_mknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_mnnn_instance.cpp)
 
-# FP64
-list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp)
+    list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_kknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_knnn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_compute_f16_mnnn_instance.cpp)
 
-list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp)
+    list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_kknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_knnn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f32_f32_f32_f32_compute_bf16_mnnn_instance.cpp)
 
-# FP16
-list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp)
+    # FP64
+    list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES ${PREFIX}_xdl_c_shuffle_f64_f64_f64_f64_kknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f64_f64_f64_f64_knnn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f64_f64_f64_f64_mknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp)
 
-# BF16
-list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
-                                                    device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp)
+    list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES ${PREFIX}_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_kknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_knnn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f64_f64_f64_f64_compute_f32_mnnn_instance.cpp)
+
+    # FP16
+    list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES ${PREFIX}_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_kknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_knnn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_f16_f16_f16_f16_compute_f32_mnnn_instance.cpp)
+
+    # BF16
+    list(APPEND DEVICE_CONTRACTION_BILINEAR_INSTANCES ${PREFIX}_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_kknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_knnn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mknn_instance.cpp
+                                                        ${PREFIX}_xdl_c_shuffle_bf16_bf16_bf16_bf16_compute_f32_mnnn_instance.cpp)
+endforeach()
 
 add_instance_library(device_contraction_bilinear_instance ${DEVICE_CONTRACTION_BILINEAR_INSTANCES})
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
index ace4d4a330..eb8f3641b9 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
index d2a1d45f7a..780127613b 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
index aa4e4df9d9..cb782238e0 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp
index a59bce0a1f..612475e03e 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
index 876acd3f69..74bacfdf60 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kk
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
index 96e961d3bb..3877b631a2 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kn
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
index 69cae90889..4e51939936 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mk
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp
index e6ac98273a..45b7938569 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mn
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
index da1df4c43e..6fe43236d1 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_k
                                    BF16,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
index 21c4e87fb1..b45837fc58 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_k
                                    BF16,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
index 9d22f8feff..d1c3c74a5d 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_m
                                    BF16,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp
index 8e861f74c9..22cdf19798 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_m
                                    BF16,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
index 390b5526e3..ddcb65fef3 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kk
                                    F16,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
index 385a13cc25..2a16c40427 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kn
                                    F16,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
index 1afd432bac..317f75b03b 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mk
                                    F16,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp
index bfb4d4630f..f07e3ffa11 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mn
                                    F16,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
index 51bc64eaad..2334cd2278 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance =
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
index 4cd28d52dd..adf1ec6bee 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance =
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
index 82bba62715..7e9c4e752c 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance =
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
index 06bf4f34b5..979792265a 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance =
                                    F32,
                                    PassThrough,
                                    PassThrough,
-                                   Scale>;
+                                   Scale,
+                                   2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
index c047153bd9..9e7d6a817a 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kk
                                        F32,
                                        PassThrough,
                                        PassThrough,
-                                       Scale>;
+                                       Scale,
+                                       2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
index ee33c56223..cd84521abf 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kn
                                        F32,
                                        PassThrough,
                                        PassThrough,
-                                       Scale>;
+                                       Scale,
+                                       2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
index b62d4bb2aa..4b928075f8 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mk
                                        F32,
                                        PassThrough,
                                        PassThrough,
-                                       Scale>;
+                                       Scale,
+                                       2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp
index 346f91beb0..620454c877 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mn
                                        F32,
                                        PassThrough,
                                        PassThrough,
-                                       Scale>;
+                                       Scale,
+                                       2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
index 30cc3b9652..c7d08bad93 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance =
                                        F64,
                                        PassThrough,
                                        PassThrough,
-                                       Scale>;
+                                       Scale,
+                                       2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
index 0a36ef7ffd..4c9d85a6fc 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance =
                                        F64,
                                        PassThrough,
                                        PassThrough,
-                                       Scale>;
+                                       Scale,
+                                       2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
index 617332c2c7..c38752729f 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance =
                                        F64,
                                        PassThrough,
                                        PassThrough,
-                                       Scale>;
+                                       Scale,
+                                       2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
similarity index 94%
rename from library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
rename to library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
index 1df7f31201..c78d8d6c2f 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/2D/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 // This (ifndef) is a hack to use customized behavior for buffer load rather than using default
 // setting Don't use this hack unless absolutely necessary!
@@ -31,7 +31,8 @@ using device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance =
                                        F64,
                                        PassThrough,
                                        PassThrough,
-                                       Scale>;
+                                       Scale,
+                                       2>;
 
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
new file mode 100644
index 0000000000..b1d42c083b
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance =
+    device_contraction_kk_instance<BF16,
+                                   BF16,
+                                   F32,
+                                   BF16,
+                                   Empty_Tuple,
+                                   BF16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
new file mode 100644
index 0000000000..a7e5aad3bb
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance =
+    device_contraction_kn_instance<BF16,
+                                   BF16,
+                                   F32,
+                                   BF16,
+                                   Empty_Tuple,
+                                   BF16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
new file mode 100644
index 0000000000..0f80e2e5d5
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance =
+    device_contraction_mk_instance<BF16,
+                                   BF16,
+                                   F32,
+                                   BF16,
+                                   Empty_Tuple,
+                                   BF16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp
new file mode 100644
index 0000000000..e335b2062d
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance =
+    device_contraction_mn_instance<BF16,
+                                   BF16,
+                                   F32,
+                                   BF16,
+                                   Empty_Tuple,
+                                   BF16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
new file mode 100644
index 0000000000..729fcf9777
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance =
+    device_contraction_kk_instance<F16,
+                                   F16,
+                                   F32,
+                                   F16,
+                                   Empty_Tuple,
+                                   F16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
new file mode 100644
index 0000000000..1c0b0084c2
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance =
+    device_contraction_kn_instance<F16,
+                                   F16,
+                                   F32,
+                                   F16,
+                                   Empty_Tuple,
+                                   F16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
new file mode 100644
index 0000000000..4cae7d5f7f
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance =
+    device_contraction_mk_instance<F16,
+                                   F16,
+                                   F32,
+                                   F16,
+                                   Empty_Tuple,
+                                   F16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp
new file mode 100644
index 0000000000..263442e39a
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance =
+    device_contraction_mn_instance<F16,
+                                   F16,
+                                   F32,
+                                   F16,
+                                   Empty_Tuple,
+                                   F16,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
new file mode 100644
index 0000000000..24965e16f7
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance =
+    device_contraction_kk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   BF16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
new file mode 100644
index 0000000000..3859cfebc1
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance =
+    device_contraction_kn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   BF16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
new file mode 100644
index 0000000000..f7c6f77c56
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance =
+    device_contraction_mk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   BF16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp
new file mode 100644
index 0000000000..1eb71160c0
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance =
+    device_contraction_mn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   BF16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
new file mode 100644
index 0000000000..5a32dd09ac
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance =
+    device_contraction_kk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   F16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
new file mode 100644
index 0000000000..e773803f9d
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance =
+    device_contraction_kn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   F16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
new file mode 100644
index 0000000000..c30460d0a5
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance =
+    device_contraction_mk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   F16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp
new file mode 100644
index 0000000000..f5cde5ac20
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance =
+    device_contraction_mn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   F16,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
new file mode 100644
index 0000000000..d21b75645a
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance =
+    device_contraction_kk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
new file mode 100644
index 0000000000..d2ef6eb83f
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance =
+    device_contraction_kn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
new file mode 100644
index 0000000000..718c81c754
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance =
+    device_contraction_mk_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
new file mode 100644
index 0000000000..6ef8dae23c
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance =
+    device_contraction_mn_instance<F32,
+                                   F32,
+                                   F32,
+                                   F32,
+                                   Empty_Tuple,
+                                   F32,
+                                   F32,
+                                   PassThrough,
+                                   PassThrough,
+                                   Scale,
+                                   6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
new file mode 100644
index 0000000000..ff7d2ddb9e
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance =
+    device_contraction_f64_kk_instance<F64,
+                                       F64,
+                                       F32,
+                                       F64,
+                                       Empty_Tuple,
+                                       F64,
+                                       F32,
+                                       PassThrough,
+                                       PassThrough,
+                                       Scale,
+                                       6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
new file mode 100644
index 0000000000..19ae892505
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance =
+    device_contraction_f64_kn_instance<F64,
+                                       F64,
+                                       F32,
+                                       F64,
+                                       Empty_Tuple,
+                                       F64,
+                                       F32,
+                                       PassThrough,
+                                       PassThrough,
+                                       Scale,
+                                       6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
new file mode 100644
index 0000000000..4af5853c23
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance =
+    device_contraction_f64_mk_instance<F64,
+                                       F64,
+                                       F32,
+                                       F64,
+                                       Empty_Tuple,
+                                       F64,
+                                       F32,
+                                       PassThrough,
+                                       PassThrough,
+                                       Scale,
+                                       6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp
new file mode 100644
index 0000000000..65f0c84a99
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance =
+    device_contraction_f64_mn_instance<F64,
+                                       F64,
+                                       F32,
+                                       F64,
+                                       Empty_Tuple,
+                                       F64,
+                                       F32,
+                                       PassThrough,
+                                       PassThrough,
+                                       Scale,
+                                       6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
new file mode 100644
index 0000000000..6bf5c34970
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance =
+    device_contraction_f64_kk_instance<F64,
+                                       F64,
+                                       F64,
+                                       F64,
+                                       Empty_Tuple,
+                                       F64,
+                                       F64,
+                                       PassThrough,
+                                       PassThrough,
+                                       Scale,
+                                       6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
new file mode 100644
index 0000000000..4324914df8
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// k/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance =
+    device_contraction_f64_kn_instance<F64,
+                                       F64,
+                                       F64,
+                                       F64,
+                                       Empty_Tuple,
+                                       F64,
+                                       F64,
+                                       PassThrough,
+                                       PassThrough,
+                                       Scale,
+                                       6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
new file mode 100644
index 0000000000..0ae7215e5f
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/k/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance =
+    device_contraction_f64_mk_instance<F64,
+                                       F64,
+                                       F64,
+                                       F64,
+                                       Empty_Tuple,
+                                       F64,
+                                       F64,
+                                       PassThrough,
+                                       PassThrough,
+                                       Scale,
+                                       6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
new file mode 100644
index 0000000000..2a56a88fba
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/6D/device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/gpu/contraction/device_contraction_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance =
+    device_contraction_f64_mn_instance<F64,
+                                       F64,
+                                       F64,
+                                       F64,
+                                       Empty_Tuple,
+                                       F64,
+                                       F64,
+                                       PassThrough,
+                                       PassThrough,
+                                       Scale,
+                                       6>;
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/contraction_scale/CMakeLists.txt b/library/src/tensor_operation_instance/gpu/contraction_scale/CMakeLists.txt
index b91de832e4..dd36f88c43 100644
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/CMakeLists.txt
@@ -1,44 +1,50 @@
 # ONLY XDL_KERNELS
 set(DEVICE_CONTRACTION_SCALE_INSTANCES)
 
-# FP32
-list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp)
+list(APPEND DIMS 2 6)
 
-list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp)
+foreach(idx IN LISTS DIMS)
+    set(PREFIX ${idx}D/device_contraction_scale_m${idx}_n${idx}_k${idx})
 
-list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp)
+    # FP32
+    list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES ${PREFIX}_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp)
 
-# FP64
-list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp)
+    list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES ${PREFIX}_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance.cpp)
 
-list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp)
+    list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES ${PREFIX}_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance.cpp)
 
-# FP16
-list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp)
+    # FP64
+    list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES ${PREFIX}_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp)
 
-# BF16
-list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
-                                                device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp)
+    list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES ${PREFIX}_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance.cpp)
+
+    # FP16
+    list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES ${PREFIX}_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance.cpp)
+
+    # BF16
+    list(APPEND DEVICE_CONTRACTION_SCALE_INSTANCES ${PREFIX}_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance.cpp
+                                                    ${PREFIX}_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance.cpp)
+endforeach()
 
 add_instance_library(device_contraction_scale_instance ${DEVICE_CONTRACTION_SCALE_INSTANCES})
 
diff --git a/profiler/README.md b/profiler/README.md
index f26c90d0b3..a4daefba92 100644
--- a/profiler/README.md
+++ b/profiler/README.md
@@ -52,21 +52,23 @@ Best Perf: 1.42509 ms, 102.988 TFlops, 234.086 GB/s
 #arg1: tensor operation (contraction_bilinear=CONTRACTION+Bilinear)
 #arg2: data type (0: fp32; 1: f64; 2: f16; 3: bf16)
 #arg3: compute data type (0: fp32; 1: f64; 2: f16; 3: bf16)
-#arg4: matrix layout (0: A[m0, m1, k0, k1] * B[k0, k1, n0, n1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1];
+#arg4: Number of dimension for M, N and K (one for all)
+#arg5: matrix layout (0: A[m0, m1, k0, k1] * B[k0, k1, n0, n1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1];
 #                     1: A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1];
 #                     2: A[k0, k1, m0, m1] * B[k0, k1, n0, n1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1];
 #                     3: A[k0, k1, m0, m1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1])
-#arg5: verification (0: no; 1: yes)
-#arg6: initialization (0: no init; 1: integer value; 2: decimal value)
-#arg7: print tensor value (0: no; 1: yes)
-#arg8: time kernel (0: no, 1: yes)
-#arg9: alpha
-#arg10: beta
-#arg11 to 16: M0, M1, N0, N1, K0, K1
-#arg17 to 32: Strides for A, B, D and E (skip for default)
+#arg6: verification (0: no; 1: yes)
+#arg7: initialization (0: no init; 1: integer value; 2: decimal 
+#      value)
+#arg8: print tensor value (0: no; 1: yes)
+#arg9: time kernel (0: no, 1: yes)
+#arg10: alpha
+#arg11: beta
+#arg12 to 17/29: M0, M1, N0, N1, K0, K1
+#arg18/30 to 33/77: Strides for A, B, D and E (skip for default)
 
-################                   op  datatype  compute_datatype  layout  verify  init  log  time  alpha  beta  M0  M1  N0  N1  K0  K1
-./bin/ckProfiler contraction_bilinear         0                 0       1       0     0    0     1    1.0   1.0 128 128 128 128 128 128
+################                   op  datatype  compute_datatype  num_dim layout  verify  init  log  time  alpha  beta  M0  M1  N0  N1  K0  K1
+./bin/ckProfiler contraction_bilinear         0                 0        2      1       0     0    0     1    1.0   1.0 128 128 128 128 128 128
 ```
 
 Result (MI100)
diff --git a/profiler/include/profiler/profile_contraction_impl.hpp b/profiler/include/profiler/profile_contraction_impl.hpp
index f6e4b3f395..604032a01d 100644
--- a/profiler/include/profiler/profile_contraction_impl.hpp
+++ b/profiler/include/profiler/profile_contraction_impl.hpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -22,6 +22,7 @@
 #include "ck/library/utility/host_tensor_generator.hpp"
 #include "ck/library/utility/literals.hpp"
 #include "ck/library/reference_tensor_operation/cpu/reference_contraction.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 #include "ck/host_utility/io.hpp"
 
@@ -34,7 +35,8 @@ using Scale    = ck::tensor_operation::element_wise::Scale;
 using F32 = float;
 using F64 = double;
 
-template <typename ALayout,
+template <index_t NumDimMNK,
+          typename ALayout,
           typename BLayout,
           typename CDELayout,
           typename DataType,
@@ -104,18 +106,24 @@ int profile_contraction_impl(ck::index_t do_verification,
     e_device_buf.SetZero();
     d_device_buf.ToDevice(d_m_n.mData.data());
 
-    const std::vector<index_t> a_ms_ks_lengths = {M[0], M[1], K[0], K[1]};
-    const std::vector<index_t> b_ns_ks_lengths = {N[0], N[1], K[0], K[1]};
-    const std::vector<index_t> e_ms_ns_lengths = {M[0], M[1], N[0], N[1]};
-    const std::vector<index_t> d_m_n_lengths   = {M[0], M[1], N[0], N[1]};
+    auto merge_dims = [](const std::vector<ck::index_t>& dims01,
+                         const std::vector<ck::index_t>& dims23) {
+        std::vector<ck::index_t> dims_szt(dims01.begin(), dims01.end());
+        dims_szt.insert(dims_szt.end(), dims23.begin(), dims23.end());
+        return dims_szt;
+    };
+
+    const std::vector<index_t> a_ms_ks_lengths = merge_dims(M, K);
+    const std::vector<index_t> b_ns_ks_lengths = merge_dims(N, K);
+    const std::vector<index_t> e_ms_ns_lengths = merge_dims(M, N);
+    const std::vector<index_t> d_m_n_lengths   = merge_dims(M, N);
 
     const auto a_element_op = AElementOp{};
     const auto b_element_op = BElementOp{};
 
-    constexpr ck::index_t NumDim = 2;
-    using DeviceOp               = ck::tensor_operation::device::DeviceContractionMultipleD<NumDim,
-                                                                              NumDim,
-                                                                              NumDim,
+    using DeviceOp = ck::tensor_operation::device::DeviceContractionMultipleD<NumDimMNK,
+                                                                              NumDimMNK,
+                                                                              NumDimMNK,
                                                                               DataType,
                                                                               DataType,
                                                                               DTupleDataType,
@@ -138,9 +146,9 @@ int profile_contraction_impl(ck::index_t do_verification,
     if(do_verification)
     {
         using ReferenceGemmInstance =
-            ck::tensor_operation::host::ReferenceContraction_M2_N2_K2<NumDim,
-                                                                      NumDim,
-                                                                      NumDim,
+            ck::tensor_operation::host::ReferenceContraction_M2_N2_K2<NumDimMNK,
+                                                                      NumDimMNK,
+                                                                      NumDimMNK,
                                                                       DataType,
                                                                       DataType,
                                                                       DataType,
@@ -159,33 +167,20 @@ int profile_contraction_impl(ck::index_t do_verification,
 
         ref_invoker.Run(ref_argument);
 
-        for(size_t m0 = 0; m0 < e_m_n_host_result.mDesc.GetLengths()[0]; ++m0)
-        {
-            for(size_t m1 = 0; m1 < e_m_n_host_result.mDesc.GetLengths()[1]; ++m1)
+        e_m_n_host_result.ForEach([&](auto& self, auto idx) {
+            if constexpr(is_same<CDElementOp, Bilinear>::value)
             {
-                for(size_t n0 = 0; n0 < e_m_n_host_result.mDesc.GetLengths()[2]; ++n0)
-                {
-                    for(size_t n1 = 0; n1 < e_m_n_host_result.mDesc.GetLengths()[3]; ++n1)
-                    {
-                        if constexpr(is_same<CDElementOp, Bilinear>::value)
-                        {
-                            cde_element_op(e_m_n_host_result(m0, m1, n0, n1),
-                                           c_m_n_host_result(m0, m1, n0, n1),
-                                           d_m_n(m0, m1, n0, n1));
-                        }
-                        else if constexpr(is_same<CDElementOp, Scale>::value)
-                        {
-                            cde_element_op(e_m_n_host_result(m0, m1, n0, n1),
-                                           c_m_n_host_result(m0, m1, n0, n1));
-                        }
-                        else
-                        {
-                            static_assert("Unsupported CDElementOp in contraction profiler.");
-                        }
-                    }
-                }
+                cde_element_op(self(idx), c_m_n_host_result(idx), d_m_n(idx));
             }
-        }
+            else if constexpr(is_same<CDElementOp, Scale>::value)
+            {
+                cde_element_op(self(idx), c_m_n_host_result(idx));
+            }
+            else
+            {
+                static_assert("Unsupported CDElementOp in contraction profiler.");
+            }
+        });
     }
 
     std::string best_op_name;
@@ -242,9 +237,12 @@ int profile_contraction_impl(ck::index_t do_verification,
 
         auto invoker_ptr = op_ptr->MakeInvokerPointer();
 
-        auto nelems_m = M[0] * M[1];
-        auto nelems_n = N[0] * N[1];
-        auto nelems_k = K[0] * K[1];
+        auto nelems_m = ck::accumulate_n<ck::index_t>(
+            a_ms_ks_lengths.begin(), NumDimMNK, 1, std::multiplies<>{});
+        auto nelems_n = ck::accumulate_n<ck::index_t>(
+            b_ns_ks_lengths.begin(), NumDimMNK, 1, std::multiplies<>{});
+        auto nelems_k = ck::accumulate_n<ck::index_t>(
+            a_ms_ks_lengths.begin() + NumDimMNK, NumDimMNK, 1, std::multiplies<>{});
 
         if(op_ptr->IsSupportedArgument(argument_ptr.get()))
         {
diff --git a/profiler/include/profiler/profile_contraction_utils.hpp b/profiler/include/profiler/profile_contraction_utils.hpp
index 05ec7daf9d..adfd98a37e 100644
--- a/profiler/include/profiler/profile_contraction_utils.hpp
+++ b/profiler/include/profiler/profile_contraction_utils.hpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -48,14 +48,36 @@ inline void collect_index_params(char* argv[],
 
 // Defualt strides for row-major: {Dim1 * Dim2 * Dim3, Dim2 * Dim3, Dim3, 1}
 // Defualt strides for column-major: {Dim1, 1, Dim0 * Dim1 * Dim3, Dim0 * Dim1}
+
+// M1, 1, M0 * M1 * K1, M0 * M1
+// K0, K1, M0, M1
 inline void
 assign_default_strides(Row, std::vector<ck::index_t>& strides, std::vector<ck::index_t> dims)
 {
-    strides = {dims[1] * dims[2] * dims[3], dims[2] * dims[3], dims[3], 1};
+    ck::index_t stride = 1;
+    for(ck::index_t s = strides.size() - 1; s >= 0; s--)
+    {
+        strides[s] = stride;
+        stride *= dims[s];
+    }
 }
 
 inline void
 assign_default_strides(Col, std::vector<ck::index_t>& strides, std::vector<ck::index_t> dims)
 {
-    strides = {dims[1], 1, dims[0] * dims[1] * dims[3], dims[0] * dims[1]};
+    // Assign second half of strides
+    ck::index_t stride = 1;
+    for(ck::index_t s = strides.size() / 2 - 1; s >= 0; s--)
+    {
+        strides[s] = stride;
+        stride *= dims[s];
+    }
+
+    // Assign first half of strides
+    for(ck::index_t s = strides.size() - 1; s > static_cast<ck::index_t>(strides.size()) / 2 - 1;
+        s--)
+    {
+        strides[s] = stride;
+        stride *= dims[s];
+    }
 }
diff --git a/profiler/src/profile_contraction_bilinear.cpp b/profiler/src/profile_contraction_bilinear.cpp
index 8cb34ccc6c..990e1e1196 100644
--- a/profiler/src/profile_contraction_bilinear.cpp
+++ b/profiler/src/profile_contraction_bilinear.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <iostream>
 #include <numeric>
@@ -19,7 +19,8 @@ static void print_helper_msg()
     std::cout << "arg1: tensor operation (" OP_NAME ": " OP_DESC ")\n"
               << "arg2: data type (0: fp32; 1: f64; 2: f16; 3: bf16)\n"
               << "arg3: compute data type (0: fp32; 1: f64; 2: f16; 3: bf16)\n"
-              << "arg4: matrix layout (0: A[m0, m1, k0, k1] * B[k0, k1, n0, n1] + "
+              << "arg4: Number of dimension for M, N and K (one for all)\n"
+              << "arg5: matrix layout (0: A[m0, m1, k0, k1] * B[k0, k1, n0, n1] + "
                  "D[m0, m1, n0, n1] = E[m0, m1, n0, n1];\n"
               << "                     1: A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + "
                  "D[m0, m1, n0, n1] = E[m0, m1, n0, n1];\n"
@@ -27,23 +28,23 @@ static void print_helper_msg()
                  "D[m0, m1, n0, n1] = E[m0, m1, n0, n1];\n"
               << "                     3: A[k0, k1, m0, m1] * B[n0, n1, k0, k1] + "
                  "D[m0, m1, n0, n1] = E[m0, m1, n0, n1])\n"
-              << "arg5: verification (0: no; 1: yes)\n"
-              << "arg6: initialization (0: no init; 1: integer value; 2: decimal "
+              << "arg6: verification (0: no; 1: yes)\n"
+              << "arg7: initialization (0: no init; 1: integer value; 2: decimal "
               << "value)\n"
-              << "arg7: print tensor value (0: no; 1: yes)\n"
-              << "arg8: time kernel (0: no, 1: yes)\n"
-              << "arg9: alpha\n"
-              << "arg10: beta\n"
-              << "arg11 to 16: M0, M1, N0, N1, K0, K1\n"
-              << "arg17 to 32: Strides for A, B, D and E (skip for default)\n"
+              << "arg8: print tensor value (0: no; 1: yes)\n"
+              << "arg9: time kernel (0: no, 1: yes)\n"
+              << "arg10: alpha\n"
+              << "arg11: beta\n"
+              << "arg12 to 17/29: M0, M1, N0, N1, K0, K1\n"
+              << "arg18/30 to 33/77: Strides for A, B, D and E (skip for default)\n"
               << std::endl;
 }
 
 int profile_contraction_bilinear(int argc, char* argv[])
 {
-    const bool default_strides = argc == 17;
+    const bool default_strides = argc == 18 || 30;
 
-    if(argc != 33 && argc != 17)
+    if(argc != 34 && argc != 78 && !default_strides)
     {
         print_helper_msg();
         exit(1);
@@ -51,32 +52,33 @@ int profile_contraction_bilinear(int argc, char* argv[])
 
     const auto data_type          = static_cast<ContractionDataType>(std::stoi(argv[2]));
     const auto compute_data_type  = static_cast<ContractionComputeDataType>(std::stoi(argv[3]));
-    const auto layout             = static_cast<ContractionMatrixLayout>(std::stoi(argv[4]));
-    const bool do_verification    = std::stoi(argv[5]);
-    const ck::index_t init_method = std::stoi(argv[6]);
-    const bool do_log             = std::stoi(argv[7]);
-    const bool time_kernel        = std::stoi(argv[8]);
-    const float alpha             = std::stof(argv[9]);
-    const float beta              = std::stof(argv[10]);
+    const ck::index_t NumDimMNK   = std::stoi(argv[4]);
+    const auto layout             = static_cast<ContractionMatrixLayout>(std::stoi(argv[5]));
+    const bool do_verification    = std::stoi(argv[6]);
+    const ck::index_t init_method = std::stoi(argv[7]);
+    const bool do_log             = std::stoi(argv[8]);
+    const bool time_kernel        = std::stoi(argv[9]);
+    const float alpha             = std::stof(argv[10]);
+    const float beta              = std::stof(argv[11]);
 
     std::vector<ck::index_t> M;
     std::vector<ck::index_t> N;
     std::vector<ck::index_t> K;
-    const ck::index_t dims_arg_num = 11;
-    collect_index_params(argv, M, dims_arg_num, 2);
-    collect_index_params(argv, N, dims_arg_num + 2, 2);
-    collect_index_params(argv, K, dims_arg_num + 4, 2);
+    const ck::index_t dims_arg_num = 12;
+    collect_index_params(argv, M, dims_arg_num, NumDimMNK);
+    collect_index_params(argv, N, dims_arg_num + NumDimMNK, NumDimMNK);
+    collect_index_params(argv, K, dims_arg_num + NumDimMNK * 2, NumDimMNK);
 
-    std::vector<ck::index_t> StridesA;
-    std::vector<ck::index_t> StridesB;
-    std::vector<ck::index_t> StridesE;
-    std::vector<ck::index_t> StridesD;
+    std::vector<ck::index_t> StridesA(NumDimMNK * 2);
+    std::vector<ck::index_t> StridesB(NumDimMNK * 2);
+    std::vector<ck::index_t> StridesE(NumDimMNK * 2);
+    std::vector<ck::index_t> StridesD(NumDimMNK * 2);
     if(!default_strides)
     {
-        collect_index_params(argv, StridesA, dims_arg_num + 6, 4);
-        collect_index_params(argv, StridesB, dims_arg_num + 10, 4);
-        collect_index_params(argv, StridesE, dims_arg_num + 14, 4);
-        collect_index_params(argv, StridesD, dims_arg_num + 18, 4);
+        collect_index_params(argv, StridesA, dims_arg_num + NumDimMNK * 3, NumDimMNK * 2);
+        collect_index_params(argv, StridesB, dims_arg_num + NumDimMNK * 5, NumDimMNK * 2);
+        collect_index_params(argv, StridesE, dims_arg_num + NumDimMNK * 7, NumDimMNK * 2);
+        collect_index_params(argv, StridesD, dims_arg_num + NumDimMNK * 9, NumDimMNK * 2);
     }
 
     using F16  = ck::half_t;
@@ -95,31 +97,71 @@ int profile_contraction_bilinear(int argc, char* argv[])
 
             if(default_strides)
             {
-                assign_default_strides(a_layout, StridesA, {M[0], M[1], K[0], K[1]});
-                assign_default_strides(b_layout, StridesB, {N[0], N[1], K[0], K[1]});
-                assign_default_strides(cde_layout, StridesE, {M[0], M[1], N[0], N[1]});
-                assign_default_strides(cde_layout, StridesD, {M[0], M[1], N[0], N[1]});
-            }
-            bool pass = ck::profiler::profile_contraction_impl<ALayout,
-                                                               BLayout,
-                                                               CDELayout,
-                                                               DataType,
-                                                               ComputeDataType,
-                                                               ck::Tuple<DataType>,
-                                                               Bilinear>(do_verification,
-                                                                         init_method,
-                                                                         do_log,
-                                                                         time_kernel,
-                                                                         Bilinear{alpha, beta},
-                                                                         M,
-                                                                         N,
-                                                                         K,
-                                                                         StridesA,
-                                                                         StridesB,
-                                                                         StridesE,
-                                                                         StridesD);
+                auto merge_dims = [](const std::vector<ck::index_t>& dims01,
+                                     const std::vector<ck::index_t>& dims23) {
+                    std::vector<ck::index_t> dims_szt(dims01.begin(), dims01.end());
+                    dims_szt.insert(dims_szt.end(), dims23.begin(), dims23.end());
+                    return dims_szt;
+                };
 
-            return pass;
+                assign_default_strides(a_layout, StridesA, merge_dims(M, K));
+                assign_default_strides(b_layout, StridesB, merge_dims(N, K));
+                assign_default_strides(cde_layout, StridesE, merge_dims(M, N));
+                assign_default_strides(cde_layout, StridesD, merge_dims(M, N));
+            }
+            if(NumDimMNK == 2)
+            {
+                bool pass = ck::profiler::profile_contraction_impl<2,
+                                                                   ALayout,
+                                                                   BLayout,
+                                                                   CDELayout,
+                                                                   DataType,
+                                                                   ComputeDataType,
+                                                                   ck::Tuple<DataType>,
+                                                                   Bilinear>(do_verification,
+                                                                             init_method,
+                                                                             do_log,
+                                                                             time_kernel,
+                                                                             Bilinear{alpha, beta},
+                                                                             M,
+                                                                             N,
+                                                                             K,
+                                                                             StridesA,
+                                                                             StridesB,
+                                                                             StridesE,
+                                                                             StridesD);
+
+                return pass;
+            }
+            else if(NumDimMNK == 6)
+            {
+                bool pass = ck::profiler::profile_contraction_impl<6,
+                                                                   ALayout,
+                                                                   BLayout,
+                                                                   CDELayout,
+                                                                   DataType,
+                                                                   ComputeDataType,
+                                                                   ck::Tuple<DataType>,
+                                                                   Bilinear>(do_verification,
+                                                                             init_method,
+                                                                             do_log,
+                                                                             time_kernel,
+                                                                             Bilinear{alpha, beta},
+                                                                             M,
+                                                                             N,
+                                                                             K,
+                                                                             StridesA,
+                                                                             StridesB,
+                                                                             StridesE,
+                                                                             StridesD);
+
+                return pass;
+            }
+            else
+            {
+                throw std::runtime_error("Not supported NumDimMNK");
+                return false;
+            }
         };
 
     auto run_profile_for_datatype = [&](auto type, auto compute_type) {
diff --git a/profiler/src/profile_contraction_scale.cpp b/profiler/src/profile_contraction_scale.cpp
index ca9c199988..85252eaa37 100644
--- a/profiler/src/profile_contraction_scale.cpp
+++ b/profiler/src/profile_contraction_scale.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <iostream>
 #include <numeric>
@@ -19,7 +19,8 @@ static void print_helper_msg()
     std::cout << "arg1: tensor operation (" OP_NAME ": " OP_DESC ")\n"
               << "arg2: data type (0: fp32; 1: f64; 2: f16; 3: bf16)\n"
               << "arg3: compute data type (0: fp32; 1: f64; 2: f16; 3: bf16)\n"
-              << "arg4: matrix layout (0: A[m0, m1, k0, k1] * B[k0, k1, n0, n1] + "
+              << "arg4: Number of dimension for M, N and K (one for all)\n"
+              << "arg5: matrix layout (0: A[m0, m1, k0, k1] * B[k0, k1, n0, n1] + "
                  "D[m0, m1, n0, n1] = E[m0, m1, n0, n1];\n"
               << "                     1: A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + "
                  "D[m0, m1, n0, n1] = E[m0, m1, n0, n1];\n"
@@ -27,22 +28,22 @@ static void print_helper_msg()
                  "D[m0, m1, n0, n1] = E[m0, m1, n0, n1];\n"
               << "                     3: A[k0, k1, m0, m1] * B[n0, n1, k0, k1] + "
                  "D[m0, m1, n0, n1] = E[m0, m1, n0, n1])\n"
-              << "arg5: verification (0: no; 1: yes)\n"
-              << "arg6: initialization (0: no init; 1: integer value; 2: decimal "
+              << "arg6: verification (0: no; 1: yes)\n"
+              << "arg7: initialization (0: no init; 1: integer value; 2: decimal "
               << "value)\n"
-              << "arg7: print tensor value (0: no; 1: yes)\n"
-              << "arg8: time kernel (0: no, 1: yes)\n"
-              << "arg9: alpha\n"
-              << "arg10 to 15: M0, M1, N0, N1, K0, K1\n"
-              << "arg16 to 31: Strides for A, B, D and E (skip for default)\n"
+              << "arg8: print tensor value (0: no; 1: yes)\n"
+              << "arg9: time kernel (0: no, 1: yes)\n"
+              << "arg10: alpha\n"
+              << "arg11 to 16/28: M0, M1, N0, N1, K0, K1\n"
+              << "arg17/29 to 32/63: Strides for A, B, E (skip for default)\n"
               << std::endl;
 }
 
 int profile_contraction_scale(int argc, char* argv[])
 {
-    const bool default_strides = argc == 16;
+    const bool default_strides = argc == 17 || argc == 29;
 
-    if(argc != 32 && argc != 16)
+    if(argc != 29 && argc != 65 && !default_strides)
     {
         print_helper_msg();
         exit(1);
@@ -50,31 +51,30 @@ int profile_contraction_scale(int argc, char* argv[])
 
     const auto data_type          = static_cast<ContractionDataType>(std::stoi(argv[2]));
     const auto compute_data_type  = static_cast<ContractionComputeDataType>(std::stoi(argv[3]));
-    const auto layout             = static_cast<ContractionMatrixLayout>(std::stoi(argv[4]));
-    const bool do_verification    = std::stoi(argv[5]);
-    const ck::index_t init_method = std::stoi(argv[6]);
-    const bool do_log             = std::stoi(argv[7]);
-    const bool time_kernel        = std::stoi(argv[8]);
-    const float alpha             = std::stof(argv[9]);
+    const ck::index_t NumDimMNK   = std::stoi(argv[4]);
+    const auto layout             = static_cast<ContractionMatrixLayout>(std::stoi(argv[5]));
+    const bool do_verification    = std::stoi(argv[6]);
+    const ck::index_t init_method = std::stoi(argv[7]);
+    const bool do_log             = std::stoi(argv[8]);
+    const bool time_kernel        = std::stoi(argv[9]);
+    const float alpha             = std::stof(argv[10]);
 
     std::vector<ck::index_t> M;
     std::vector<ck::index_t> N;
     std::vector<ck::index_t> K;
-    const ck::index_t dims_arg_num = 10;
-    collect_index_params(argv, M, dims_arg_num, 2);
-    collect_index_params(argv, N, dims_arg_num + 2, 2);
-    collect_index_params(argv, K, dims_arg_num + 4, 2);
+    const ck::index_t dims_arg_num = 11;
+    collect_index_params(argv, M, dims_arg_num, NumDimMNK);
+    collect_index_params(argv, N, dims_arg_num + NumDimMNK, NumDimMNK);
+    collect_index_params(argv, K, dims_arg_num + NumDimMNK * 2, NumDimMNK);
 
-    std::vector<ck::index_t> StridesA;
-    std::vector<ck::index_t> StridesB;
-    std::vector<ck::index_t> StridesE;
-    std::vector<ck::index_t> StridesD;
+    std::vector<ck::index_t> StridesA(NumDimMNK * 2);
+    std::vector<ck::index_t> StridesB(NumDimMNK * 2);
+    std::vector<ck::index_t> StridesE(NumDimMNK * 2);
     if(!default_strides)
     {
-        collect_index_params(argv, StridesA, dims_arg_num + 6, 4);
-        collect_index_params(argv, StridesB, dims_arg_num + 10, 4);
-        collect_index_params(argv, StridesE, dims_arg_num + 14, 4);
-        collect_index_params(argv, StridesD, dims_arg_num + 18, 4);
+        collect_index_params(argv, StridesA, dims_arg_num + NumDimMNK * 3, NumDimMNK * 2);
+        collect_index_params(argv, StridesB, dims_arg_num + NumDimMNK * 5, NumDimMNK * 2);
+        collect_index_params(argv, StridesE, dims_arg_num + NumDimMNK * 7, NumDimMNK * 2);
     }
 
     using F16  = ck::half_t;
@@ -93,32 +93,71 @@ int profile_contraction_scale(int argc, char* argv[])
 
             if(default_strides)
             {
-                assign_default_strides(a_layout, StridesA, {M[0], M[1], K[0], K[1]});
-                assign_default_strides(b_layout, StridesB, {N[0], N[1], K[0], K[1]});
-                assign_default_strides(cde_layout, StridesE, {M[0], M[1], N[0], N[1]});
-                assign_default_strides(cde_layout, StridesD, {M[0], M[1], N[0], N[1]});
+                auto merge_dims = [](const std::vector<ck::index_t>& dims01,
+                                     const std::vector<ck::index_t>& dims23) {
+                    std::vector<ck::index_t> dims_szt(dims01.begin(), dims01.end());
+                    dims_szt.insert(dims_szt.end(), dims23.begin(), dims23.end());
+                    return dims_szt;
+                };
+
+                assign_default_strides(a_layout, StridesA, merge_dims(M, K));
+                assign_default_strides(b_layout, StridesB, merge_dims(N, K));
+                assign_default_strides(cde_layout, StridesE, merge_dims(M, N));
             }
 
-            bool pass = ck::profiler::profile_contraction_impl<ALayout,
-                                                               BLayout,
-                                                               CDELayout,
-                                                               DataType,
-                                                               ComputeDataType,
-                                                               ck::Tuple<>,
-                                                               Scale>(do_verification,
-                                                                      init_method,
-                                                                      do_log,
-                                                                      time_kernel,
-                                                                      Scale{alpha},
-                                                                      M,
-                                                                      N,
-                                                                      K,
-                                                                      StridesA,
-                                                                      StridesB,
-                                                                      StridesE,
-                                                                      StridesD);
+            if(NumDimMNK == 2)
+            {
+                bool pass = ck::profiler::profile_contraction_impl<2,
+                                                                   ALayout,
+                                                                   BLayout,
+                                                                   CDELayout,
+                                                                   DataType,
+                                                                   ComputeDataType,
+                                                                   ck::Tuple<>,
+                                                                   Scale>(do_verification,
+                                                                          init_method,
+                                                                          do_log,
+                                                                          time_kernel,
+                                                                          Scale{alpha},
+                                                                          M,
+                                                                          N,
+                                                                          K,
+                                                                          StridesA,
+                                                                          StridesB,
+                                                                          StridesE,
+                                                                          StridesE);
 
-            return pass;
+                return pass;
+            }
+            else if(NumDimMNK == 6)
+            {
+                bool pass = ck::profiler::profile_contraction_impl<6,
+                                                                   ALayout,
+                                                                   BLayout,
+                                                                   CDELayout,
+                                                                   DataType,
+                                                                   ComputeDataType,
+                                                                   ck::Tuple<>,
+                                                                   Scale>(do_verification,
+                                                                          init_method,
+                                                                          do_log,
+                                                                          time_kernel,
+                                                                          Scale{alpha},
+                                                                          M,
+                                                                          N,
+                                                                          K,
+                                                                          StridesA,
+                                                                          StridesB,
+                                                                          StridesE,
+                                                                          StridesE);
+
+                return pass;
+            }
+            else
+            {
+                throw std::runtime_error("Not supported NumDimMNK");
+                return false;
+            }
         };
 
     auto run_profile_for_datatype = [&](auto type, auto compute_type) {
diff --git a/test/contraction/test_contraction_interface_xdl.cpp b/test/contraction/test_contraction_interface_xdl.cpp
index d6b290e20a..58232d209c 100644
--- a/test/contraction/test_contraction_interface_xdl.cpp
+++ b/test/contraction/test_contraction_interface_xdl.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <stdexcept>
 #include <vector>
@@ -125,18 +125,6 @@ class ContractionDeviceOpWrapper
     }
 };
 
-TEST(TestContractionInterface, IncorrectNumDims)
-{
-    std::vector<std::vector<ck::index_t>> Dims    = {{4, 4}, {4, 4, 4, 4}, {4, 4, 4, 4, 4, 4}};
-    std::vector<std::vector<ck::index_t>> Strides = {{1, 1}, {1, 1, 1, 1}, {1, 1, 1, 1, 1, 1}};
-    ContractionDeviceOpWrapper<F32, F32, F32, F32, 1> wrapper_1d;
-    ContractionDeviceOpWrapper<F32, F32, F32, F32, 2> wrapper_2d;
-    ContractionDeviceOpWrapper<F32, F32, F32, F32, 3> wrapper_3d;
-    EXPECT_FALSE(wrapper_1d.IsSupportedInstance(Dims[0], Strides[0]));
-    EXPECT_TRUE(wrapper_2d.IsSupportedInstance(Dims[1], Strides[1]));
-    EXPECT_FALSE(wrapper_3d.IsSupportedInstance(Dims[2], Strides[2]));
-}
-
 TEST(TestContractionInterface, IncorrectDataTypes)
 {
     std::vector<ck::index_t> Dims    = {4, 4, 4, 4};
diff --git a/test/contraction/test_contraction_xdl.cpp b/test/contraction/test_contraction_xdl.cpp
index 958d5be38c..c84375b1db 100644
--- a/test/contraction/test_contraction_xdl.cpp
+++ b/test/contraction/test_contraction_xdl.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <cstdlib>
 #include <iostream>
@@ -23,8 +23,11 @@ using Col = ck::tensor_layout::gemm::ColumnMajor;
 using Bilinear = ck::tensor_operation::element_wise::Bilinear;
 using Scale    = ck::tensor_operation::element_wise::Scale;
 
+template <ck::index_t NDims>
 struct Dimensions
 {
+    constexpr static ck::index_t NumDimMNK = NDims;
+
     std::vector<ck::index_t> M;
     std::vector<ck::index_t> N;
     std::vector<ck::index_t> K;
@@ -42,53 +45,58 @@ class TestContraction : public ::testing::Test
     using ComputeDataType = std::tuple_element_t<5, Tuple>;
     using CDElementOp     = std::tuple_element_t<6, Tuple>;
 
-    std::vector<Dimensions> dimension_list = {{{32, 32}, {32, 32}, {32, 32}},
-                                              {{16, 16}, {32, 32}, {16, 16}}};
-
     std::vector<ck::index_t> init_methods = {1, 2};
     std::unique_ptr<CDElementOp> p_cd_element_op;
 
-    void Run()
+    template <ck::index_t NumDim>
+    void Run(Dimensions<NumDim> dimension_params)
     {
-        for(auto& dimension_params : dimension_list)
+        constexpr ck::index_t NumDimMNK = ck::remove_cvref_t<decltype(dimension_params)>::NumDimMNK;
+
+        std::vector<ck::index_t> StridesA(2 * NumDim);
+        std::vector<ck::index_t> StridesB(2 * NumDim);
+        std::vector<ck::index_t> StridesC(2 * NumDim);
+        std::vector<ck::index_t> StridesD(2 * NumDim);
+
+        const auto& M = dimension_params.M;
+        const auto& N = dimension_params.N;
+        const auto& K = dimension_params.K;
+
+        auto merge_dims = [](const std::vector<ck::index_t>& dims01,
+                             const std::vector<ck::index_t>& dims23) {
+            std::vector<ck::index_t> dims_szt(dims01.begin(), dims01.end());
+            dims_szt.insert(dims_szt.end(), dims23.begin(), dims23.end());
+            return dims_szt;
+        };
+
+        assign_default_strides(ALayout{}, StridesA, merge_dims(M, K));
+        assign_default_strides(BLayout{}, StridesB, merge_dims(N, K));
+        assign_default_strides(CDLayout{}, StridesC, merge_dims(M, N));
+        assign_default_strides(CDLayout{}, StridesD, merge_dims(M, N));
+
+        for(const ck::index_t init_method : init_methods)
         {
-            std::vector<ck::index_t> StridesA;
-            std::vector<ck::index_t> StridesB;
-            std::vector<ck::index_t> StridesC;
-            std::vector<ck::index_t> StridesD;
-
-            const auto& M = dimension_params.M;
-            const auto& N = dimension_params.N;
-            const auto& K = dimension_params.K;
-
-            assign_default_strides(ALayout{}, StridesA, {M[0], M[1], K[0], K[1]});
-            assign_default_strides(BLayout{}, StridesB, {N[0], N[1], K[0], K[1]});
-            assign_default_strides(CDLayout{}, StridesC, {M[0], M[1], N[0], N[1]});
-            assign_default_strides(CDLayout{}, StridesD, {M[0], M[1], N[0], N[1]});
-
-            for(const ck::index_t init_method : init_methods)
-            {
-                bool pass =
-                    ck::profiler::profile_contraction_impl<ALayout,
-                                                           BLayout,
-                                                           CDLayout,
-                                                           DataType,
-                                                           ComputeDataType,
-                                                           DTupleDataType,
-                                                           CDElementOp>(true /*do_verification*/,
-                                                                        init_method,
-                                                                        false /*do_logs*/,
-                                                                        false /*time_kernel*/,
-                                                                        *p_cd_element_op,
-                                                                        dimension_params.M,
-                                                                        dimension_params.N,
-                                                                        dimension_params.K,
-                                                                        StridesA,
-                                                                        StridesB,
-                                                                        StridesC,
-                                                                        StridesD);
-                EXPECT_TRUE(pass);
-            }
+            bool pass =
+                ck::profiler::profile_contraction_impl<NumDimMNK,
+                                                       ALayout,
+                                                       BLayout,
+                                                       CDLayout,
+                                                       DataType,
+                                                       ComputeDataType,
+                                                       DTupleDataType,
+                                                       CDElementOp>(true /*do_verification*/,
+                                                                    init_method,
+                                                                    false /*do_logs*/,
+                                                                    false /*time_kernel*/,
+                                                                    *p_cd_element_op,
+                                                                    dimension_params.M,
+                                                                    dimension_params.N,
+                                                                    dimension_params.K,
+                                                                    StridesA,
+                                                                    StridesB,
+                                                                    StridesC,
+                                                                    StridesD);
+            EXPECT_TRUE(pass);
         }
     }
 };
@@ -122,17 +130,31 @@ TYPED_TEST_SUITE(TestContractionScale, ScaleKernelTypes);
 TYPED_TEST(TestContractionBilinear, bilinear)
 {
     this->p_cd_element_op = std::make_unique<Bilinear>(1.f, 1.f);
-    this->Run();
+    this->template Run<6>({{2, 3, 2, 3, 2, 3}, {2, 3, 2, 3, 2, 3}, {2, 2, 2, 2, 2, 4}});
+    this->template Run<6>({{1, 1, 1, 3, 2, 3}, {1, 1, 1, 3, 2, 3}, {1, 1, 1, 2, 2, 4}});
+    this->template Run<2>({{16, 8}, {16, 8}, {16, 8}});
+    this->template Run<2>({{8, 16}, {16, 8}, {8, 16}});
+
     this->p_cd_element_op = std::make_unique<Bilinear>(-0.5f, 0.5f);
-    this->Run();
+    this->template Run<6>({{2, 3, 2, 3, 2, 3}, {2, 3, 2, 3, 2, 3}, {2, 2, 2, 2, 2, 4}});
+    this->template Run<6>({{1, 1, 1, 3, 2, 3}, {1, 1, 1, 3, 2, 3}, {1, 1, 1, 2, 2, 4}});
+    this->template Run<2>({{16, 8}, {16, 8}, {16, 8}});
+    this->template Run<2>({{8, 16}, {16, 8}, {8, 16}});
 }
 
 TYPED_TEST(TestContractionScale, scale)
 {
     this->p_cd_element_op = std::make_unique<Scale>(1.f);
-    this->Run();
+    this->template Run<6>({{2, 3, 2, 3, 2, 3}, {2, 3, 2, 3, 2, 3}, {2, 2, 2, 2, 2, 4}});
+    this->template Run<6>({{1, 1, 1, 3, 2, 3}, {1, 1, 1, 3, 2, 3}, {1, 1, 1, 2, 2, 4}});
+    this->template Run<2>({{16, 8}, {16, 8}, {16, 8}});
+    this->template Run<2>({{8, 16}, {16, 8}, {8, 16}});
+
     this->p_cd_element_op = std::make_unique<Scale>(0.5f);
-    this->Run();
+    this->template Run<6>({{2, 3, 2, 3, 2, 3}, {2, 3, 2, 3, 2, 3}, {2, 2, 2, 2, 2, 4}});
+    this->template Run<6>({{1, 1, 1, 3, 2, 3}, {1, 1, 1, 3, 2, 3}, {1, 1, 1, 2, 2, 4}});
+    this->template Run<2>({{16, 8}, {16, 8}, {16, 8}});
+    this->template Run<2>({{8, 16}, {16, 8}, {8, 16}});
 }
 
 template <typename Tuple>
@@ -165,15 +187,29 @@ TYPED_TEST_SUITE(TestContractionScaleMixedPrecision, ScaleKernelTypesMixedPrecis
 TYPED_TEST(TestContractionBilinearMixedPrecision, bilinear)
 {
     this->p_cd_element_op = std::make_unique<Bilinear>(1.f, 1.f);
-    this->Run();
+    this->template Run<6>({{2, 3, 2, 3, 2, 3}, {2, 3, 2, 3, 2, 3}, {2, 2, 2, 2, 2, 4}});
+    this->template Run<6>({{1, 1, 1, 3, 2, 3}, {1, 1, 1, 3, 2, 3}, {1, 1, 1, 2, 2, 4}});
+    this->template Run<2>({{16, 8}, {16, 8}, {16, 8}});
+    this->template Run<2>({{8, 16}, {16, 8}, {8, 16}});
+
     this->p_cd_element_op = std::make_unique<Bilinear>(-0.5f, 0.5f);
-    this->Run();
+    this->template Run<6>({{2, 3, 2, 3, 2, 3}, {2, 3, 2, 3, 2, 3}, {2, 2, 2, 2, 2, 4}});
+    this->template Run<6>({{1, 1, 1, 3, 2, 3}, {1, 1, 1, 3, 2, 3}, {1, 1, 1, 2, 2, 4}});
+    this->template Run<2>({{16, 8}, {16, 8}, {16, 8}});
+    this->template Run<2>({{8, 16}, {16, 8}, {8, 16}});
 }
 
 TYPED_TEST(TestContractionScaleMixedPrecision, scale)
 {
     this->p_cd_element_op = std::make_unique<Scale>(1.f);
-    this->Run();
+    this->template Run<6>({{2, 3, 2, 3, 2, 3}, {2, 3, 2, 3, 2, 3}, {2, 2, 2, 2, 2, 4}});
+    this->template Run<6>({{1, 1, 1, 3, 2, 3}, {1, 1, 1, 3, 2, 3}, {1, 1, 1, 2, 2, 4}});
+    this->template Run<2>({{16, 8}, {16, 8}, {16, 8}});
+    this->template Run<2>({{8, 16}, {16, 8}, {8, 16}});
+
     this->p_cd_element_op = std::make_unique<Scale>(0.5f);
-    this->Run();
+    this->template Run<6>({{2, 3, 2, 3, 2, 3}, {2, 3, 2, 3, 2, 3}, {2, 2, 2, 2, 2, 4}});
+    this->template Run<6>({{1, 1, 1, 3, 2, 3}, {1, 1, 1, 3, 2, 3}, {1, 1, 1, 2, 2, 4}});
+    this->template Run<2>({{16, 8}, {16, 8}, {16, 8}});
+    this->template Run<2>({{8, 16}, {16, 8}, {8, 16}});
 }

From 381d44aa60501098baef69bc78add2c24adb4ba4 Mon Sep 17 00:00:00 2001
From: zjing14 <zhangjing14@gmail.com>
Date: Tue, 9 Apr 2024 21:32:02 -0500
Subject: [PATCH 4/7] add yigex (#1230)

---
 .github/CODEOWNERS | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/CODEOWNERS b/.github/CODEOWNERS
index 37407cebf1..01e3bee0b6 100644
--- a/.github/CODEOWNERS
+++ b/.github/CODEOWNERS
@@ -1,4 +1,4 @@
-* @zjing14 @junliume @illsilin @carlushuang @aosewski
+* @zjing14 @junliume @illsilin @carlushuang @aosewski @yigex
 # Documentation files
 docs/* @ROCm/rocm-documentation
 *.md @ROCm/rocm-documentation

From b2735caf465fd33dc900169f0be5588fe07cfa55 Mon Sep 17 00:00:00 2001
From: "dependabot[bot]" <49699333+dependabot[bot]@users.noreply.github.com>
Date: Thu, 11 Apr 2024 07:39:44 -0700
Subject: [PATCH 5/7] Bump rocm-docs-core from 0.38.0 to 0.38.1 in /docs/sphinx
 (#1232)

Bumps [rocm-docs-core](https://github.com/RadeonOpenCompute/rocm-docs-core) from 0.38.0 to 0.38.1.
- [Release notes](https://github.com/RadeonOpenCompute/rocm-docs-core/releases)
- [Changelog](https://github.com/ROCm/rocm-docs-core/blob/develop/CHANGELOG.md)
- [Commits](https://github.com/RadeonOpenCompute/rocm-docs-core/compare/v0.38.0...v0.38.1)

---
updated-dependencies:
- dependency-name: rocm-docs-core
  dependency-type: direct:production
  update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
---
 docs/sphinx/requirements.in  | 2 +-
 docs/sphinx/requirements.txt | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/sphinx/requirements.in b/docs/sphinx/requirements.in
index 2b28fcdd3c..a854542439 100644
--- a/docs/sphinx/requirements.in
+++ b/docs/sphinx/requirements.in
@@ -1,2 +1,2 @@
-rocm-docs-core==0.38.0
+rocm-docs-core==0.38.1
 sphinxcontrib-bibtex==2.6.2
diff --git a/docs/sphinx/requirements.txt b/docs/sphinx/requirements.txt
index 335d6e5e00..801726ed60 100644
--- a/docs/sphinx/requirements.txt
+++ b/docs/sphinx/requirements.txt
@@ -111,7 +111,7 @@ requests==2.31.0
     # via
     #   pygithub
     #   sphinx
-rocm-docs-core==0.38.0
+rocm-docs-core==0.38.1
     # via -r requirements.in
 six==1.16.0
     # via

From bbefc12a261b0cbd6efb3e036db1d9a0c0fefe4b Mon Sep 17 00:00:00 2001
From: Rostyslav Geyyer <46627076+geyyer@users.noreply.github.com>
Date: Thu, 11 Apr 2024 10:35:00 -0500
Subject: [PATCH 6/7] Add instances for conv_scale with bf8@fp8->fp8 (#1231)

* Add instances

* Add example

* Add profiler mode

* Add client example
---
 client_example/16_convnd_fwd/CMakeLists.txt   |  3 +
 .../16_convnd_fwd/conv3d_fwd_bf8_fp8.cpp      | 50 +++++++++++
 example/09_convnd_fwd/CMakeLists.txt          |  1 +
 .../09_convnd_fwd/convnd_fwd_xdl_bf8_fp8.cpp  | 83 +++++++++++++++++++
 .../device_grouped_conv_fwd_xdl_instance.hpp  | 36 ++++++++
 .../gpu/grouped_convolution_forward.hpp       |  8 ++
 .../gpu/grouped_convolution_forward_xdl.inc   | 18 ++++
 .../gpu/grouped_conv3d_fwd/CMakeLists.txt     |  2 +
 ..._ndhwgc_gkzyxc_ndhwgk_bf8_fp8_instance.cpp | 54 ++++++++++++
 profiler/src/profile_grouped_conv_fwd.cpp     |  8 +-
 10 files changed, 262 insertions(+), 1 deletion(-)
 create mode 100644 client_example/16_convnd_fwd/conv3d_fwd_bf8_fp8.cpp
 create mode 100644 example/09_convnd_fwd/convnd_fwd_xdl_bf8_fp8.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_fp8_instance.cpp

diff --git a/client_example/16_convnd_fwd/CMakeLists.txt b/client_example/16_convnd_fwd/CMakeLists.txt
index 808693b632..23311b4024 100644
--- a/client_example/16_convnd_fwd/CMakeLists.txt
+++ b/client_example/16_convnd_fwd/CMakeLists.txt
@@ -20,6 +20,9 @@ endif()
 if((DTYPES MATCHES "fp8" AND DTYPES MATCHES "bf8") OR NOT DEFINED DTYPES)
     add_executable(client_conv3d_fwd_fp8_bf8 conv3d_fwd_fp8_bf8.cpp)
     target_link_libraries(client_conv3d_fwd_fp8_bf8 PRIVATE composable_kernel::device_conv_operations)
+
+    add_executable(client_conv3d_fwd_bf8_fp8 conv3d_fwd_bf8_fp8.cpp)
+    target_link_libraries(client_conv3d_fwd_bf8_fp8 PRIVATE composable_kernel::device_conv_operations)
 endif()
 
 if((DTYPES MATCHES "fp32") OR NOT DEFINED DTYPES)
diff --git a/client_example/16_convnd_fwd/conv3d_fwd_bf8_fp8.cpp b/client_example/16_convnd_fwd/conv3d_fwd_bf8_fp8.cpp
new file mode 100644
index 0000000000..b195d87bbc
--- /dev/null
+++ b/client_example/16_convnd_fwd/conv3d_fwd_bf8_fp8.cpp
@@ -0,0 +1,50 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "common.hpp"
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+
+using InDataType  = ck::bf8_t;
+using WeiDataType = ck::f8_t;
+using OutDataType = ck::f8_t;
+
+using InLayout  = ck::tensor_layout::convolution::NDHWGC;
+using WeiLayout = ck::tensor_layout::convolution::GKZYXC;
+using OutLayout = ck::tensor_layout::convolution::NDHWGK;
+
+using AComputeType = ck::bf8_t;
+using BComputeType = ck::f8_t;
+
+static constexpr ck::index_t NumDimSpatial = 3;
+static constexpr ck::index_t G             = 1;
+static constexpr ck::index_t N             = 64;
+static constexpr ck::index_t K             = 128;
+static constexpr ck::index_t C             = 64;
+static constexpr ck::index_t Z             = 3;
+static constexpr ck::index_t Y             = 3;
+static constexpr ck::index_t X             = 3;
+static constexpr ck::index_t Di            = 28;
+static constexpr ck::index_t Hi            = 28;
+static constexpr ck::index_t Wi            = 3;
+static constexpr ck::index_t Do            = 28;
+static constexpr ck::index_t Ho            = 28;
+static constexpr ck::index_t Wo            = 3;
+
+int main()
+{
+    return run_grouped_conv_fwd<NumDimSpatial,
+                                InDataType,
+                                WeiDataType,
+                                OutDataType,
+                                InLayout,
+                                WeiLayout,
+                                OutLayout,
+                                3,
+                                AComputeType,
+                                BComputeType>(
+               {N, Di, Hi, Wi, G, C}, {G, K, Z, Y, X, C}, {N, Do, Ho, Wo, G, K})
+               ? EXIT_SUCCESS
+               : EXIT_FAILURE;
+}
diff --git a/example/09_convnd_fwd/CMakeLists.txt b/example/09_convnd_fwd/CMakeLists.txt
index 778e81872d..8a295d14c4 100644
--- a/example/09_convnd_fwd/CMakeLists.txt
+++ b/example/09_convnd_fwd/CMakeLists.txt
@@ -7,6 +7,7 @@ add_example_executable(example_convnd_fwd_xdl_fp64 convnd_fwd_xdl_fp64.cpp)
 add_example_executable(example_convnd_fwd_xdl_bf8 convnd_fwd_xdl_bf8.cpp)
 add_example_executable(example_convnd_fwd_xdl_fp16_comp_fp8 convnd_fwd_xdl_fp16_comp_fp8.cpp)
 add_example_executable(example_convnd_fwd_xdl_fp8_bf8 convnd_fwd_xdl_fp8_bf8.cpp)
+add_example_executable(example_convnd_fwd_xdl_bf8_fp8 convnd_fwd_xdl_bf8_fp8.cpp)
 add_example_executable(example_convnd_fwd_dl_fp16 convnd_fwd_dl_fp16.cpp)
 add_example_executable(example_convnd_fwd_dl_fp32 convnd_fwd_dl_fp32.cpp)
 add_example_executable(example_convnd_fwd_dl_int8 convnd_fwd_dl_int8.cpp)
diff --git a/example/09_convnd_fwd/convnd_fwd_xdl_bf8_fp8.cpp b/example/09_convnd_fwd/convnd_fwd_xdl_bf8_fp8.cpp
new file mode 100644
index 0000000000..9eba00993a
--- /dev/null
+++ b/example/09_convnd_fwd/convnd_fwd_xdl_bf8_fp8.cpp
@@ -0,0 +1,83 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "convnd_fwd_common.hpp"
+
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_abd_xdl_cshuffle.hpp"
+
+#include "ck/library/utility/convolution_host_tensor_descriptor_helper.hpp"
+
+using InDataType       = ck::bf8_t;
+using WeiDataType      = ck::f8_t;
+using AccDataType      = float;
+using CShuffleDataType = ck::f8_t;
+using OutDataType      = ck::f8_t;
+using AComputeType     = ck::bf8_t;
+using BComputeType     = ck::f8_t;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using InElementOp  = ck::tensor_operation::element_wise::PassThrough;
+using WeiElementOp = ck::tensor_operation::element_wise::PassThrough;
+using OutElementOp = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto ConvSpec =
+    ck::tensor_operation::device::ConvolutionForwardSpecialization::Default;
+
+static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+template <ck::index_t NDimSpatial, typename InLayout, typename WeiLayout, typename OutLayout>
+using DeviceGroupedConvNDFwdInstance =
+    ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<
+        NDimSpatial,
+        InLayout,
+        WeiLayout,
+        ck::Tuple<>,
+        OutLayout,
+        InDataType,
+        WeiDataType,
+        AccDataType,
+        CShuffleDataType,
+        ck::Tuple<>,
+        OutDataType,
+        InElementOp,
+        WeiElementOp,
+        OutElementOp,
+        ConvSpec,    // ConvForwardSpecialization
+        GemmSpec,    // GemmSpecialization
+        1,           //
+        256,         // BlockSize
+        128,         // MPerBlock
+        256,         // NPerBlock
+        32,          // KPerBlock
+        8,           // AK1
+        8,           // BK1
+        32,          // MPerXdl
+        32,          // NPerXdl
+        2,           // MXdlPerWave
+        4,           // NXdlPerWave
+        S<4, 64, 1>, // ABlockTransferThreadClusterLengths_AK0_M_AK1
+        S<1, 0, 2>,  // ABlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,  // ABlockTransferSrcAccessOrder
+        2,           // ABlockTransferSrcVectorDim
+        8,           // ABlockTransferSrcScalarPerVector
+        8,           // ABlockTransferDstScalarPerVector_AK1
+        1,           // ABlockLdsExtraM
+        S<4, 64, 1>, // BBlockTransferThreadClusterLengths_BK0_N_BK1
+        S<1, 0, 2>,  // BBlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,  // BBlockTransferSrcAccessOrder
+        2,           // BBlockTransferSrcVectorDim
+        8,           // BBlockTransferSrcScalarPerVector
+        8,           // BBlockTransferDstScalarPerVector_BK1
+        1,           // BBlockLdsExtraN
+        1,
+        1,
+        S<1, 32, 1, 8>,
+        8,
+        AComputeType,
+        BComputeType>;
+
+#include "run_convnd_fwd_example.inc"
+
+int main(int argc, char* argv[]) { return run_convnd_fwd_example(argc, argv) ? 0 : 1; }
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp
index 40878e4f0e..af79eefa1f 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp
@@ -326,6 +326,42 @@ using device_grouped_conv_fwd_xdl_f8_bf8_instances = std::tuple<
     // clang-format on
     >;
 
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_bf8_f8_instances = std::tuple<
+// clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           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|AComputeType|BComputeType|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| 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|            |            |
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |            |            |
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+        // generic instance
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1,         BF8,          F8>,
+        // instances for small conv.K and conv.C
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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>,               1,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8,         F8>,
+
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        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,         BF8,         F8>
+#endif
+    // clang-format on
+    >;
+
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp b/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp
index e61ec28284..8602a82ff8 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp
@@ -301,6 +301,14 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f8_bf8_instances(op_ptrs);
             }
 #endif
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+            if constexpr(is_same_v<InDataType, ck::bf8_t> && is_same_v<WeiDataType, ck::f8_t> &&
+                         is_same_v<OutDataType, ck::f8_t> && is_same_v<AComputeType, ck::bf8_t> &&
+                         is_same_v<BComputeType, ck::f8_t>)
+            {
+                add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_f8_instances(op_ptrs);
+            }
+#endif
 #ifdef CK_ENABLE_FP16
             if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
                          is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_xdl.inc b/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_xdl.inc
index 691414ebcb..e627d428de 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_xdl.inc
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_xdl.inc
@@ -369,6 +369,24 @@ void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f8_bf8_instances(
                                                                 BF8>>>& instances);
 #endif
 
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_f8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                BF8,
+                                                                F8,
+                                                                Empty_Tuple,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                BF8,
+                                                                F8>>>& instances);
+#endif
+
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/CMakeLists.txt b/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/CMakeLists.txt
index 50a6ec9a45..579bea00d8 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/CMakeLists.txt
@@ -44,6 +44,8 @@ endif()
 if((DTYPES MATCHES "fp8" AND DTYPES MATCHES "bf8") OR NOT DEFINED DTYPES)
     list(APPEND GROUPED_CONV3D_FWD
       xdl/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_fp8_bf8_instance.cpp)
+    list(APPEND GROUPED_CONV3D_FWD
+      xdl/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_fp8_instance.cpp)
 endif()
 
 add_instance_library(device_grouped_conv3d_fwd_instance ${GROUPED_CONV3D_FWD})
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_fp8_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_fp8_instance.cpp
new file mode 100644
index 0000000000..08c45d7dd9
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_fp8_instance.cpp
@@ -0,0 +1,54 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_f8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                BF8,
+                                                                F8,
+                                                                Empty_Tuple,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                BF8,
+                                                                F8>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_fwd_xdl_bf8_f8_instances<3,
+                                                                                NDHWGC,
+                                                                                GKZYXC,
+                                                                                Empty_Tuple,
+                                                                                NDHWGK,
+                                                                                ConvFwdDefault>{});
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_fwd_xdl_bf8_f8_instances<3,
+                                                                                NDHWGC,
+                                                                                GKZYXC,
+                                                                                Empty_Tuple,
+                                                                                NDHWGK,
+                                                                                ConvFwd1x1P0>{});
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_fwd_xdl_bf8_f8_instances<3,
+                                                                                NDHWGC,
+                                                                                GKZYXC,
+                                                                                Empty_Tuple,
+                                                                                NDHWGK,
+                                                                                ConvFwd1x1S1P0>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/profiler/src/profile_grouped_conv_fwd.cpp b/profiler/src/profile_grouped_conv_fwd.cpp
index a847999b56..577efafb1e 100644
--- a/profiler/src/profile_grouped_conv_fwd.cpp
+++ b/profiler/src/profile_grouped_conv_fwd.cpp
@@ -26,6 +26,7 @@ enum struct ConvDataType
     F8_F8_F8,       // 4
     BF8_BF8_F8,     // 5
     F8_BF8_F8,      // 6
+    BF8_F8_F8,      // 7
 };
 
 #define OP_NAME "grouped_conv_fwd"
@@ -42,7 +43,8 @@ static void print_helper_msg()
         << "                 3: Input int8, Weight int8, Output int8\n"
         << "                 4: Input fp8, Weight fp8, Output fp8\n"
         << "                 5: Input bf8, Weight bf8, Output fp8\n"
-        << "                 6: Input fp8, Weight bf8, Output fp8)\n"
+        << "                 6: Input fp8, Weight bf8, Output fp8\n"
+        << "                 7: Input bf8, Weight fp8, Output fp8)\n"
         << "arg3: tensor layout (0: Input[G, N, Hi, Wi, C], Weight[G, K, Y, X, C], Output[G, N, Ho, Wo, K]\n"
         << "                     1: Input[N, Hi, Wi, G, C], Weight[G, K, Y, X, C], Output[N, Ho, Wo, G, K])\n"
         << "arg4: verification (0: no, 1: yes)\n"
@@ -281,6 +283,10 @@ int profile_grouped_conv_fwd(int argc, char* argv[])
         {
             return profile(I3, NDHWGC{}, GKZYXC{}, NDHWGK{}, F8{}, BF8{}, F8{}, F8{}, BF8{});
         }
+        else if(data_type == ConvDataType::BF8_F8_F8)
+        {
+            return profile(I3, NDHWGC{}, GKZYXC{}, NDHWGK{}, BF8{}, F8{}, F8{}, BF8{}, F8{});
+        }
     }
 
     std::cout << "this data_type & layout is not implemented" << std::endl;

From d7f05fb996a55a37bb005f27d0c09c7595e5aee2 Mon Sep 17 00:00:00 2001
From: Illia Silin <98187287+illsilin@users.noreply.github.com>
Date: Thu, 11 Apr 2024 16:40:45 -0700
Subject: [PATCH 7/7] [HotFix] pass XDL and WMMA macros to libs that use CK
 (#1234)

---
 include/ck/config.h.in | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/include/ck/config.h.in b/include/ck/config.h.in
index 1748344756..eb9049b599 100644
--- a/include/ck/config.h.in
+++ b/include/ck/config.h.in
@@ -104,6 +104,20 @@
 #cmakedefine CK_ENABLE_INSTANCES_ONLY @CK_ENABLE_INSTANCES_ONLY@
 #endif
 
+//
+// CK kernels which support XDL (MI series)
+//
+#ifndef CK_USE_XDL
+#cmakedefine CK_USE_XDL @CK_USE_XDL@
+#endif
+
+//
+// CK Kernels which support WMMA (recent Navi series)
+//
+#ifndef CK_USE_WMMA
+#cmakedefine CK_USE_WMMA @CK_USE_WMMA@
+#endif
+
 // clang-format on
 
 #endif // CK_CONFIG_H_IN