Add more memory-bound instances.

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

@@ -83,7 +83,7 @@ void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_nk_mn_mem_pv
                                                   PassThrough,
                                                   PassThrough>>>& instances);
 
-void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v1(
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv1(
     std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
                                                   Row,
                                                   Empty_Tuple,
@@ -96,7 +96,7 @@ void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregu
                                                   PassThrough,
                                                   PassThrough>>>& instances);
 
-void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v1_interwave(
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv1_interwave(
     std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
                                                   Row,
                                                   Empty_Tuple,
@@ -109,7 +109,7 @@ void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregu
                                                   PassThrough,
                                                   PassThrough>>>& instances);
 
-void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v2(
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv2(
     std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
                                                   Row,
                                                   Empty_Tuple,
@@ -121,6 +121,59 @@ void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregu
                                                   PassThrough,
                                                   PassThrough,
                                                   PassThrough>>>& instances);
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv1_pf1_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv1_pf2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv2_pf1_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv2_pf2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances);
+
 #endif
 
 template <typename ALayout,
@@ -168,11 +221,20 @@ struct DeviceOperationInstanceFactory<
                          is_same_v<ELayout, Row>)
             {
 #if defined(CK_ENABLE_FP16)
-                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v1(
+                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv1(
+                    op_ptrs);
+                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv1_interwave(
+                    op_ptrs);
+                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv2(
+                    op_ptrs);
+
+                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv1_pf1_instances(
+                    op_ptrs);
+                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv1_pf2_instances(
                     op_ptrs);
-                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v1_interwave(
+                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv2_pf1_instances(
                     op_ptrs);
-                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v2(
+                add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv2_pf2_instances(
                     op_ptrs);
 #endif
             }

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

 # ONLY XDL_KERNELS
 add_instance_library(device_grouped_gemm_multiple_d_instance
+   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance_pv1.cpp
+   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance_pv1_interwave.cpp
+   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance_pv2.cpp
+
    device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_nk_mn_irregular_instance.cpp
    device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_nk_mn_memory_pv1_pf1_instance.cpp
    device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_nk_mn_memory_pv1_pf2_instance.cpp
    device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_nk_mn_memory_pv2_pf1_instance.cpp
    device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_nk_mn_memory_pv2_pf2_instance.cpp
-   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance_pipeline_v1.cpp
-   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance_pipeline_v1_interwave.cpp
-   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance_pipeline_v2.cpp
+
+   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_pv1_pf1_instance.cpp
+   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_pv1_pf2_instance.cpp
+   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_pv2_pf1_instance.cpp
+   device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_pv2_pf2_instance.cpp
 )

library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance_pipeline_v1_interwave.cpp → library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance_pv1.cpp

@@ -6,19 +6,20 @@
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_tile_loop.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
 
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
 using GemmInstances =
-    device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_tile_instances<
-        PipelineVersion::v1,
-        LoopScheduler::Interwave>;
+    device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_tile_instances<GemmMNKPadding,
+                                                                            PipelineVersion::v1>;
 
-void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v1_interwave(
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv1(
     std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
                                                   Row,
                                                   Empty_Tuple,

library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance_pipeline_v1.cpp → library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance_pv1_interwave.cpp

@@ -6,18 +6,21 @@
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_tile_loop.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
 
-using GemmInstances =
-    device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_tile_instances<
-        PipelineVersion::v1>;
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
 
-void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v1(
+using GemmInstances = device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_tile_instances<
+    GemmMNKPadding,
+    PipelineVersion::v1,
+    LoopScheduler::Interwave>;
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv1_interwave(
     std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
                                                   Row,
                                                   Empty_Tuple,

library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance_pipeline_v2.cpp → library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance_pv2.cpp

@@ -6,18 +6,20 @@
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_tile_loop.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
 namespace instance {
 
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
 using GemmInstances =
-    device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_irregular_tile_instances<
-        PipelineVersion::v2>;
+    device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_tile_instances<GemmMNKPadding,
+                                                                            PipelineVersion::v2>;
 
-void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_irregular_instances_pipeline_v2(
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_instances_pv2(
     std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
                                                   Row,
                                                   Empty_Tuple,

library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_pv1_pf1_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_tile_loop.hpp"
+#include "ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_selector.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F16         = ck::half_t;
+using Row         = ck::tensor_layout::gemm::RowMajor;
+using Empty_Tuple = ck::Tuple<>;
+
+using PassThrough                    = ck::tensor_operation::element_wise::PassThrough;
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+static constexpr ck::index_t NumPrefetchK = 1;
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv1_pf1_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_instances<
+            GemmMNKPadding,
+            NumPrefetchK,
+            ck::PipelineVersion::v1>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_pv1_pf2_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_tile_loop.hpp"
+#include "ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_selector.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F16         = ck::half_t;
+using Row         = ck::tensor_layout::gemm::RowMajor;
+using Empty_Tuple = ck::Tuple<>;
+
+using PassThrough                    = ck::tensor_operation::element_wise::PassThrough;
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+static constexpr ck::index_t NumPrefetchK = 2;
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv1_pf2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_instances<
+            GemmMNKPadding,
+            NumPrefetchK,
+            ck::PipelineVersion::v1>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_pv2_pf1_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_tile_loop.hpp"
+#include "ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_selector.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F16         = ck::half_t;
+using Row         = ck::tensor_layout::gemm::RowMajor;
+using Empty_Tuple = ck::Tuple<>;
+
+using PassThrough                    = ck::tensor_operation::element_wise::PassThrough;
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+static constexpr ck::index_t NumPrefetchK = 1;
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv2_pf1_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_instances<
+            GemmMNKPadding,
+            NumPrefetchK,
+            ck::PipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_pv2_pf2_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_multiple_d/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_instance.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_splitk_xdl_cshuffle_tile_loop.hpp"
+#include "ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_selector.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using F16         = ck::half_t;
+using Row         = ck::tensor_layout::gemm::RowMajor;
+using Empty_Tuple = ck::Tuple<>;
+
+using PassThrough                    = ck::tensor_operation::element_wise::PassThrough;
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+static constexpr ck::index_t NumPrefetchK = 2;
+
+void add_device_grouped_gemm_multi_d_splitk_cshuffle_f16_f16_f16_mk_kn_mn_mem_pv2_pf2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemm<Row,
+                                                  Row,
+                                                  Empty_Tuple,
+                                                  Row,
+                                                  F16,
+                                                  F16,
+                                                  Empty_Tuple,
+                                                  F16,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_ggemm_md_splitk_xdl_cshuffle_f16_f16_f16_mk_kn_mn_memory_instances<
+            GemmMNKPadding,
+            NumPrefetchK,
+            ck::PipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck