Merge remote-tracking branch 'origin/develop' into contraction

library/src/tensor_operation_instance/gpu/convnd_bwd_data/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f32_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 F32 = float;
 
@@ -82,7 +82,7 @@ void add_device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f32_instances(
         instances, device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f32_instances{});
 }
 
-} // namespace device_conv2d_bwd_data_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/convnd_bwd_data/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_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_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_int8_instances(
         instances, device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_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/convnd_bwd_data/device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_bf16_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 BF16 = bhalf_t;
 using F32  = float;
@@ -83,7 +83,7 @@ void add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_bf16_instances(
         instances, device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_bf16_instances{});
 }
 
-} // namespace device_conv2d_bwd_data_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/convnd_bwd_data/device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f16_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 F16 = ck::half_t;
 using F32 = float;
@@ -83,7 +83,7 @@ void add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f16_instances(
         instances, device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_f16_instances{});
 }
 
-} // namespace device_conv2d_bwd_data_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

library/src/tensor_operation_instance/gpu/convnd_bwd_data/device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f32_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 F32 = float;
 
@@ -82,7 +82,7 @@ void add_device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f32_instances(
         instances, device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_f32_instances{});
 }
 
-} // namespace device_conv2d_bwd_data_instance
+} // namespace instance
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck

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/CMakeLists.txt

+set(DEVICE_ELEMENTWISE_INSTANCE_SOURCE
+    device_normalize_instance.cpp
+)
+
+add_instance_library(device_elementwise_instance ${DEVICE_ELEMENTWISE_INSTANCE_SOURCE})
+
+target_compile_features(device_elementwise_instance PUBLIC)
+set_target_properties(device_elementwise_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
+
+clang_tidy_check(device_elementwise_instance)

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

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.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/device_5ary_elementwise.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;
+
+using inputType      = F16;
+using MeanType       = F32;
+using SquareMeanType = F32;
+using GammaDataType  = F16;
+using BetaDataType   = F16;
+using outputType     = F16;
+
+using Normalize = ck::tensor_operation::element_wise::Normalize;
+using device_normalize_from_mean_squaremean_f16_f32_f32_f16_f16_instances = std::tuple<
+    // clang-format off
+    //###################|in | mean| square_mean| gamma| beta| out| ComputeDataType|  functor| NDim| MPerThread| in, mean, square_mean, gamma, beta, out ScalarPerVector|
+    //###################|in | mean| square_mean| gamma| beta| out| ComputeDataType|  functor| NDim| MPerThread| in, mean, square_mean, gamma, beta, out ScalarPerVector|
+    //###################|in | mean| square_mean| gamma| beta| out| ComputeDataType|  functor| NDim| MPerThread| in, mean, square_mean, gamma, beta, out ScalarPerVector|
+    //###################|in | mean| square_mean| gamma| beta| out| ComputeDataType|  functor| NDim| MPerThread| in, mean, square_mean, gamma, beta, out ScalarPerVector|
+    Device5AryElementwise<F16,  F32,         F32,   F16,  F16, F16,            F32, Normalize,    2,          8,  8,    1,           1,     8,    8,   8                >,
+    Device5AryElementwise<F16,  F32,         F32,   F16,  F16, F16,            F32, Normalize,    2,          4,  4,    1,           1,     4,    4,   4                >,
+    Device5AryElementwise<F16,  F32,         F32,   F16,  F16, F16,            F32, Normalize,    2,          2,  2,    1,           1,     2,    2,   2                >,
+    Device5AryElementwise<F16,  F32,         F32,   F16,  F16, F16,            F32, Normalize,    2,          1,  1,    1,           1,     1,    1,   1                >
+    // clang-format on
+    >;
+
+void add_device_normalize_from_mean_squaremean_f16_f32_f32_f16_f16_instances(
+    std::vector<DeviceElementwisePtr<5, 1, 2, Normalize>>& instances)
+{
+    add_device_operation_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/CMakeLists.txt

@@ -28,14 +28,6 @@ set(DEVICE_GEMM_INSTANCE_SOURCE
    device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instance.cpp;
    device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instance.cpp;
    device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp;
-   device_gemm_xdl_splitk_f32_f32_f32_mk_kn_mn_instance.cpp;
-   device_gemm_xdl_splitk_f32_f32_f32_mk_nk_mn_instance.cpp;
-   device_gemm_xdl_splitk_f32_f32_f32_km_kn_mn_instance.cpp;
-   device_gemm_xdl_splitk_f32_f32_f32_km_nk_mn_instance.cpp;
-   device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instance.cpp;
-   device_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_instance.cpp;
-   device_gemm_xdl_splitk_f16_f16_f16_km_kn_mn_instance.cpp;
-   device_gemm_xdl_splitk_f16_f16_f16_km_nk_mn_instance.cpp;
    device_gemm_dl_f32_f32_f32_mk_kn_mn_instance.cpp;
    device_gemm_dl_f32_f32_f32_mk_nk_mn_instance.cpp;
    device_gemm_dl_f32_f32_f32_km_kn_mn_instance.cpp;

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