Jetson TK1 cuda opencv caffe 安装

该平台目前驱动支持cuda 6.5版本
(1)sudodpkg -i cuda-repo-l4t-r21.2-6-5-prod_6.5-34_armhf.deb
(2)sudo apt-get update
(3)sudo apt-get installcuda-toolkit-6-5


(4)
echo "# Add CUDA bin & library paths:" >> ~/.bashrc 
echo "export PATH=/usr/local/cuda-6.5/bin:$PATH" >> ~/.bashrc 
echo "export LD_LIBRARY_PATH=/usr/local/cuda-6.5/lib:$LD_LIBRARY_PATH" >> ~/.bashrc 
source ~/.bashrc
(5)
nvcc -V
(6)测试
cd /usr/local/cuda-6.5/samples/5_Simulations/
cd nbody
make
./nbody -benchmark -numbodies=65536


cd smokeParticles
make
./smokeParticles

Jetson TK1 opencv安装，mainfold

https://developer.nvidia.com/linux-tegra-rel-21

直接参考上面的OpenCV4Tegra README的安装即可

参考测试方法

http://www.kancloud.cn/digest/jetsontk1-fred/149671  opencv

http://www.kancloud.cn/digest/jetsontk1-fred/149673  opnecv+gpu

假如open自己编译的话，也可以参考下面的，但是可能会GPU调用有问题

问题：

cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local .. 

nvcc fatal : Unsupported gpu architecture'compute_11'

解决方法：

cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D CUDA_GENERATION=Kepler ..

opencv-2.4.9/modules/gpu/src/nvidia/core/NCVPixelOperations.hpp(51): error: a storage class is not allowed in an explicit specialization

解决方法在此：http://code.opencv.org/issues/3814  下载 NCVPixelOperations.hpp 替换掉opencv2.4.9内的文件， 重新build。

NCVPixelOperations.hpp

妙算上cuda、opencv以及qt的配置安装过程可以参考另外参考：http://blog.csdn.net/sinat_21699465/article/details/50750464

caffe 安装参考

http://www.gpus.cn/qianduan/front/getNewsArc?articleid=83