Undo some changes

include/ck/tensor_operation/gpu/device/device_base.hpp

@@ -12,6 +12,7 @@ namespace ck {
 namespace tensor_operation {
 namespace device {
 
+#ifndef __HIPCC_RTC__
 struct BaseArgument
 {
     BaseArgument()                    = default;
@@ -23,7 +24,6 @@ struct BaseArgument
     void* p_workspace_ = nullptr;
 };
 
-#ifndef __HIPCC_RTC__
 struct BaseInvoker
 {
     BaseInvoker()                   = default;
@@ -45,9 +45,9 @@ struct BaseOperator
     BaseOperator(const BaseOperator&) = default;
     BaseOperator& operator=(const BaseOperator&) = default;
 
+#ifndef __HIPCC_RTC__
     virtual bool IsSupportedArgument(const BaseArgument*) { return false; }
 
-#ifndef __HIPCC_RTC__
     virtual std::string GetTypeString() const { return ""; }
 
     virtual std::string GetTypeIdName() const { return typeid(*this).name(); }
@@ -60,15 +60,14 @@ struct BaseOperator
 
         return oss.str();
     };
-#endif
     virtual size_t GetWorkSpaceSize(const BaseArgument*) const { return 0; }
 
     virtual void SetWorkSpacePointer(BaseArgument* p_arg, void* p_workspace) const
     {
-        // assert(p_arg);
+        assert(p_arg);
         p_arg->p_workspace_ = p_workspace;
     }
-
+#endif
     virtual ~BaseOperator() {}
 };
 

include/ck/tensor_operation/gpu/device/device_gemm_multiple_d.hpp

@@ -36,6 +36,7 @@ struct DeviceGemmMultipleD : public BaseOperator
 {
     static constexpr index_t NumDTensor = DsDataType::Size();
 
+#ifndef __HIPCC_RTC__
     virtual std::unique_ptr<BaseArgument>
     MakeArgumentPointer(const void* p_a,
                         const void* p_b,
@@ -51,7 +52,6 @@ struct DeviceGemmMultipleD : public BaseOperator
                         AElementwiseOperation a_element_op,
                         BElementwiseOperation b_element_op,
                         CDEElementwiseOperation cde_element_op) = 0;
-#ifndef __HIPCC_RTC__
     virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer()   = 0;
 #endif
 };

include/ck/tensor_operation/gpu/device/impl/device_batched_contraction_multiple_d_wmma_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"
@@ -19,6 +15,8 @@
 #include "ck/tensor_operation/gpu/device/tensor_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
 #include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_wmma_cshuffle.hpp"
+#include "ck/host_utility/device_prop.hpp"
+#include "ck/host_utility/kernel_launch.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -689,7 +687,7 @@ struct DeviceBatchedContractionMultipleD_Wmma_CShuffle
         // Batch Offset
         ComputePtrOffsetOfStridedBatch compute_ptr_offset_of_batch_;
     };
-#ifndef __HIPCC_RTC__
+
     // Invoker
     struct Invoker : public BaseInvoker
     {
@@ -763,14 +761,13 @@ struct DeviceBatchedContractionMultipleD_Wmma_CShuffle
             return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
         }
     };
-#endif
 
     static constexpr bool IsValidCompilationParameter()
     {
         // TODO: properly implement this check
         return true;
     }
-#ifndef __HIPCC_RTC__
+
     static bool IsSupportedArgument(const Argument& arg)
     {
         if(ck::get_device_name() == "gfx1100" || ck::get_device_name() == "gfx1101" ||
@@ -872,7 +869,7 @@ struct DeviceBatchedContractionMultipleD_Wmma_CShuffle
     {
         return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
     }
-#endif
+
     static auto
     MakeArgument(const void* p_a,
                  const void* p_b,
@@ -946,7 +943,6 @@ struct DeviceBatchedContractionMultipleD_Wmma_CShuffle
                                           cde_element_op);
     }
 
-#ifndef __HIPCC_RTC__
     static auto MakeInvoker() { return Invoker{}; }
 
     // polymorphic
@@ -989,7 +985,6 @@ struct DeviceBatchedContractionMultipleD_Wmma_CShuffle
 
         return str.str();
     }
-#endif
 };
 
 } // namespace device

include/ck/tensor_operation/gpu/device/impl/device_batched_contraction_multiple_d_xdl_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"
@@ -19,6 +15,8 @@
 #include "ck/tensor_operation/gpu/device/tensor_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
 #include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
+#include "ck/host_utility/device_prop.hpp"
+#include "ck/host_utility/kernel_launch.hpp"
 
 namespace ck {
 
@@ -761,7 +759,6 @@ struct DeviceBatchedContractionMultipleD_Xdl_CShuffle
         ComputePtrOffsetOfStridedBatch compute_ptr_offset_of_batch_;
     };
 
-#ifndef __HIPCC_RTC__
     // Invoker
     struct Invoker : public BaseInvoker
     {
@@ -844,6 +841,7 @@ struct DeviceBatchedContractionMultipleD_Xdl_CShuffle
             return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
         }
     };
+
     static bool IsSupportedArgument(const Argument& arg)
     {
         if(!ck::is_xdl_supported())
@@ -937,7 +935,7 @@ struct DeviceBatchedContractionMultipleD_Xdl_CShuffle
     {
         return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
     }
-#endif
+
     static auto
     MakeArgument(const void* p_a,
                  const void* p_b,
@@ -972,9 +970,7 @@ struct DeviceBatchedContractionMultipleD_Xdl_CShuffle
                         cde_element_op};
     }
 
-#ifndef __HIPCC_RTC__
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>
@@ -1010,7 +1006,7 @@ struct DeviceBatchedContractionMultipleD_Xdl_CShuffle
                                           b_element_op,
                                           cde_element_op);
     }
-#ifndef __HIPCC_RTC__
+
     // polymorphic
     std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
     {
@@ -1042,7 +1038,6 @@ struct DeviceBatchedContractionMultipleD_Xdl_CShuffle
 
         return str.str();
     }
-#endif
 };
 
 } // namespace device

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

 #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"
@@ -15,6 +11,8 @@
 #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_multiple_d_xdl_cshuffle.hpp"
+#include "ck/host_utility/device_prop.hpp"
+#include "ck/host_utility/kernel_launch.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -492,7 +490,6 @@ struct DeviceBatchedGemmEPermuteXdl : public DeviceBatchedGemmEPermute<ALayout,
         CDEElementwiseOperation cde_element_op_;
     };
 
-#ifndef __HIPCC_RTC__
     // Invoker
     struct Invoker : public BaseInvoker
     {
@@ -568,14 +565,13 @@ struct DeviceBatchedGemmEPermuteXdl : public DeviceBatchedGemmEPermute<ALayout,
             return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
         }
     };
-#endif
 
     static constexpr bool IsValidCompilationParameter()
     {
         // TODO: properly implement this check
         return true;
     }
-#ifndef __HIPCC_RTC__
+
     static bool IsSupportedArgument(const Argument& arg)
     {
         if(!ck::is_xdl_supported())
@@ -595,7 +591,7 @@ struct DeviceBatchedGemmEPermuteXdl : public DeviceBatchedGemmEPermute<ALayout,
     {
         return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
     }
-#endif
+
     static auto MakeArgument(const ADataType* p_a,
                              const BDataType* p_b,
                              EDataType* p_e,
@@ -629,9 +625,7 @@ struct DeviceBatchedGemmEPermuteXdl : public DeviceBatchedGemmEPermute<ALayout,
                         cde_element_op};
     }
 
-#ifndef __HIPCC_RTC__
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>
@@ -668,7 +662,6 @@ struct DeviceBatchedGemmEPermuteXdl : public DeviceBatchedGemmEPermute<ALayout,
                                           cde_element_op);
     }
 
-#ifndef __HIPCC_RTC__
     // polymorphic
     std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
     {
@@ -692,7 +685,6 @@ struct DeviceBatchedGemmEPermuteXdl : public DeviceBatchedGemmEPermute<ALayout,
 
         return str.str();
     }
-#endif
 };
 
 } // namespace device

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

@@ -3,13 +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"
-#include "ck/host_utility/io.hpp"
-#endif
 
 #include "ck/utility/common_header.hpp"
 #include "ck/tensor_description/tensor_descriptor.hpp"
@@ -19,6 +14,9 @@
 #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_batched_gemm_gemm_xdl_cshuffle_v1.hpp"
+#include "ck/host_utility/device_prop.hpp"
+#include "ck/host_utility/kernel_launch.hpp"
+#include "ck/host_utility/io.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -467,7 +465,6 @@ struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout
             }
         }
 
-#ifndef __HIPCC_RTC__
         void Print() const
         {
             std::cout << "A[AK0, M, AK1]: " << a_grid_desc_ak0_m_ak1_ << std::endl;
@@ -475,7 +472,6 @@ struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout
             std::cout << "B1[BK0, N, BK1]: " << b1_grid_desc_bk0_n_bk1_ << std::endl;
             std::cout << "C[M, N]: " << c_grid_desc_m_n_ << std::endl;
         }
-#endif
 
         //  private:
         const ADataType* p_a_grid_;
@@ -501,7 +497,6 @@ struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout
         std::vector<index_t> raw_lengths_m_n_k_o_;
     };
 
-#ifndef __HIPCC_RTC__
     // Invoker
     struct Invoker : public BaseInvoker
     {
@@ -585,13 +580,13 @@ struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout
             return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
         }
     };
-#endif
+
     static constexpr bool IsValidCompilationParameter()
     {
         // TODO: properly implement this check
         return true;
     }
-#ifndef __HIPCC_RTC__
+
     static bool IsSupportedArgument(const Argument& arg)
     {
         if(!ck::is_xdl_supported())
@@ -636,7 +631,6 @@ struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout
     {
         return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
     }
-#endif
 
     static auto MakeArgument(const ADataType* p_a,
                              const BDataType* p_b,
@@ -668,9 +662,7 @@ struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout
                         b1_element_op, c_element_op};
     }
 
-#ifndef __HIPCC_RTC__
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
@@ -720,7 +712,6 @@ struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout
                                           c_element_op);
     }
 
-#ifndef __HIPCC_RTC__
     // polymorphic
     std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
     {
@@ -750,7 +741,6 @@ struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout
 
         return str.str();
     }
-#endif
 };
 
 } // namespace device

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

@@ -642,9 +642,8 @@ struct DeviceBatchedGemmMultiD_Xdl : public DeviceBatchedGemmMultiD<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_batched_gemm_multiple_d_dl.hpp

@@ -715,9 +715,8 @@ struct DeviceBatchedGemmMultipleD_Dl : public DeviceBatchedGemmMultiD<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_batched_gemm_multiple_d_gemm_multiple_d_xdl_cshuffle.hpp

@@ -894,9 +894,8 @@ struct DeviceBatchedGemmMultipleDGemmMultipleD_Xdl_CShuffle
                         b0_element_op, cde0_element_op,
                         b1_element_op, cde1_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>

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

@@ -898,9 +898,8 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<0, ReduceO
                         reduce_out_element_ops,
                         Batch};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>

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

@@ -357,9 +357,8 @@ struct DeviceBatchedGemmXdl : public DeviceBatchedGemm<ALayout,
                         BatchStrideC,
                         Batch};
     }
-#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_cgemm_4gemm_xdl_cshuffle.hpp

@@ -500,9 +500,8 @@ struct DeviceCGemm_4Gemm_Xdl_CShuffle
                         StrideB,
                         StrideC};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a_real,

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

@@ -714,9 +714,8 @@ struct DeviceContractionMultipleD_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_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk.hpp

@@ -694,9 +694,8 @@ struct DeviceConv2dBwdDataXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
                         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_conv2d_fwd_xdl_c_shuffle_bias_activation_add_nhwc_kyxc_nhwk.hpp

@@ -903,9 +903,8 @@ struct
                         wei_element_op,
                         out_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument>
     MakeArgumentPointer(const void* p_in_grid,

include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_nhwc_kyxc_nhwk.hpp

@@ -860,9 +860,8 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X
                         wei_element_op,
                         out_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument>
     MakeArgumentPointer(const void* p_in_grid,

include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp

@@ -830,9 +830,8 @@ struct DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_W
                         wei_element_op,
                         out_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument>
     MakeArgumentPointer(const void* p_in_grid,

include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_nhwc_kyxc_nhwk.hpp

@@ -605,9 +605,8 @@ struct DeviceConv2dFwdXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
                         input_left_pads,
                         input_right_pads};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     std::unique_ptr<BaseArgument>
     MakeArgumentPointer(const void* p_in_grid,

include/ck/tensor_operation/gpu/device/impl/device_conv3d_fwd_naive_ndhwc_kzyxc_ndhwk.hpp

@@ -203,9 +203,8 @@ struct DeviceConv3dFwdNaive_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_W
                         wei_element_op,
                         out_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>

include/ck/tensor_operation/gpu/device/impl/device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp

@@ -577,9 +577,8 @@ struct DeviceConv3dFwdXdl_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_
                         wei_element_op,
                         out_element_op};
     }
-#ifndef __HIPCC_RTC__
+
     static auto MakeInvoker() { return Invoker{}; }
-#endif
 
     // polymorphic
     std::unique_ptr<BaseArgument>