Problems with linking CUDA for running fastHOG algorithm - c

Question

I am trying to set up the fastHOG codebase in my linux server, but unable to finish it as i am getting heap of link errors. I am giving here a sample portion of the link errors
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0x910): undefined reference to 'cudaGetErrorString'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0x944): undefined reference to 'cudaGetErrorString'
/home/n8385106/bin/contrib/qut/qutanomaly/matlab/person- detector/fasthog/obj/release/HOG/HOGHistogram.cu.o: In function 'InitHistograms(int, int, int, int, int, float)':
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xac0): undefined reference to 'cudaCreateChannelDesc'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xafe): undefined reference to 'cudaMallocArray'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xb46): undefined reference to 'cudaMemcpyToArray'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xc3d): undefined reference to 'cudaMemcpyToSymbol'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xc63): undefined reference to 'cudaMemcpyToSymbol'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xc89): undefined reference to 'cudaMemcpyToSymbol'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xcae): undefined reference to 'cudaMemcpyToSymbol'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xccd): undefined reference to 'cudaGetErrorString'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xd01): undefined reference to 'cudaGetErrorString'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xd15): undefined reference to 'cudaGetErrorString'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xd29): undefined reference to 'cudaGetErrorString'
tmpxft_00006d91_00000000-1_HOGHistogram.cudafe1.cpp:(.text+0xd3d): undefined reference to 'cudaGetErrorString'
It seems the linker didn't realize any of the linker libraries provided in the make file. It did not say anything like "was unable to find this library" but it avoided all the information provided about the libraries to be linked such as
# Libs
LIB := -L$(CUDA_INSTALL_PATH)/cuda/lib64 -L$(LIBDIR) - L$(COMMONDIR)/lib/$(OSLOWER)
# If dynamically linking to CUDA and CUDART, we exclude the libraries from the LIB
ifeq ($(USECUDADYNLIB),1)
LIB += ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB}
else
# static linking, we will statically link against CUDA and CUDART
ifeq ($(USEDRVAPI),1)
LIB += -lcuda ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB}
else
LIB += -lcudart ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB}
endif
endif
ifeq ($(USECUFFT),1)
ifeq ($(emu),1)
LIB += -lcufftemu
else
LIB += -lcufft
endif
endif
ifeq ($(USECUBLAS),1)
ifeq ($(emu),1)
LIB += -lcublasemu
else
LIB += -lcublas
endif
endif
I was able to create all the object files but problem is comming during the linking process, I am a newbie to CUDA, Can anyone please help me to sort this out?
My OS is 64 bit linux
Edited : My MakeFile
# Add source files here
EXECUTABLE := fastHOG
# C/C++ source files (compiled with gcc / c++)
CCFILES := \
fastHOG.cpp \
# HOG UTILS
CCUTILS := \
ImageWindow.cpp \
# CC HOG
CCHOG := \
HOGImage.cpp \
HOGEngine.cpp \
HOGNMS.cpp \
# CUDA HOG
CUFILES := \
HOGEngineDevice.cu \
HOGConvolution.cu \
HOGHistogram.cu \
HOGPadding.cu \
HOGScale.cu \
HOGSVMSlider.cu \
HOGUtils.cu \
################################################################################
# Rules and targets
include common.mk
Contents of the common.mk file
################################################################################
#
# Copyright 1993-2006 NVIDIA Corporation. All rights reserved.
#
################################################################################
#
# Common build script
#
################################################################################
.SUFFIXES : .cu .cu_dbg.o .c_dbg.o .cpp_dbg.o .cu_rel.o .c_rel.o .cpp_rel.o .cubin
#Add new SM Versions here as devices with new Compute Capability are released
SM_VERSIONS := sm_10 sm_11 sm_12 sm_13
CUDA_INSTALL_PATH ?= fasthoglib/cuda/cuda
ifdef cuda-install
CUDA_INSTALL_PATH := $(cuda-install)
endif
# detect OS
OSUPPER = $(shell uname -s 2>/dev/null | tr [:lower:] [:upper:])
OSLOWER = $(shell uname -s 2>/dev/null | tr [:upper:] [:lower:])
# 'linux' is output for Linux system, 'darwin' for OS X
DARWIN = $(strip $(findstring DARWIN, $(OSUPPER)))
# Basic directory setup for SDK
# (override directories only if they are not already defined)
SRCDIR ?=
SRCDIRUTILS ?= Utils
SRCDIRHOG ?= HOG
ROOTDIR ?= /home/n8385106/src/saivt-vxl/contrib/qut/qutanomaly/matlab/pedestrian-detector/fasthoglib/cuda/sdk
ROOTBINDIR ?= /home/n8385106/bin/contrib/qut/qutanomaly/matlab/person-detector/fasthog/bin
BINDIR ?= $(ROOTBINDIR)
ROOTOBJDIR ?= /home/n8385106/bin/contrib/qut/qutanomaly/matlab/person-detector/fasthog/obj
LIBDIR := $(ROOTDIR)/C/lib
COMMONDIR := $(ROOTDIR)/C/common
# Compilers
NVCC := $(CUDA_INSTALL_PATH)/bin/nvcc
CXX := g++
CC := gcc
LINK := g++ -fPIC
# Includes
INCLUDES += -I. -I$(CUDA_INSTALL_PATH)/cuda/include -I$(COMMONDIR)/inc
# architecture flag for cubin build
CUBIN_ARCH_FLAG := -m32
# Warning flags
CXXWARN_FLAGS := \
-W -Wall \
-Wimplicit \
-Wswitch \
-Wformat \
-Wchar-subscripts \
-Wparentheses \
-Wmultichar \
-Wtrigraphs \
-Wpointer-arith \
-Wcast-align \
-Wreturn-type \
-Wno-unused-function \
$(SPACE)
CWARN_FLAGS := $(CXXWARN_FLAGS) \
-Wstrict-prototypes \
-Wmissing-prototypes \
-Wmissing-declarations \
-Wnested-externs \
-Wmain \
# Compiler-specific flags
NVCCFLAGS :=
CXXFLAGS := $(CXXWARN_FLAGS)
CFLAGS := $(CWARN_FLAGS)
# Common flags
COMMONFLAGS += $(INCLUDES) -DUNIX
# Debug/release configuration
ifeq ($(dbg),1)
COMMONFLAGS += -g
NVCCFLAGS += -D_DEBUG
BINSUBDIR := debug
LIBSUFFIX := D
else
COMMONFLAGS += -O3
BINSUBDIR := release
LIBSUFFIX :=
NVCCFLAGS += --compiler-options -fno-strict-aliasing
CXXFLAGS += -fno-strict-aliasing
CFLAGS += -fno-strict-aliasing
endif
# append optional arch/SM version flags (such as -arch sm_11)
#NVCCFLAGS += $(SMVERSIONFLAGS)
# architecture flag for cubin build
CUBIN_ARCH_FLAG := -m32
# detect if 32 bit or 64 bit system
HP_64 = $(shell uname -m | grep 64)
# OpenGL is used or not (if it is used, then it is necessary to include GLEW)
ifeq ($(USEGLLIB),1)
ifneq ($(DARWIN),)
OPENGLLIB := -L/System/Library/Frameworks/OpenGL.framework/Libraries -lGL -lGLU $(COMMONDIR)/lib/$(OSLOWER)/libGLEW.a
else
OPENGLLIB := -lGL -lGLU -lX11 -lXi -lXmu
ifeq "$(strip $(HP_64))" ""
OPENGLLIB += -lGLEW -L/usr/X11R6/lib
else
OPENGLLIB += -lGLEW_x86_64 -L/usr/X11R6/lib64
endif
endif
CUBIN_ARCH_FLAG := -m64
endif
ifeq ($(USEGLUT),1)
ifneq ($(DARWIN),)
OPENGLLIB += -framework GLUT
else
OPENGLLIB += -lglut
endif
endif
ifeq ($(USEPARAMGL),1)
PARAMGLLIB := -lparamgl$(LIBSUFFIX)
endif
ifeq ($(USERENDERCHECKGL),1)
RENDERCHECKGLLIB := -lrendercheckgl$(LIBSUFFIX)
endif
USECUDPP := 1;
ifeq ($(USECUDPP), 1)
ifeq "$(strip $(HP_64))" ""
CUDPPLIB := -lcudpp
else
CUDPPLIB := -lcudpp64
endif
CUDPPLIB := $(CUDPPLIB)$(LIBSUFFIX)
ifeq ($(emu), 1)
CUDPPLIB := $(CUDPPLIB)_emu
endif
endif
# Libs
LIB := -L$(CUDA_INSTALL_PATH)/cuda/lib64 -L$(LIBDIR) -L$(COMMONDIR)/lib/$(OSLOWER)
# If dynamically linking to CUDA and CUDART, we exclude the libraries from the LIB
ifeq ($(USECUDADYNLIB),1)
LIB += ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB}
else
# static linking, we will statically link against CUDA and CUDART
ifeq ($(USEDRVAPI),1)
LIB += -lcuda ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB}
else
LIB += -lcudart ${OPENGLLIB} $(PARAMGLLIB) $(RENDERCHECKGLLIB) $(CUDPPLIB) ${LIB}
endif
endif
ifeq ($(USECUFFT),1)
ifeq ($(emu),1)
LIB += -lcufftemu
else
LIB += -lcufft
endif
endif
ifeq ($(USECUBLAS),1)
ifeq ($(emu),1)
LIB += -lcublasemu
else
LIB += -lcublas
endif
endif
# Lib/exe configuration
ifneq ($(STATIC_LIB),)
TARGETDIR := $(LIBDIR)
TARGET := $(subst .a,$(LIBSUFFIX).a,$(LIBDIR)/$(STATIC_LIB))
LINKLINE = ar rucv $(TARGET) $(OBJS)
else
LIB += -lcutil$(LIBSUFFIX)
# Device emulation configuration
ifeq ($(emu), 1)
NVCCFLAGS += -deviceemu
CUDACCFLAGS +=
BINSUBDIR := emu$(BINSUBDIR)
# consistency, makes developing easier
CXXFLAGS += -D__DEVICE_EMULATION__
CFLAGS += -D__DEVICE_EMULATION__
endif
TARGETDIR := $(BINDIR)/$(BINSUBDIR)
TARGET := $(TARGETDIR)/$(EXECUTABLE)
#fltk
LIB += -lfltk2 -lXft -lfltk2_images -lXext -lXinerama -lXi
#boost thread for interface
LIB += -lboost_thread
#read images in HOGImage from file
LIB := /pkg/suse11/cuda/4.2/sdk/CUDALibraries/common/FreeImage/lib/linux/libfreeimage64.a
LINKLINE = $(LINK) -o $(TARGET) $(OBJS) $(LIB)
endif
# check if verbose
ifeq ($(verbose), 1)
VERBOSE :=
else
VERBOSE := #
endif
################################################################################
# Check for input flags and set compiler flags appropriately
################################################################################
ifeq ($(fastmath), 1)
NVCCFLAGS += -use_fast_math
endif
ifeq ($(keep), 1)
NVCCFLAGS += -keep
NVCC_KEEP_CLEAN := *.i* *.cubin *.cu.c *.cudafe* *.fatbin.c *.ptx
endif
ifdef maxregisters
NVCCFLAGS += -maxrregcount $(maxregisters)
endif
# Add cudacc flags
NVCCFLAGS += $(CUDACCFLAGS)
# workaround for mac os x cuda 1.1 compiler issues
ifneq ($(DARWIN),)
NVCCFLAGS += --host-compilation C
endif
# Add common flags
NVCCFLAGS += $(COMMONFLAGS)
CXXFLAGS += $(COMMONFLAGS)
CFLAGS += $(COMMONFLAGS)
ifeq ($(nvcc_warn_verbose),1)
NVCCFLAGS += $(addprefix --compiler-options ,$(CXXWARN_FLAGS))
NVCCFLAGS += --compiler-options -fno-strict-aliasing
endif
################################################################################
# Set up object files
################################################################################
OBJDIR := $(ROOTOBJDIR)/$(BINSUBDIR)
OBJS += $(patsubst %.cpp,$(OBJDIR)/%.cpp.o,$(notdir $(CCFILES)))
OBJDIRUTILS := $(ROOTOBJDIR)/$(BINSUBDIR)/$(SRCDIRUTILS)
OBJS += $(patsubst %.cpp,$(OBJDIRUTILS)/%.cpp.o,$(notdir $(CCUTILS)))
OBJDIRHOG := $(ROOTOBJDIR)/$(BINSUBDIR)/$(SRCDIRHOG)
OBJS += $(patsubst %.cu,$(OBJDIRHOG)/%.cu.o,$(notdir $(CUFILES)))
OBJS += $(patsubst %.cpp,$(OBJDIRHOG)/%.cpp.o,$(notdir $(CCHOG)))
################################################################################
# Set up cubin files
################################################################################
CUBINDIR := $(SRCDIR)data
CUBINS += $(patsubst %.cu,$(CUBINDIR)/%.cubin,$(notdir $(CUBINFILES)))
################################################################################
# Rules
################################################################################
$(OBJDIR)/%.c.o : $(SRCDIR)%.c $(C_DEPS)
$(VERBOSE)$(CC) $(CFLAGS) -o $# -c $<
$(OBJDIRUTILS)/%.cpp.o : $(SRCDIRUTILS)%.cpp $(C_DEPS)
$(VERBOSE)$(CXX) $(CXXFLAGS) -o $# -c $<
$(OBJDIR)/%.cpp.o : $(SRCDIR)%.cpp $(C_DEPS)
$(VERBOSE)$(CXX) $(CXXFLAGS) -o $# -c $<
$(OBJDIRHOG)/%.cpp.o : $(SRCDIRHOG)%.cpp $(C_DEPS)
$(VERBOSE)$(CXX) $(CXXFLAGS) -o $# -c $<
$(OBJDIR)/%.cu.o : $(SRCDIR)%.cu $(CU_DEPS)
$(VERBOSE)$(NVCC) $(NVCCFLAGS) $(SMVERSIONFLAGS) -o $# -c $<
$(OBJDIRHOG)/%.cu.o : $(SRCDIRHOG)%.cu $(CU_DEPS)
$(VERBOSE)$(NVCC) $(NVCCFLAGS) $(SMVERSIONFLAGS) -o $# -c $<
$(CUBINDIR)/%.cubin : $(SRCDIR)%.cu cubindirectory
$(VERBOSE)$(NVCC) $(CUBIN_ARCH_FLAG) $(NVCCFLAGS) $(SMVERSIONFLAGS) -o $# -cubin $<
#
# The following definition is a template that gets instantiated for each SM
# version (sm_10, sm_13, etc.) stored in SMVERSIONS. It does 2 things:
# 1. It adds to OBJS a .cu_sm_XX.o for each .cu file it finds in CUFILES_sm_XX.
# 2. It generates a rule for building .cu_sm_XX.o files from the corresponding
# .cu file.
#
# The intended use for this is to allow Makefiles that use common.mk to compile
# files to different Compute Capability targets (aka SM arch version). To do
# so, in the Makefile, list files for each SM arch separately, like so:
#
# CUFILES_sm_10 := mycudakernel_sm10.cu app.cu
# CUFILES_sm_12 := anothercudakernel_sm12.cu
#
define SMVERSION_template
OBJS += $(patsubst %.cu,$(OBJDIR)/%.cu_$(1).o,$(notdir $(CUFILES_$(1))))
$(OBJDIR)/%.cu_$(1).o : $(SRCDIR)%.cu $(CU_DEPS)
$(VERBOSE)$(NVCC) -o $$# -c $$< $(NVCCFLAGS) -arch $(1)
endef
# This line invokes the above template for each arch version stored in
# SM_VERSIONS. The call function invokes the template, and the eval
# function interprets it as make commands.
$(foreach smver,$(SM_VERSIONS),$(eval $(call SMVERSION_template,$(smver))))
$(TARGET): makedirectories $(OBJS) $(CUBINS) Makefile
$(VERBOSE)$(LINKLINE)
cubindirectory:
$(VERBOSE)mkdir -p $(CUBINDIR)
makedirectories:
$(VERBOSE)mkdir -p $(LIBDIR)
$(VERBOSE)mkdir -p $(OBJDIR)
$(VERBOSE)mkdir -p $(OBJDIRUTILS)
$(VERBOSE)mkdir -p $(OBJDIRHOG)
$(VERBOSE)mkdir -p $(TARGETDIR)
tidy :-lboost_thread
$(VERBOSE)find . | egrep "#" | xargs rm -f
$(VERBOSE)find . | egrep "\~" | xargs rm -f
clean : tidy
$(VERBOSE)rm -f $(OBJS)
$(VERBOSE)rm -f $(CUBINS)
$(VERBOSE)rm -f $(TARGET)
$(VERBOSE)rm -f $(NVCC_KEEP_CLEAN)
clobber : clean
$(VERBOSE)rm -rf $(ROOTOBJDIR)

Answer 1

The common.mk has been broken (presumably by a modification you made to it) here:
LIB := /pkg/suse11/cuda/4.2/sdk/CUDALibraries/common/FreeImage/lib/linux/libfreeimage64.a
This is undoubtedly not what you want, it is destroying all of the carefully crafted library dependencies assigned above it. I guess you intended something like
LIB += /pkg/suse11/cuda/4.2/sdk/CUDALibraries/common/FreeImage/lib/linux/libfreeimage64.a