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

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
0dcb3496 · Chao Liu · GitHub · fa9a0a5c · 0dcb3496 · 0dcb3496
Unverified Commit 0dcb3496 authored Jun 30, 2022 by Chao Liu Committed by GitHub Jun 30, 2022
20 changed files
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_bf16_bf16_bf16_gmk_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_bf16_bf16_bf16_gmk_gnk_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using BF16 = ck::bhalf_t;
 using F32  = float;
@@ -49,13 +49,15 @@ using device_batched_gemm_xdl_bf16_bf16_bf16_gmk_gnk_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_bf16_bf16_bf16_gmk_gnk_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Row, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_bf16_bf16_bf16_gmk_gnk_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -44,13 +44,15 @@ using device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Col, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -44,13 +44,15 @@ using device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Col, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -53,13 +53,15 @@ using device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Row, Row, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -49,13 +49,15 @@ using device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Row, Col, Row, F16, F16, F16, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f32_f32_f32_gkm_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f32_f32_f32_gkm_gkn_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -44,13 +44,15 @@ using device_batched_gemm_xdl_f32_f32_f32_gkm_gkn_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_f32_f32_f32_gkm_gkn_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Col, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_f32_f32_f32_gkm_gkn_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f32_f32_f32_gkm_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f32_f32_f32_gkm_gnk_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -44,13 +44,15 @@ using device_batched_gemm_xdl_f32_f32_f32_gkm_gnk_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_f32_f32_f32_gkm_gnk_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Col, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_f32_f32_f32_gkm_gnk_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f32_f32_f32_gmk_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f32_f32_f32_gmk_gkn_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -44,13 +44,15 @@ using device_batched_gemm_xdl_f32_f32_f32_gmk_gkn_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_f32_f32_f32_gmk_gkn_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Row, Row, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_f32_f32_f32_gmk_gkn_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f32_f32_f32_gmk_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f32_f32_f32_gmk_gnk_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -49,13 +49,15 @@ using device_batched_gemm_xdl_f32_f32_f32_gmk_gnk_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_f32_f32_f32_gmk_gnk_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<
+        DeviceBatchedGemm<Row, Col, Row, F32, F32, F32, PassThrough, PassThrough, PassThrough>>>&
+        instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_f32_f32_f32_gmk_gnk_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -59,13 +59,21 @@ using device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<DeviceBatchedGemm<Col,
+                                                  Row,
+                                                  Row,
+                                                  int8_t,
+                                                  int8_t,
+                                                  int8_t,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_int8_int8_int8_gkm_gkn_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -59,13 +59,21 @@ using device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<DeviceBatchedGemm<Col,
+                                                  Col,
+                                                  Row,
+                                                  int8_t,
+                                                  int8_t,
+                                                  int8_t,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_int8_int8_int8_gkm_gnk_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -59,13 +59,21 @@ using device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<DeviceBatchedGemm<Row,
+                                                  Row,
+                                                  Row,
+                                                  int8_t,
+                                                  int8_t,
+                                                  int8_t,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_int8_int8_int8_gmk_gkn_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gmk_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_int8_int8_int8_gmk_gnk_gmn_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_batched_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_batched_gemm_instance {
+namespace instance {

 using Row = ck::tensor_layout::gemm::RowMajor;
 using Col = ck::tensor_layout::gemm::ColumnMajor;
@@ -51,13 +51,21 @@ using device_batched_gemm_xdl_int8_int8_int8_gmk_gnk_gmn_instances = std::tuple<
    >;

 void add_device_batched_gemm_xdl_int8_int8_int8_gmk_gnk_gmn_instances(
-    std::vector<DeviceBatchedGemmPtr<PassThrough, PassThrough, PassThrough>>& instances)
+    std::vector<std::unique_ptr<DeviceBatchedGemm<Row,
+                                                  Col,
+                                                  Row,
+                                                  int8_t,
+                                                  int8_t,
+                                                  int8_t,
+                                                  PassThrough,
+                                                  PassThrough,
+                                                  PassThrough>>>& instances)
 {
    add_device_operation_instances(instances,
                                   device_batched_gemm_xdl_int8_int8_int8_gmk_gnk_gmn_instances{});
 }

-} // namespace device_batched_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gkn_gmn_instance.cpp
@@ -8,12 +8,12 @@
 #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_batched_gemm_reduce_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;
@@ -74,7 +74,7 @@ void add_device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gkn_gmn
        device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gkn_gmn_instances{});
 }

-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gnk_gmn_instance.cpp
@@ -8,12 +8,12 @@
 #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_batched_gemm_reduce_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;
@@ -74,7 +74,7 @@ void add_device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gnk_gmn
        device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gnk_gmn_instances{});
 }

-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gkn_gmn_instance.cpp
@@ -8,12 +8,12 @@
 #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_batched_gemm_reduce_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;
@@ -74,7 +74,7 @@ void add_device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gkn_gmn
        device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gkn_gmn_instances{});
 }

-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gnk_gmn_instance.cpp
@@ -8,12 +8,12 @@
 #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_batched_gemm_reduce_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;
@@ -71,7 +71,7 @@ void add_device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gnk_gmn
        device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gnk_gmn_instances{});
 }

-} // namespace device_gemm_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/conv1d_fwd/device_conv1d_fwd_xdl_nwc_kxc_nwk_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/conv1d_fwd/device_conv1d_fwd_xdl_nwc_kxc_nwk_bf16_instance.cpp
@@ -8,12 +8,12 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_convnd_fwd_xdl_nhwc_kyxc_nhwk.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_conv1d_fwd_instance {
+namespace instance {

 using F32  = float;
 using BF16 = bhalf_t;
@@ -109,7 +109,7 @@ void add_device_conv1d_fwd_xdl_nwc_kxc_nwk_bf16_instances(
                                   device_conv1d_fwd_xdl_nwc_kxc_nwk_1x1_s1_p0_bf16_instances{});
 }

-} // namespace device_conv1d_fwd_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/conv1d_fwd/device_conv1d_fwd_xdl_nwc_kxc_nwk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/conv1d_fwd/device_conv1d_fwd_xdl_nwc_kxc_nwk_f16_instance.cpp
@@ -8,12 +8,12 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_convnd_fwd_xdl_nhwc_kyxc_nhwk.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_conv1d_fwd_instance {
+namespace instance {

 using F16 = ck::half_t;
 using F32 = float;
@@ -109,7 +109,7 @@ void add_device_conv1d_fwd_xdl_nwc_kxc_nwk_f16_instances(
                                   device_conv1d_fwd_xdl_nwc_kxc_nwk_1x1_s1_p0_f16_instances{});
 }

-} // namespace device_conv1d_fwd_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
--- a/library/src/tensor_operation_instance/gpu/conv1d_fwd/device_conv1d_fwd_xdl_nwc_kxc_nwk_f32_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/conv1d_fwd/device_conv1d_fwd_xdl_nwc_kxc_nwk_f32_instance.cpp
@@ -8,12 +8,12 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/device_convnd_fwd_xdl_nhwc_kyxc_nhwk.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_conv1d_fwd_instance {
+namespace instance {

 using F32 = float;

@@ -112,7 +112,7 @@ void add_device_conv1d_fwd_xdl_nwc_kxc_nwk_f32_instances(
                                   device_conv1d_fwd_xdl_nwc_kxc_nwk_1x1_s1_p0_f32_instances{});
 }

-} // namespace device_conv1d_fwd_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck