Caffe OpenCL support

CMakeLists.txt

@@ -15,8 +15,10 @@ include(cmake/Summary.cmake)
 include(cmake/ConfigGen.cmake)
 
 # ---[ Options
-caffe_option(CPU_ONLY  "Build Caffe wihtout CUDA support" OFF) # TODO: rename to USE_CUDA
-caffe_option(USE_CUDNN "Build Caffe with cuDNN libary support" ON IF NOT CPU_ONLY)
+caffe_option(CPU_ONLY  "Build Caffe wihtout CUDA and OpenCL support" OFF) # TODO: rename to USE_CUDA
+caffe_option(USE_CUDA "Build Caffe with CUDA support" ON)
+caffe_option(USE_GREENTEA "Build Caffe with OpenCL support" ON)
+caffe_option(USE_CUDNN "Build Caffe with cuDNN libary support" ON IF USE_CUDA AND NOT CPU_ONLY)
 caffe_option(BUILD_SHARED_LIBS "Build shared libraries" ON)
 caffe_option(BUILD_python "Build Python wrapper" ON)
 set(python_version "2" CACHE STRING "Specify which python version to use")
@@ -29,11 +31,11 @@ include(cmake/Dependencies.cmake)
 
 # ---[ Flags
 if(UNIX OR APPLE)
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fPIC -Wall")
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fPIC -Wall -std=c++11")
 endif()
 
 if(USE_libstdcpp)
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -stdlib=libstdc++")
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -stdlib=libstdc++ -std=c++11")
   message("-- Warning: forcing libstdc++ (controlled by USE_libstdcpp option in cmake)")
 endif()
 

Greentea_Building_Blocks.txt

@@ -0,0 +1,25 @@
+GREENTEA BUILDING BLOCKS:
+
+viennacl::ocl::context &ctx = viennacl::ocl::get_context(
+        this->device_context_.id());
+viennacl::ocl::program &program = Caffe::Get().GetDeviceProgram(
+        this->device_context_.id());
+
+viennacl::ocl::kernel &oclk_kernel = program.get_kernel(
+    CL_KERNEL_SELECT("kernel"));
+viennacl::ocl::enqueue(
+    oclk_kernel(WrapHandle((cl_mem) data, ctx)),
+    ctx.get_queue());
+
+if (this->device_context_.backend() == BACKEND_CUDA) {
+#ifdef USE_CUDA
+#endif // USE_CUDA
+    } else {
+#ifdef USE_GREENTEA
+#endif // USE_GREENTEA
+}
+
+#ifdef USE_GREENTEA
+#include "caffe/greentea/greentea.hpp"
+#include "caffe/greentea/greentea_math_functions.hpp"
+#endif

Makefile

@@ -7,6 +7,15 @@ $(error $(CONFIG_FILE) not found. See $(CONFIG_FILE).example.)
 endif
 include $(CONFIG_FILE)
 
+ifeq ($(CPU_ONLY),1)
+	USE_CUDA := 0
+	USE_GREENTEA := 0
+endif
+
+CXXFLAGS += -std=c++11 -Wno-deprecated-declarations
+LINKFLAGS += -std=c++11 -Wno-deprecated-declarations
+NVCCFLAGS += -Xcompiler "-Wno-deprecated-declarations" -Xlinker "-Wno-deprecated-declarations" -Xarchive "-Wno-deprecated-declarations" -Xnvlink "-Wno-deprecated-declarations"
+
 BUILD_DIR_LINK := $(BUILD_DIR)
 ifeq ($(RELEASE_BUILD_DIR),)
 	RELEASE_BUILD_DIR := .$(BUILD_DIR)_release
@@ -24,6 +33,7 @@ else
 	OTHER_BUILD_DIR := $(DEBUG_BUILD_DIR)
 endif
 
+
 # All of the directories containing code.
 SRC_DIRS := $(shell find * -type d -exec bash -c "find {} -maxdepth 1 \
 	\( -name '*.cpp' -o -name '*.proto' \) | grep -q ." \; -print)
@@ -151,6 +161,78 @@ ALL_WARNS := $(ALL_CXX_WARNS) $(ALL_CU_WARNS)
 EMPTY_WARN_REPORT := $(BUILD_DIR)/.$(WARNS_EXT)
 NONEMPTY_WARN_REPORT := $(BUILD_DIR)/$(WARNS_EXT)
 
+##############################
+# GreenTea backend related include and lib
+##############################
+
+ifeq ($(USE_GREENTEA),1)
+	# Find a valid OpenCL library
+	# TODO: Validate and complete this based on different SDKs
+	ifdef OPENCL_INC
+		CLLINC = '$(OPENCL_INC)'
+	endif
+
+	ifdef OPENCL_LIB
+		CLLIBS = '$(OPENCL_LIB)'
+	endif
+
+	ifdef OPENCLROOT
+		CLLIBS = '$(OPENCLROOT)'
+	endif
+
+	ifdef CUDA_PATH
+		CLLIBS = '$(CUDA_PATH)/lib/x64'
+	endif
+
+	ifdef INTELOCLSDKROOT
+		CLLIBS = '$(INTELOCLSDKROOT)/lib/x64'
+	endif
+
+	ifdef AMDAPPSDKROOT
+		CLLIBS = '$(AMDAPPSDKROOT)/lib/x86_64'
+		CLLINC = '$(AMDAPPSDKROOT)/include'
+	endif
+
+	# Use AMD clBLAS
+	ifeq ($(USE_CLBLAS), 1)
+		LIBRARIES += clBLAS
+		COMMON_FLAGS += -DUSE_CLBLAS
+	endif
+
+	# Use ViennaCL BLAS
+	ifeq ($(USE_VIENNACLBLAS), 1)
+		LIBRARIES += viennacl
+		COMMON_FLAGS += -DUSE_VIENNACLBLAS
+	endif
+
+	# Requires valid OpenCL library
+	LIBRARY_DIRS += $(CLLIBS)
+	# Requires valid OpenCL headers and valid ViennaCL
+	INCLUDE_DIRS += $(CLLINC) $(VIENNACL_DIR)
+	# Requires OpenCL compile library flag and librt
+	LIBRARIES += OpenCL rt
+	# Additional flags
+	COMMON_FLAGS += -DUSE_GREENTEA -DVIENNACL_WITH_OPENCL
+
+	# Viennacl runtime debug output
+	ifeq ($(DEBUG), 1)
+		COMMON_FLAGS += -DVIENNACL_DEBUG_ALL
+	endif
+
+	ifeq ($(GREENTEA_DOUBLE_SUPPORT), 1)
+		COMMON_FLAGS += -DGREENTEA_DOUBLE_SUPPORT
+	endif
+
+	CL_KERNELS_CPP = src/caffe/greentea/cl_kernels.cpp
+	CL_KERNELS = src/caffe/greentea/cl_kernels/*.cl
+	CL_HEADERS = src/caffe/greentea/cl_headers/*.cl
+	CL_KERNELS_SH = src/caffe/greentea/cl_kernels.sh
+endif
+
+ifeq ($(USE_CUDA), 1)
+	COMMON_FLAGS += -DUSE_CUDA
+endif
+
 ##############################
 # Derive include and lib directories
 ##############################
@@ -167,7 +249,7 @@ INCLUDE_DIRS += $(BUILD_INCLUDE_DIR) ./src ./include
 ifneq ($(CPU_ONLY), 1)
 	INCLUDE_DIRS += $(CUDA_INCLUDE_DIR)
 	LIBRARY_DIRS += $(CUDA_LIB_DIR)
-	LIBRARIES := cudart cublas curand
+	LIBRARIES += cudart cublas curand
 endif
 LIBRARIES += glog gflags protobuf leveldb snappy \
 	lmdb boost_system hdf5_hl hdf5 m \
@@ -179,7 +261,6 @@ WARNINGS := -Wall -Wno-sign-compare
 # Set build directories
 ##############################
 
-DISTRIBUTE_DIR ?= distribute
 DISTRIBUTE_SUBDIRS := $(DISTRIBUTE_DIR)/bin $(DISTRIBUTE_DIR)/lib
 DIST_ALIASES := dist
 ifneq ($(strip $(DISTRIBUTE_DIR)),distribute)
@@ -241,15 +322,13 @@ endif
 # libstdc++ for NVCC compatibility on OS X >= 10.9 with CUDA < 7.0
 ifeq ($(OSX), 1)
 	CXX := /usr/bin/clang++
-	ifneq ($(CPU_ONLY), 1)
-		CUDA_VERSION := $(shell $(CUDA_DIR)/bin/nvcc -V | grep -o 'release \d' | grep -o '\d')
-		ifeq ($(shell echo $(CUDA_VERSION) \< 7.0 | bc), 1)
-			CXXFLAGS += -stdlib=libstdc++
-			LINKFLAGS += -stdlib=libstdc++
-		endif
-		# clang throws this warning for cuda headers
-		WARNINGS += -Wno-unneeded-internal-declaration
+	CUDA_VERSION := $(shell $(CUDA_DIR)/bin/nvcc -V | grep -o 'release \d' | grep -o '\d')
+	ifeq ($(shell echo $(CUDA_VERSION) \< 7.0 | bc), 1)
+		CXXFLAGS += -stdlib=libstdc++
+		LINKFLAGS += -stdlib=libstdc++
 	endif
+	# clang throws this warning for cuda headers
+	WARNINGS += -Wno-unneeded-internal-declaration
 	# gtest needs to use its own tuple to not conflict with clang
 	COMMON_FLAGS += -DGTEST_USE_OWN_TR1_TUPLE=1
 	# boost::thread is called boost_thread-mt to mark multithreading on OS X
@@ -390,7 +469,7 @@ endif
 	py mat py$(PROJECT) mat$(PROJECT) proto runtest \
 	superclean supercleanlist supercleanfiles warn everything
 
-all: $(STATIC_NAME) $(DYNAMIC_NAME) tools examples
+all: $(CL_KERNELS_CPP) $(STATIC_NAME) $(DYNAMIC_NAME) tools examples
 
 everything: $(EVERYTHING_TARGETS)
 
@@ -526,12 +605,19 @@ $(PROTO_BUILD_DIR)/%.pb.o: $(PROTO_BUILD_DIR)/%.pb.cc $(PROTO_GEN_HEADER) \
 	@ cat $@.$(WARNS_EXT)
 
 $(BUILD_DIR)/cuda/%.o: %.cu | $(ALL_BUILD_DIRS)
+ifeq ($(USE_CUDA), 1)
 	@ echo NVCC $<
 	$(Q)$(CUDA_DIR)/bin/nvcc $(NVCCFLAGS) $(CUDA_ARCH) -M $< -o ${@:.o=.d} \
 		-odir $(@D)
 	$(Q)$(CUDA_DIR)/bin/nvcc $(NVCCFLAGS) $(CUDA_ARCH) -c $< -o $@ 2> $@.$(WARNS_EXT) \
 		|| (cat $@.$(WARNS_EXT); exit 1)
 	@ cat $@.$(WARNS_EXT)
+else
+	@ echo CXX $<
+	$(Q)$(CXX) $(CXXFLAGS) -c -x c++ $< -o $@ 2> $@.$(WARNS_EXT) \
+		|| (cat $@.$(WARNS_EXT); exit 1)
+	@ cat $@.$(WARNS_EXT)
+endif
 
 $(TEST_ALL_BIN): $(TEST_MAIN_SRC) $(TEST_OBJS) $(GTEST_OBJ) \
 		| $(DYNAMIC_NAME) $(TEST_BIN_DIR)
@@ -566,6 +652,10 @@ $(EXAMPLE_BINS): %.bin : %.o | $(DYNAMIC_NAME)
 	$(Q)$(CXX) $< -o $@ $(LINKFLAGS) -l$(PROJECT) $(LDFLAGS) \
 		-Wl,-rpath,$(ORIGIN)/../../lib
 
+# Copy the OpenCL kernels into C++ char strings
+$(CL_KERNELS_CPP) : $(CL_HEADERS) $(CL_KERNELS)
+	$(CL_KERNELS_SH)
+
 proto: $(PROTO_GEN_CC) $(PROTO_GEN_HEADER)
 
 $(PROTO_BUILD_DIR)/%.pb.cc $(PROTO_BUILD_DIR)/%.pb.h : \

Makefile.config.example

@@ -1,6 +1,28 @@
 ## Refer to http://caffe.berkeleyvision.org/installation.html
 # Contributions simplifying and improving our build system are welcome!
 
+# GreenTea (ViennaCL/OpenCL) backend switch
+
+# Enable the CUDA backend
+USE_CUDA := 1
+
+# Enable the OpenCL/Greentea backend
+USE_GREENTEA := 1
+
+# Folder of the ViennaCL header-only library
+VIENNACL_DIR = ../ViennaCL
+
+# Either set CLBLAS or VIENNACLBLAS to 1, not both.
+# If you want to use OpenCL/Greentea on the CPU only, you can also disable both.
+# When both are disabled, GPUs won't work. CPUs always use CBLAS (Atlas, MKL or OpenBLAS).
+# The chosen BLAS library needs to be compiled and installed from source.
+# CLBLAS should be faster, especially on AMD cards.
+USE_CLBLAS := 1
+USE_VIENNACLBLAS := 0
+
+# Enable or disable double precision support for OpenCL/Greentea
+GREENTEA_DOUBLE_SUPPORT := 1
+
 # cuDNN acceleration switch (uncomment to build with cuDNN).
 # USE_CUDNN := 1
 
@@ -30,7 +52,7 @@ CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \
 # atlas for ATLAS (default)
 # mkl for MKL
 # open for OpenBlas
-BLAS := atlas
+BLAS := open
 # Custom (MKL/ATLAS/OpenBLAS) include and lib directories.
 # Leave commented to accept the defaults for your choice of BLAS
 # (which should work)!

cmake/Dependencies.cmake

@@ -46,7 +46,7 @@ list(APPEND Caffe_LINKER_LIBS ${Snappy_LIBRARIES})
 # ---[ CUDA
 include(cmake/Cuda.cmake)
 if(NOT HAVE_CUDA)
-  if(CPU_ONLY)
+  if(CPU_ONLY OR NOT USE_CUDA)
     message("-- CUDA is disabled. Building without it...")
   else()
     message("-- CUDA is not detected by cmake. Building without it...")