TX2+PCL1.8.0+CUDA8的环境编译遇到的问题

先是编译的PCL1.7.2，结果发现没有GPU模块，1.8.0以后的版本应该都是有的。

安装PCL之前需要先安装一些环境依赖：

sudo apt-get update  
sudo apt-get install git build-essential linux-libc-dev  
sudo apt-get install cmake cmake-gui   
sudo apt-get install libusb-1.0-0-dev libusb-dev libudev-dev  
sudo apt-get install mpi-default-dev openmpi-bin openmpi-common    
sudo apt-get install libflann1.8 libflann-dev  
sudo apt-get install libeigen3-dev  
sudo apt-get install libboost-all-dev  
sudo apt-get install libvtk5.10-qt4 libvtk5.10 libvtk5-dev  
sudo apt-get install libqhull* libgtest-dev  
sudo apt-get install freeglut3-dev pkg-config  
sudo apt-get install libxmu-dev libxi-dev   
sudo apt-get install mono-complete  
sudo apt-get install qt-sdk openjdk-8-jdk openjdk-8-jre

sudo apt-get install libopenni-dev   
sudo apt-get install libopenni2-dev 
sudo apt-get install libx11-dev libxext-dev libxtst-dev libxrender-dev libxmu-dev libxmuu-dev
sudo apt-get install build-essential libgl1-mesa-dev libglu1-mesa-dev
sudo apt-get install cmake cmake-gui 

然后就是安装PCL：

git clone --recursive https://github.com/PointCloudLibrary/pcl

git checkout (pcl version)

mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_GPU=ON -DBUILD_apps=ON -DPCL_QT_VERSION=4 -DBUILD_examples=ON -DCMAKE_INSTALL_PREFIX=/usr/local/pcl1.8.0 -DCUDA_ARCH_BIN=6.2 -DCUDA_ARCH_PTX=6.2 -DCUDA_NVCC_FLAGS=-arch=sm_62 .. 
make -j2  
sudo make install  

在编译的过程中遇到的一些问题：

1. 

[ 57%] Building CXX object features/CMakeFiles/pcl_features.dir/src/our_cvfh.cpp.o
[ 57%] Building CXX object features/CMakeFiles/pcl_features.dir/src/crh.cpp.o
[ 57%] Building CXX object features/CMakeFiles/pcl_features.dir/src/don.cpp.o
[ 58%] Building CXX object features/CMakeFiles/pcl_features.dir/src/fpfh.cpp.o
c++: internal compiler error: Segmentation fault (program cc1plus)

解决：将make -j2换成make 或者是 make VERBOSE=1

2. 使用PCL GPU加速分割速度不如CPU的速度，可能的原因在于计算量不大，其GPU的优势没有体现出来，反而花费过多时间在数据向显存中传递的过程（内存->显存）。

 

Inference: https://blog.youkuaiyun.com/zhou_x11/article/details/99560680