Undo some changes

include/ck/tensor_operation/gpu/device/impl/device_convnd_bwd_data_nwc_kxc_nwk_dl.hpp

@@ -1515,9 +1515,8 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Dl
                         wei_element_op,
                         out_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument>
     MakeArgumentPointer(void* p_in_grid,

include/ck/tensor_operation/gpu/device/impl/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp

@@ -1399,9 +1399,8 @@ struct DeviceConvNdBwdDataNwcKxcNwk_Xdl
                         input_left_pads,
                         input_right_pads};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument>
     MakeArgumentPointer(void* p_in_grid,

include/ck/tensor_operation/gpu/device/impl/device_elementwise_2d_impl.hpp

@@ -325,9 +325,8 @@ struct DeviceElementwise2dImpl : public DeviceElementwise<InDataTypeTuple,
                                           out_dev_buffers,
                                           elementwise_op);
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
     std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
     {
         return std::make_unique<Invoker>();

include/ck/tensor_operation/gpu/device/impl/device_elementwise_impl.hpp

@@ -290,9 +290,8 @@ struct DeviceElementwiseImpl
                                           out_dev_buffers,
                                           elementwise_op);
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
     std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
     {
         return std::make_unique<Invoker>();

include/ck/tensor_operation/gpu/device/impl/device_gemm_bias_add_reduce_xdl_cshuffle.hpp

@@ -771,9 +771,8 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle : public DeviceGemmReduce<1, ReduceO
                         reduce_in_element_ops,
                         reduce_out_element_ops};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>

include/ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp

@@ -546,9 +546,8 @@ struct DeviceGemmDl : public DeviceGemm<ALayout,
                         b_element_op,
                         c_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_dl.hpp

@@ -603,9 +603,8 @@ struct DeviceGemmMultipleD_Dl : public DeviceGemmMultipleD<ALayout,
                         b_element_op,
                         cde_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_layernorm_xdl_cshuffle.hpp

@@ -996,9 +996,8 @@ struct DeviceGemmMultipleDLayernorm_Xdl_CShuffle
                         cde_element_op,
                         h_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp

@@ -611,9 +611,8 @@ struct DeviceGemmMultipleDMultipleR_Xdl_CShuffle
                         qs_element_op,
                         rs_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_wmma_cshuffle.hpp

@@ -656,9 +656,8 @@ struct DeviceGemmMultipleD_Wmma_CShuffle : public DeviceGemmMultipleD<ALayout,
                                           b_element_op,
                                           cde_element_op);
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseInvoker> MakeInvokerPointer() override

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp

@@ -306,7 +306,7 @@ struct DeviceGemmMultipleD_Xdl_CShuffle : public DeviceGemmMultipleD<ALayout,
     // block-to-e-tile map
     using Block2ETileMap =
         remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>;
-
+#ifndef __HIPCC_RTC__
     // Argument
     struct Argument : public BaseArgument
     {
@@ -376,7 +376,6 @@ struct DeviceGemmMultipleD_Xdl_CShuffle : public DeviceGemmMultipleD<ALayout,
             }
         }
 
-#ifndef __HIPCC_RTC__
         void Print() const
         {
             std::cout << "A[M, K]: " << a_grid_desc_m_k_ << std::endl;
@@ -385,7 +384,6 @@ struct DeviceGemmMultipleD_Xdl_CShuffle : public DeviceGemmMultipleD<ALayout,
                 [&](auto i) { std::cout << "Ds[M, N]: " << ds_grid_desc_m_n_[i] << std::endl; });
             std::cout << "E[M, N]: " << e_grid_desc_m_n_ << std::endl;
         }
-#endif
 
         //  private:
         // pointers
@@ -420,7 +418,6 @@ struct DeviceGemmMultipleD_Xdl_CShuffle : public DeviceGemmMultipleD<ALayout,
         index_t NRaw_;
         index_t KRaw_;
     };
-#ifndef __HIPCC_RTC__
     // Invoker
     struct Invoker : public BaseInvoker
     {
@@ -601,7 +598,6 @@ struct DeviceGemmMultipleD_Xdl_CShuffle : public DeviceGemmMultipleD<ALayout,
     {
         return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
     }
-#endif
     static auto MakeArgument(const void* p_a,
                              const void* p_b,
                              ck::Array<const void*, NumDTensor> p_ds,
@@ -633,9 +629,7 @@ struct DeviceGemmMultipleD_Xdl_CShuffle : public DeviceGemmMultipleD<ALayout,
                         cde_element_op};
     }
 
-#ifndef __HIPCC_RTC__
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>
@@ -670,7 +664,6 @@ struct DeviceGemmMultipleD_Xdl_CShuffle : public DeviceGemmMultipleD<ALayout,
                                           cde_element_op);
     }
 
-#ifndef __HIPCC_RTC__
     // polymorphic
     std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
     {

include/ck/tensor_operation/gpu/device/impl/device_gemm_reduce_xdl_cshuffle.hpp

@@ -732,9 +732,8 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<0, ReduceOperatio
                         reduce_in_element_ops,
                         reduce_out_element_ops};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>

include/ck/tensor_operation/gpu/device/impl/device_gemm_wmma.hpp

@@ -526,9 +526,8 @@ struct DeviceGemmWmma_CShuffle : public DeviceGemm<ALayout,
                         b_element_op,
                         c_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,

include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_layernorm_cshuffle.hpp

@@ -701,9 +701,8 @@ struct DeviceGemmLayerNorm_Xdl_CShuffle : public BaseOperator
                         acc_element_op,
                         c_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
                                                       const void* p_b,

include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_skip_b_lds.hpp

@@ -462,9 +462,8 @@ struct DeviceGemmXdlSkipBLds : public DeviceGemm<ALayout,
                         b_element_op,
                         c_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,

include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_splitk_c_shuffle.hpp

@@ -280,9 +280,8 @@ struct DeviceGemmXdlSplitKCShuffle : public DeviceGemmSplitK<ALayout,
                         GridwiseGemm::CalculateK0(K, KBatch),
                         KBatch};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,

include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_waveletmodel_cshuffle.hpp

@@ -3,12 +3,8 @@
 
 #pragma once
 
-#ifndef __HIPCC_RTC__
 #include <iostream>
 #include <sstream>
-#include "ck/host_utility/device_prop.hpp"
-#include "ck/host_utility/kernel_launch.hpp"
-#endif
 
 #include "ck/utility/common_header.hpp"
 #include "ck/tensor_description/tensor_descriptor.hpp"
@@ -18,6 +14,8 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
 #include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_waveletmodel_cshuffle.hpp"
+#include "ck/host_utility/device_prop.hpp"
+#include "ck/host_utility/kernel_launch.hpp"
 
 namespace ck {
 
@@ -419,7 +417,6 @@ struct DeviceGemm_Xdl_WaveletModel_CShuffle : public DeviceGemm<ALayout,
         }
     };
 
-#ifndef __HIPCC_RTC__
     static bool IsSupportedArgument(const Argument& arg)
     {
         if(!ck::is_xdl_supported())
@@ -438,7 +435,6 @@ struct DeviceGemm_Xdl_WaveletModel_CShuffle : public DeviceGemm<ALayout,
     {
         return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
     }
-#endif
 
     static auto MakeArgument(const ADataType* p_a,
                              const BDataType* p_b,
@@ -466,9 +462,8 @@ struct DeviceGemm_Xdl_WaveletModel_CShuffle : public DeviceGemm<ALayout,
                         b_element_op,
                         cde_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>

include/ck/tensor_operation/gpu/device/impl/device_grouped_contraction_multiple_d_xdl_cshuffle.hpp

@@ -846,9 +846,8 @@ struct DeviceGroupedContractionMultipleD_Xdl_CShuffle
                         b_element_op,
                         cde_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_data_multiple_d_xdl_cshuffle_v1.hpp

@@ -993,9 +993,8 @@ struct DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
                         b_element_op,
                         cde_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument> MakeArgumentPointer(
         const void* p_a,                                                 // output image

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_gnwc_gkxc_gnwk_dl.hpp

@@ -1155,9 +1155,8 @@ struct DeviceGroupedConvBwdWeightGnwcGkxcGnwk_Dl
                         out_element_op,
                         split_k};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument>
     MakeArgumentPointer(const void* p_in_grid,