Improve external interface for GEMM and GEMM+add+add+fastgelu (#311)

* interface for GEMM and GEMM+add+add+fastgelu

* rename namespace

* instance factory

* fix build

* fix build; add GEMM client example

* clean

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_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_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 F32 = float;
 
@@ -54,13 +54,15 @@ using device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instances =
         >;
 
 void add_device_gemm_xdl_c_shuffle_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_xdl_c_shuffle_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_xdl_c_shuffle_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_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 F32 = float;
 
@@ -54,13 +54,15 @@ using device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instances =
         >;
 
 void add_device_gemm_xdl_c_shuffle_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_xdl_c_shuffle_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_xdl_c_shuffle_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_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 F32 = float;
 
@@ -54,13 +54,15 @@ using device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instances =
         >;
 
 void add_device_gemm_xdl_c_shuffle_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_xdl_c_shuffle_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_xdl_c_shuffle_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_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 F32 = float;
 
@@ -51,13 +51,15 @@ using device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances =
         >;
 
 void add_device_gemm_xdl_c_shuffle_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_xdl_c_shuffle_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_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.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;
@@ -47,12 +47,14 @@ using device_gemm_xdl_f16_f16_f16_km_kn_mn_instances =
         >;
 
 void add_device_gemm_xdl_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_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_xdl_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_xdl.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;
@@ -47,12 +47,14 @@ using device_gemm_xdl_f16_f16_f16_km_nk_mn_instances =
         >;
 
 void add_device_gemm_xdl_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_xdl_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_xdl_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_xdl.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;
@@ -56,12 +56,14 @@ using device_gemm_xdl_f16_f16_f16_mk_kn_mn_instances =
         >;
 
 void add_device_gemm_xdl_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_xdl_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_xdl_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.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;
@@ -66,14 +66,16 @@ using device_gemm_xdl_f16_f16_f16_mk_nk_mn_irregular_tile_instances =
         >;
 
 void add_device_gemm_xdl_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_f16_f16_f16_mk_nk_mn_instances{});
     add_device_operation_instances(instances,
                                    device_gemm_xdl_f16_f16_f16_mk_nk_mn_irregular_tile_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_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_xdl.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;
@@ -47,12 +47,14 @@ using device_gemm_xdl_f32_f32_f32_km_kn_mn_instances =
         >;
 
 void add_device_gemm_xdl_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_xdl_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_xdl_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_xdl.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;
@@ -47,12 +47,14 @@ using device_gemm_xdl_f32_f32_f32_km_nk_mn_instances =
         >;
 
 void add_device_gemm_xdl_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_xdl_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_xdl_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_xdl.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;
@@ -47,12 +47,14 @@ using device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances =
         >;
 
 void add_device_gemm_xdl_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_xdl_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_xdl_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_xdl.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;
@@ -52,12 +52,14 @@ using device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances =
         >;
 
 void add_device_gemm_xdl_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_xdl_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_xdl_f64_f64_f64_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.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 F64 = double;
 
@@ -43,12 +43,14 @@ using device_gemm_xdl_f64_f64_f64_km_kn_mn_instances =
         >;
 
 void add_device_gemm_xdl_f64_f64_f64_km_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Row, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_xdl_f64_f64_f64_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_f64_f64_f64_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.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 F64 = double;
 
@@ -43,12 +43,14 @@ using device_gemm_xdl_f64_f64_f64_km_nk_mn_instances =
         >;
 
 void add_device_gemm_xdl_f64_f64_f64_km_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Col, Col, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_xdl_f64_f64_f64_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_f64_f64_f64_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.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 F64 = double;
 
@@ -43,12 +43,14 @@ using device_gemm_xdl_f64_f64_f64_mk_kn_mn_instances =
         >;
 
 void add_device_gemm_xdl_f64_f64_f64_mk_kn_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Row, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_xdl_f64_f64_f64_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_f64_f64_f64_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.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 F64 = double;
 
@@ -48,12 +48,14 @@ using device_gemm_xdl_f64_f64_f64_mk_nk_mn_instances =
         >;
 
 void add_device_gemm_xdl_f64_f64_f64_mk_nk_mn_instances(
-    std::vector<DeviceGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceGemm<Row, Col, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
     add_device_operation_instances(instances, device_gemm_xdl_f64_f64_f64_mk_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp

@@ -6,12 +6,13 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_xdl_cshuffle.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_gemm_instance {
+namespace instance {
 
 using F16     = ck::half_t;
 using F32     = float;
@@ -57,13 +58,22 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances
     >;
 
 void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
-    std::vector<DeviceGemmMultipleDPtr<2, PassThrough, PassThrough, AddAddFastGelu>>& instances)
+    std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
+                                                    Row,
+                                                    Row,
+                                                    F16,
+                                                    F16,
+                                                    F16_F16,
+                                                    F16,
+                                                    PassThrough,
+                                                    PassThrough,
+                                                    AddAddFastGelu>>>& instances)
 {
     add_device_operation_instances(
         instances, device_gemm_add_add_fastgelu_xdl_c_shuffle_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_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp

@@ -6,12 +6,13 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_xdl_cshuffle.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_gemm_instance {
+namespace instance {
 
 using F16     = ck::half_t;
 using F32     = float;
@@ -57,13 +58,22 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances
     >;
 
 void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(
-    std::vector<DeviceGemmMultipleDPtr<2, PassThrough, PassThrough, AddAddFastGelu>>& instances)
+    std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
+                                                    Col,
+                                                    Row,
+                                                    F16,
+                                                    F16,
+                                                    F16_F16,
+                                                    F16,
+                                                    PassThrough,
+                                                    PassThrough,
+                                                    AddAddFastGelu>>>& instances)
 {
     add_device_operation_instances(
         instances, device_gemm_add_add_fastgelu_xdl_c_shuffle_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_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp

@@ -6,12 +6,13 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_xdl_cshuffle.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_gemm_instance {
+namespace instance {
 
 using F16     = ck::half_t;
 using F32     = float;
@@ -57,13 +58,22 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances
     >;
 
 void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(
-    std::vector<DeviceGemmMultipleDPtr<2, PassThrough, PassThrough, AddAddFastGelu>>& instances)
+    std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
+                                                    Row,
+                                                    Row,
+                                                    F16,
+                                                    F16,
+                                                    F16_F16,
+                                                    F16,
+                                                    PassThrough,
+                                                    PassThrough,
+                                                    AddAddFastGelu>>>& instances)
 {
     add_device_operation_instances(
         instances, device_gemm_add_add_fastgelu_xdl_c_shuffle_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_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp

@@ -6,12 +6,13 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_xdl_cshuffle.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_gemm_instance {
+namespace instance {
 
 using F16     = ck::half_t;
 using F32     = float;
@@ -54,13 +55,22 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances
     >;
 
 void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(
-    std::vector<DeviceGemmMultipleDPtr<2, PassThrough, PassThrough, AddAddFastGelu>>& instances)
+    std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
+                                                    Col,
+                                                    Row,
+                                                    F16,
+                                                    F16,
+                                                    F16_F16,
+                                                    F16,
+                                                    PassThrough,
+                                                    PassThrough,
+                                                    AddAddFastGelu>>>& instances)
 {
     add_device_operation_instances(
         instances, device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances{});
 }
 
-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck