merge develop

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

@@ -166,11 +166,22 @@ void add_device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_mem_v1_kpadding_instances
     std::vector<std::unique_ptr<
         DeviceGemmV2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
         instances);
+
 void add_device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_mem_v2_default_instances(
     std::vector<std::unique_ptr<
         DeviceGemmV2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
         instances);
 
+void add_device_gemm_xdl_universal_f16_i4_f16_mk_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmV2<Row, Col, Row, F16, I4, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
+void add_device_gemm_xdl_universal_bf16_i4_bf16_mk_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<
+        DeviceGemmV2<Row, Col, Row, BF16, I4, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances);
+
 void add_device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_mem_v2_kpadding_instances(
     std::vector<std::unique_ptr<
         DeviceGemmV2<Row, Col, Row, F16, F8, F16, PassThrough, PassThrough, PassThrough>>>&
@@ -810,6 +821,28 @@ struct DeviceOperationInstanceFactory<
             }
         }
 #endif
+
+        if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, pk_i4_t> &&
+                     is_same_v<CDataType, half_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_universal_f16_i4_f16_mk_nk_mn_mem_v2_default_instances(op_ptrs);
+            }
+        }
+
+        if constexpr(is_same_v<ADataType, bhalf_t> && is_same_v<BDataType, pk_i4_t> &&
+                     is_same_v<CDataType, bhalf_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_universal_bf16_i4_bf16_mk_nk_mn_mem_v2_default_instances(
+                    op_ptrs);
+            }
+        }
+
         return op_ptrs;
     }
 };

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

@@ -238,6 +238,403 @@ void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_mnkpaddin
                                                       PassThrough>>>& instances);
 #endif
 
+#ifdef CK_ENABLE_BF16
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v1_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v2_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_comp_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_comp_mnpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_comp_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v1_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v2_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Row,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_comp_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_comp_mpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_comp_mkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v1_mkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+
+void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v2_mkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Col,
+                                                      Col,
+                                                      Row,
+                                                      BF16,
+                                                      BF16,
+                                                      BF16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances);
+#endif
+
 #if(defined(CK_ENABLE_FP8))
 void add_device_gemm_xdl_universal_streamk_f16_f8_f16_mk_kn_mn_comp_default_instances(
     std::vector<std::unique_ptr<
@@ -527,6 +924,109 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGemm_S
         }
 #endif
 
+#ifdef CK_ENABLE_BF16
+        if constexpr(is_same_v<ADataType, bhalf_t> && is_same_v<BDataType, bhalf_t> &&
+                     is_same_v<CDataType, bhalf_t>)
+        {
+            if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
+                         is_same_v<CLayout, Row>)
+            {
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v1_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v1_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v1_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v2_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v2_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_v2_mnkpadding_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_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_v1_kpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_v2_kpadding_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_universal_streamk_bf16_bf16_bf16_km_kn_mn_comp_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_comp_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_comp_mnpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_comp_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v1_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v1_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v1_mnkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v2_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v2_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_kn_mn_mem_v2_mnkpadding_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_universal_streamk_bf16_bf16_bf16_km_nk_mn_comp_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_comp_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_comp_mpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_comp_mkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v1_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v1_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v1_mkpadding_instances(
+                    op_ptrs);
+
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v2_default_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v2_kpadding_instances(
+                    op_ptrs);
+                add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_km_nk_mn_mem_v2_mkpadding_instances(
+                    op_ptrs);
+            }
+        }
+#endif
+
 #if(defined(CK_ENABLE_FP8))
         if constexpr(is_same_v<ADataType, half_t> && is_same_v<BDataType, f8_t> &&
                      is_same_v<CDataType, half_t>)

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

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
@@ -41,11 +41,13 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_f16_generic_instances =
     std::tuple<
-        // clang-format off
+// clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+#if defined(CK_USE_AMD_MFMA_GFX950)
+#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>
         // clang-format on
         >;
@@ -58,11 +60,13 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_f16_instances = std::tuple<
-    // clang-format off
+// clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+#if defined(CK_USE_AMD_MFMA_GFX950)
+#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
 
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              2,              2,      false,  S<4, 16,  1>, S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 2>,
@@ -82,13 +86,37 @@ template <ck::index_t NDimSpatial,
           ConvolutionBackwardWeightSpecialization ConvSpec,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
-using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_bf16_generic_instances =
+using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_f16_irregular_instances =
     std::tuple<
         // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    48,    64,     32,   8,   16,   16,    3,    4,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              3,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    48,     32,   8,   16,   16,    4,    3,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              4,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,              3,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    80,     32,   8,   16,   16,    4,    5,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              4,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,              5,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,   112,     32,   8,   16,   16,    4,    7,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              4,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,              7,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,   208,     32,   8,   16,   16,    4,   13,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              4,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,             13,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>
+        // clang-format on
+        >;
+
+template <ck::index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename ELayout,
+          ConvolutionBackwardWeightSpecialization ConvSpec,
+          BlockGemmPipelineScheduler Scheduler,
+          BlockGemmPipelineVersion PipelineVersion>
+using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_bf16_generic_instances =
+    std::tuple<
+// clang-format off
+        //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
+        //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
+        //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
+        //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+#if defined(CK_USE_AMD_MFMA_GFX950)
+#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>
         // clang-format on
         >;
@@ -101,11 +129,13 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_bf16_instances = std::tuple<
-    // clang-format off
+// clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+#if defined(CK_USE_AMD_MFMA_GFX950)
+#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
 
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              2,              2,      false,  S<4, 16,  1>, S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 2>,
@@ -125,13 +155,37 @@ template <ck::index_t NDimSpatial,
           ConvolutionBackwardWeightSpecialization ConvSpec,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
-using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_f16_generic_instances =
+using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_bf16_irregular_instances =
     std::tuple<
         // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     BF16,     BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    48,    64,     32,   8,   16,   16,    3,    4,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              3,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     BF16,     BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    48,     32,   8,   16,   16,    4,    3,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              4,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,              3,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     BF16,     BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    80,     32,   8,   16,   16,    4,    5,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              4,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,              5,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     BF16,     BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,   112,     32,   8,   16,   16,    4,    7,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              4,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,              7,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     BF16,     BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,   208,     32,   8,   16,   16,    4,   13,  S<4, 16, 1>,  S<2, 0, 1>,  S<2, 0, 1>,                  1,              4,              4,      false,  S<4, 16,  1>,  S<2, 0, 1>,  S<2, 0, 1>,                1,             13,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>
+        // clang-format on
+        >;
+
+template <ck::index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename ELayout,
+          ConvolutionBackwardWeightSpecialization ConvSpec,
+          BlockGemmPipelineScheduler Scheduler,
+          BlockGemmPipelineVersion PipelineVersion>
+using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_f16_generic_instances =
+    std::tuple<
+// clang-format off
+        //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
+        //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
+        //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
+        //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+#if defined(CK_USE_AMD_MFMA_GFX950)
+#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1, F16, F16, 1, 1>
         // clang-format on
         >;
@@ -145,11 +199,13 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_f16_instances = std::tuple<
-    // clang-format off
+// clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+#if defined(CK_USE_AMD_MFMA_GFX950)
+#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1, F16, F16, 1, 1>,
     
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              2,              2,      false,  S<4, 16,  1>, S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 2, F16, F16, 2, 2>,
@@ -185,11 +241,13 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_bf16_generic_instances =
     std::tuple<
-        // clang-format off
+// clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+#if defined(CK_USE_AMD_MFMA_GFX950)
+#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,    F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1, BF16, BF16, 1, 1>
         // clang-format on
         >;
@@ -202,11 +260,13 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_bf16_instances = std::tuple<
-    // clang-format off
+// clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+#if defined(CK_USE_AMD_MFMA_GFX950)
+#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,    F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1, BF16, BF16, 1, 1>,
     
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,    F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              2,              2,      false,  S<4, 16,  1>, S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 2, BF16, BF16, 2, 2>,

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

@@ -56,11 +56,13 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_bf16_comp_instances = std::tuple<
-    // clang-format off
+// 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|
         //########################################|    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(__gfx950__)
+#else
         // Compute friendly
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
@@ -79,7 +81,7 @@ using device_grouped_conv_fwd_xdl_bf16_comp_instances = std::tuple<
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,    64,    64,   8,   8,  32,   32,    2,    1,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,    64,   128,    64,   8,   8,  32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,    64,    64,    64,   8,   8,  32,   32,    1,    1,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
-
+#endif // defined(__gfx950__)
     // clang-format on
     >;
 
@@ -90,11 +92,13 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_f16_comp_instances = std::tuple<
-    // clang-format off
+// 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|
         //########################################|    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(__gfx950__)
+#else
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   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,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
@@ -109,6 +113,7 @@ using device_grouped_conv_fwd_xdl_f16_comp_instances = std::tuple<
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   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,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   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,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
+#endif // defined(__gfx950__)
     // clang-format on
     >;
 
@@ -138,11 +143,13 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_int8_comp_instances = std::tuple<
-    // clang-format off
+// 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|
         //########################################|    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(__gfx950__)
+#else
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   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,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
@@ -153,6 +160,7 @@ using device_grouped_conv_fwd_xdl_int8_comp_instances = std::tuple<
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   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,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   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,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
+#endif // defined(__gfx950__)
     // clang-format on
     >;
 

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

@@ -40,15 +40,18 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_merged_groups_bf16_instances = std::tuple<
-    // clang-format off
+// 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| ACompute| BCompute| BlockGemm| NumGroups|
         //########################################|    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|     Type|     Type|  Pipeline|   ToMerge|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |   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|         |         | Scheduler|          |
         //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |         |         |          |          |
+#if defined(__gfx950__)
+#else
         // Instances with NumGroupsPerBatch > 1
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 8>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 16>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 32>
+#endif // defined(__gfx950__)
     // clang-format on
     >;
 
@@ -59,15 +62,18 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_merged_groups_f16_instances = std::tuple<
-    // clang-format off
+// 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|
         //########################################|    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(__gfx950__)
+#else
         // Instances with NumGroupsPerBatch > 1
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 8>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 16>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 32>
+#endif // defined(__gfx950__)
     // clang-format on
     >;