Clean paddle/fluid/platform/dynload [fluid_ops] (#66386)

* Fix

* Fix

* Fix

* Fix

Author: co63oc

paddle/fluid/framework/CMakeLists.txt

@@ -144,12 +144,6 @@ cc_library(
   var_type_traits
   SRCS var_type_traits.cc
   DEPS framework_proto scope phi common)
-if(WITH_GPU)
-  target_link_libraries(var_type_traits dynload_cuda)
-endif()
-if(WITH_XPU)
-  target_link_libraries(var_type_traits dynload_xpti)
-endif()
 
 # every source file that includes "dnnl.h" must depends on onednn
 # or, the first one should depends on onednn
@@ -947,7 +941,6 @@ cc_library(
        framework_proto
        op_registry
        operator
-       dynamic_loader
        string_helper
        phi
        common
@@ -971,7 +964,8 @@ set(FLUID_FRAMEWORK_MODULES
     executor
     data_feed_proto
     layer
-    dynamic_loader
+    phi
+    common
     custom_operator)
 
 cc_library(paddle_framework DEPS ${FLUID_FRAMEWORK_MODULES})

paddle/fluid/inference/tensorrt/CMakeLists.txt

@@ -27,22 +27,22 @@ nv_library(
 nv_test(
   test_tensorrt
   SRCS test_tensorrt.cc
-  DEPS dynload_cuda device_context dynamic_loader)
+  DEPS device_context phi common)
 if(WIN32)
   nv_test(
     test_tensorrt_engine
     SRCS test_engine.cc test_dynamic_engine.cc
-    DEPS dynload_cuda tensorrt_engine tensorrt_plugin)
+    DEPS phi common tensorrt_engine tensorrt_plugin)
 elseif(WITH_CINN)
   nv_test(
     test_tensorrt_engine
     SRCS test_engine.cc test_dynamic_engine.cc
-    DEPS fleet_executor dynload_cuda tensorrt_engine tensorrt_plugin python)
+    DEPS fleet_executor phi common tensorrt_engine tensorrt_plugin python)
 else()
   nv_test(
     test_tensorrt_engine
     SRCS test_engine.cc test_dynamic_engine.cc
-    DEPS fleet_executor dynload_cuda tensorrt_engine tensorrt_plugin python)
+    DEPS fleet_executor phi common tensorrt_engine tensorrt_plugin python)
 endif()
 nv_test(
   test_arg_mapping_context

paddle/fluid/memory/allocation/CMakeLists.txt

@@ -30,7 +30,7 @@ if(WITH_GPU OR WITH_ROCM)
     pinned_allocator.cc
     stream_safe_cuda_allocator.cc
     thread_local_allocator.cc)
-  list(APPEND ALLOCATOR_DEPS cuda_device_guard gpu_info dynload_cuda)
+  list(APPEND ALLOCATOR_DEPS cuda_device_guard gpu_info phi common)
 endif()
 
 if(WITH_ROCM)

paddle/fluid/operators/CMakeLists.txt

@@ -111,7 +111,7 @@ cc_library(ops_extra_info SRCS ops_extra_info.cc DEPS attribute phi common)
 set(COMMON_OP_DEPS ${COMMON_OP_DEPS} phi common)
 set(COMMON_OP_DEPS ${COMMON_OP_DEPS} selected_rows_utils
 lod_tensor lod_rank_table executor static_prim_api)
-set(COMMON_OP_DEPS ${COMMON_OP_DEPS} dynload_warpctc static_prim_api static_utils static_global_utils prim_utils)
+set(COMMON_OP_DEPS ${COMMON_OP_DEPS} static_prim_api static_utils static_global_utils prim_utils)
 set(COMMON_OP_DEPS ${COMMON_OP_DEPS} box_wrapper ps_gpu_wrapper)
 set(COMMON_OP_DEPS ${COMMON_OP_DEPS} processgroup_comm_utils)
 if(WITH_NCCL OR WITH_RCCL)

paddle/fluid/platform/CMakeLists.txt

@@ -44,7 +44,6 @@ else()
 endif()
 
 add_subdirectory(device)
-add_subdirectory(dynload)
 
 cc_library(
   cpu_helper
@@ -61,11 +60,11 @@ if(WITH_DGC)
 endif()
 
 if(WITH_GPU OR WITH_ROCM)
-  set(GPU_CTX_DEPS dynload_cuda dynamic_loader)
+  set(GPU_CTX_DEPS phi common)
 endif()
 
 if(WITH_XPU)
-  set(XPU_CTX_DEPS dynload_xpti dynamic_loader)
+  set(XPU_CTX_DEPS phi common)
 endif()
 
 if(WITH_IPU)
@@ -258,13 +257,7 @@ if(WITH_GPU)
   nv_library(
     profiler
     SRCS profiler.cc profiler.cu
-    DEPS phi
-         common
-         gpu_info
-         enforce
-         dynload_cuda
-         op_proto_maker
-         shape_inference)
+    DEPS phi common gpu_info enforce op_proto_maker shape_inference)
 elseif(WITH_ROCM)
   hip_library(
     profiler
@@ -274,7 +267,7 @@ elseif(WITH_XPU)
   cc_library(
     profiler
     SRCS profiler.cc
-    DEPS phi common enforce dynload_xpti op_proto_maker shape_inference)
+    DEPS phi common enforce op_proto_maker shape_inference)
 else()
   cc_library(
     profiler

paddle/fluid/platform/device/gpu/CMakeLists.txt

@@ -3,25 +3,25 @@ if(WITH_GPU)
   nv_library(
     gpu_info
     SRCS gpu_info.cc
-    DEPS phi common glog enforce monitor dynload_cuda)
+    DEPS phi common glog enforce monitor)
 
   nv_test(cuda_helper_test SRCS cuda_helper_test.cu)
   nv_test(
     cudnn_desc_test
     SRCS cudnn_desc_test.cc
-    DEPS dynload_cuda)
+    DEPS phi common)
 elseif(WITH_ROCM)
   add_subdirectory(rocm)
   hip_library(
     gpu_info
     SRCS gpu_info.cc
-    DEPS phi common glog enforce monitor dynload_cuda)
+    DEPS phi common glog enforce monitor)
 
   hip_test(cuda_helper_test SRCS cuda_helper_test.cu)
   hip_test(
     cudnn_desc_test
     SRCS cudnn_desc_test.cc
-    DEPS dynload_cuda)
+    DEPS phi common)
 endif()
 
 cc_library(

paddle/fluid/platform/device/gpu/cuda/CMakeLists.txt

@@ -6,4 +6,4 @@ nv_library(
 nv_test(
   cudnn_helper_test
   SRCS cudnn_helper_test.cc
-  DEPS dynload_cuda)
+  DEPS phi common)

paddle/fluid/platform/device/gpu/rocm/CMakeLists.txt

@@ -1,4 +1,4 @@
 hip_test(
   miopen_helper_test
   SRCS miopen_helper_test.cc
-  DEPS dynload_cuda)
+  DEPS phi common)

paddle/fluid/platform/device/xpu/CMakeLists.txt

@@ -19,8 +19,7 @@ cc_library(
        xpulib
        device_context
        phi
-       common
-       dynload_xpti)
+       common)
 cc_library(
   xpu_op_list
   SRCS xpu_op_list.cc
@@ -30,8 +29,7 @@ cc_library(
        device_context
        op_kernel_type
        phi
-       common
-       dynload_xpti)
+       common)
 cc_library(
   xpu_resource_pool
   SRCS xpu_resource_pool.cc