Merge branch 'develop' into standalone-layernorm

library/src/tensor_operation_instance/gpu/convnd_bwd_data/device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_int8_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_convnd_bwd_data_xdl_ndhwc_kzyxc_ndhwk.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_conv2d_bwd_data_instance {
+namespace instance {
 
 using DataType = int8_t;
 using AccType  = int32_t;
@@ -83,7 +83,7 @@ void add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_int8_instances(
         instances, device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_int8_instances{});
 }
 
-} // namespace device_conv2d_bwd_data_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/elementwise/device_normalize_instance.cpp

@@ -7,11 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_5ary_elementwise.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -44,6 +45,7 @@ void add_device_normalize_from_mean_squaremean_f16_f32_f32_f16_f16_instances(
         instances, device_normalize_from_mean_squaremean_f16_f32_f32_f16_f16_instances{});
 }
 
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f16_f16_f16_km_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -39,12 +39,14 @@ using device_gemm_dl_f16_f16_f16_km_kn_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_dl_f16_f16_f16_km_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_f16_f16_f16_km_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f16_f16_f16_km_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -39,12 +39,14 @@ using device_gemm_dl_f16_f16_f16_km_nk_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_dl_f16_f16_f16_km_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_f16_f16_f16_km_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f16_f16_f16_mk_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -39,12 +39,14 @@ using device_gemm_dl_f16_f16_f16_mk_kn_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_dl_f16_f16_f16_mk_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_f16_f16_f16_mk_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f16_f16_f16_mk_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -40,12 +40,14 @@ using device_gemm_dl_f16_f16_f16_mk_nk_mn_instances =
         >;
 
 void add_device_gemm_dl_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_f16_f16_f16_mk_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f32_f32_f32_km_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -39,12 +39,14 @@ using device_gemm_dl_f32_f32_f32_km_kn_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_dl_f32_f32_f32_km_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_f32_f32_f32_km_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f32_f32_f32_km_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -40,12 +40,14 @@ using device_gemm_dl_f32_f32_f32_km_nk_mn_instances =
         >;
 
 void add_device_gemm_dl_f32_f32_f32_km_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_f32_f32_f32_km_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f32_f32_f32_mk_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -40,12 +40,14 @@ using device_gemm_dl_f32_f32_f32_mk_kn_mn_instances =
         >;
 
 void add_device_gemm_dl_f32_f32_f32_mk_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_f32_f32_f32_mk_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_f32_f32_f32_mk_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -40,12 +40,14 @@ using device_gemm_dl_f32_f32_f32_mk_nk_mn_instances =
         >;
 
 void add_device_gemm_dl_f32_f32_f32_mk_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_f32_f32_f32_mk_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_km_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -36,12 +36,14 @@ using device_gemm_dl_i8_i8_i8_km_kn_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_dl_i8_i8_i8_km_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_i8_i8_i8_km_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_km_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -36,12 +36,14 @@ using device_gemm_dl_i8_i8_i8_km_nk_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_i8_i8_i8_km_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_mk_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -36,12 +36,14 @@ using device_gemm_dl_i8_i8_i8_mk_kn_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_dl_i8_i8_i8_mk_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_i8_i8_i8_mk_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_mk_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_dl.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -36,12 +36,14 @@ using device_gemm_dl_i8_i8_i8_mk_nk_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_dl_i8_i8_i8_mk_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, int8_t, int8_t, int8_t, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_dl_i8_i8_i8_mk_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_xdl_cshuffle.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -51,13 +51,15 @@ using device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances = std::tu
     >;
 
 void add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_xdl_cshuffle.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using BF16 = ck::bhalf_t;
 using F32  = float;
@@ -54,13 +54,15 @@ using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances,
                                    device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_xdl_cshuffle.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using BF16 = ck::bhalf_t;
 using F32  = float;
@@ -54,13 +54,15 @@ using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances,
                                    device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_xdl_cshuffle.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using BF16 = ck::bhalf_t;
 using F32  = float;
@@ -54,13 +54,15 @@ using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances,
                                    device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_xdl_cshuffle.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using BF16 = ck::bhalf_t;
 using F32  = float;
@@ -51,13 +51,15 @@ using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances,
                                    device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp

@@ -7,12 +7,12 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_xdl_cshuffle.hpp"
-#include "ck/library/tensor_operation_instance/device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
-namespace device_gemm_instance {
+namespace instance {
 
 using F16 = ck::half_t;
 using F32 = float;
@@ -54,13 +54,15 @@ using device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances = std::tuple<
     >;
 
 void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances,
                                    device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck