Improve brpccmake (#11842)

CMakeLists.txt

@@ -64,6 +64,7 @@ option(WITH_CONTRIB     "Compile the third-party contributation"        OFF)
 option(REPLACE_ENFORCE_GLOG "Replace PADDLE_ENFORCE with glog/CHECK for better debug." OFF)
 option(WITH_ANAKIN      "Compile with Anakin library"                   OFF)
 option(WITH_GRPC     "Use grpc as the default rpc framework"            ${WITH_DISTRIBUTE})
+option(WITH_BRPC_RDMA     "Use brpc rdma as the rpc protocal"           OFF)
 
 # CMAKE_BUILD_TYPE
 if(NOT CMAKE_BUILD_TYPE)
@@ -158,12 +159,24 @@ include(external/cares)
 if(WITH_DISTRIBUTE)
     if(WITH_GRPC)
         include(external/grpc)
+        message(STATUS "Use grpc framework.")
     else()
+        message(STATUS "Use brpc framework.")
         include(external/leveldb)
         include(external/brpc)
     endif()
 endif()
 
+if(WITH_BRPC_RDMA)
+    message(STATUS "Use brpc with rdma.")
+    if(WITH_GRPC)
+        message(FATAL_ERROR "Can't use grpc with brpc rdma.")
+    endif()
+    if(NOT WITH_DISTRIBUTE)
+        message(FATAL_ERROR "Can't use brpc rdma in no distribute env.")
+    endif()
+endif()
+
 include(external/snappy)    # download snappy
 include(external/snappystream)
 include(external/threadpool)

cmake/configure.cmake

@@ -174,3 +174,7 @@ endif(WITH_GOLANG)
 if(WITH_GRPC)
     add_definitions(-DPADDLE_WITH_GRPC)
 endif(WITH_GRPC)
+
+if(WITH_BRPC_RDMA)
+    add_definitions(-DPADDLE_WITH_BRPC_RDMA)
+endif(WITH_BRPC_RDMA)

cmake/external/brpc.cmake

@@ -14,6 +14,15 @@
 
 INCLUDE(ExternalProject)
 
+find_library(SSL_LIBRARY NAMES ssl)
+ADD_LIBRARY(ssl SHARED IMPORTED GLOBAL)
+SET_PROPERTY(TARGET ssl PROPERTY IMPORTED_LOCATION ${SSL_LIBRARY})
+
+find_library(CRYPTO_LIBRARY NAMES crypto)
+ADD_LIBRARY(crypto SHARED IMPORTED GLOBAL)
+SET_PROPERTY(TARGET crypto PROPERTY IMPORTED_LOCATION ${CRYPTO_LIBRARY})
+
+
 SET(BRPC_SOURCES_DIR ${THIRD_PARTY_PATH}/brpc)
 SET(BRPC_INSTALL_DIR ${THIRD_PARTY_PATH}/install/brpc)
 SET(BRPC_INCLUDE_DIR "${BRPC_INSTALL_DIR}/include" CACHE PATH "brpc include directory." FORCE)
@@ -22,14 +31,14 @@ SET(BRPC_LIBRARIES "${BRPC_INSTALL_DIR}/lib/libbrpc.a" CACHE FILEPATH "brpc libr
 INCLUDE_DIRECTORIES(${BRPC_INCLUDE_DIR})
 
 # Reference https://stackoverflow.com/questions/45414507/pass-a-list-of-prefix-paths-to-externalproject-add-in-cmake-args
-set(prefix_path "${THIRD_PARTY_PATH}/install/gflags|${THIRD_PARTY_PATH}/install/leveldb|${THIRD_PARTY_PATH}/install/snappy|${THIRD_PARTY_PATH}/install/gtest|${THIRD_PARTY_PATH}/install/protobuf")
+set(prefix_path "${THIRD_PARTY_PATH}/install/gflags|${THIRD_PARTY_PATH}/install/leveldb|${THIRD_PARTY_PATH}/install/snappy|${THIRD_PARTY_PATH}/install/gtest|${THIRD_PARTY_PATH}/install/protobuf|${THIRD_PARTY_PATH}/install/zlib")
 
 # If minimal .a is need, you can set  WITH_DEBUG_SYMBOLS=OFF
 ExternalProject_Add(
     extern_brpc
     ${EXTERNAL_PROJECT_LOG_ARGS}
-    GIT_REPOSITORY  "https://github.com/brpc/brpc"
-    GIT_TAG         "6d153dd7ff00f960ae6895c9c5fff0ce9f07aff2"
+    GIT_REPOSITORY  "https://github.com/gongweibao/brpc"
+    GIT_TAG         "7dc04defad1fd4173aae170c3fcbde131b65155a"
     PREFIX          ${BRPC_SOURCES_DIR}
     UPDATE_COMMAND  ""
     CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
@@ -42,6 +51,8 @@ ExternalProject_Add(
                     -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
                     -DCMAKE_PREFIX_PATH=${prefix_path}
                     -DBRPC_WITH_GLOG=ON
+                    -DIOBUF_WITH_HUGE_BLOCK=ON
+                    -DBRPC_WITH_RDMA=${WITH_BRPC_RDMA}
                     ${EXTERNAL_OPTIONAL_ARGS}
     LIST_SEPARATOR |
     CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${BRPC_INSTALL_DIR}
@@ -49,7 +60,7 @@ ExternalProject_Add(
                      -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
                      -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
 )
-ADD_DEPENDENCIES(extern_brpc protobuf leveldb gflags glog gtest snappy)
+ADD_DEPENDENCIES(extern_brpc protobuf ssl crypto leveldb gflags glog gtest snappy)
 ADD_LIBRARY(brpc STATIC IMPORTED GLOBAL)
 SET_PROPERTY(TARGET brpc PROPERTY IMPORTED_LOCATION ${BRPC_LIBRARIES})
 ADD_DEPENDENCIES(brpc extern_brpc)

paddle/fluid/framework/executor.cc

@@ -20,9 +20,7 @@ limitations under the License. */
 #include "paddle/fluid/framework/lod_tensor_array.h"
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/framework/reader.h"
-#ifdef PADDLE_WITH_DISTRIBUTE
-#include "paddle/fluid/operators/distributed/grpc_client.h"
-#endif
+#include "paddle/fluid/operators/detail/macros.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/platform/profiler.h"
 
@@ -49,8 +47,7 @@ Executor::Executor(const platform::Place& place) : place_(place) {}
 
 #ifdef PADDLE_WITH_DISTRIBUTE
 void Executor::Complete() {
-  ::paddle::operators::distributed::RPCClient::GetInstance<
-      ::paddle::operators::distributed::GRPCClient>()
+  ::paddle::operators::distributed::RPCClient::GetInstance<RPCCLIENT_T>()
       ->SendComplete();
 }
 #endif

paddle/fluid/operators/CMakeLists.txt

@@ -184,6 +184,7 @@ else()
     set(DEPS_OPS ${DEPS_OPS} nccl_op)
 endif()
 
+set(DISTRIBUTE_DEPS "")
 if(WITH_DISTRIBUTE)
     add_subdirectory(distributed)
     
@@ -192,6 +193,18 @@ if(WITH_DISTRIBUTE)
         set(DISTRIBUTE_DEPS sendrecvop_grpc grpc++_unsecure grpc_unsecure gpr cares zlib protobuf)
     else()
         set(DISTRIBUTE_DEPS sendrecvop_brpc brpc leveldb snappystream snappy protobuf ssl crypto zlib)
+        if(WITH_BRPC_RDMA)
+            find_library(IBVERBS_LIBRARY NAMES ibverbs)
+            ADD_LIBRARY(ibverbs SHARED IMPORTED GLOBAL)
+            SET_PROPERTY(TARGET ibverbs PROPERTY IMPORTED_LOCATION ${IBVERBS_LIBRARY})
+
+
+            find_library(RDMACM_LIBRARY NAMES rdmacm)
+            ADD_LIBRARY(rdmacm SHARED IMPORTED GLOBAL)
+            SET_PROPERTY(TARGET rdmacm PROPERTY IMPORTED_LOCATION ${RDMACM_LIBRARY})
+
+            set(DISTRIBUTE_DEPS ${DISTRIBUTE_DEPS} ibverbs rdmacm)
+        endif()
     endif()
 
     set(DISTRIBUTE_COMPILE_FLAGS "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor")
@@ -205,7 +218,7 @@ if(WITH_DISTRIBUTE)
     #        listen_and_serv_op sum_op executor SERIAL)
     if(WITH_GPU)
         set_source_files_properties(test_send_nccl_id.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
-        cc_test(test_send_nccl_id SRCS test_send_nccl_id.cc DEPS listen_and_serv_op executor SERIAL)
+        cc_test(test_send_nccl_id SRCS test_send_nccl_id.cc DEPS listen_and_serv_op ${DISTRIBUTE_DEPS} executor SERIAL)
         if(WITH_GRPC)
             op_library(gen_nccl_id_op DEPS nccl_common sendrecvop_grpc)
         else()
@@ -297,6 +310,7 @@ foreach(src ${DETECTION_LIBRARY})
 endforeach()
 
 set(GLOB_OP_LIB ${OP_LIBRARY} CACHE INTERNAL "Global OP library")
+set(GLOB_DISTRIBUTE_DEPS ${DISTRIBUTE_DEPS} CACHE INTERNAL "distributed dependency")
 
 cc_test(gather_test SRCS gather_test.cc DEPS tensor)
 cc_test(scatter_test SRCS scatter_test.cc DEPS tensor)

paddle/fluid/operators/detail/macros.h

@@ -14,14 +14,22 @@
 
 #pragma once
 
+#ifdef PADDLE_WITH_DISTRIBUTE
+
 #ifdef PADDLE_WITH_GRPC
+
 #include "paddle/fluid/operators/distributed/grpc_client.h"
 #include "paddle/fluid/operators/distributed/grpc_server.h"
-#define RPCSERVER_T distributed::AsyncGRPCServer
-#define RPCCLIENT_T distributed::GRPCClient
-#else
+#define RPCSERVER_T paddle::operators::distributed::AsyncGRPCServer
+#define RPCCLIENT_T paddle::operators::distributed::GRPCClient
+
+#else  // PADDLE_WITH_GRPC
+
 #include "paddle/fluid/operators/distributed/brpc_client.h"
 #include "paddle/fluid/operators/distributed/brpc_server.h"
-#define RPCSERVER_T distributed::AsyncBRPCServer
-#define RPCCLIENT_T distributed::BRPCClient
-#endif
+#define RPCSERVER_T paddle::operators::distributed::AsyncBRPCServer
+#define RPCCLIENT_T paddle::operators::distributed::BRPCClient
+
+#endif  // PADDLE_WITH_GRPC
+
+#endif  // PADDLE_WITH_DISTRIBUTE