change add device instances function name

example/47_gemm_bias_softmax_gemm_permute/gemm_bias_softmax_gemm_permute_nolib.cpp

@@ -134,8 +134,8 @@ int main()
         MaskingSpec>;
 
     // get device op instances
-    const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
-        DeviceOp>::GetInstances();
+    std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
+    ck::tensor_operation::device::instance::add_device_instances(op_ptrs);
 
     std::cout << "found " << op_ptrs.size() << " instances" << std::endl;
 

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

@@ -11,7 +11,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
-#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_multiple_d_softmax_gemm_permute_xdl_cshuffle_half_gmk_gnk_gno_gmo_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/batched_gemm_softmax_gemm_permute.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -19,9 +19,8 @@ namespace device {
 namespace instance {
 
 extern template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<
-        DeviceBatchedGemmSoftmaxGemmPermute<2,
+add_device_instances(std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<
+                         2,
                          1,
                          1,
                          1,
@@ -37,11 +36,9 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
                          ScaleAdd,
                          PassThrough,
                          PassThrough,
-                                            MaskingSpecialization::MaskOutUpperTriangle>>>&
-        instances);
+                         MaskingSpecialization::MaskOutUpperTriangle>>>& instances);
 
-extern template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
+extern template void add_device_instances(
     std::vector<
         std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<2,
                                                             1,
@@ -63,9 +60,8 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
         instances);
 
 extern template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<
-        DeviceBatchedGemmSoftmaxGemmPermute<2,
+add_device_instances(std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<
+                         2,
                          1,
                          1,
                          1,
@@ -81,11 +77,9 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
                          ScaleAdd,
                          PassThrough,
                          PassThrough,
-                                            MaskingSpecialization::MaskOutUpperTriangle>>>&
-        instances);
+                         MaskingSpecialization::MaskOutUpperTriangle>>>& instances);
 
-extern template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
+extern template void add_device_instances(
     std::vector<
         std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<2,
                                                             1,
@@ -107,9 +101,8 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
         instances);
 
 extern template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<
-        DeviceBatchedGemmSoftmaxGemmPermute<2,
+add_device_instances(std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<
+                         2,
                          1,
                          1,
                          1,
@@ -125,11 +118,9 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
                          Scale,
                          PassThrough,
                          PassThrough,
-                                            MaskingSpecialization::MaskOutUpperTriangle>>>&
-        instances);
+                         MaskingSpecialization::MaskOutUpperTriangle>>>& instances);
 
-extern template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
+extern template void add_device_instances(
     std::vector<
         std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<2,
                                                             1,
@@ -151,9 +142,8 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
         instances);
 
 extern template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<
-        DeviceBatchedGemmSoftmaxGemmPermute<2,
+add_device_instances(std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<
+                         2,
                          1,
                          1,
                          1,
@@ -169,11 +159,9 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
                          Scale,
                          PassThrough,
                          PassThrough,
-                                            MaskingSpecialization::MaskOutUpperTriangle>>>&
-        instances);
+                         MaskingSpecialization::MaskOutUpperTriangle>>>& instances);
 
-extern template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
+extern template void add_device_instances(
     std::vector<
         std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<2,
                                                             1,
@@ -249,8 +237,7 @@ struct DeviceOperationInstanceFactory<
     static auto GetInstances()
     {
         std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
-        add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-            op_ptrs);
+        add_device_instances(op_ptrs);
         return op_ptrs;
     }
 };

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

@@ -35,8 +35,8 @@ template <index_t NumDimG,
           typename B1ElementwiseOperation,
           typename C1DEElementwiseOperation,
           MaskingSpecialization MaskingSpec>
-struct DeviceOperationInstanceFactory<
-    ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute<NumDimG,
+void add_device_instances(
+    std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<NumDimG,
                                                                     NumDimM,
                                                                     NumDimN,
                                                                     NumDimK,
@@ -52,9 +52,10 @@ struct DeviceOperationInstanceFactory<
                                                                     C0DEElementwiseOperation,
                                                                     B1ElementwiseOperation,
                                                                     C1DEElementwiseOperation,
-                                                                      MaskingSpec>>
+                                                                    MaskingSpec>>>& instances)
 {
-    using DeviceOp = DeviceBatchedGemmSoftmaxGemmPermute<NumDimG,
+    add_device_operation_instances(instances,
+                                   create_device_instances<NumDimG,
                                                            NumDimM,
                                                            NumDimN,
                                                            NumDimK,
@@ -70,15 +71,8 @@ struct DeviceOperationInstanceFactory<
                                                            C0DEElementwiseOperation,
                                                            B1ElementwiseOperation,
                                                            C1DEElementwiseOperation,
-                                                         MaskingSpec>;
-    static auto GetInstances()
-    {
-        std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
-        add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-            op_ptrs);
-        return op_ptrs;
-    }
-};
+                                                           MaskingSpec>());
+}
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation

library/include/ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_multiple_d_softmax_gemm_permute_xdl_cshuffle_half_gmk_gnk_gno_gmo_instance.hpp

@@ -85,28 +85,9 @@ template <index_t NumDimG,
           typename enable_if<is_same<remove_cvref_t<ADataType>, ck::half_t>::value ||
                                  is_same<remove_cvref_t<ADataType>, ck::bhalf_t>::value,
                              bool>::type = false>
-void add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<NumDimG,
-                                                                    NumDimM,
-                                                                    NumDimN,
-                                                                    NumDimK,
-                                                                    NumDimO,
-                                                                    ADataType,
-                                                                    B0DataType,
-                                                                    B1DataType,
-                                                                    CDataType,
-                                                                    Acc0BiasDataType,
-                                                                    Acc1BiasDataType,
-                                                                    AElementwiseOperation,
-                                                                    B0ElementwiseOperation,
-                                                                    C0DEElementwiseOperation,
-                                                                    B1ElementwiseOperation,
-                                                                    C1DEElementwiseOperation,
-                                                                    MaskingSpec>>>& instances)
+auto create_device_instances()
 {
-    add_device_operation_instances(
-        instances,
-        device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances<
+    return device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances<
         NumDimG,
         NumDimM,
         NumDimN,
@@ -116,7 +97,7 @@ void add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk
         F32,
         Acc0BiasDataType,
         C0DEElementwiseOperation,
-            MaskingSpec>{});
+        MaskingSpec>{};
 }
 
 } // namespace instance

library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp

@@ -6,7 +6,7 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_multiple_d_softmax_gemm_permute_xdl_cshuffle_half_gmk_gnk_gno_gmo_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/batched_gemm_softmax_gemm_permute.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
@@ -26,10 +26,8 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 using ScaleAdd    = ck::tensor_operation::element_wise::ScaleAdd;
 
-template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<
-        DeviceBatchedGemmSoftmaxGemmPermute<2,
+template void add_device_instances(std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<
+                                       2,
                                        1,
                                        1,
                                        1,
@@ -45,11 +43,9 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
                                        ScaleAdd,
                                        PassThrough,
                                        PassThrough,
-                                            MaskingSpecialization::MaskOutUpperTriangle>>>&
-        instances);
+                                       MaskingSpecialization::MaskOutUpperTriangle>>>& instances);
 
-template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
+template void add_device_instances(
     std::vector<
         std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<2,
                                                             1,

library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp

@@ -6,7 +6,7 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_multiple_d_softmax_gemm_permute_xdl_cshuffle_half_gmk_gnk_gno_gmo_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/batched_gemm_softmax_gemm_permute.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
@@ -26,10 +26,8 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 using ScaleAdd    = ck::tensor_operation::element_wise::ScaleAdd;
 
-template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<
-        DeviceBatchedGemmSoftmaxGemmPermute<2,
+template void add_device_instances(std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<
+                                       2,
                                        1,
                                        1,
                                        1,
@@ -45,11 +43,9 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
                                        ScaleAdd,
                                        PassThrough,
                                        PassThrough,
-                                            MaskingSpecialization::MaskOutUpperTriangle>>>&
-        instances);
+                                       MaskingSpecialization::MaskOutUpperTriangle>>>& instances);
 
-template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
+template void add_device_instances(
     std::vector<
         std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<2,
                                                             1,

library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp

@@ -6,7 +6,7 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_multiple_d_softmax_gemm_permute_xdl_cshuffle_half_gmk_gnk_gno_gmo_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/batched_gemm_softmax_gemm_permute.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
@@ -26,10 +26,8 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 using Scale       = ck::tensor_operation::element_wise::Scale;
 
-template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<
-        DeviceBatchedGemmSoftmaxGemmPermute<2,
+template void add_device_instances(std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<
+                                       2,
                                        1,
                                        1,
                                        1,
@@ -45,11 +43,9 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
                                        Scale,
                                        PassThrough,
                                        PassThrough,
-                                            MaskingSpecialization::MaskOutUpperTriangle>>>&
-        instances);
+                                       MaskingSpecialization::MaskOutUpperTriangle>>>& instances);
 
-template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
+template void add_device_instances(
     std::vector<
         std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<2,
                                                             1,

library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp

@@ -6,7 +6,7 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_multiple_d_softmax_gemm_permute_xdl_cshuffle_half_gmk_gnk_gno_gmo_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/batched_gemm_softmax_gemm_permute.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 
 namespace ck {
@@ -26,10 +26,8 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 using Scale       = ck::tensor_operation::element_wise::Scale;
 
-template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
-    std::vector<std::unique_ptr<
-        DeviceBatchedGemmSoftmaxGemmPermute<2,
+template void add_device_instances(std::vector<std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<
+                                       2,
                                        1,
                                        1,
                                        1,
@@ -45,11 +43,9 @@ add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_
                                        Scale,
                                        PassThrough,
                                        PassThrough,
-                                            MaskingSpecialization::MaskOutUpperTriangle>>>&
-        instances);
+                                       MaskingSpecialization::MaskOutUpperTriangle>>>& instances);
 
-template void
-add_device_batched_gemm_mutiple_d_softmax_gemm_permute_xdl_cshuffle_gmk_gnk_gno_gmo_instances(
+template void add_device_instances(
     std::vector<
         std::unique_ptr<DeviceBatchedGemmSoftmaxGemmPermute<2,
                                                             1,