Merge branch 'amd-develop-staging-0423' into amd-master

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

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -16,6 +16,7 @@ namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
+// 2D
 #ifdef CK_ENABLE_FP32
 void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance(
     std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
@@ -387,6 +388,378 @@ void add_device_contraction_scale_m2_n2_k2_xdl_c_shuffle_bf16_bf16_bf16_compute_
                                                            Scale,
                                                            F32>>>& instances);
 #endif // CK_ENABLE_FP16
+// 6D
+#ifdef CK_ENABLE_FP32
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F32,
+                                                           F32,
+                                                           Empty_Tuple,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           BF16>>>& instances);
+#endif // CK_ENABLE_FP32
+
+#ifdef CK_ENABLE_FP64
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F64>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F64,
+                                                           F64,
+                                                           Empty_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP64
+
+#ifdef CK_ENABLE_FP16
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           F16,
+                                                           F16,
+                                                           Empty_Tuple,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP16
+
+#ifdef CK_ENABLE_BF16
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+
+void add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<6,
+                                                           6,
+                                                           6,
+                                                           BF16,
+                                                           BF16,
+                                                           Empty_Tuple,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Scale,
+                                                           F32>>>& instances);
+#endif // CK_ENABLE_FP16
 
 // Contraction + Scale
 template <index_t NumDimM,
@@ -464,6 +837,42 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_mnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, ck::half_t>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_f16_mnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, ck::bhalf_t>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f32_f32_f32_compute_bf16_mnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP32
 #ifdef CK_ENABLE_FP64
@@ -495,6 +904,31 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, double>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_mnn_instance(
+                        op_ptrs);
+                }
+                else if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f64_f64_f64_compute_f32_mnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP64
 #ifdef CK_ENABLE_FP16
@@ -515,6 +949,20 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_f16_f16_f16_compute_f32_mnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_FP16
 #ifdef CK_ENABLE_BF16
@@ -535,6 +983,20 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceContra
                         op_ptrs);
                 }
             }
+            else if constexpr(NumDimM == 6 && NumDimN == 6 && NumDimK == 6)
+            {
+                if constexpr(is_same_v<ComputeDataType, float>)
+                {
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_kkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_knn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mkn_instance(
+                        op_ptrs);
+                    add_device_contraction_scale_m6_n6_k6_xdl_c_shuffle_bf16_bf16_bf16_compute_f32_mnn_instance(
+                        op_ptrs);
+                }
+            }
         }
 #endif // CK_ENABLE_BF16
         return op_ptrs;

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

@@ -12,397 +12,20 @@
 
 #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
 
-namespace ck {
-namespace tensor_operation {
-namespace device {
-namespace instance {
-
-#if defined(CK_ENABLE_FP16) && defined(DL_KERNELS)
-void add_device_gemm_dl_f16_f16_f16_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f16_f16_f16_km_kn_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dpp_f16_f16_f16_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dpp_f16_f16_f16_km_kn_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f16_f16_f16_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f16_f16_f16_km_nk_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dpp_f16_f16_f16_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dpp_f16_f16_f16_km_nk_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f16_f16_f16_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f16_f16_f16_mk_kn_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dpp_f16_f16_f16_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dpp_f16_f16_f16_mk_kn_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f16_f16_f16_mk_nk_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dpp_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dpp_f16_f16_f16_mk_nk_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-#endif
-#if defined(CK_ENABLE_FP32) && defined(DL_KERNELS)
-void add_device_gemm_dl_f32_f32_f32_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f32_f32_f32_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f32_f32_f32_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_f32_f32_f32_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-#endif
-#if defined(CK_ENABLE_INT8) && defined(DL_KERNELS)
-void add_device_gemm_dl_i8_i8_i8_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_i8_i8_i8_km_kn_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_i8_i8_i8_km_nk_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_i8_i8_i8_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_i8_i8_i8_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_dl_i8_i8_i8_mk_nk_mn_irregular_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-#endif
-#ifdef CK_ENABLE_INT8
-void add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f16_f16_f16_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f16_f16_f16_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f16_f16_f16_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_lds_direct_load_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
-        instances);
+#ifdef DL_KERNELS
+#include "gemm_dl.inc"
 #endif
-#ifdef CK_ENABLE_FP64
-void add_device_gemm_xdl_f64_f64_f64_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-
-        DeviceGemm<Col, Row, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f64_f64_f64_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f64_f64_f64_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_f64_f64_f64_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
-        instances);
+#ifdef CK_USE_WMMA
+#include "gemm_wmma.inc"
 #endif
-#ifdef CK_ENABLE_FP8
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v1_default_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v1_interwave_default_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v2_default_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v1_padded_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v1_interwave_padded_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v2_padded_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
-
-void add_device_gemm_xdl_c_shuffle_f16_f8_f16_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_xdl_c_shuffle_f16_f8_f16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
+#ifdef CK_USE_XDL
+#include "gemm_xdl.inc"
 #endif
 
-void add_device_gemm_wmma_f16_f16_f16_km_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_wmma_f16_f16_f16_km_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_wmma_f16_f16_f16_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
-
-void add_device_gemm_wmma_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<std::unique_ptr<
-        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
-        instances);
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
 
 template <typename ALayout,
           typename BLayout,
@@ -435,52 +58,29 @@ struct DeviceOperationInstanceFactory<
     {
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
 
+#ifdef DL_KERNELS
         if constexpr(is_same_v<ADataType, float> && is_same_v<BDataType, float> &&
                      is_same_v<CDataType, float>)
         {
             if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
                          is_same_v<CLayout, Row>)
             {
-                /// add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
                 add_device_gemm_dl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
-#endif
-                add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
-                add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_mk_kn_mn_instances(
-                    op_ptrs);
             }
             else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
                               is_same_v<CLayout, Row>)
             {
-                /// add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
                 add_device_gemm_dl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
-#endif
-                add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
-                add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_mk_nk_mn_instances(
-                    op_ptrs);
             }
             else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
                               is_same_v<CLayout, Row>)
             {
-                /// add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
                 add_device_gemm_dl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
-#endif
-                add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instances(op_ptrs);
-                add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_km_kn_mn_instances(
-                    op_ptrs);
             }
             else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
                               is_same_v<CLayout, Row>)
             {
-                /// add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
                 add_device_gemm_dl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
-#endif
-                add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(op_ptrs);
-                add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_km_nk_mn_instances(
-                    op_ptrs);
             }
         }
 #ifdef CK_ENABLE_FP16
@@ -490,60 +90,160 @@ struct DeviceOperationInstanceFactory<
             if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
                          is_same_v<CLayout, Row>)
             {
-                /// add_device_gemm_xdl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
                 add_device_gemm_dl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
                 add_device_gemm_dl_f16_f16_f16_mk_kn_mn_irregular_instances(op_ptrs);
                 add_device_gemm_dpp_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
                 add_device_gemm_dpp_f16_f16_f16_mk_kn_mn_irregular_instances(op_ptrs);
-#endif
-                add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
-                add_device_gemm_wmma_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
             }
             else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
                               is_same_v<CLayout, Row>)
             {
-                /// add_device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
                 add_device_gemm_dl_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
                 add_device_gemm_dl_f16_f16_f16_mk_nk_mn_irregular_instances(op_ptrs);
                 add_device_gemm_dpp_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
                 add_device_gemm_dpp_f16_f16_f16_mk_nk_mn_irregular_instances(op_ptrs);
-#endif
-                add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
-                add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
-                add_device_gemm_xdl_c_shuffle_lds_direct_load_f16_f16_f16_mk_nk_mn_instances(
-                    op_ptrs);
-                add_device_gemm_wmma_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
             }
             else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
                               is_same_v<CLayout, Row>)
             {
-                /// add_device_gemm_xdl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
                 add_device_gemm_dl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
                 add_device_gemm_dl_f16_f16_f16_km_kn_mn_irregular_instances(op_ptrs);
                 add_device_gemm_dpp_f16_f16_f16_km_kn_mn_instances(op_ptrs);
                 add_device_gemm_dpp_f16_f16_f16_km_kn_mn_irregular_instances(op_ptrs);
-#endif
-                add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(op_ptrs);
-                add_device_gemm_wmma_f16_f16_f16_km_kn_mn_instances(op_ptrs);
             }
             else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
                               is_same_v<CLayout, Row>)
             {
-                /// add_device_gemm_xdl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
                 add_device_gemm_dl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
                 add_device_gemm_dl_f16_f16_f16_km_nk_mn_irregular_instances(op_ptrs);
                 add_device_gemm_dpp_f16_f16_f16_km_nk_mn_instances(op_ptrs);
                 add_device_gemm_dpp_f16_f16_f16_km_nk_mn_irregular_instances(op_ptrs);
+            }
+        }
 #endif
-                add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(op_ptrs);
+#ifdef CK_ENABLE_INT8
+        else if constexpr(is_same_v<ADataType, int8_t> && is_same_v<BDataType, int8_t> &&
+                          is_same_v<CDataType, int8_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_dl_i8_i8_i8_mk_kn_mn_instances(op_ptrs);
+                add_device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_dl_i8_i8_i8_mk_nk_mn_instances(op_ptrs);
+                add_device_gemm_dl_i8_i8_i8_mk_nk_mn_irregular_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_dl_i8_i8_i8_km_kn_mn_instances(op_ptrs);
+                add_device_gemm_dl_i8_i8_i8_km_kn_mn_irregular_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(op_ptrs);
+                add_device_gemm_dl_i8_i8_i8_km_nk_mn_irregular_instances(op_ptrs);
+            }
+        }
+#endif
+#endif // DL_KERNELS
+
+#ifdef CK_USE_WMMA
+#ifdef CK_ENABLE_FP16
+        if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
+                     is_same_v<CDataType, half_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_wmma_f16_f16_f16_km_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
                 add_device_gemm_wmma_f16_f16_f16_km_nk_mn_instances(op_ptrs);
             }
         }
 #endif
+#endif
+
+#ifdef CK_USE_XDL
+        if constexpr(is_same_v<ADataType, float> && is_same_v<BDataType, float> &&
+                     is_same_v<CDataType, float>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
+                add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_mk_kn_mn_instances(
+                    op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
+                add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_mk_nk_mn_instances(
+                    op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instances(op_ptrs);
+                add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_km_kn_mn_instances(
+                    op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(op_ptrs);
+                add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_km_nk_mn_instances(
+                    op_ptrs);
+            }
+        }
+#ifdef CK_ENABLE_FP16
+        else if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
+                          is_same_v<CDataType, half_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
+                add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
+                add_device_gemm_xdl_c_shuffle_lds_direct_load_f16_f16_f16_mk_nk_mn_instances(
+                    op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(op_ptrs);
+            }
+            else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
+                              is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(op_ptrs);
+            }
+        }
+#endif
 #ifdef CK_ENABLE_BF16
         else if constexpr(is_same_v<ADataType, ck::bhalf_t> && is_same_v<BDataType, ck::bhalf_t> &&
                           is_same_v<CDataType, ck::bhalf_t>)
@@ -578,37 +278,21 @@ struct DeviceOperationInstanceFactory<
                          is_same_v<CLayout, Row>)
             {
                 add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
-                add_device_gemm_dl_i8_i8_i8_mk_kn_mn_instances(op_ptrs);
-                add_device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instances(op_ptrs);
-#endif
             }
             else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
                               is_same_v<CLayout, Row>)
             {
                 add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
-                add_device_gemm_dl_i8_i8_i8_mk_nk_mn_instances(op_ptrs);
-                add_device_gemm_dl_i8_i8_i8_mk_nk_mn_irregular_instances(op_ptrs);
-#endif
             }
             else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
                               is_same_v<CLayout, Row>)
             {
                 add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
-                add_device_gemm_dl_i8_i8_i8_km_kn_mn_instances(op_ptrs);
-                add_device_gemm_dl_i8_i8_i8_km_kn_mn_irregular_instances(op_ptrs);
-#endif
             }
             else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
                               is_same_v<CLayout, Row>)
             {
                 add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instances(op_ptrs);
-#ifdef DL_KERNELS
-                add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(op_ptrs);
-                add_device_gemm_dl_i8_i8_i8_km_nk_mn_irregular_instances(op_ptrs);
-#endif
             }
         }
 #endif
@@ -658,6 +342,7 @@ struct DeviceOperationInstanceFactory<
                 add_device_gemm_xdl_c_shuffle_f16_f8_f16_mk_nk_mn_instances(op_ptrs);
             }
         }
+#endif
 #endif
         return op_ptrs;
     }

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

@@ -16,7 +16,7 @@ namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
-#ifdef CK_ENABLE_FP16
+#if defined(CK_ENABLE_FP16) && defined(CK_USE_XDL)
 void add_device_gemm_bilinear_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
                                                     Row,
@@ -69,7 +69,7 @@ void add_device_gemm_bilinear_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instance
                                                     PassThrough,
                                                     Bilinear>>>& instances);
 #endif
-#ifdef CK_ENABLE_INT8
+#if defined(CK_ENABLE_INT8) && defined(CK_USE_WMMA)
 void add_device_gemm_bilinear_wmma_c_shuffle_i8_i8_i8_i8_mk_kn_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
                                                     Row,
@@ -159,7 +159,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGemmMu
     static auto GetInstances()
     {
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
-#ifdef CK_ENABLE_FP16
+#if defined(CK_ENABLE_FP16) && defined(CK_USE_XDL)
         if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
                      is_same_v<DDataType, half_t> && is_same_v<EDataType, half_t>)
         {
@@ -189,7 +189,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGemmMu
             }
         }
 #endif
-#ifdef CK_ENABLE_INT8
+#if defined(CK_ENABLE_INT8) && defined(CK_USE_WMMA)
         if constexpr(is_same_v<ADataType, std::int8_t> && is_same_v<BDataType, std::int8_t> &&
                      is_same_v<DDataType, std::int8_t> && is_same_v<EDataType, std::int8_t>)
         {

library/include/ck/library/tensor_operation_instance/gpu/gemm_dl.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <memory>
+#include <vector>
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/device_gemm.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 {
+
+#if defined(CK_ENABLE_FP16)
+void add_device_gemm_dl_f16_f16_f16_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f16_f16_f16_km_kn_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dpp_f16_f16_f16_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dpp_f16_f16_f16_km_kn_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f16_f16_f16_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f16_f16_f16_km_nk_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dpp_f16_f16_f16_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dpp_f16_f16_f16_km_nk_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f16_f16_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f16_f16_f16_mk_kn_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dpp_f16_f16_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dpp_f16_f16_f16_mk_kn_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f16_f16_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f16_f16_f16_mk_nk_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dpp_f16_f16_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dpp_f16_f16_f16_mk_nk_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+#if defined(CK_ENABLE_FP32)
+void add_device_gemm_dl_f32_f32_f32_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f32_f32_f32_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f32_f32_f32_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_f32_f32_f32_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+#if defined(CK_ENABLE_INT8)
+void add_device_gemm_dl_i8_i8_i8_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_i8_i8_i8_km_kn_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_i8_i8_i8_km_nk_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_i8_i8_i8_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_i8_i8_i8_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_dl_i8_i8_i8_mk_nk_mn_irregular_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/gemm_wmma.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_wmma_f16_f16_f16_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_f16_f16_f16_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_f16_f16_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_wmma_f16_f16_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/gemm_xdl.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+#ifdef CK_ENABLE_INT8
+void add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f16_f16_f16_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f16_f16_f16_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f16_f16_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_lds_direct_load_f16_f16_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+#ifdef CK_ENABLE_BF16
+void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_lds_direct_load_f32_f32_f32_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+#ifdef CK_ENABLE_FP64
+void add_device_gemm_xdl_f64_f64_f64_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+
+        DeviceGemm<Col, Row, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f64_f64_f64_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f64_f64_f64_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_f64_f64_f64_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+#ifdef CK_ENABLE_FP8
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_km_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_km_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v1_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v1_interwave_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v2_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v1_padded_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v1_interwave_padded_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_v2_padded_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
+
+void add_device_gemm_xdl_c_shuffle_f16_f8_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_c_shuffle_f16_f8_f16_mk_nk_mn_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_instance.hpp

@@ -17,6 +17,10 @@ namespace instance {
 using F8 = ck::f8_t;
 #endif
 
+#ifdef CK_ENABLE_BF8
+using BF8 = ck::bf8_t;
+#endif
+
 using BF16 = ck::bhalf_t;
 using F16  = ck::half_t;
 using F32  = float;
@@ -250,6 +254,114 @@ using device_grouped_conv_fwd_xdl_f8_instances = std::tuple<
     // clang-format on
     >;
 
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_bf8_instances = std::tuple<
+// clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| ComputeType|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            |
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            |
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |            |
+#ifdef CK_ENABLE_BF8
+        // generic instance
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1,         BF8>,
+        // instances for small conv.K and conv.C
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8>,
+
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   BF8,   BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         BF8>
+#endif
+    // clang-format on
+    >;
+
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_f8_bf8_instances = std::tuple<
+// clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|AComputeType|BComputeType|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            |            |
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            |            |
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |            |            |
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+        // generic instance
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1,         F8,         BF8>,
+        // instances for small conv.K and conv.C
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         F8,         BF8>,
+
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         F8,         BF8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F8,  BF8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         F8,         BF8>
+#endif
+    // clang-format on
+    >;
+
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_bf8_f8_instances = std::tuple<
+// clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|AComputeType|BComputeType|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            |            |
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            |            |
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |            |            |
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+        // generic instance
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1,         BF8,          F8>,
+        // instances for small conv.K and conv.C
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8,         F8>,
+
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,   128,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,   128,    32,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         BF8,         F8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF8,   F8,     F32,      F8,    DsLayout,   F8, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,               8,         BF8,         F8>
+#endif
+    // clang-format on
+    >;
+
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation

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

@@ -10,439 +10,18 @@
 
 #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
 
+#ifdef CK_USE_XDL
+#include "grouped_convolution_backward_data_xdl.inc"
+#endif
+#ifdef CK_USE_WMMA
+#include "grouped_convolution_backward_data_wmma.inc"
+#endif
+
 namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
 
-// conv2d backward data
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  GNHWK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  GNHWC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  GNHWK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  GNHWC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  GNHWK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  GNHWC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  GNHWK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  GNHWC,
-                                                                  F32,
-                                                                  F32,
-                                                                  Empty_Tuple,
-                                                                  F32,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  GNHWK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  GNHWC,
-                                                                  BF16,
-                                                                  BF16,
-                                                                  Empty_Tuple,
-                                                                  BF16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  GNHWK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  GNHWC,
-                                                                  int8_t,
-                                                                  int8_t,
-                                                                  Empty_Tuple,
-                                                                  int8_t,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  GNHWK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  GNHWC,
-                                                                  int8_t,
-                                                                  int8_t,
-                                                                  Empty_Tuple,
-                                                                  int8_t,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  NHWGK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  NHWGC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  NHWGK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  NHWGC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  NHWGK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  NHWGC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  NHWGK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  NHWGC,
-                                                                  F32,
-                                                                  F32,
-                                                                  Empty_Tuple,
-                                                                  F32,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  NHWGK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  NHWGC,
-                                                                  BF16,
-                                                                  BF16,
-                                                                  Empty_Tuple,
-                                                                  BF16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  NHWGK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  NHWGC,
-                                                                  int8_t,
-                                                                  int8_t,
-                                                                  Empty_Tuple,
-                                                                  int8_t,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
-                                                                  NHWGK,
-                                                                  GKYXC,
-                                                                  Empty_Tuple,
-                                                                  NHWGC,
-                                                                  int8_t,
-                                                                  int8_t,
-                                                                  Empty_Tuple,
-                                                                  int8_t,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-// conv3d backward data
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  GNDHWK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  GNDHWC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  GNDHWK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  GNDHWC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  GNDHWK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  GNDHWC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  GNDHWK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  GNDHWC,
-                                                                  F32,
-                                                                  F32,
-                                                                  Empty_Tuple,
-                                                                  F32,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  GNDHWK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  GNDHWC,
-                                                                  BF16,
-                                                                  BF16,
-                                                                  Empty_Tuple,
-                                                                  BF16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  GNDHWK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  GNDHWC,
-                                                                  int8_t,
-                                                                  int8_t,
-                                                                  Empty_Tuple,
-                                                                  int8_t,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  GNDHWK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  GNDHWC,
-                                                                  int8_t,
-                                                                  int8_t,
-                                                                  Empty_Tuple,
-                                                                  int8_t,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  NDHWGK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  NDHWGC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  NDHWGK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  NDHWGC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  NDHWGK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  NDHWGC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  NDHWGK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  NDHWGC,
-                                                                  F32,
-                                                                  F32,
-                                                                  Empty_Tuple,
-                                                                  F32,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  NDHWGK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  NDHWGC,
-                                                                  BF16,
-                                                                  BF16,
-                                                                  Empty_Tuple,
-                                                                  BF16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  NDHWGK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  NDHWGC,
-                                                                  int8_t,
-                                                                  int8_t,
-                                                                  Empty_Tuple,
-                                                                  int8_t,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  NDHWGK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  NDHWGC,
-                                                                  int8_t,
-                                                                  int8_t,
-                                                                  Empty_Tuple,
-                                                                  int8_t,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough>>>& instances);
-#endif
-#if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-void add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_input_f16_comp_bf8f8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
-                                                                  NDHWGK,
-                                                                  GKZYXC,
-                                                                  Empty_Tuple,
-                                                                  NDHWGC,
-                                                                  F16,
-                                                                  F16,
-                                                                  Empty_Tuple,
-                                                                  F16,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  PassThrough,
-                                                                  BF8,
-                                                                  F8>>>& instances);
-#endif
 template <ck::index_t NumDimSpatial,
           typename OutLayout,
           typename WeiLayout,
@@ -488,9 +67,10 @@ struct DeviceOperationInstanceFactory<
     static auto GetInstances()
     {
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
+
+#ifdef CK_USE_XDL
         if constexpr(NumDimSpatial == 2)
         {
-
             if constexpr(is_same_v<InLayout, GNHWC> && is_same_v<WeiLayout, GKYXC> &&
                          is_same_v<OutLayout, GNHWK>)
             {
@@ -500,43 +80,28 @@ struct DeviceOperationInstanceFactory<
                              is_same_v<ComputeTypeB, F16>)
                 {
                     add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_f16_instances(op_ptrs);
-                    add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_instances(
-                        op_ptrs);
-                    add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_1x1s1p0_instances(
-                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP32
-                else if constexpr(is_same_v<InDataType, F32> && is_same_v<WeiDataType, F32> &&
-                                  is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
-                                  is_same_v<ComputeTypeB, F32>)
+                if constexpr(is_same_v<InDataType, F32> && is_same_v<WeiDataType, F32> &&
+                             is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
+                             is_same_v<ComputeTypeB, F32>)
                 {
                     add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_f32_instances(op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, BF16> && is_same_v<WeiDataType, BF16> &&
-                                  is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
-                                  is_same_v<ComputeTypeB, BF16>)
+                if constexpr(is_same_v<InDataType, BF16> && is_same_v<WeiDataType, BF16> &&
+                             is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
+                             is_same_v<ComputeTypeB, BF16>)
                 {
                     add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_bf16_instances(
                         op_ptrs);
                 }
-#endif
-#ifdef CK_ENABLE_INT8
-                else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                                  is_same_v<OutDataType, int8_t> &&
-                                  is_same_v<ComputeTypeA, int8_t> &&
-                                  is_same_v<ComputeTypeB, int8_t>)
-                {
-                    add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_instances(op_ptrs);
-                    add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_1x1s1p0_instances(
-                        op_ptrs);
-                }
 #endif
             }
-            else if constexpr(is_same_v<InLayout, NHWGC> && is_same_v<WeiLayout, GKYXC> &&
-                              is_same_v<OutLayout, NHWGK>)
+            if constexpr(is_same_v<InLayout, NHWGC> && is_same_v<WeiLayout, GKYXC> &&
+                         is_same_v<OutLayout, NHWGK>)
             {
 #ifdef CK_ENABLE_FP16
                 if constexpr(is_same_v<InDataType, F16> && is_same_v<WeiDataType, F16> &&
@@ -544,45 +109,29 @@ struct DeviceOperationInstanceFactory<
                              is_same_v<ComputeTypeB, F16>)
                 {
                     add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_f16_instances(op_ptrs);
-                    add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_instances(
-                        op_ptrs);
-                    add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_1x1s1p0_instances(
-                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP32
-                else if constexpr(is_same_v<InDataType, F32> && is_same_v<WeiDataType, F32> &&
-                                  is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
-                                  is_same_v<ComputeTypeB, F32>)
+                if constexpr(is_same_v<InDataType, F32> && is_same_v<WeiDataType, F32> &&
+                             is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
+                             is_same_v<ComputeTypeB, F32>)
                 {
                     add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_f32_instances(op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, BF16> && is_same_v<WeiDataType, BF16> &&
-                                  is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
-                                  is_same_v<ComputeTypeB, BF16>)
+                if constexpr(is_same_v<InDataType, BF16> && is_same_v<WeiDataType, BF16> &&
+                             is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
+                             is_same_v<ComputeTypeB, BF16>)
                 {
                     add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_bf16_instances(
                         op_ptrs);
                 }
-#endif
-#ifdef CK_ENABLE_INT8
-                else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                                  is_same_v<OutDataType, int8_t> &&
-                                  is_same_v<ComputeTypeA, int8_t> &&
-                                  is_same_v<ComputeTypeB, int8_t>)
-                {
-                    add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_instances(op_ptrs);
-                    add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_1x1s1p0_instances(
-                        op_ptrs);
-                }
 #endif
             }
         }
-        else if constexpr(NumDimSpatial == 3)
+        if constexpr(NumDimSpatial == 3)
         {
-
             if constexpr(is_same_v<InLayout, GNDHWC> && is_same_v<WeiLayout, GKZYXC> &&
                          is_same_v<OutLayout, GNDHWK>)
             {
@@ -593,45 +142,29 @@ struct DeviceOperationInstanceFactory<
                 {
                     add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_f16_instances(
                         op_ptrs);
-                    add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_instances(
-                        op_ptrs);
-                    add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_1x1s1p0_instances(
-                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP32
-                else if constexpr(is_same_v<InDataType, F32> && is_same_v<WeiDataType, F32> &&
-                                  is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
-                                  is_same_v<ComputeTypeB, F32>)
+                if constexpr(is_same_v<InDataType, F32> && is_same_v<WeiDataType, F32> &&
+                             is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
+                             is_same_v<ComputeTypeB, F32>)
                 {
                     add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_f32_instances(
                         op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, BF16> && is_same_v<WeiDataType, BF16> &&
-                                  is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
-                                  is_same_v<ComputeTypeB, BF16>)
+                if constexpr(is_same_v<InDataType, BF16> && is_same_v<WeiDataType, BF16> &&
+                             is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
+                             is_same_v<ComputeTypeB, BF16>)
                 {
                     add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_bf16_instances(
                         op_ptrs);
                 }
-#endif
-#ifdef CK_ENABLE_INT8
-                else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                                  is_same_v<OutDataType, int8_t> &&
-                                  is_same_v<ComputeTypeA, int8_t> &&
-                                  is_same_v<ComputeTypeB, int8_t>)
-                {
-                    add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_instances(
-                        op_ptrs);
-                    add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_1x1s1p0_instances(
-                        op_ptrs);
-                }
 #endif
             }
-            else if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
-                              is_same_v<OutLayout, NDHWGK>)
+            if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
+                         is_same_v<OutLayout, NDHWGK>)
             {
 #ifdef CK_ENABLE_FP16
                 if constexpr(is_same_v<InDataType, F16> && is_same_v<WeiDataType, F16> &&
@@ -640,44 +173,139 @@ struct DeviceOperationInstanceFactory<
                 {
                     add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_f16_instances(
                         op_ptrs);
-                    add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_instances(
-                        op_ptrs);
-                    add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_1x1s1p0_instances(
-                        op_ptrs);
                 }
 #endif
 #if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-                else if constexpr(is_same_v<InDataType, F16> && is_same_v<WeiDataType, F16> &&
-                                  is_same_v<OutDataType, F16> && is_same_v<ComputeTypeA, bf8_t> &&
-                                  is_same_v<ComputeTypeB, f8_t>)
+                if constexpr(is_same_v<InDataType, F16> && is_same_v<WeiDataType, F16> &&
+                             is_same_v<OutDataType, F16> && is_same_v<ComputeTypeA, bf8_t> &&
+                             is_same_v<ComputeTypeB, f8_t>)
                 {
                     add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_input_f16_comp_bf8f8_instances(
                         op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP32
-                else if constexpr(is_same_v<InDataType, F32> && is_same_v<WeiDataType, F32> &&
-                                  is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
-                                  is_same_v<ComputeTypeB, F32>)
+                if constexpr(is_same_v<InDataType, F32> && is_same_v<WeiDataType, F32> &&
+                             is_same_v<OutDataType, F32> && is_same_v<ComputeTypeA, F32> &&
+                             is_same_v<ComputeTypeB, F32>)
                 {
                     add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_f32_instances(
                         op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
-                else if constexpr(is_same_v<InDataType, BF16> && is_same_v<WeiDataType, BF16> &&
-                                  is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
-                                  is_same_v<ComputeTypeB, BF16>)
+                if constexpr(is_same_v<InDataType, BF16> && is_same_v<WeiDataType, BF16> &&
+                             is_same_v<OutDataType, BF16> && is_same_v<ComputeTypeA, BF16> &&
+                             is_same_v<ComputeTypeB, BF16>)
                 {
                     add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_bf16_instances(
                         op_ptrs);
                 }
+#endif
+            }
+        }
+#endif
+
+#ifdef CK_USE_WMMA
+        if constexpr(NumDimSpatial == 2)
+        {
+            if constexpr(is_same_v<InLayout, GNHWC> && is_same_v<WeiLayout, GKYXC> &&
+                         is_same_v<OutLayout, GNHWK>)
+            {
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, F16> && is_same_v<WeiDataType, F16> &&
+                             is_same_v<OutDataType, F16> && is_same_v<ComputeTypeA, F16> &&
+                             is_same_v<ComputeTypeB, F16>)
+                {
+                    add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_1x1s1p0_instances(
+                        op_ptrs);
+                }
 #endif
 #ifdef CK_ENABLE_INT8
-                else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                                  is_same_v<OutDataType, int8_t> &&
-                                  is_same_v<ComputeTypeA, int8_t> &&
-                                  is_same_v<ComputeTypeB, int8_t>)
+                if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                             is_same_v<OutDataType, int8_t> && is_same_v<ComputeTypeA, int8_t> &&
+                             is_same_v<ComputeTypeB, int8_t>)
+                {
+                    add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_instances(op_ptrs);
+                    add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_1x1s1p0_instances(
+                        op_ptrs);
+                }
+#endif
+            }
+            if constexpr(is_same_v<InLayout, NHWGC> && is_same_v<WeiLayout, GKYXC> &&
+                         is_same_v<OutLayout, NHWGK>)
+            {
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, F16> && is_same_v<WeiDataType, F16> &&
+                             is_same_v<OutDataType, F16> && is_same_v<ComputeTypeA, F16> &&
+                             is_same_v<ComputeTypeB, F16>)
+                {
+                    add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_1x1s1p0_instances(
+                        op_ptrs);
+                }
+#endif
+#ifdef CK_ENABLE_INT8
+                if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                             is_same_v<OutDataType, int8_t> && is_same_v<ComputeTypeA, int8_t> &&
+                             is_same_v<ComputeTypeB, int8_t>)
+                {
+                    add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_instances(op_ptrs);
+                    add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_1x1s1p0_instances(
+                        op_ptrs);
+                }
+#endif
+            }
+        }
+        if constexpr(NumDimSpatial == 3)
+        {
+            if constexpr(is_same_v<InLayout, GNDHWC> && is_same_v<WeiLayout, GKZYXC> &&
+                         is_same_v<OutLayout, GNDHWK>)
+            {
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, F16> && is_same_v<WeiDataType, F16> &&
+                             is_same_v<OutDataType, F16> && is_same_v<ComputeTypeA, F16> &&
+                             is_same_v<ComputeTypeB, F16>)
+                {
+                    add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_1x1s1p0_instances(
+                        op_ptrs);
+                }
+#endif
+#ifdef CK_ENABLE_INT8
+                if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                             is_same_v<OutDataType, int8_t> && is_same_v<ComputeTypeA, int8_t> &&
+                             is_same_v<ComputeTypeB, int8_t>)
+                {
+                    add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_1x1s1p0_instances(
+                        op_ptrs);
+                }
+#endif
+            }
+            else if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
+                              is_same_v<OutLayout, NDHWGK>)
+            {
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, F16> && is_same_v<WeiDataType, F16> &&
+                             is_same_v<OutDataType, F16> && is_same_v<ComputeTypeA, F16> &&
+                             is_same_v<ComputeTypeB, F16>)
+                {
+                    add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_1x1s1p0_instances(
+                        op_ptrs);
+                }
+#endif
+#ifdef CK_ENABLE_INT8
+                if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                             is_same_v<OutDataType, int8_t> && is_same_v<ComputeTypeA, int8_t> &&
+                             is_same_v<ComputeTypeB, int8_t>)
                 {
                     add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_i8_instances(
                         op_ptrs);
@@ -687,6 +315,7 @@ struct DeviceOperationInstanceFactory<
 #endif
             }
         }
+#endif
 
         return op_ptrs;
     }

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_data_wmma.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// conv2d backward data
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  GNHWK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  GNHWC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  GNHWK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  GNHWC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  NHWGK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  NHWGK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  GNDHWK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  GNDHWC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  GNDHWK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  GNDHWC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  NDHWGK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  NDHWGK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_INT8
+void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  GNHWK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  GNHWC,
+                                                                  int8_t,
+                                                                  int8_t,
+                                                                  Empty_Tuple,
+                                                                  int8_t,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_data_wmma_gnhwk_gkyxc_gnhwc_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  GNHWK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  GNHWC,
+                                                                  int8_t,
+                                                                  int8_t,
+                                                                  Empty_Tuple,
+                                                                  int8_t,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  NHWGK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGC,
+                                                                  int8_t,
+                                                                  int8_t,
+                                                                  Empty_Tuple,
+                                                                  int8_t,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_data_wmma_nhwgk_gkyxc_nhwgc_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  NHWGK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGC,
+                                                                  int8_t,
+                                                                  int8_t,
+                                                                  Empty_Tuple,
+                                                                  int8_t,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  GNDHWK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  GNDHWC,
+                                                                  int8_t,
+                                                                  int8_t,
+                                                                  Empty_Tuple,
+                                                                  int8_t,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_data_wmma_gndhwk_gkzyxc_gndhwc_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  GNDHWK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  GNDHWC,
+                                                                  int8_t,
+                                                                  int8_t,
+                                                                  Empty_Tuple,
+                                                                  int8_t,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  NDHWGK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGC,
+                                                                  int8_t,
+                                                                  int8_t,
+                                                                  Empty_Tuple,
+                                                                  int8_t,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_data_wmma_ndhwgk_gkzyxc_ndhwgc_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  NDHWGK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGC,
+                                                                  int8_t,
+                                                                  int8_t,
+                                                                  Empty_Tuple,
+                                                                  int8_t,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_data_xdl.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  GNHWK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  GNHWC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  GNHWK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  GNHWC,
+                                                                  F32,
+                                                                  F32,
+                                                                  Empty_Tuple,
+                                                                  F32,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv2d_bwd_data_xdl_gnhwk_gkyxc_gnhwc_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  GNHWK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  GNHWC,
+                                                                  BF16,
+                                                                  BF16,
+                                                                  Empty_Tuple,
+                                                                  BF16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  NHWGK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  NHWGK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGC,
+                                                                  F32,
+                                                                  F32,
+                                                                  Empty_Tuple,
+                                                                  F32,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv2d_bwd_data_xdl_nhwgk_gkyxc_nhwgc_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<2,
+                                                                  NHWGK,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGC,
+                                                                  BF16,
+                                                                  BF16,
+                                                                  Empty_Tuple,
+                                                                  BF16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+
+// conv3d backward data
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  GNDHWK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  GNDHWC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  GNDHWK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  GNDHWC,
+                                                                  F32,
+                                                                  F32,
+                                                                  Empty_Tuple,
+                                                                  F32,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv3d_bwd_data_xdl_gndhwk_gkzyxc_gndhwc_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  GNDHWK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  GNDHWC,
+                                                                  BF16,
+                                                                  BF16,
+                                                                  Empty_Tuple,
+                                                                  BF16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  NDHWGK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  NDHWGK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGC,
+                                                                  F32,
+                                                                  F32,
+                                                                  Empty_Tuple,
+                                                                  F32,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  NDHWGK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGC,
+                                                                  BF16,
+                                                                  BF16,
+                                                                  Empty_Tuple,
+                                                                  BF16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough>>>& instances);
+#endif
+#if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
+void add_device_grouped_conv3d_bwd_data_xdl_ndhwgk_gkzyxc_ndhwgc_input_f16_comp_bf8f8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdDataMultipleD<3,
+                                                                  NDHWGK,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGC,
+                                                                  F16,
+                                                                  F16,
+                                                                  Empty_Tuple,
+                                                                  F16,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  BF8,
+                                                                  F8>>>& instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

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

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -12,564 +12,19 @@
 
 #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
 
-namespace ck {
-namespace tensor_operation {
-namespace device {
-namespace instance {
-
-// xdl
-// conv1d backward weight
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           GNWC,
-                                                           GKXC,
-                                                           GNWK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           GNWC,
-                                                           GKXC,
-                                                           GNWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           GNWC,
-                                                           GKXC,
-                                                           GNWK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-// conv2d backward weight
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           GNHWC,
-                                                           GKYXC,
-                                                           GNHWK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           GNHWC,
-                                                           GKYXC,
-                                                           GNHWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           GNHWC,
-                                                           GKYXC,
-                                                           GNHWK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-// conv3d backward weight
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           int8_t,
-                                                           int8_t,
-                                                           int8_t,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           int8_t,
-                                                           int8_t,
-                                                           int8_t,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           BF8,
-                                                           F8>>>& instances);
-#endif
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           int8_t,
-                                                           int8_t,
-                                                           int8_t,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           int8_t,
-                                                           int8_t,
-                                                           int8_t,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-
 #ifdef DL_KERNELS
-// dl
-// conv1d backward weight
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           GNWC,
-                                                           GKXC,
-                                                           GNWK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           GNWC,
-                                                           GKXC,
-                                                           GNWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           GNWC,
-                                                           GKXC,
-                                                           GNWK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           NWGC,
-                                                           GKXC,
-                                                           NWGK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           NWGC,
-                                                           GKXC,
-                                                           NWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
+#include "grouped_convolution_backward_weight_dl.inc"
 #endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
-                                                           NWGC,
-                                                           GKXC,
-                                                           NWGK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-// conv2d backward weight
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           GNHWC,
-                                                           GKYXC,
-                                                           GNHWK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           GNHWC,
-                                                           GKYXC,
-                                                           GNHWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           GNHWC,
-                                                           GKYXC,
-                                                           GNHWK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-// conv3d backward weight
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           GNDHWC,
-                                                           GKZYXC,
-                                                           GNDHWK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           BF16,
-                                                           F32,
-                                                           BF16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
-                                                           NDHWGC,
-                                                           GKZYXC,
-                                                           NDHWGK,
-                                                           F32,
-                                                           F32,
-                                                           F32,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
+#ifdef CK_USE_XDL
+#include "grouped_convolution_backward_weight_xdl.inc"
 #endif
+#ifdef CK_USE_WMMA
+#include "grouped_convolution_backward_weight_wmma.inc"
 #endif
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
 
 template <ck::index_t NumDimSpatial,
           typename InLayout,
@@ -611,6 +66,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
     {
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
 
+#ifdef DL_KERNELS
         if constexpr(NumDimSpatial == 1)
         {
             if constexpr(is_same_v<InLayout, GNWC> && is_same_v<WeiLayout, GKXC> &&
@@ -621,10 +77,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_f32_instances(op_ptrs);
-#endif
-                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP16
@@ -632,10 +85,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_f16_instances(op_ptrs);
-#endif
-                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
@@ -644,19 +94,14 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
                         op_ptrs);
-#endif
-                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
-                        op_ptrs);
                 }
 #endif
             }
             if constexpr(is_same_v<InLayout, NWGC> && is_same_v<WeiLayout, GKXC> &&
                          is_same_v<OutLayout, NWGK>)
             {
-#ifdef DL_KERNELS
 #ifdef CK_ENABLE_FP32
                 if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
@@ -682,7 +127,6 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                     add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_bf16_f32_bf16_instances(
                         op_ptrs);
                 }
-#endif
 #endif
             }
         }
@@ -696,12 +140,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_f32_instances(
                         op_ptrs);
-#endif
-                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
-                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP16
@@ -709,12 +149,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_f16_instances(
                         op_ptrs);
-#endif
-                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
-                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
@@ -723,12 +159,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
                         op_ptrs);
-#endif
-                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
-                        op_ptrs);
                 }
 #endif
             }
@@ -740,12 +172,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_f32_instances(
                         op_ptrs);
-#endif
-                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
-                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP16
@@ -753,12 +181,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_f16_instances(
                         op_ptrs);
-#endif
-                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
-                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
@@ -767,12 +191,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
                         op_ptrs);
-#endif
-                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
-                        op_ptrs);
                 }
 #endif
             }
@@ -787,12 +207,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_f32_instances(
                         op_ptrs);
-#endif
-                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
-                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP16
@@ -800,15 +216,39 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-#ifdef DL_KERNELS
                     add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_f16_instances(
                         op_ptrs);
+                }
 #endif
-                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
+#ifdef CK_ENABLE_BF16
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeA, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeB, ck::bhalf_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
                         op_ptrs);
-                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
+                }
+#endif
+            }
+            if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
+                         is_same_v<OutLayout, NDHWGK>)
+            {
+#ifdef CK_ENABLE_FP32
+                if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
+                             is_same_v<ComputeTypeB, float>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
                         op_ptrs);
-                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
+                }
+#endif
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
+                             is_same_v<ComputeTypeB, half_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
                         op_ptrs);
                 }
 #endif
@@ -818,40 +258,125 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-#ifdef DL_KERNELS
-                    add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
+                    add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
                         op_ptrs);
+                }
 #endif
-                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
+            }
+        }
+#endif // DL_KERNELS
+#ifdef CK_USE_XDL
+        if constexpr(NumDimSpatial == 1)
+        {
+            if constexpr(is_same_v<InLayout, GNWC> && is_same_v<WeiLayout, GKXC> &&
+                         is_same_v<OutLayout, GNWK>)
+            {
+#ifdef CK_ENABLE_FP32
+                if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
+                             is_same_v<ComputeTypeB, float>)
+                {
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(op_ptrs);
+                }
+#endif
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
+                             is_same_v<ComputeTypeB, half_t>)
+                {
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(op_ptrs);
+                }
+#endif
+#ifdef CK_ENABLE_BF16
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeA, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeB, ck::bhalf_t>)
+                {
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
                         op_ptrs);
                 }
 #endif
-#ifdef CK_ENABLE_INT8
-                else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                                  is_same_v<OutDataType, int8_t> &&
-                                  is_same_v<ComputeTypeA, int8_t> &&
-                                  is_same_v<ComputeTypeB, int8_t>)
+            }
+        }
+        if constexpr(NumDimSpatial == 2)
+        {
+            if constexpr(is_same_v<InLayout, GNHWC> && is_same_v<WeiLayout, GKYXC> &&
+                         is_same_v<OutLayout, GNHWK>)
+            {
+#ifdef CK_ENABLE_FP32
+                if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
+                             is_same_v<ComputeTypeB, float>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
                         op_ptrs);
-                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
+                }
+#endif
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
+                             is_same_v<ComputeTypeB, half_t>)
+                {
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
+                        op_ptrs);
+                }
+#endif
+#ifdef CK_ENABLE_BF16
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeA, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeB, ck::bhalf_t>)
+                {
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
                         op_ptrs);
                 }
 #endif
             }
-            if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
-                         is_same_v<OutLayout, NDHWGK>)
+            if constexpr(is_same_v<InLayout, NHWGC> && is_same_v<WeiLayout, GKYXC> &&
+                         is_same_v<OutLayout, NHWGK>)
             {
 #ifdef CK_ENABLE_FP32
                 if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-#ifdef DL_KERNELS
-                    add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
                         op_ptrs);
+                }
 #endif
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
+                             is_same_v<ComputeTypeB, half_t>)
+                {
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
+                        op_ptrs);
+                }
+#endif
+#ifdef CK_ENABLE_BF16
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeA, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeB, ck::bhalf_t>)
+                {
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
+                        op_ptrs);
+                }
+#endif
+            }
+        }
+        if constexpr(NumDimSpatial == 3)
+        {
+            if constexpr(is_same_v<InLayout, GNDHWC> && is_same_v<WeiLayout, GKZYXC> &&
+                         is_same_v<OutLayout, GNDHWK>)
+            {
+#ifdef CK_ENABLE_FP32
+                if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
+                             is_same_v<ComputeTypeB, float>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
                         op_ptrs);
                 }
 #endif
@@ -860,15 +385,39 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-#ifdef DL_KERNELS
-                    add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
                         op_ptrs);
+                }
 #endif
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+#ifdef CK_ENABLE_BF16
+                if constexpr(is_same_v<InDataType, ck::bhalf_t> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeA, ck::bhalf_t> &&
+                             is_same_v<ComputeTypeB, ck::bhalf_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
                         op_ptrs);
-                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+                }
+#endif
+            }
+            if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
+                         is_same_v<OutLayout, NDHWGK>)
+            {
+#ifdef CK_ENABLE_FP32
+                if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                             is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
+                             is_same_v<ComputeTypeB, float>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
                         op_ptrs);
-                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
+                }
+#endif
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
+                             is_same_v<ComputeTypeB, half_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
                         op_ptrs);
                 }
 #endif
@@ -878,11 +427,36 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-#ifdef DL_KERNELS
-                    add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
                         op_ptrs);
+                }
 #endif
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
+#if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, bf8_t> &&
+                             is_same_v<ComputeTypeB, f8_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_instances(
+                        op_ptrs);
+                }
+#endif
+            }
+        }
+#endif
+#ifdef CK_USE_WMMA
+        if constexpr(NumDimSpatial == 3)
+        {
+            if constexpr(is_same_v<InLayout, GNDHWC> && is_same_v<WeiLayout, GKZYXC> &&
+                         is_same_v<OutLayout, GNDHWK>)
+            {
+#ifdef CK_ENABLE_FP16
+                if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
+                             is_same_v<ComputeTypeB, half_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
                         op_ptrs);
                 }
 #endif
@@ -892,23 +466,42 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                                   is_same_v<ComputeTypeA, int8_t> &&
                                   is_same_v<ComputeTypeB, int8_t>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
+                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
                         op_ptrs);
-                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
+                    add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
                         op_ptrs);
                 }
 #endif
-#if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
+            }
+            if constexpr(is_same_v<InLayout, NDHWGC> && is_same_v<WeiLayout, GKZYXC> &&
+                         is_same_v<OutLayout, NDHWGK>)
+            {
+#ifdef CK_ENABLE_FP16
                 if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, bf8_t> &&
-                             is_same_v<ComputeTypeB, f8_t>)
+                             is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
+                             is_same_v<ComputeTypeB, half_t>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_instances(
+                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
+                        op_ptrs);
+                }
+#endif
+#ifdef CK_ENABLE_INT8
+                else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                                  is_same_v<OutDataType, int8_t> &&
+                                  is_same_v<ComputeTypeA, int8_t> &&
+                                  is_same_v<ComputeTypeB, int8_t>)
+                {
+                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
                         op_ptrs);
                 }
 #endif
             }
         }
+#endif
 
         return op_ptrs;
     }

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight_dl.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// conv1d backward weight
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           NWGC,
+                                                           GKXC,
+                                                           NWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           NWGC,
+                                                           GKXC,
+                                                           NWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv1d_bwd_weight_dl_gnwc_gkxc_gnwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv1d_bwd_weight_dl_nwgc_gkxc_nwgk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           NWGC,
+                                                           GKXC,
+                                                           NWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+// conv2d backward weight
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_bwd_weight_dl_gnhwc_gkyxc_gnhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_weight_dl_nhwgc_gkyxc_nhwgk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+// conv3d backward weight
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv3d_bwd_weight_dl_gndhwc_gkzyxc_gndhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_dl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight_wmma.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// conv3d backward weight
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_INT8
+void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           int8_t,
+                                                           int8_t,
+                                                           int8_t,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           int8_t,
+                                                           int8_t,
+                                                           int8_t,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           int8_t,
+                                                           int8_t,
+                                                           int8_t,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_bwd_weight_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           int8_t,
+                                                           int8_t,
+                                                           int8_t,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight_xdl.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// conv1d backward weight
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+// conv2d backward weight
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+// conv3d backward weight
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           BF8,
+                                                           F8>>>& instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

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

@@ -12,890 +12,20 @@
 
 #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
 
-namespace ck {
-namespace tensor_operation {
-namespace device {
-namespace instance {
-
-#ifdef CK_ENABLE_BF16
-// grouped conv1d forward, GNWC/GKXC/GNWK
-void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
-                                                                GNWC,
-                                                                GKXC,
-                                                                Empty_Tuple,
-                                                                GNWK,
-                                                                BF16,
-                                                                BF16,
-                                                                Empty_Tuple,
-                                                                BF16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
-                                                                GNWC,
-                                                                GKXC,
-                                                                Empty_Tuple,
-                                                                GNWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
-                                                                GNWC,
-                                                                GKXC,
-                                                                Empty_Tuple,
-                                                                GNWK,
-                                                                F32,
-                                                                F32,
-                                                                Empty_Tuple,
-                                                                F32,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_int8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
-                                                                GNWC,
-                                                                GKXC,
-                                                                Empty_Tuple,
-                                                                GNWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_1x1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
+#ifdef DL_KERNELS
+#include "grouped_convolution_forward_dl.inc"
 #endif
-
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
+#ifdef CK_USE_XDL
+#include "grouped_convolution_forward_xdl.inc"
 #endif
-
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_oddc_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
+#ifdef CK_USE_WMMA
+#include "grouped_convolution_forward_wmma.inc"
 #endif
 
-#ifdef CK_ENABLE_BF16
-// grouped conv2d forward, GNHWC/GKYXC/GNHWK
-void add_device_grouped_conv1d_fwd_xdl_gnhwc_gkyxc_gnhwk_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                BF16,
-                                                                BF16,
-                                                                Empty_Tuple,
-                                                                BF16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_fwd_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_fwd_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F32,
-                                                                F32,
-                                                                Empty_Tuple,
-                                                                F32,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_oddc_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-// grouped conv2d forward, NHWGC/GKYXC/NHWGK
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                BF16,
-                                                                BF16,
-                                                                Empty_Tuple,
-                                                                BF16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_f16_1x1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_f16_oddc_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                F32,
-                                                                F32,
-                                                                Empty_Tuple,
-                                                                F32,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_BF16
-// grouped conv3d forward, GNDHWC/GKZYXC/GNDHWK
-void add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                BF16,
-                                                                BF16,
-                                                                Empty_Tuple,
-                                                                BF16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_oddc_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                F32,
-                                                                F32,
-                                                                Empty_Tuple,
-                                                                F32,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_int8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_oddc_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                GNDHWC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                GNDHWK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_BF16
-// grouped conv3d forward, NDHWGC/GKZYXC/NDHWGK
-void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                BF16,
-                                                                BF16,
-                                                                Empty_Tuple,
-                                                                BF16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_oddc_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_oddc_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP8
-void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_f8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                F8>>>& instances);
-
-void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                F8,
-                                                                F8,
-                                                                Empty_Tuple,
-                                                                F8,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                F8>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                F32,
-                                                                F32,
-                                                                Empty_Tuple,
-                                                                F32,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#ifdef CK_ENABLE_INT8
-void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_int8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_oddc_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
-                                                                NDHWGC,
-                                                                GKZYXC,
-                                                                Empty_Tuple,
-                                                                NDHWGK,
-                                                                int8_t,
-                                                                int8_t,
-                                                                Empty_Tuple,
-                                                                int8_t,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#if(defined(CK_ENABLE_FP32) && defined(DL_KERNELS))
-void add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                F32,
-                                                                F32,
-                                                                Empty_Tuple,
-                                                                F32,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-
-#endif
-
-#if(defined(CK_ENABLE_FP16) && defined(DL_KERNELS))
-void add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#if(defined(CK_ENABLE_FP16) && defined(DL_KERNELS))
-void add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F16,
-                                                                F16,
-                                                                Empty_Tuple,
-                                                                F16,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
-
-#if(defined(CK_ENABLE_FP32) && defined(DL_KERNELS))
-void add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
-                                                                GNHWC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                GNHWK,
-                                                                F32,
-                                                                F32,
-                                                                Empty_Tuple,
-                                                                F32,
-                                                                PassThrough,
-                                                                PassThrough,
-                                                                PassThrough>>>& instances);
-#endif
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
 
 template <ck::index_t NumDimSpatial,
           typename InLayout,
@@ -904,7 +34,8 @@ template <ck::index_t NumDimSpatial,
           typename InDataType,
           typename WeiDataType,
           typename OutDataType,
-          typename ComputeType>
+          typename AComputeType,
+          typename BComputeType>
 struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
     NumDimSpatial,
     InLayout,
@@ -918,7 +49,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
     ck::tensor_operation::element_wise::PassThrough,
     ck::tensor_operation::element_wise::PassThrough,
     ck::tensor_operation::element_wise::PassThrough,
-    ComputeType>>
+    AComputeType,
+    BComputeType>>
 {
     using DeviceOp =
         DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
@@ -933,25 +65,73 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                                         ck::tensor_operation::element_wise::PassThrough,
                                         ck::tensor_operation::element_wise::PassThrough,
                                         ck::tensor_operation::element_wise::PassThrough,
-                                        ComputeType>;
+                                        AComputeType,
+                                        BComputeType>;
 
     static auto GetInstances()
     {
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
 
+#ifdef DL_KERNELS
+        if constexpr(NumDimSpatial == 2 && is_same_v<InLayout, GNHWC> &&
+                     is_same_v<WeiLayout, GKYXC> && is_same_v<OutLayout, GNHWK>)
+        {
+#ifdef CK_ENABLE_FP32
+            if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                         is_same_v<OutDataType, float> && is_same_v<AComputeType, float> &&
+                         is_same_v<BComputeType, float>)
+            {
+                add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f32_instances(op_ptrs);
+            }
+#endif
+#ifdef CK_ENABLE_FP16
+            if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
+            {
+                add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instances(op_ptrs);
+            }
+#endif
+        }
+        if constexpr(NumDimSpatial == 2 && is_same_v<InLayout, NHWGC> &&
+                     is_same_v<WeiLayout, GKYXC> && is_same_v<OutLayout, NHWGK>)
+        {
+
+#ifdef CK_ENABLE_FP32
+            if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
+                         is_same_v<OutDataType, float> && is_same_v<AComputeType, float> &&
+                         is_same_v<BComputeType, float>)
+            {
+                add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f32_instances(op_ptrs);
+            }
+#endif
+#ifdef CK_ENABLE_FP16
+            if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
+            {
+                add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f16_instances(op_ptrs);
+            }
+#endif
+        }
+#endif // DL_KERNELS
+
+#ifdef CK_USE_XDL
         if constexpr(NumDimSpatial == 1 && is_same_v<InLayout, GNWC> &&
                      is_same_v<WeiLayout, GKXC> && is_same_v<OutLayout, GNWK>)
         {
 #ifdef CK_ENABLE_FP32
             if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
-                         is_same_v<OutDataType, float> && is_same_v<ComputeType, float>)
+                         is_same_v<OutDataType, float> && is_same_v<AComputeType, float> &&
+                         is_same_v<BComputeType, float>)
             {
                 add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f32_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_FP16
             if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                         is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
             {
                 add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f16_instances(op_ptrs);
             }
@@ -959,14 +139,17 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
 #ifdef CK_ENABLE_BF16
             if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
                          is_same_v<WeiDataType, ck::bhalf_t> &&
-                         is_same_v<OutDataType, ck::bhalf_t> && is_same_v<ComputeType, bhalf_t>)
+                         is_same_v<OutDataType, ck::bhalf_t> &&
+                         is_same_v<AComputeType, ck::bhalf_t> &&
+                         is_same_v<BComputeType, ck::bhalf_t>)
             {
                 add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_bf16_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_INT8
             if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                         is_same_v<OutDataType, int8_t> && is_same_v<ComputeType, int8_t>)
+                         is_same_v<OutDataType, int8_t> && is_same_v<AComputeType, int8_t> &&
+                         is_same_v<BComputeType, int8_t>)
             {
                 add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_int8_instances(op_ptrs);
             }
@@ -978,114 +161,60 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
         {
 #ifdef CK_ENABLE_FP32
             if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
-                         is_same_v<OutDataType, float> && is_same_v<ComputeType, float>)
+                         is_same_v<OutDataType, float> && is_same_v<AComputeType, float> &&
+                         is_same_v<BComputeType, float>)
             {
                 add_device_grouped_conv2d_fwd_xdl_gnhwc_gkyxc_gnhwk_f32_instances(op_ptrs);
             }
 #endif
-
-#if(defined(CK_ENABLE_FP32) && defined(DL_KERNELS))
-            if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
-                         is_same_v<OutDataType, float> && is_same_v<ComputeType, float>)
-            {
-                add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f32_instances(op_ptrs);
-            }
-#endif
-
 #ifdef CK_ENABLE_FP16
             if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                         is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
             {
                 add_device_grouped_conv2d_fwd_xdl_gnhwc_gkyxc_gnhwk_f16_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1p0_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1s1p0_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_oddc_instances(op_ptrs);
-            }
-#endif
-
-#if(defined(CK_ENABLE_FP16) && defined(DL_KERNELS))
-            if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                         is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
-            {
-                add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instances(op_ptrs);
             }
 #endif
-
 #ifdef CK_ENABLE_BF16
             if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
                          is_same_v<WeiDataType, ck::bhalf_t> &&
-                         is_same_v<OutDataType, ck::bhalf_t> && is_same_v<ComputeType, bhalf_t>)
+                         is_same_v<OutDataType, ck::bhalf_t> &&
+                         is_same_v<AComputeType, ck::bhalf_t> &&
+                         is_same_v<BComputeType, ck::bhalf_t>)
             {
                 add_device_grouped_conv1d_fwd_xdl_gnhwc_gkyxc_gnhwk_bf16_instances(op_ptrs);
             }
 #endif
-
-#ifdef CK_ENABLE_INT8
-            if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                         is_same_v<OutDataType, int8_t> && is_same_v<ComputeType, int8_t>)
-            {
-                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1p0_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1s1p0_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_oddc_instances(op_ptrs);
-            }
-#endif
         }
 
         if constexpr(NumDimSpatial == 2 && is_same_v<InLayout, NHWGC> &&
                      is_same_v<WeiLayout, GKYXC> && is_same_v<OutLayout, NHWGK>)
         {
-
 #ifdef CK_ENABLE_FP32
             if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
-                         is_same_v<OutDataType, float> && is_same_v<ComputeType, float>)
+                         is_same_v<OutDataType, float> && is_same_v<AComputeType, float> &&
+                         is_same_v<BComputeType, float>)
             {
                 add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f32_instances(op_ptrs);
             }
 #endif
-
-#if(defined(CK_ENABLE_FP32) && defined(DL_KERNELS))
-            if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
-                         is_same_v<OutDataType, float> && is_same_v<ComputeType, float>)
-            {
-                add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f32_instances(op_ptrs);
-            }
-#endif
-
 #ifdef CK_ENABLE_FP16
             if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                         is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
             {
                 add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_instances(op_ptrs);
             }
 #endif
-
-#if(defined(CK_ENABLE_FP16) && defined(DL_KERNELS))
-            if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                         is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
-            {
-                add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f16_instances(op_ptrs);
-            }
-#endif
-
 #ifdef CK_ENABLE_BF16
             if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
                          is_same_v<WeiDataType, ck::bhalf_t> &&
-                         is_same_v<OutDataType, ck::bhalf_t> && is_same_v<ComputeType, bhalf_t>)
+                         is_same_v<OutDataType, ck::bhalf_t> &&
+                         is_same_v<AComputeType, ck::bhalf_t> &&
+                         is_same_v<BComputeType, ck::bhalf_t>)
             {
                 add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_instances(op_ptrs);
             }
-#endif
-#ifdef CK_ENABLE_INT8
-            else if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                              is_same_v<OutDataType, int8_t> && is_same_v<ComputeType, int8_t>)
-            {
-                add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_1x1p0_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_1x1s1p0_instances(op_ptrs);
-                add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_oddc_instances(op_ptrs);
-            }
 #endif
         }
 
@@ -1094,42 +223,36 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
         {
 #ifdef CK_ENABLE_FP32
             if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
-                         is_same_v<OutDataType, float> && is_same_v<ComputeType, float>)
+                         is_same_v<OutDataType, float> && is_same_v<AComputeType, float> &&
+                         is_same_v<BComputeType, float>)
             {
                 add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_FP16
             if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                         is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
             {
                 add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1p0_instances(
-                    op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
-                    op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_oddc_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_BF16
             if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
                          is_same_v<WeiDataType, ck::bhalf_t> &&
-                         is_same_v<OutDataType, ck::bhalf_t> && is_same_v<ComputeType, bhalf_t>)
+                         is_same_v<OutDataType, ck::bhalf_t> &&
+                         is_same_v<AComputeType, ck::bhalf_t> &&
+                         is_same_v<BComputeType, ck::bhalf_t>)
             {
                 add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_bf16_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_INT8
             if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                         is_same_v<OutDataType, int8_t> && is_same_v<ComputeType, int8_t>)
+                         is_same_v<OutDataType, int8_t> && is_same_v<AComputeType, int8_t> &&
+                         is_same_v<BComputeType, int8_t>)
             {
                 add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_int8_instances(op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1p0_instances(op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
-                    op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_oddc_instances(op_ptrs);
             }
 #endif
         }
@@ -1139,7 +262,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
         {
 #ifdef CK_ENABLE_FP32
             if constexpr(is_same_v<InDataType, float> && is_same_v<WeiDataType, float> &&
-                         is_same_v<OutDataType, float> && is_same_v<ComputeType, float>)
+                         is_same_v<OutDataType, float> && is_same_v<AComputeType, float> &&
+                         is_same_v<BComputeType, float>)
             {
                 add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(op_ptrs);
             }
@@ -1147,44 +271,168 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
 
 #ifdef CK_ENABLE_FP8
             if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                         is_same_v<OutDataType, half_t> && is_same_v<ComputeType, ck::f8_t>)
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, ck::f8_t> &&
+                         is_same_v<BComputeType, ck::f8_t>)
             {
                 add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_f8_instances(
                     op_ptrs);
             }
 
             if constexpr(is_same_v<InDataType, ck::f8_t> && is_same_v<WeiDataType, ck::f8_t> &&
-                         is_same_v<OutDataType, ck::f8_t> && is_same_v<ComputeType, ck::f8_t>)
+                         is_same_v<OutDataType, ck::f8_t> && is_same_v<AComputeType, ck::f8_t> &&
+                         is_same_v<BComputeType, ck::f8_t>)
             {
                 add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f8_instances(op_ptrs);
             }
 #endif
+#ifdef CK_ENABLE_BF8
+            if constexpr(is_same_v<InDataType, ck::bf8_t> && is_same_v<WeiDataType, ck::bf8_t> &&
+                         is_same_v<OutDataType, ck::f8_t> && is_same_v<AComputeType, ck::bf8_t> &&
+                         is_same_v<BComputeType, ck::bf8_t>)
+            {
+                add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_instances(op_ptrs);
+            }
+#endif
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+            if constexpr(is_same_v<InDataType, ck::f8_t> && is_same_v<WeiDataType, ck::bf8_t> &&
+                         is_same_v<OutDataType, ck::f8_t> && is_same_v<AComputeType, ck::f8_t> &&
+                         is_same_v<BComputeType, ck::bf8_t>)
+            {
+                add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f8_bf8_instances(op_ptrs);
+            }
+#endif
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+            if constexpr(is_same_v<InDataType, ck::bf8_t> && is_same_v<WeiDataType, ck::f8_t> &&
+                         is_same_v<OutDataType, ck::f8_t> && is_same_v<AComputeType, ck::bf8_t> &&
+                         is_same_v<BComputeType, ck::f8_t>)
+            {
+                add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_f8_instances(op_ptrs);
+            }
+#endif
 #ifdef CK_ENABLE_FP16
             if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
-                         is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
             {
                 add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1p0_instances(
-                    op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
-                    op_ptrs);
-                add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_oddc_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_BF16
             if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
                          is_same_v<WeiDataType, ck::bhalf_t> &&
-                         is_same_v<OutDataType, ck::bhalf_t> && is_same_v<ComputeType, bhalf_t>)
+                         is_same_v<OutDataType, ck::bhalf_t> &&
+                         is_same_v<AComputeType, ck::bhalf_t> &&
+                         is_same_v<BComputeType, ck::bhalf_t>)
             {
                 add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instances(op_ptrs);
             }
 #endif
 #ifdef CK_ENABLE_INT8
             if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
-                         is_same_v<OutDataType, int8_t> && is_same_v<ComputeType, int8_t>)
+                         is_same_v<OutDataType, int8_t> && is_same_v<AComputeType, int8_t> &&
+                         is_same_v<BComputeType, int8_t>)
             {
                 add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_int8_instances(op_ptrs);
+            }
+#endif
+        }
+#endif
+
+#ifdef CK_USE_WMMA
+        if constexpr(NumDimSpatial == 2 && is_same_v<InLayout, GNHWC> &&
+                     is_same_v<WeiLayout, GKYXC> && is_same_v<OutLayout, GNHWK>)
+        {
+#ifdef CK_ENABLE_FP16
+            if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
+            {
+                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1p0_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1s1p0_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_oddc_instances(op_ptrs);
+            }
+#endif
+#ifdef CK_ENABLE_INT8
+            if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                         is_same_v<OutDataType, int8_t> && is_same_v<AComputeType, int8_t> &&
+                         is_same_v<BComputeType, int8_t>)
+            {
+                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1p0_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1s1p0_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_oddc_instances(op_ptrs);
+            }
+#endif
+        }
+
+        if constexpr(NumDimSpatial == 2 && is_same_v<InLayout, NHWGC> &&
+                     is_same_v<WeiLayout, GKYXC> && is_same_v<OutLayout, NHWGK>)
+        {
+#ifdef CK_ENABLE_INT8
+            if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                         is_same_v<OutDataType, int8_t> && is_same_v<AComputeType, int8_t> &&
+                         is_same_v<BComputeType, int8_t>)
+            {
+                add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_1x1p0_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_1x1s1p0_instances(op_ptrs);
+                add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_oddc_instances(op_ptrs);
+            }
+#endif
+        }
+
+        if constexpr(NumDimSpatial == 3 && is_same_v<InLayout, GNDHWC> &&
+                     is_same_v<WeiLayout, GKZYXC> && is_same_v<OutLayout, GNDHWK>)
+        {
+#ifdef CK_ENABLE_FP16
+            if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
+            {
+                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1p0_instances(
+                    op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
+                    op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_oddc_instances(op_ptrs);
+            }
+#endif
+#ifdef CK_ENABLE_INT8
+            if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                         is_same_v<OutDataType, int8_t> && is_same_v<AComputeType, int8_t> &&
+                         is_same_v<BComputeType, int8_t>)
+            {
+                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1p0_instances(op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
+                    op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_oddc_instances(op_ptrs);
+            }
+#endif
+        }
+
+        if constexpr(NumDimSpatial == 3 && is_same_v<InLayout, NDHWGC> &&
+                     is_same_v<WeiLayout, GKZYXC> && is_same_v<OutLayout, NDHWGK>)
+        {
+#ifdef CK_ENABLE_FP16
+            if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
+                         is_same_v<OutDataType, half_t> && is_same_v<AComputeType, half_t> &&
+                         is_same_v<BComputeType, half_t>)
+            {
+                add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1p0_instances(
+                    op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
+                    op_ptrs);
+                add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_oddc_instances(op_ptrs);
+            }
+#endif
+#ifdef CK_ENABLE_INT8
+            if constexpr(is_same_v<InDataType, int8_t> && is_same_v<WeiDataType, int8_t> &&
+                         is_same_v<OutDataType, int8_t> && is_same_v<AComputeType, int8_t> &&
+                         is_same_v<BComputeType, int8_t>)
+            {
                 add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(op_ptrs);
                 add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1p0_instances(op_ptrs);
                 add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
@@ -1193,6 +441,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
             }
 #endif
         }
+#endif
 
         return op_ptrs;
     }

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_dl.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                F32,
+                                                                F32,
+                                                                Empty_Tuple,
+                                                                F32,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F32,
+                                                                F32,
+                                                                Empty_Tuple,
+                                                                F32,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_wmma.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+#ifdef CK_ENABLE_INT8
+void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_1x1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_i8_oddc_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_oddc_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_i8_oddc_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_i8_oddc_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_f16_1x1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_nhwgc_gkyxc_nhwgk_f16_oddc_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_oddc_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv3d_fwd_wmma_ndhwgc_gkzyxc_ndhwgk_f16_oddc_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1s1p0_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_oddc_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_xdl.inc

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+#ifdef CK_ENABLE_BF16
+// grouped conv1d forward, GNWC/GKXC/GNWK
+void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
+                                                                GNWC,
+                                                                GKXC,
+                                                                Empty_Tuple,
+                                                                GNWK,
+                                                                BF16,
+                                                                BF16,
+                                                                Empty_Tuple,
+                                                                BF16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
+                                                                GNWC,
+                                                                GKXC,
+                                                                Empty_Tuple,
+                                                                GNWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
+                                                                GNWC,
+                                                                GKXC,
+                                                                Empty_Tuple,
+                                                                GNWK,
+                                                                F32,
+                                                                F32,
+                                                                Empty_Tuple,
+                                                                F32,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_INT8
+void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_int8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
+                                                                GNWC,
+                                                                GKXC,
+                                                                Empty_Tuple,
+                                                                GNWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_BF16
+// grouped conv2d forward, GNHWC/GKYXC/GNHWK
+void add_device_grouped_conv1d_fwd_xdl_gnhwc_gkyxc_gnhwk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                BF16,
+                                                                BF16,
+                                                                Empty_Tuple,
+                                                                BF16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_fwd_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_fwd_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                GNHWC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                GNHWK,
+                                                                F32,
+                                                                F32,
+                                                                Empty_Tuple,
+                                                                F32,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+// grouped conv2d forward, NHWGC/GKYXC/NHWGK
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                BF16,
+                                                                BF16,
+                                                                Empty_Tuple,
+                                                                BF16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
+                                                                NHWGC,
+                                                                GKYXC,
+                                                                Empty_Tuple,
+                                                                NHWGK,
+                                                                F32,
+                                                                F32,
+                                                                Empty_Tuple,
+                                                                F32,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_BF16
+// grouped conv3d forward, GNDHWC/GKZYXC/GNDHWK
+void add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                BF16,
+                                                                BF16,
+                                                                Empty_Tuple,
+                                                                BF16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+
+#endif
+
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                F32,
+                                                                F32,
+                                                                Empty_Tuple,
+                                                                F32,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_INT8
+void add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_int8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                GNDHWC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                GNDHWK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_BF16
+// grouped conv3d forward, NDHWGC/GKZYXC/NDHWGK
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                BF16,
+                                                                BF16,
+                                                                Empty_Tuple,
+                                                                BF16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP8
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_f8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F16,
+                                                                F16,
+                                                                Empty_Tuple,
+                                                                F16,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                F8>>>& instances);
+
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F8,
+                                                                F8,
+                                                                Empty_Tuple,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                F8>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F32,
+                                                                F32,
+                                                                Empty_Tuple,
+                                                                F32,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_INT8
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_int8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                int8_t,
+                                                                int8_t,
+                                                                Empty_Tuple,
+                                                                int8_t,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough>>>& instances);
+#endif
+
+#ifdef CK_ENABLE_BF8
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                BF8,
+                                                                BF8,
+                                                                Empty_Tuple,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                BF8>>>& instances);
+#endif
+
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f8_bf8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                F8,
+                                                                BF8,
+                                                                Empty_Tuple,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                F8,
+                                                                BF8>>>& instances);
+#endif
+
+#if(defined(CK_ENABLE_FP8) && defined(CK_ENABLE_BF8))
+void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf8_f8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                Empty_Tuple,
+                                                                NDHWGK,
+                                                                BF8,
+                                                                F8,
+                                                                Empty_Tuple,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                BF8,
+                                                                F8>>>& instances);
+#endif
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

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

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -146,6 +146,32 @@ void add_device_grouped_gemm_xdl_splitk_f8_f16_f16_mk_kn_mn_irregular_instances(
                                                   PassThrough,
                                                   PassThrough>>>& instances);
 
+void add_device_grouped_gemm_multiple_d_xdl_two_stage_f16_f16_f16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_multiple_d_xdl_two_stage_bf16_i8_bf16_mk_kn_mn_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  BF16,
+                                                  I8,
+                                                  Empty_Tuple,
+                                                  BF16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
 template <typename ALayout,
           typename BLayout,
           typename ELayout,
@@ -180,6 +206,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
     {
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
 
+#if defined(CK_ENABLE_FP16)
         if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, half_t> &&
                      is_same_v<EDataType, half_t>)
         {
@@ -190,6 +217,8 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
                 add_device_grouped_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_instances(
                     op_ptrs);
+                add_device_grouped_gemm_multiple_d_xdl_two_stage_f16_f16_f16_mk_kn_mn_instances(
+                    op_ptrs);
             }
             else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
                               is_same_v<ELayout, Row>)
@@ -210,8 +239,10 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 add_device_grouped_gemm_xdl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
             }
         }
-        else if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, f8_t> &&
-                          is_same_v<EDataType, half_t>)
+#endif
+#if defined(CK_ENABLE_FP16) && defined(CK_ENABLE_FP8)
+        if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, f8_t> &&
+                     is_same_v<EDataType, half_t>)
         {
             if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
                          is_same_v<ELayout, Row>)
@@ -228,6 +259,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                 add_device_grouped_gemm_xdl_splitk_f8_f16_f16_mk_kn_mn_irregular_instances(op_ptrs);
             }
         }
+#endif
+#if defined(CK_ENABLE_BF16) && defined(CK_ENABLE_INT8)
+        if constexpr(is_same_v<ADataType, bhalf_t> && is_same_v<BDataType, int8_t> &&
+                     is_same_v<EDataType, bhalf_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<ELayout, Row>)
+            {
+                add_device_grouped_gemm_multiple_d_xdl_two_stage_bf16_i8_bf16_mk_kn_mn_instances(
+                    op_ptrs);
+            }
+        }
+#endif
         return op_ptrs;
     }
 };

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

@@ -7,7 +7,7 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
-#include "ck/tensor_operation/gpu/device/device_elementwise_scale.hpp"
+#include "ck/tensor_operation/gpu/device/device_elementwise.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
@@ -19,125 +19,67 @@ namespace instance {
 
 #ifdef CK_ENABLE_FP16
 void add_device_permute_scale_1d_f16_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F16>,
-                                                  ck::Tuple<F16>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  1>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F16>, ck::Tuple<F16>, element_wise::Scale, 1>>>&);
 
 void add_device_permute_scale_2d_f16_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F16>,
-                                                  ck::Tuple<F16>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  2>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F16>, ck::Tuple<F16>, element_wise::Scale, 2>>>&);
 
 void add_device_permute_scale_3d_f16_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F16>,
-                                                  ck::Tuple<F16>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  3>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F16>, ck::Tuple<F16>, element_wise::Scale, 3>>>&);
 
 void add_device_permute_scale_4d_f16_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F16>,
-                                                  ck::Tuple<F16>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  4>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F16>, ck::Tuple<F16>, element_wise::Scale, 4>>>&);
 
 void add_device_permute_scale_5d_f16_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F16>,
-                                                  ck::Tuple<F16>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  5>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F16>, ck::Tuple<F16>, element_wise::Scale, 5>>>&);
 
 void add_device_permute_scale_6d_f16_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F16>,
-                                                  ck::Tuple<F16>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  6>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F16>, ck::Tuple<F16>, element_wise::Scale, 6>>>&);
 
 #endif
 
 #ifdef CK_ENABLE_FP32
 void add_device_permute_scale_1d_f32_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F32>,
-                                                  ck::Tuple<F32>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  1>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F32>, ck::Tuple<F32>, element_wise::Scale, 1>>>&);
 
 void add_device_permute_scale_2d_f32_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F32>,
-                                                  ck::Tuple<F32>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  2>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F32>, ck::Tuple<F32>, element_wise::Scale, 2>>>&);
 
 void add_device_permute_scale_3d_f32_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F32>,
-                                                  ck::Tuple<F32>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  3>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F32>, ck::Tuple<F32>, element_wise::Scale, 3>>>&);
 
 void add_device_permute_scale_4d_f32_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F32>,
-                                                  ck::Tuple<F32>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  4>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F32>, ck::Tuple<F32>, element_wise::Scale, 4>>>&);
 
 void add_device_permute_scale_5d_f32_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F32>,
-                                                  ck::Tuple<F32>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  5>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F32>, ck::Tuple<F32>, element_wise::Scale, 5>>>&);
 
 void add_device_permute_scale_6d_f32_instances(
-    std::vector<std::unique_ptr<DeviceElementwise<ck::Tuple<F32>,
-                                                  ck::Tuple<F32>,
-                                                  PassThrough,
-                                                  element_wise::UnarySquare,
-                                                  Scale,
-                                                  6>>>&);
+    std::vector<std::unique_ptr<
+        DeviceElementwise<ck::Tuple<F32>, ck::Tuple<F32>, element_wise::Scale, 6>>>&);
 #endif
 
 template <typename InDataTypeTuple,
           typename OutDataTypeTuple,
           typename ElementwiseOperation,
-          typename UnaryOperation,
-          typename Scale,
           index_t NumDim>
 struct DeviceOperationInstanceFactory<
-    ck::tensor_operation::device::DeviceElementwise<InDataTypeTuple,
-                                                    OutDataTypeTuple,
-                                                    ElementwiseOperation,
-                                                    UnaryOperation,
-                                                    Scale,
-                                                    NumDim>>
+    ck::tensor_operation::device::
+        DeviceElementwise<InDataTypeTuple, OutDataTypeTuple, ElementwiseOperation, NumDim>>
 {
-    using DeviceOp = DeviceElementwise<InDataTypeTuple,
-                                       OutDataTypeTuple,
-                                       ElementwiseOperation,
-                                       UnaryOperation,
-                                       Scale,
-                                       NumDim>;
+    using DeviceOp =
+        DeviceElementwise<InDataTypeTuple, OutDataTypeTuple, ElementwiseOperation, NumDim>;
 
     static auto GetInstances()
     {