Merge branch 'develop' into gemm_v2r3_kpad_fix

library/src/tensor_operation_instance/gpu/grouped_gemm_fixed_nk/device_grouped_gemm_xdl_fixed_nk_f16_f16_f16_mk_nk_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_fixed_nk.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using DsDataType = ck::Tuple<>;
+
+using DsLayout = ck::Tuple<>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+using device_grouped_gemm_xdl_fixed_nk_f16_f16_f16_mk_nk_mn_irregular_tile_instances = std::tuple<
+    // clang-format off
+        //############################|      A|      B|          Ds|      E| 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|
+        //############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    64,   8,   8,   32,   32,    4,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   256,    64,   8,   8,   32,   32,    1,    4,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    64,   8,   8,   32,   32,    2,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>
+    // clang-format on
+    >;
+
+void add_device_grouped_gemm_xdl_fixed_nk_f16_f16_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmFixedNK<Row,
+                                                         Col,
+                                                         DsLayout,
+                                                         Row,
+                                                         F16,
+                                                         F16,
+                                                         DsDataType,
+                                                         F16,
+                                                         PassThrough,
+                                                         PassThrough,
+                                                         PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_fixed_nk_f16_f16_f16_mk_nk_mn_irregular_tile_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_fixed_nk/device_grouped_gemm_xdl_fixed_nk_f16_f8_f16_mk_kn_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_fixed_nk.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using DsDataType = ck::Tuple<>;
+
+using DsLayout = ck::Tuple<>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+using device_grouped_gemm_xdl_fixed_nk_f16_f8_f16_mk_kn_mn_irregular_tile_instances = std::tuple<
+    // clang-format off
+        //############################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 16,16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2, S< 1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2, S< 1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2, S< 1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2, S< 1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>
+    // clang-format on
+    >;
+
+void add_device_grouped_gemm_xdl_fixed_nk_f16_f8_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmFixedNK<Row,
+                                                         Row,
+                                                         DsLayout,
+                                                         Row,
+                                                         F16,
+                                                         F8,
+                                                         DsDataType,
+                                                         F16,
+                                                         PassThrough,
+                                                         PassThrough,
+                                                         PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_grouped_gemm_xdl_fixed_nk_f16_f8_f16_mk_kn_mn_irregular_tile_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_fixed_nk/device_grouped_gemm_xdl_fixed_nk_f16_f8_f16_mk_nk_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_fixed_nk.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F8  = ck::f8_t;
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using DsDataType = ck::Tuple<>;
+
+using DsLayout = ck::Tuple<>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+using device_grouped_gemm_xdl_fixed_nk_f16_f8_f16_mk_nk_mn_irregular_tile_instances = std::tuple<
+    // clang-format off
+        //############################|      A|      B|          Ds|      E| 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|
+        //############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    64,   8,   8,   32,   32,    4,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   256,    64,   8,   8,   32,   32,    1,    4,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    64,   8,   8,   32,   32,    2,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    F8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>
+    // clang-format on
+    >;
+
+void add_device_grouped_gemm_xdl_fixed_nk_f16_f8_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmFixedNK<Row,
+                                                         Col,
+                                                         DsLayout,
+                                                         Row,
+                                                         F16,
+                                                         F8,
+                                                         DsDataType,
+                                                         F16,
+                                                         PassThrough,
+                                                         PassThrough,
+                                                         PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_grouped_gemm_xdl_fixed_nk_f16_f8_f16_mk_nk_mn_irregular_tile_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_fixed_nk/device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_kn_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_fixed_nk.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using I8  = int8_t;
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using DsDataType = ck::Tuple<>;
+
+using DsLayout = ck::Tuple<>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+using device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_kn_mn_irregular_tile_instances = std::tuple<
+    // clang-format off
+        //############################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 16,16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2, S< 1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2, S< 1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2, S< 1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2, S< 1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Row,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2, S< 1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>
+    // clang-format on
+    >;
+
+void add_device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmFixedNK<Row,
+                                                         Row,
+                                                         DsLayout,
+                                                         Row,
+                                                         F16,
+                                                         I8,
+                                                         DsDataType,
+                                                         F16,
+                                                         PassThrough,
+                                                         PassThrough,
+                                                         PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_kn_mn_irregular_tile_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_fixed_nk/device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_nk_mn_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_fixed_nk.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using I8  = int8_t;
+using F16 = ck::half_t;
+using F32 = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using DsDataType = ck::Tuple<>;
+
+using DsLayout = ck::Tuple<>;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+using device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_nk_mn_irregular_tile_instances = std::tuple<
+    // clang-format off
+        //############################|      A|      B|          Ds|      E| 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|
+        //############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    64,   8,   8,   32,   32,    4,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    64,   8,   8,   32,   32,    2,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   256,    64,   8,   8,   32,   32,    1,    4,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    64,   8,   8,   32,   32,    2,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    64,   8,   8,   32,   32,    2,    1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
+        DeviceGroupedGemm_Xdl_Fixed_NK<    Row,    Col,    DsLayout,    Row,   F16,    I8,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 8,  8, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,             3,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>
+    // clang-format on
+    >;
+
+void add_device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmFixedNK<Row,
+                                                         Col,
+                                                         DsLayout,
+                                                         Row,
+                                                         F16,
+                                                         I8,
+                                                         DsDataType,
+                                                         F16,
+                                                         PassThrough,
+                                                         PassThrough,
+                                                         PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances, device_grouped_gemm_xdl_fixed_nk_f16_i8_f16_mk_nk_mn_irregular_tile_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

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

+add_instance_library(device_image_to_column_instance
+   device_image_to_column_nhwc_1d_instance.cpp
+   device_image_to_column_nhwc_2d_instance.cpp
+   device_image_to_column_nhwc_3d_instance.cpp
+)

library/src/tensor_operation_instance/gpu/image_to_column/device_image_to_column_nhwc_1d_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/image_to_column/device_image_to_column_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_image_to_column_nhwc_1d_bf16_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<1, GNWC, BF16, BF16>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_bf16_instances<1, GNWC>{});
+}
+
+void add_device_image_to_column_nhwc_1d_f16_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<1, GNWC, F16, F16>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_f16_instances<1, GNWC>{});
+}
+
+void add_device_image_to_column_nhwc_1d_f32_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<1, GNWC, F32, F32>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_f32_instances<1, GNWC>{});
+}
+
+void add_device_image_to_column_nhwc_1d_i8_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<1, GNWC, int8_t, int8_t>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_i8_instances<1, GNWC>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/image_to_column/device_image_to_column_nhwc_2d_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/image_to_column/device_image_to_column_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_image_to_column_nhwc_2d_bf16_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<2, GNHWC, BF16, BF16>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_bf16_instances<2, GNHWC>{});
+}
+
+void add_device_image_to_column_nhwc_2d_f16_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<2, GNHWC, F16, F16>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_f16_instances<2, GNHWC>{});
+}
+
+void add_device_image_to_column_nhwc_2d_f32_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<2, GNHWC, F32, F32>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_f32_instances<2, GNHWC>{});
+}
+
+void add_device_image_to_column_nhwc_2d_i8_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<2, GNHWC, int8_t, int8_t>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_i8_instances<2, GNHWC>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/image_to_column/device_image_to_column_nhwc_3d_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/image_to_column/device_image_to_column_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_image_to_column_nhwc_3d_bf16_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<3, GNDHWC, BF16, BF16>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_bf16_instances<3, GNDHWC>{});
+}
+
+void add_device_image_to_column_nhwc_3d_f16_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<3, GNDHWC, F16, F16>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_f16_instances<3, GNDHWC>{});
+}
+
+void add_device_image_to_column_nhwc_3d_f32_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<3, GNDHWC, F32, F32>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_f32_instances<3, GNDHWC>{});
+}
+
+void add_device_image_to_column_nhwc_3d_i8_instances(
+    std::vector<std::unique_ptr<DeviceImageToColumn<3, GNDHWC, int8_t, int8_t>>>& instances)
+{
+    add_device_operation_instances(instances, device_image_to_column_i8_instances<3, GNDHWC>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

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

 set(DEVICE_MAXPOOL_BWD_INSTANCES)
-if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_MAXPOOL_BWD_INSTANCES device_max_pool_bwd_f16_instance.cpp)
-endif()
-if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_MAXPOOL_BWD_INSTANCES device_max_pool_bwd_bf16_instance.cpp)
-endif()
-if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_MAXPOOL_BWD_INSTANCES device_max_pool_bwd_f32_instance.cpp)
-endif()
+list(APPEND DEVICE_MAXPOOL_BWD_INSTANCES device_max_pool_bwd_f16_instance.cpp
+                                         device_max_pool_bwd_bf16_instance.cpp
+                                         device_max_pool_bwd_f32_instance.cpp)
 add_instance_library(device_max_pool_bwd_instance ${DEVICE_MAXPOOL_BWD_INSTANCES})

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

 set(DEVICE_NORMALIZATION_INSTANCES)
-if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_NORMALIZATION_INSTANCES device_layernorm2d_f16_instance.cpp
+
+list(APPEND DEVICE_NORMALIZATION_INSTANCES 
+    device_layernorm2d_f16_instance.cpp
     device_layernorm4d_f16_instance.cpp
     device_groupnorm_f16_instance.cpp
     device_groupnorm_swish_f16_instance.cpp
-    device_groupnorm_swish_f16_f32_f32_f16_instance.cpp)
-endif()
-if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_NORMALIZATION_INSTANCES device_layernorm2d_f32_instance.cpp
+    device_groupnorm_swish_f16_f32_f32_f16_instance.cpp
+    device_layernorm2d_f32_instance.cpp
     device_layernorm4d_f32_instance.cpp
     device_groupnorm_f32_instance.cpp
     device_groupnorm_swish_f32_instance.cpp)
-endif()
+
 add_instance_library(device_normalization_instance ${DEVICE_NORMALIZATION_INSTANCES})

library/src/tensor_operation_instance/gpu/pool3d_fwd/CMakeLists.txt

 set(DEVICE_POOL3D_FWD_INSTANCES)
-if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_POOL3D_FWD_INSTANCES device_avg_pool3d_fwd_ndhwc_f16_instance.cpp
-                                          device_max_pool3d_fwd_ndhwc_f16_instance.cpp)
-endif()
-if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_POOL3D_FWD_INSTANCES device_avg_pool3d_fwd_ndhwc_bf16_instance.cpp
-                                          device_max_pool3d_fwd_ndhwc_bf16_instance.cpp)
-endif()
-if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_POOL3D_FWD_INSTANCES device_avg_pool3d_fwd_ndhwc_f32_instance.cpp
-                                          device_max_pool3d_fwd_ndhwc_f32_instance.cpp)
-endif()
+list(APPEND DEVICE_POOL3D_FWD_INSTANCES device_avg_pool3d_fwd_ndhwc_f16_instance.cpp
+                                        device_max_pool3d_fwd_ndhwc_f16_instance.cpp
+                                        device_avg_pool3d_fwd_ndhwc_f32_instance.cpp
+                                        device_max_pool3d_fwd_ndhwc_f32_instance.cpp
+                                        device_avg_pool3d_fwd_ndhwc_bf16_instance.cpp
+                                        device_max_pool3d_fwd_ndhwc_bf16_instance.cpp)
 add_instance_library(device_pool3d_fwd_instance ${DEVICE_POOL3D_FWD_INSTANCES})

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

-if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
-
 set(CONV2D_PERLAYER_QUANT_SRC conv2d_fwd/device_conv2d_xdl_perlayer_quantization_int8_instance.cpp)
 set(CONV2D_PERCHANNEL_QUANT_SRC conv2d_fwd/device_conv2d_xdl_perchannel_quantization_int8_instance.cpp)
 set(CONV2D_BIAS_PERLAYER_QUANT_SRC conv2d_fwd/device_conv2d_xdl_bias_perlayer_quantization_int8_instance.cpp)
@@ -10,17 +8,16 @@ set(GEMM_QUANT_SRC
     gemm/device_gemm_quantization_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp
     gemm/device_gemm_quantization_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
 )
-if(DL_KERNELS)
-    list(APPEND CONV2D_PERLAYER_QUANT_SRC conv2d_fwd/device_conv2d_dl_perlayer_quantization_int8_instance.cpp)
-    list(APPEND CONV2D_PERCHANNEL_QUANT_SRC conv2d_fwd/device_conv2d_dl_perchannel_quantization_int8_instance.cpp)
-    list(APPEND CONV2D_BIAS_PERLAYER_QUANT_SRC conv2d_fwd/device_conv2d_dl_bias_perlayer_quantization_int8_instance.cpp)
-    list(APPEND CONV2D_BIAS_PERCHANNEL_QUANT_SRC conv2d_fwd/device_conv2d_dl_bias_perchannel_quantization_int8_instance.cpp)
-    list(APPEND GEMM_QUANT_SRC
-        gemm/device_gemm_quantization_dl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp
-        gemm/device_gemm_quantization_dl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp
-        gemm/device_gemm_quantization_dl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp
-        gemm/device_gemm_quantization_dl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp)
-endif()
+
+list(APPEND CONV2D_PERLAYER_QUANT_SRC conv2d_fwd/device_conv2d_dl_perlayer_quantization_int8_instance.cpp)
+list(APPEND CONV2D_PERCHANNEL_QUANT_SRC conv2d_fwd/device_conv2d_dl_perchannel_quantization_int8_instance.cpp)
+list(APPEND CONV2D_BIAS_PERLAYER_QUANT_SRC conv2d_fwd/device_conv2d_dl_bias_perlayer_quantization_int8_instance.cpp)
+list(APPEND CONV2D_BIAS_PERCHANNEL_QUANT_SRC conv2d_fwd/device_conv2d_dl_bias_perchannel_quantization_int8_instance.cpp)
+list(APPEND GEMM_QUANT_SRC
+    gemm/device_gemm_quantization_dl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp
+    gemm/device_gemm_quantization_dl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp
+    gemm/device_gemm_quantization_dl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp
+    gemm/device_gemm_quantization_dl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp)
 
 add_instance_library(device_quantization_instance
     ${CONV2D_PERLAYER_QUANT_SRC}
@@ -29,4 +26,3 @@ add_instance_library(device_quantization_instance
     ${CONV2D_BIAS_PERCHANNEL_QUANT_SRC}
     ${GEMM_QUANT_SRC}
 )
-endif()
\ No newline at end of file

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

 set(DEVICE_SOFTMAX_INSTANCES)
-if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_SOFTMAX_INSTANCES device_softmax_f16_f16_instance_rank3_reduce1.cpp
+list(APPEND DEVICE_SOFTMAX_INSTANCES 
+    device_softmax_f16_f16_instance_rank3_reduce1.cpp
     device_softmax_f16_f16_instance_rank3_reduce2.cpp
     device_softmax_f16_f16_instance_rank3_reduce3.cpp
     device_softmax_f16_f16_instance_rank4_reduce1.cpp
     device_softmax_f16_f16_instance_rank4_reduce2.cpp
     device_softmax_f16_f16_instance_rank4_reduce3.cpp
-    device_softmax_f16_f16_instance_rank4_reduce4.cpp)
-endif()
-if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
-    list(APPEND DEVICE_SOFTMAX_INSTANCES device_softmax_f32_f32_instance_rank3_reduce1.cpp
+    device_softmax_f16_f16_instance_rank4_reduce4.cpp
+    device_softmax_f32_f32_instance_rank3_reduce1.cpp
     device_softmax_f32_f32_instance_rank3_reduce2.cpp
     device_softmax_f32_f32_instance_rank3_reduce3.cpp
     device_softmax_f32_f32_instance_rank4_reduce1.cpp
     device_softmax_f32_f32_instance_rank4_reduce2.cpp
     device_softmax_f32_f32_instance_rank4_reduce3.cpp
     device_softmax_f32_f32_instance_rank4_reduce4.cpp)
-endif()
 add_instance_library(device_softmax_instance ${DEVICE_SOFTMAX_INSTANCES})

profiler/README.md

@@ -184,3 +184,41 @@ tflops: 95.337
 GB/s: 69.2301
 ```
 Note: This kernel use atomic add, this will cause output buffer to be accumulated multiple times, causing verification failure. To work around it, do not use CK's own timer and do verification at the same time.
+
+## Profile image to column kernels
+```bash
+# arg1: tensor operation (" OP_NAME ": " OP_DESC ")
+# arg2: data type (0: Input fp32, Weight fp32, Output fp32
+#                  1: Input fp16, Weight fp16, Output fp16
+#                  2: Input bf16, Weight bf16, Output bf16
+#                  3: Input int8, Weight int8, Output int8)
+# arg3: tensor layout (0: Input[N, Hi, Wi, C], Output[N * Ho * Wo, Y * X * C])
+# arg4: verification (0: no, 1: yes)
+# arg5: initialization (0: no init, 1: integer value, 2: decimal value)
+# arg6: print tensor value (0: no; 1: yes)
+# arg7: time kernel (0: no, 1: yes)
+# Following arguments (depending on number of spatial dims):
+#  Number of spatial dimensions (1=Conv1d, 2=Conv2d, 3=Conv3d)
+#  G, N, K, C, 
+#  <filter spatial dimensions>, (ie Y, X for 2D)
+#  <input image spatial dimensions>, (ie Hi, Wi for 2D)
+#  <strides>, (ie Sy, Sx for 2D)
+#  <dilations>, (ie Dy, Dx for 2D)
+#  <left padding>, (ie LeftPy, LeftPx for 2D)
+#  <right padding>, (ie RightPy, RightPx for 2D)
+
+ ################             op   datatype  layout  verify  init  log  time  Ndims  G   N   K   C  Y  X  Hi  Wi  Sy  Sx  Dy  Dx  LeftPy  LeftPx  RightPy  RightPx
+./bin/ckProfiler image_to_column          0       0       1     1    0     1      2  1 256   1 512  3  3   28  28   1   1   1   1        0       0       0        0
+
+ ```
+
+Result (MI210, FP32, NHWC)
+```
+input: dim 5, lengths {1, 256, 512, 28, 28}, strides {102760448, 401408, 1, 14336, 512}
+output: dim 2, lengths {173056, 4608}, strides {4608, 1}
+....
+Best configuration parameters:
+name: DeviceImageToColumn<128, 32, 64, 4>
+avg_time: 3.12326
+GB/s: 2042.59
+```

profiler/include/profiler/profile_gemm_impl.hpp

@@ -223,6 +223,12 @@ int profile_gemm_impl(int do_verification,
     {
         std::cout << "Best Perf for datatype = int8";
     }
+#if defined CK_ENABLE_FP8
+    else if constexpr(is_same<CDataType, f8_t>::value)
+    {
+        std::cout << "Best Perf for datatype = fp8";
+    }
+#endif
 
     if constexpr(is_same<ALayout, tensor_layout::gemm::RowMajor>::value)
     {

profiler/include/profiler/profile_gemm_multiply_add_impl.hpp

@@ -83,8 +83,8 @@ bool profile_gemm_multiply_add_impl(int do_verification,
         d1_m_n.GenerateTensorValue(GeneratorTensor_2<D1DataType>{-1, 1});
         break;
     default:
-        a_m_k.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
-        b_k_n.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
+        a_m_k.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 0.2});
+        b_k_n.GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.1, 0.1});
         d0_m_n.GenerateTensorValue(GeneratorTensor_3<D0DataType>{0.0, 1.0});
         d1_m_n.GenerateTensorValue(GeneratorTensor_3<D1DataType>{0.0, 1.0});
     }

profiler/include/profiler/profile_gemm_splitk_impl.hpp

@@ -214,6 +214,7 @@ bool profile_gemm_splitk_impl(int do_verification,
                           << " TFlops, " << gb_per_sec << " GB/s, " << op_name << ", KBatch "
                           << kbatch_curr << std::endl;
 
+#if defined CK_ENABLE_FP8
                 // set softer tolerances for fp8
                 if constexpr(is_same_v<ADataType, f8_t> || is_same_v<BDataType, f8_t> ||
                              is_same_v<CDataType, f8_t>)
@@ -226,8 +227,11 @@ bool profile_gemm_splitk_impl(int do_verification,
                 }
                 else
                 {
+#endif
                     pass = pass & ck::utils::check_err(c_m_n_device_result, c_m_n_host_result);
+#if defined CK_ENABLE_FP8
                 }
+#endif
 
                 if(tflops > best_tflops)
                 {

profiler/include/profiler/profile_grouped_conv_fwd_impl.hpp

@@ -215,7 +215,7 @@ bool profile_grouped_conv_fwd_impl(int do_verification,
     const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
         DeviceOp>::GetInstances();
 
-    std::cout << "xdl found " << op_ptrs.size() << " instances" << std::endl;
+    std::cout << "ckProfiler found " << op_ptrs.size() << " instances" << std::endl;
 
     for(auto& op_ptr : op_ptrs)
     {

profiler/include/profiler/profile_image_to_column_impl.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <iomanip>
+#include <iostream>
+#include <typeinfo>
+#include <limits>
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/device_image_to_column.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_image_to_column_impl.hpp"
+#include "ck/library/tensor_operation_instance/gpu/image_to_column.hpp"
+#include "ck/library/utility/check_err.hpp"
+#include "ck/library/utility/device_memory.hpp"
+#include "ck/library/utility/host_tensor.hpp"
+#include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/convolution_parameter.hpp"
+#include "ck/library/utility/convolution_host_tensor_descriptor_helper.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_image_to_column.hpp"
+
+namespace ck {
+namespace profiler {
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+template <index_t NDimSpatial,
+          typename InputLayout,
+          typename InputDataType,
+          typename OutputDataType>
+bool profile_image_to_column_impl(int do_verification,
+                                  int init_method,
+                                  bool do_log,
+                                  bool time_kernel,
+                                  const ck::utils::conv::ConvParam& conv_param)
+{
+    const ck::index_t NDoHoWo =
+        conv_param.N_ *
+        ck::accumulate_n<ck::index_t>(
+            conv_param.output_spatial_lengths_.begin(), NDimSpatial, 1, std::multiplies<>());
+    const ck::index_t CZYX =
+        conv_param.C_ *
+        ck::accumulate_n<ck::index_t>(
+            conv_param.filter_spatial_lengths_.begin(), NDimSpatial, 1, std::multiplies<>());
+
+    const auto in_desc =
+        ck::utils::conv::make_input_host_tensor_descriptor_g_n_c_wis_packed<InputLayout>(
+            conv_param);
+    const auto out_desc = HostTensorDescriptor({NDoHoWo, CZYX});
+
+    std::array<ck::index_t, NDimSpatial> input_spatial_lengths{};
+    std::array<ck::index_t, NDimSpatial> filter_spatial_lengths{};
+    std::array<ck::index_t, NDimSpatial> output_spatial_lengths{};
+    std::array<ck::index_t, NDimSpatial + 3> input_g_n_c_wis_strides{};
+    std::array<ck::index_t, 2> output_m_k_strides{};
+    std::array<ck::index_t, NDimSpatial> conv_filter_strides{};
+    std::array<ck::index_t, NDimSpatial> conv_filter_dilations{};
+    std::array<ck::index_t, NDimSpatial> input_left_pads{};
+    std::array<ck::index_t, NDimSpatial> input_right_pads{};
+
+    auto copy = [](const auto& x, auto& y) { std::copy(x.begin(), x.end(), y.begin()); };
+
+    copy(conv_param.input_spatial_lengths_, input_spatial_lengths);
+    copy(conv_param.filter_spatial_lengths_, filter_spatial_lengths);
+    copy(conv_param.output_spatial_lengths_, output_spatial_lengths);
+    copy(in_desc.GetStrides(), input_g_n_c_wis_strides);
+    copy(out_desc.GetStrides(), output_m_k_strides);
+    copy(conv_param.conv_filter_strides_, conv_filter_strides);
+    copy(conv_param.conv_filter_dilations_, conv_filter_dilations);
+    copy(conv_param.input_left_pads_, input_left_pads);
+    copy(conv_param.input_right_pads_, input_right_pads);
+
+    Tensor<InputDataType> input(in_desc);
+    Tensor<OutputDataType> host_output(out_desc);
+    Tensor<OutputDataType> device_output(out_desc);
+
+    std::cout << "input: " << input.mDesc << std::endl;
+    std::cout << "output: " << host_output.mDesc << std::endl;
+
+    switch(init_method)
+    {
+    case 0: break;
+    case 1: input.GenerateTensorValue(GeneratorTensor_2<InputDataType>{-5, 5}); break;
+    default: input.GenerateTensorValue(GeneratorTensor_3<InputDataType>{0.0, 1.0});
+    }
+
+    DeviceMem in_device_buf(sizeof(InputDataType) * input.mDesc.GetElementSpaceSize());
+    DeviceMem out_device_buf(sizeof(OutputDataType) * device_output.mDesc.GetElementSpaceSize());
+
+    in_device_buf.ToDevice(input.mData.data());
+
+    // run reference op
+    if(do_verification)
+    {
+        auto ref_image_to_column = ck::tensor_operation::host::
+            ReferenceImageToColumn<NDimSpatial, InputLayout, InputDataType, OutputDataType>{};
+
+        auto ref_invoker  = ref_image_to_column.MakeInvoker();
+        auto ref_argument = ref_image_to_column.MakeArgument(input,
+                                                             host_output,
+                                                             conv_param.filter_spatial_lengths_,
+                                                             conv_param.conv_filter_strides_,
+                                                             conv_param.conv_filter_dilations_,
+                                                             conv_param.input_left_pads_,
+                                                             conv_param.input_right_pads_);
+
+        // init host output to zero
+        host_output.SetZero();
+
+        ref_invoker.Run(ref_argument);
+    }
+
+    using DeviceOp = ck::tensor_operation::device::
+        DeviceImageToColumn<NDimSpatial, InputLayout, InputDataType, OutputDataType>;
+
+    // get device op instances
+    const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
+        DeviceOp>::GetInstances();
+
+    std::cout << "found " << op_ptrs.size() << " instances" << std::endl;
+
+    std::string best_op_name;
+    float best_avg_time   = std::numeric_limits<float>::max();
+    float best_gb_per_sec = 0;
+
+    // profile device op instances
+    bool pass                   = true;
+    bool is_supporting_instance = false;
+
+    for(auto& op_ptr : op_ptrs)
+    {
+        auto argument_ptr = op_ptr->MakeArgumentPointer(
+            static_cast<InputDataType*>(in_device_buf.GetDeviceBuffer()),
+            static_cast<OutputDataType*>(out_device_buf.GetDeviceBuffer()),
+            conv_param.N_,
+            conv_param.C_,
+            input_spatial_lengths,
+            filter_spatial_lengths,
+            output_spatial_lengths,
+            input_g_n_c_wis_strides,
+            output_m_k_strides,
+            conv_filter_strides,
+            conv_filter_dilations,
+            input_left_pads,
+            input_right_pads);
+
+        if(op_ptr->IsSupportedArgument(argument_ptr.get()))
+        {
+            is_supporting_instance = true;
+            // re-init output to zero before profiling next kernel
+            out_device_buf.SetZero();
+            std::string op_name = op_ptr->GetTypeString();
+            auto invoker_ptr    = op_ptr->MakeInvokerPointer();
+            float avg_time =
+                invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, time_kernel});
+            std::size_t num_btype =
+                NDoHoWo * CZYX * (sizeof(OutputDataType) + sizeof(InputDataType));
+            float gb_per_sec = num_btype / 1.E6 / avg_time;
+            std::cout << "Perf: " << std::setw(10) << avg_time << " ms, " << gb_per_sec << " GB/s, "
+                      << op_name << std::endl;
+
+            if(avg_time < best_avg_time)
+            {
+                best_op_name    = op_name;
+                best_avg_time   = avg_time;
+                best_gb_per_sec = gb_per_sec;
+            }
+
+            if(do_verification)
+            {
+                out_device_buf.FromDevice(device_output.mData.data());
+                pass = pass & ck::utils::check_err(device_output, host_output);
+
+                if(do_log)
+                {
+                    LogRangeAsType<float>(std::cout << "input : ", input.mData, ",") << std::endl;
+                    LogRangeAsType<float>(std::cout << "host_output  : ", host_output.mData, ",")
+                        << std::endl;
+                    LogRangeAsType<float>(std::cout << "device_output: ", device_output.mData, ",")
+                        << std::endl;
+                }
+            }
+        }
+        else
+        {
+            std::cout << op_ptr->GetTypeString() << " does not support this problem" << std::endl;
+        }
+    }
+
+    std::cout << "Best configuration parameters:"
+              << "\nname: " << best_op_name << "\navg_time: " << best_avg_time
+              << "\nGB/s: " << best_gb_per_sec << std::endl;
+
+    return is_supporting_instance && pass;
+}
+
+} // namespace profiler
+} // namespace ck