update ckprofiler for conv bwd data

example/17_convnd_bwd_data/CMakeLists.txt

+add_example_executable(example_convnd_bwd_data_xdl_fp16 convnd_bwd_data_xdl_fp16.cpp)
+target_link_libraries(example_convnd_bwd_data_xdl_fp16 PRIVATE utility)

example/17_convnd_bwd_data_xdl/CMakeLists.txt

-add_example_executable(example_convnd_bwd_data_xdl convnd_bwd_data_xdl.cpp)
-target_link_libraries(example_convnd_bwd_data_xdl PRIVATE utility)

example/CMakeLists.txt

@@ -33,10 +33,10 @@ add_subdirectory(13_pool2d_fwd)
 add_subdirectory(14_gemm_xdl_requant_relu_requant)
 add_subdirectory(15_grouped_gemm)
 add_subdirectory(16_gemm_reduce)
-add_subdirectory(17_convnd_bwd_data_xdl)
+add_subdirectory(17_convnd_bwd_data)
 add_subdirectory(18_batched_gemm_reduce)
 add_subdirectory(19_binary_elementwise)
-add_subdirectory(20_convnd_bwd_weight_xdl)
+add_subdirectory(20_convnd_bwd_weight)
 add_subdirectory(21_gemm_layernorm)
 add_subdirectory(22_cgemm)
 add_subdirectory(23_softmax)

include/ck/ck.hpp

@@ -146,7 +146,7 @@
 
 // workaround: verifaction failure, due to compiler regression, for conv bwd-data fp16 using some
 // tuning parameter
-#define CK_WORKAROUND_SWDEV_325164 1
+#define CK_WORKAROUND_SWDEV_325164 0
 
 namespace ck {
 

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

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/device_conv_bwd_data.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// conv1d backward data
+void add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<1,
+                                                  NWC,
+                                                  KXC,
+                                                  NWK,
+                                                  BF16,
+                                                  BF16,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f16_instances(
+    std::vector<std::unique_ptr<
+        DeviceConvBwdData<1, NWC, KXC, NWK, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f32_instances(
+    std::vector<std::unique_ptr<
+        DeviceConvBwdData<1, NWC, KXC, NWK, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_int8_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<1,
+                                                  NWC,
+                                                  KXC,
+                                                  NWK,
+                                                  int8_t,
+                                                  int8_t,
+                                                  int8_t,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+// conv2d backward data
+void add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<2,
+                                                  NHWC,
+                                                  KYXC,
+                                                  NHWK,
+                                                  BF16,
+                                                  BF16,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<2,
+                                                  NHWC,
+                                                  KYXC,
+                                                  NHWK,
+                                                  F16,
+                                                  F16,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<2,
+                                                  NHWC,
+                                                  KYXC,
+                                                  NHWK,
+                                                  F32,
+                                                  F32,
+                                                  F32,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_int8_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<2,
+                                                  NHWC,
+                                                  KYXC,
+                                                  NHWK,
+                                                  int8_t,
+                                                  int8_t,
+                                                  int8_t,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+// conv3d backward data
+void add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<3,
+                                                  NDHWC,
+                                                  KZYXC,
+                                                  NDHWK,
+                                                  BF16,
+                                                  BF16,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<3,
+                                                  NDHWC,
+                                                  KZYXC,
+                                                  NDHWK,
+                                                  F16,
+                                                  F16,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<3,
+                                                  NDHWC,
+                                                  KZYXC,
+                                                  NDHWK,
+                                                  F32,
+                                                  F32,
+                                                  F32,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_int8_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<3,
+                                                  NDHWC,
+                                                  KZYXC,
+                                                  NDHWK,
+                                                  int8_t,
+                                                  int8_t,
+                                                  int8_t,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+template <ck::index_t NumDimSpatial,
+          typename InLayout,
+          typename WeiLayout,
+          typename OutLayout,
+          typename InDataType,
+          typename WeiDataType,
+          typename OutDataType>
+struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceConvBwdData<
+    NumDimSpatial,
+    InLayout,
+    WeiLayout,
+    OutLayout,
+    InDataType,
+    WeiDataType,
+    OutDataType,
+    ck::tensor_operation::element_wise::PassThrough,
+    ck::tensor_operation::element_wise::PassThrough,
+    ck::tensor_operation::element_wise::PassThrough>>
+{
+    using DeviceOp = DeviceConvBwdData<NumDimSpatial,
+                                       InLayout,
+                                       WeiLayout,
+                                       OutLayout,
+                                       InDataType,
+                                       WeiDataType,
+                                       OutDataType,
+                                       ck::tensor_operation::element_wise::PassThrough,
+                                       ck::tensor_operation::element_wise::PassThrough,
+                                       ck::tensor_operation::element_wise::PassThrough>;
+
+    static auto GetInstances()
+    {
+        std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
+
+        if constexpr(NumDimSpatial == 1 && is_same_v<InLayout, NWC> && is_same_v<WeiLayout, KXC> &&
+                     is_same_v<OutLayout, NWK>)
+        {
+            if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                         is_same_v<OutDataType, float>)
+            {
+                add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f32_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                              is_same_v<OutDataType, half_t>)
+            {
+                add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f16_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
+                              is_same_v<WeiDataType, ck::bhalf_t> &&
+                              is_same_v<OutDataType, ck::bhalf_t>)
+            {
+                add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_bf16_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                              is_same_v<OutDataType, int8_t>)
+            {
+                add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_int8_instances(op_ptrs);
+            }
+        }
+        else if constexpr(NumDimSpatial == 2 && is_same_v<InLayout, NHWC> &&
+                          is_same_v<WeiLayout, KYXC> && is_same_v<OutLayout, NHWK>)
+        {
+            if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                         is_same_v<OutDataType, float>)
+            {
+                add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f32_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                              is_same_v<OutDataType, half_t>)
+            {
+                add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
+                              is_same_v<WeiDataType, ck::bhalf_t> &&
+                              is_same_v<OutDataType, ck::bhalf_t>)
+            {
+                add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                              is_same_v<OutDataType, int8_t>)
+            {
+                add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_int8_instances(op_ptrs);
+            }
+        }
+        else if constexpr(NumDimSpatial == 3 && is_same_v<InLayout, NDHWC> &&
+                          is_same_v<WeiLayout, KZYXC> && is_same_v<OutLayout, NDHWK>)
+        {
+            if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                         is_same_v<OutDataType, float>)
+            {
+                add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f32_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                              is_same_v<OutDataType, half_t>)
+            {
+                add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f16_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
+                              is_same_v<WeiDataType, ck::bhalf_t> &&
+                              is_same_v<OutDataType, ck::bhalf_t>)
+            {
+                add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_bf16_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                              is_same_v<OutDataType, int8_t>)
+            {
+                add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_int8_instances(op_ptrs);
+            }
+        }
+
+        return op_ptrs;
+    }
+};
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/CMakeLists.txt

@@ -19,13 +19,14 @@ add_subdirectory(contraction_bilinear)
 add_subdirectory(conv1d_fwd)
 add_subdirectory(conv2d_fwd)
 add_subdirectory(conv3d_fwd)
-add_subdirectory(conv2d_fwd_bias_relu)
-add_subdirectory(conv2d_fwd_bias_relu_add)
+add_subdirectory(conv1d_bwd_data)
 add_subdirectory(conv2d_bwd_data)
-add_subdirectory(convnd_bwd_data)
+add_subdirectory(conv3d_bwd_data)
 add_subdirectory(conv1d_bwd_weight)
 add_subdirectory(conv2d_bwd_weight)
 add_subdirectory(conv3d_bwd_weight)
+add_subdirectory(conv2d_fwd_bias_relu)
+add_subdirectory(conv2d_fwd_bias_relu_add)
 add_subdirectory(reduce)
 add_subdirectory(normalization)
 add_subdirectory(elementwise)
@@ -44,13 +45,14 @@ add_library(device_operations STATIC
     $<TARGET_OBJECTS:device_conv1d_fwd_instance>
     $<TARGET_OBJECTS:device_conv2d_fwd_instance>
     $<TARGET_OBJECTS:device_conv3d_fwd_instance>
-    $<TARGET_OBJECTS:device_conv2d_fwd_bias_relu_instance>
-    $<TARGET_OBJECTS:device_conv2d_fwd_bias_relu_add_instance>
+    $<TARGET_OBJECTS:device_conv1d_bwd_data_instance>
     $<TARGET_OBJECTS:device_conv2d_bwd_data_instance>
-    $<TARGET_OBJECTS:device_convnd_bwd_data_instance>
+    $<TARGET_OBJECTS:device_conv3d_bwd_data_instance>
     $<TARGET_OBJECTS:device_conv1d_bwd_weight_instance>
     $<TARGET_OBJECTS:device_conv2d_bwd_weight_instance>
     $<TARGET_OBJECTS:device_conv3d_bwd_weight_instance>
+    $<TARGET_OBJECTS:device_conv2d_fwd_bias_relu_instance>
+    $<TARGET_OBJECTS:device_conv2d_fwd_bias_relu_add_instance>
     $<TARGET_OBJECTS:device_reduce_instance>
     $<TARGET_OBJECTS:device_normalization_instance>
     $<TARGET_OBJECTS:device_elementwise_instance>

library/src/tensor_operation_instance/gpu/conv1d_bwd_data/CMakeLists.txt

+# device_conv1d_bwd_data_instance
+set(DEVICE_CONV1D_BWD_DATA_INSTANCE_SOURCE
+   device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f16_instance.cpp;
+   device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f32_instance.cpp;
+   device_conv1d_bwd_data_xdl_nwc_kxc_nwk_bf16_instance.cpp;
+   device_conv1d_bwd_data_xdl_nwc_kxc_nwk_int8_instance.cpp;
+)
+
+add_library(device_conv1d_bwd_data_instance OBJECT ${DEVICE_CONV1D_BWD_DATA_INSTANCE_SOURCE})
+target_compile_features(device_conv1d_bwd_data_instance PUBLIC)
+set_target_properties(device_conv1d_bwd_data_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
+rocm_install(TARGETS device_conv1d_bwd_data_instance)
+
+clang_tidy_check(device_conv1d_bwd_data_instance)

library/src/tensor_operation_instance/gpu/conv1d_bwd_data/device_conv1d_bwd_data_xdl_nwc_kxc_nwk_bf16_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, 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/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using BF16 = bhalf_t;
+using F32  = float;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using NWC = ck::tensor_layout::convolution::NWC;
+using KXC = ck::tensor_layout::convolution::KXC;
+using NWK = ck::tensor_layout::convolution::NWK;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto ConvBwdDataDefault =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Default;
+
+static constexpr auto ConvBwdDataFilter1x1Stride1Pad0 =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Filter1x1Stride1Pad0;
+
+// Compilation parameters for in[n, wi, c] * wei[k, x, c] = out[n, wo, k]
+using device_conv1d_bwd_data_xdl_nwc_kxc_nwk_bf16_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|       ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|               Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|     Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                   |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    64,     4,  8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    32,     4,  8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    32,   128,     4,  8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>
+    // clang-format on
+    >;
+
+using device_conv1d_bwd_data_xdl_nwc_kxc_nwk_1x1_s1_p0_bf16_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                    ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                            Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|                  Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                                |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    64,     4,  8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    32,     4,  8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    32,   128,     4,  8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>
+    // clang-format on
+    >;
+
+void add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<1,
+                                                  NWC,
+                                                  KXC,
+                                                  NWK,
+                                                  BF16,
+                                                  BF16,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_conv1d_bwd_data_xdl_nwc_kxc_nwk_bf16_instances{});
+    add_device_operation_instances(
+        instances, device_conv1d_bwd_data_xdl_nwc_kxc_nwk_1x1_s1_p0_bf16_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/conv1d_bwd_data/device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f16_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, 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/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F16 = ck::half_t;
+using F32 = float;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using NWC = ck::tensor_layout::convolution::NWC;
+using KXC = ck::tensor_layout::convolution::KXC;
+using NWK = ck::tensor_layout::convolution::NWK;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto ConvBwdDataDefault =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Default;
+
+static constexpr auto ConvBwdDataFilter1x1Stride1Pad0 =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Filter1x1Stride1Pad0;
+
+// Compilation parameters for in[n, wi, c] * wei[k, x, c] = out[n, wo, k]
+using device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f16_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|       ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|               Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|     Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                   |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   256,   128,     4,   8,  32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   256,     4,   8,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,   128,     4,   8,  32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   128,     4,   8,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,    64,   128,     4,   8,  32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    32,   128,     4,   8,  32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    64,   128,     4,   8,  32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,    64,     4,   8,  32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    64,     4,   8,  32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,       
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    64,     4,   8,  32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    32,     4,   8,  32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    32,     4,   8,  32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    32,    64,     4,   8,  32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>
+    // clang-format on
+    >;
+
+using device_conv1d_bwd_data_xdl_nwc_kxc_nwk_1x1_s1_p0_f16_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                    ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                            Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|                  Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                                |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    64,     4,  8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    32,     4,  8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    32,   128,     4,  8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>
+    // clang-format on
+    >;
+
+void add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f16_instances(
+    std::vector<std::unique_ptr<
+        DeviceConvBwdData<1, NWC, KXC, NWK, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(instances,
+                                   device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f16_instances{});
+    add_device_operation_instances(
+        instances, device_conv1d_bwd_data_xdl_nwc_kxc_nwk_1x1_s1_p0_f16_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/conv1d_bwd_data/device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f32_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, 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/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F32 = float;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using NWC = ck::tensor_layout::convolution::NWC;
+using KXC = ck::tensor_layout::convolution::KXC;
+using NWK = ck::tensor_layout::convolution::NWK;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto ConvBwdDataDefault =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Default;
+
+static constexpr auto ConvBwdDataFilter1x1Stride1Pad0 =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Filter1x1Stride1Pad0;
+
+// Compilation parameters for in[n, wi, c] * wei[k, x, c] = out[n, wo, k]
+using device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f32_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|        ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|      Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                    |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   256,   256,   128,     4,  4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   256,   128,   256,     4,  4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   128,   128,   128,     4,  4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   256,   128,   128,     4,  4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   128,   128,    64,     4,  4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   128,    64,   128,     4,  4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,    64,    64,    64,     4,  4,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   256,   128,    64,     4,  4,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   256,    64,   128,     4,  4,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   128,   128,    32,     4,  4,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,   128,    32,   128,     4,  4,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,    64,    64,    32,     4,  4,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataDefault,    64,    32,    64,     4,  4,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>
+    // clang-format on
+    >;
+
+using device_conv1d_bwd_data_xdl_nwc_kxc_nwk_1x1_s1_p0_f32_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                     ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                             Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|                   Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                                 |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   256,   256,   128,     4,  4,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   256,     4,  4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   128,   128,   128,     4,  4,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   128,     4,  4,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    64,     4,  4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   128,    64,   128,     4,  4,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    64,     4,  4,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   256,   128,    64,     4,  4,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   256,    64,   128,     4,  4,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    32,     4,  4,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,   128,    32,   128,     4,  4,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    32,     4,  4,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              4,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,  ConvBwdDataFilter1x1Stride1Pad0,    64,    32,    64,     4,  4,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              4,      true,               7,               1>
+    // clang-format on
+    >;
+
+void add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f32_instances(
+    std::vector<std::unique_ptr<
+        DeviceConvBwdData<1, NWC, KXC, NWK, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
+{
+    add_device_operation_instances(instances,
+                                   device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f32_instances{});
+    add_device_operation_instances(
+        instances, device_conv1d_bwd_data_xdl_nwc_kxc_nwk_1x1_s1_p0_f32_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/conv1d_bwd_data/device_conv1d_bwd_data_xdl_nwc_kxc_nwk_int8_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, 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/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using NWC = ck::tensor_layout::convolution::NWC;
+using KXC = ck::tensor_layout::convolution::KXC;
+using NWK = ck::tensor_layout::convolution::NWK;
+
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
+static constexpr auto ConvBwdDataDefault =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Default;
+
+static constexpr auto ConvBwdDataFilter1x1Stride1Pad0 =
+    ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Filter1x1Stride1Pad0;
+
+// Compilation parameters for in[n, wi, c] * wei[k, x, c] = out[n, wo, k]
+using device_conv1d_bwd_data_xdl_nwc_kxc_nwk_int8_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|       ConvBackward| Block|  MPer|  NPer| K0Per|  K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|               Data|  Size| Block| Block| Block|    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|     Specialization|      |      |      |      |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                   |      |      |      |      |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   256,     4,  16,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,   128,     4,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   256,   128,     4,  16,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   128,     4,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    64,     4,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    64,   128,     4,  16,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    64,     4,  16,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,    64,     4,  16,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,    64,   128,     4,  16,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    32,     4,  16,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    32,   128,     4,  16,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    32,     4,  16,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    32,    64,     4,  16,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>
+    // clang-format on
+    >;
+
+using device_conv1d_bwd_data_xdl_nwc_kxc_nwk_1x1_s1_p0_int8_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                    ConvBackward| Block|  MPer|  NPer| K0Per|  K1| MPer| NPer| MXdl| NXdl|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                            Data|  Size| Block| Block| Block|    |  XDL|  XDL|  Per|  Per|    ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|                  Specialization|      |      |      |      |    |     |     | Wave| Wave|  Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                                |      |      |      |      |    |     |     |     |     |                 |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   256,   128,     4,  16,   32,   32,    4,    2,      S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   256,     4,  16,   32,   32,    2,    4,      S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,   128,     4,  16,   32,   32,    4,    2,      S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   128,     4,  16,   32,   32,    2,    2,      S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    64,     4,  16,   32,   32,    2,    2,      S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    64,   128,     4,  16,   32,   32,    2,    2,      S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    64,     4,  16,   32,   32,    2,    2,      S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,    64,     4,  16,   32,   32,    2,    1,      S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,    64,   128,     4,  16,   32,   32,    1,    2,      S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    32,     4,  16,   32,   32,    2,    1,      S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    32,   128,     4,  16,   32,   32,    1,    2,      S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    32,     4,  16,   32,   32,    2,    1,      S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,             16,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       1, int8_t,  int8_t,  int8_t, int32_t, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    32,    64,     4,  16,   32,   32,    1,    2,      S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,             16,      true,               7,               1>
+    // clang-format on
+    >;
+
+void add_device_conv1d_bwd_data_xdl_nwc_kxc_nwk_int8_instances(
+    std::vector<std::unique_ptr<DeviceConvBwdData<1,
+                                                  NWC,
+                                                  KXC,
+                                                  NWK,
+                                                  int8_t,
+                                                  int8_t,
+                                                  int8_t,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_conv1d_bwd_data_xdl_nwc_kxc_nwk_int8_instances{});
+    add_device_operation_instances(
+        instances, device_conv1d_bwd_data_xdl_nwc_kxc_nwk_1x1_s1_p0_int8_instances{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/conv2d_bwd_data/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instance.cpp

@@ -5,8 +5,11 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp"
+// FIXME: retire dedicated 2D version
 #include "ck/tensor_operation/gpu/device/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
@@ -14,13 +17,18 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-using BF16 = ck::bhalf_t;
+using BF16 = bhalf_t;
 using F32  = float;
 
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
 
+using NHWC = ck::tensor_layout::convolution::NHWC;
+using KYXC = ck::tensor_layout::convolution::KYXC;
+using NHWK = ck::tensor_layout::convolution::NHWK;
+
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
 static constexpr auto ConvBwdDataDefault =
     ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Default;
 
@@ -29,6 +37,52 @@ static constexpr auto ConvBwdDataFilter1x1Stride1Pad0 =
 
 // Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
 using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|       ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|               Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|     Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                   |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    64,     4,  8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    32,     4,  8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    32,   128,     4,  8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>
+    // clang-format on
+    >;
+
+using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_bf16_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                    ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                            Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|                  Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                                |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    64,     4,  8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    32,     4,  8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    32,   128,     4,  8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>
+    // clang-format on
+    >;
+
+// FIXME: retire dedicated 2D version
+// Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
+using device_conv_dedicated_2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instances = std::tuple<
     // clang-format off
         //####################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|        ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
         //####################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
@@ -50,7 +104,8 @@ using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instances = std::tuple<
     // clang-format on
     >;
 
-using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_bf16_instances =
+// FIXME: retire dedicated 2D version
+using device_conv_dedidecate_2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_bf16_instances =
     std::tuple<
         // clang-format off
         //####################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                     ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
@@ -74,12 +129,26 @@ using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_bf16_instances =
         >;
 
 void add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instances(
-    std::vector<DeviceConvBwdDataPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<DeviceConvBwdData<2,
+                                                  NHWC,
+                                                  KYXC,
+                                                  NHWK,
+                                                  BF16,
+                                                  BF16,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
 {
     add_device_operation_instances(instances,
                                    device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instances{});
     add_device_operation_instances(
         instances, device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_bf16_instances{});
+    add_device_operation_instances(
+        instances, device_conv_dedicated_2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instances{});
+    add_device_operation_instances(
+        instances,
+        device_conv_dedidecate_2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_bf16_instances{});
 }
 
 } // namespace instance

library/src/tensor_operation_instance/gpu/conv2d_bwd_data/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instance.cpp

@@ -5,8 +5,11 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp"
+// FIXME: retire dedicated 2D version
 #include "ck/tensor_operation/gpu/device/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
@@ -20,7 +23,12 @@ using F32 = float;
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
 
+using NHWC = ck::tensor_layout::convolution::NHWC;
+using KYXC = ck::tensor_layout::convolution::KYXC;
+using NHWK = ck::tensor_layout::convolution::NHWK;
+
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+
 static constexpr auto ConvBwdDataDefault =
     ck::tensor_operation::device::ConvolutionBackwardDataSpecialization::Default;
 
@@ -29,6 +37,52 @@ static constexpr auto ConvBwdDataFilter1x1Stride1Pad0 =
 
 // Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
 using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|       ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|               Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|     Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                   |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   256,   128,     4,   8,  32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   256,     4,   8,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,   128,     4,   8,  32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,   128,     4,   8,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,    64,   128,     4,   8,  32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    32,   128,     4,   8,  32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,    64,   128,     4,   8,  32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   256,   128,    64,     4,   8,  32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    64,     4,   8,  32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,       
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    64,     4,   8,  32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,   128,   128,    32,     4,   8,  32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    64,    32,     4,   8,  32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataDefault,    64,    32,    64,     4,   8,  32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>
+    // clang-format on
+    >;
+
+using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances = std::tuple<
+    // clang-format off
+        //##############################|     Num| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                    ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
+        //##############################|     Dim|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                            Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
+        //##############################| Spatial|       |        |        |        |   Operation|   Operation|   Operation|                  Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|
+        //##############################|        |       |        |        |        |            |            |            |                                |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    64,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    64,   128,     4,  8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    64,     4,  8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,   128,    32,     4,  8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,   128,    32,   128,     4,  8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>,
+        DeviceConvNdBwdDataNwcKxcNwk_Xdl<       2,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvBwdDataFilter1x1Stride1Pad0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<2, 0, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1>
+    // clang-format on
+    >;
+
+// FIXME: retire dedicated 2D version
+// Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
+using device_conv_dedicated_2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instances = std::tuple<
     // clang-format off
         //####################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|       ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
         //####################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|               Data|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|
@@ -52,7 +106,8 @@ using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instances = std::tuple<
     // clang-format on
     >;
 
-using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances =
+// FIXME: retire dedicated 2D version
+using device_conv_dedicated_2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances =
     std::tuple<
         // clang-format off
         //####################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                     ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer|
@@ -76,12 +131,25 @@ using device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances =
         >;
 
 void add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instances(
-    std::vector<DeviceConvBwdDataPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<DeviceConvBwdData<2,
+                                                  NHWC,
+                                                  KYXC,
+                                                  NHWK,
+                                                  F16,
+                                                  F16,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
 {
     add_device_operation_instances(instances,
                                    device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instances{});
     add_device_operation_instances(
         instances, device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances{});
+    add_device_operation_instances(
+        instances, device_conv_dedicated_2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instances{});
+    add_device_operation_instances(
+        instances, device_conv_dedicated_2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances{});
 }
 
 } // namespace instance