Merge pull request apache#8 from precedenceguo/simple

Merge dmlc/mxnet

.gitignore

@@ -108,3 +108,10 @@ scala-package/*/*/target/
 *.project
 *.settings
 !scala-package/*/bin
+*.bak
+*/node_modules/
+
+# Eclipse project config
+.project
+.cproject
+.pydevproject

CMakeLists.txt

@@ -10,14 +10,17 @@ mxnet_option(USE_OPENMP  "Build with Openmp support" ON)
 mxnet_option(USE_CUDNN   "Build with cudnn support"  ON) # one could set CUDNN_ROOT for search path
 mxnet_option(USE_CUDA    "Build with CUDA support"   ON)
 mxnet_option(USE_DIST_KVSTORE    "Build with DIST_KVSTORE support"   OFF)
+mxnet_option(USE_PLUGINS_WARPCTC	"Use WARPCTC Plugins" OFF)
 
 SET(EXTRA_OPERATORS "" CACHE PATH "EXTRA OPERATORS PATH")
 
-
 include(mshadow/cmake/mshadow.cmake)
 include(mshadow/cmake/Utils.cmake)
 include(mshadow/cmake/Cuda.cmake)
 
+set(mxnet_LINKER_LIBS "")
+list(APPEND mxnet_LINKER_LIBS ${mshadow_LINKER_LIBS})
+
 include_directories("include")
 include_directories("mshadow")
 include_directories("dmlc-core/include")
@@ -29,7 +32,7 @@ if(MSVC)
   add_definitions(-D_CRT_SECURE_NO_WARNINGS)
   add_definitions(-DMXNET_EXPORTS)
   set(CMAKE_C_FLAGS "/MP")
-  set(CMAKE_CXX_FLAGS "${CMAKE_C_FLAGS}")
+  set(CMAKE_CXX_FLAGS "${CMAKE_C_FLAGS} /bigobj")
 else(MSVC)
   include(CheckCXXCompilerFlag)
   check_cxx_compiler_flag("-std=c++11"   SUPPORT_CXX11)
@@ -44,7 +47,7 @@ if(USE_OPENCV)
     find_package(OpenCV REQUIRED COMPONENTS core highgui imgproc)
   endif()
   include_directories(SYSTEM ${OpenCV_INCLUDE_DIRS})
-  list(APPEND mshadow_LINKER_LIBS ${OpenCV_LIBS})
+  list(APPEND mxnet_LINKER_LIBS ${OpenCV_LIBS})
   message(STATUS "OpenCV found (${OpenCV_CONFIG_PATH})")
   add_definitions(-DMXNET_USE_OPENCV=1)
 else(USE_OPENCV)
@@ -68,7 +71,7 @@ if(USE_CUDNN)
   if(HAVE_CUDNN)
     add_definitions(-DUSE_CUDNN)
     include_directories(SYSTEM ${CUDNN_INCLUDE})
-    list(APPEND mshadow_LINKER_LIBS ${CUDNN_LIBRARY})
+    list(APPEND mxnet_LINKER_LIBS ${CUDNN_LIBRARY})
 	add_definitions(-DMSHADOW_USE_CUDNN=1)
   endif()
 endif()
@@ -115,12 +118,22 @@ mxnet_source_group("Source\\resource.cc"   GLOB "src/resource.cc/*.cc")
 mxnet_source_group("Source\\storage"   GLOB "src/storage/*.cc")
 mxnet_source_group("Source\\symbol"   GLOB "src/symbol/*.cc")
 
-
-
-
-
 FILE(GLOB_RECURSE SOURCE "src/*.cc" "src/*.h")
-FILE(GLOB_RECURSE cuda "src/*.cu")
+FILE(GLOB_RECURSE CUDA "src/*.cu")
+
+if(USE_PLUGINS_WARPCTC)
+	set(WARPCTC_INCLUDE  "" CACHE PATH "WARPCTC include")
+	set(WARPCTC_LIB  "" CACHE FILEPATH "WARPCTC lib")
+	include_directories(SYSTEM ${WARPCTC_INCLUDE})
+	list(APPEND mxnet_LINKER_LIBS ${WARPCTC_LIB})
+	mxnet_source_group("Include\\plugin\\warpctc"   GLOB "plugin/warpctc/*.h")
+	mxnet_source_group("Source\\plugin\\warpctc"   GLOB "plugin/warpctc/*.cc")
+	mxnet_source_group("Cuda\\plugin\\warpctc"   GLOB "plugin/warpctc/*.cu")
+	FILE(GLOB_RECURSE PLUGINS_SOURCE "plugin/warpctc/*.cc" "plugin/warpctc/*.h")
+	FILE(GLOB_RECURSE PLUGINS_CUSRC "plugin/warpctc/*.cu")
+	list(APPEND SOURCE ${PLUGINS_SOURCE})
+	list(APPEND CUDA ${PLUGINS_CUSRC})
+endif()
 
 if (NOT (EXTRA_OPERATORS STREQUAL ""))
 	mxnet_source_group("Extra"   GLOB_RECURSE "${EXTRA_OPERATORS}/*.cc")
@@ -142,19 +155,18 @@ endif()
 
 if(USE_CUDA)
   # define preprocessor macro so that we will not include the generated forcelink header
-  mshadow_cuda_compile(cuda_objs ${cuda})
+  mshadow_cuda_compile(cuda_objs ${CUDA})
   if(MSVC)
     FIND_LIBRARY(CUDA_nvrtc_LIBRARY nvrtc "${CUDA_TOOLKIT_ROOT_DIR}/lib/x64"  "${CUDA_TOOLKIT_ROOT_DIR}/lib/win32")
-    list(APPEND mshadow_LINKER_LIBS ${CUDA_nvrtc_LIBRARY})
+    list(APPEND mxnet_LINKER_LIBS ${CUDA_nvrtc_LIBRARY})
     set(CUDA_cuda_LIBRARY "${CUDA_nvrtc_LIBRARY}/../cuda.lib")
-    list(APPEND mshadow_LINKER_LIBS ${CUDA_cuda_LIBRARY})
+    list(APPEND mxnet_LINKER_LIBS ${CUDA_cuda_LIBRARY})
   else(MSVC)
-    list(APPEND mshadow_LINKER_LIBS nvrtc cuda)
+    list(APPEND mxnet_LINKER_LIBS nvrtc cuda)
   endif()
-  list(APPEND SOURCE ${cuda_objs} ${cuda})
+  list(APPEND SOURCE ${cuda_objs} ${CUDA})
 endif()
 
-
 if(NOT MSVC)
   # Only add c++11 flags and definitions after cuda compiling
   add_definitions(-DDMLC_USE_CXX11)
@@ -170,10 +182,9 @@ if(${CMAKE_SYSTEM_NAME} STREQUAL "Darwin")
 else()
   add_library(mxnet SHARED ${SOURCE})
 endif()
-target_link_libraries(mxnet ${mshadow_LINKER_LIBS})
+target_link_libraries(mxnet ${mxnet_LINKER_LIBS})
 target_link_libraries(mxnet dmlccore)
 
-
 if(MSVC)
   set_target_properties(mxnet PROPERTIES OUTPUT_NAME "libmxnet")
 endif()

CONTRIBUTORS.md

@@ -107,3 +107,6 @@ List of Contributors
 * [Yuqi Li](https://github.com/ziyeqinghan)
 * [Depeng Liang](https://github.com/Ldpe2G)
 * [Kiko Qiu](https://github.com/kikoqiu)
+* [Yang Bo](https://github.com/Atry)
+* [Jonas Amaro](https://github.com/jonasrla)
+* [Yan Li](https://github.com/Godricly)

Makefile

@@ -22,7 +22,6 @@ endif
 include $(config)
 include mshadow/make/mshadow.mk
 include $(DMLC_CORE)/make/dmlc.mk
-unexport NO_OPENMP
 
 # all tge possible warning tread
 WARNFLAGS= -Wall
@@ -37,9 +36,9 @@ endif
 CFLAGS += -I$(ROOTDIR)/mshadow/ -I$(ROOTDIR)/dmlc-core/include -fPIC -Iinclude $(MSHADOW_CFLAGS)
 LDFLAGS = -pthread $(MSHADOW_LDFLAGS) $(DMLC_LDFLAGS)
 ifeq ($(DEBUG), 1)
-	NVCCFLAGS = -D_FORCE_INLINES -g -G -O0 -ccbin $(CXX) $(MSHADOW_NVCCFLAGS)
+	NVCCFLAGS = -std=c++11 -Xcompiler -D_FORCE_INLINES -g -G -O0 -ccbin $(CXX) $(MSHADOW_NVCCFLAGS)
 else
-	NVCCFLAGS = -D_FORCE_INLINES -g -O3 -ccbin $(CXX) $(MSHADOW_NVCCFLAGS)
+	NVCCFLAGS = -std=c++11 -Xcompiler -D_FORCE_INLINES -g -O3 -ccbin $(CXX) $(MSHADOW_NVCCFLAGS)
 endif
 
 ifndef LINT_LANG
@@ -150,8 +149,8 @@ endif
 
 build/src/%.o: src/%.cc
 	@mkdir -p $(@D)
-	$(CXX) -std=c++0x $(CFLAGS) -MM -MT build/src/$*.o $< >build/src/$*.d
-	$(CXX) -std=c++0x -c $(CFLAGS) -c $< -o $@
+	$(CXX) -std=c++11 $(CFLAGS) -MM -MT build/src/$*.o $< >build/src/$*.d
+	$(CXX) -std=c++11 -c $(CFLAGS) -c $< -o $@
 
 build/src/%_gpu.o: src/%.cu
 	@mkdir -p $(@D)
@@ -160,20 +159,20 @@ build/src/%_gpu.o: src/%.cu
 
 build/plugin/%.o: plugin/%.cc
 	@mkdir -p $(@D)
-	$(CXX) -std=c++0x $(CFLAGS) -MM -MT build/plugin/$*.o $< >build/plugin/$*.d
-	$(CXX) -std=c++0x -c $(CFLAGS) -c $< -o $@
+	$(CXX) -std=c++11 $(CFLAGS) -MM -MT build/plugin/$*.o $< >build/plugin/$*.d
+	$(CXX) -std=c++11 -c $(CFLAGS) -c $< -o $@
 
 # A nvcc bug cause it to generate "generic/xxx.h" dependencies from torch headers.
 # Use CXX to generate dependency instead.
 build/plugin/%_gpu.o: plugin/%.cu
 	@mkdir -p $(@D)
-	$(CXX) -std=c++0x $(CFLAGS) -MM -MT build/plugin/$*_gpu.o $< >build/plugin/$*_gpu.d
+	$(CXX) -std=c++11 $(CFLAGS) -MM -MT build/plugin/$*_gpu.o $< >build/plugin/$*_gpu.d
 	$(NVCC) -c -o $@ $(NVCCFLAGS) -Xcompiler "$(CFLAGS)" $<
 
 $(EXTRA_OPERATORS)/build/%.o: $(EXTRA_OPERATORS)/%.cc
 	@mkdir -p $(@D)
-	$(CXX) -std=c++0x $(CFLAGS) -Isrc/operator -MM -MT $(EXTRA_OPERATORS)/build/$*.o $< >$(EXTRA_OPERATORS)/build/$*.d
-	$(CXX) -std=c++0x -c $(CFLAGS) -Isrc/operator -c $< -o $@
+	$(CXX) -std=c++11 $(CFLAGS) -Isrc/operator -MM -MT $(EXTRA_OPERATORS)/build/$*.o $< >$(EXTRA_OPERATORS)/build/$*.d
+	$(CXX) -std=c++11 -c $(CFLAGS) -Isrc/operator -c $< -o $@
 
 $(EXTRA_OPERATORS)/build/%_gpu.o: $(EXTRA_OPERATORS)/%.cu
 	@mkdir -p $(@D)
@@ -201,7 +200,7 @@ bin/im2rec: tools/im2rec.cc $(ALL_DEP)
 
 $(BIN) :
 	@mkdir -p $(@D)
-	$(CXX) $(CFLAGS) -std=c++0x  -o $@ $(filter %.cpp %.o %.c %.a %.cc, $^) $(LDFLAGS)
+	$(CXX) $(CFLAGS) -std=c++11  -o $@ $(filter %.cpp %.o %.c %.a %.cc, $^) $(LDFLAGS)
 
 include tests/cpp/unittest.mk
 
@@ -237,26 +236,26 @@ rpkg:	roxygen
 scalapkg:
 	(cd $(ROOTDIR)/scala-package; \
 		mvn clean package -P$(SCALA_PKG_PROFILE) -Dcxx="$(CXX)" \
-											-Dcflags="$(CFLAGS)" -Dldflags="$(LDFLAGS)" \
-											-Dlddeps="$(LIB_DEP)")
+			-Dcflags="$(CFLAGS)" -Dldflags="$(LDFLAGS)" \
+			-Dlddeps="$(LIB_DEP)")
 
 scalatest:
 	(cd $(ROOTDIR)/scala-package; \
 		mvn verify -P$(SCALA_PKG_PROFILE) -Dcxx="$(CXX)" \
-							 -Dcflags="$(CFLAGS)" -Dldflags="$(LDFLAGS)" \
-							 -Dlddeps="$(LIB_DEP)" $(SCALA_TEST_ARGS))
+			-Dcflags="$(CFLAGS)" -Dldflags="$(LDFLAGS)" \
+			-Dlddeps="$(LIB_DEP)" $(SCALA_TEST_ARGS))
 
 scalainstall:
 	(cd $(ROOTDIR)/scala-package; \
 		mvn install -P$(SCALA_PKG_PROFILE) -DskipTests -Dcxx="$(CXX)" \
-							  -Dcflags="$(CFLAGS)" -Dldflags="$(LDFLAGS)" \
-								-Dlddeps="$(LIB_DEP)")
+			-Dcflags="$(CFLAGS)" -Dldflags="$(LDFLAGS)" \
+			-Dlddeps="$(LIB_DEP)")
 
 scaladeploy:
 	(cd $(ROOTDIR)/scala-package; \
 		mvn deploy -Prelease,$(SCALA_PKG_PROFILE) -DskipTests -Dcxx="$(CXX)" \
-							 -Dcflags="$(CFLAGS)" -Dldflags="$(LDFLAGS)" \
-							 -Dlddeps="$(LIB_DEP)")
+			-Dcflags="$(CFLAGS)" -Dldflags="$(LDFLAGS)" \
+			-Dlddeps="$(LIB_DEP)")
 
 jnilint:
 	python2 dmlc-core/scripts/lint.py mxnet-jnicpp cpp scala-package/native/src

R-package/NAMESPACE

@@ -26,6 +26,9 @@ export(mx.exec.update.arg.arrays)
 export(mx.exec.update.aux.arrays)
 export(mx.exec.update.grad.arrays)
 export(mx.gpu)
+export(mx.gru)
+export(mx.gru.forward)
+export(mx.gru.inference)
 export(mx.init.Xavier)
 export(mx.init.create)
 export(mx.init.normal)
@@ -80,6 +83,9 @@ export(mx.nd.transpose)
 export(mx.nd.zeros)
 export(mx.opt.create)
 export(mx.opt.get.updater)
+export(mx.rnn)
+export(mx.rnn.forward)
+export(mx.rnn.inference)
 export(mx.rnorm)
 export(mx.runif)
 export(mx.set.seed)