LInux CUDA Environment Configuration

1. Install Toolkit
(1) cd/home/cuda_train/software/cuda4.1
(2)./cudatoolkit_4.1.28_linux_64_rhel6.x.run
Specify the installation directory
(3) Configure CUDA Toolkit environment variables
(a) Vim ~/.BASHRC
(b) Add the following line to add the path to the Cuda bin to the environment variable path
Export path= $PATH:/usr/local/cuda/bin
(c) Add the following line to add Cuda/lib and cuda/lib64 paths to environment variables Ld_library_path
Export Ld_library_path= $LD _library_path:/usr/local/cuda/lib:/usr/local/cuda/lib64
(d) Make configuration effective
SOURCE ~/.BASHRC

2. Install Cuda SDK
(1) cd/home/cuda_train/software/cuda4.1
(2)./gpucomputingsdk_4.1.28_linux.run
During installation, you need to specify the installation directory and enter the CUDA Toolkit installation directory
(3) Compiling the SDK sample program
Suppose the SDK installation directory is $ (CUDA_SDK) =~/nvidia_gpu_computing_sdk/
(a) before compiling the SDK routines, the Sir becomes a static library
Enter $ (CUDA_SDK)/shared directory, execute make, build libshrutil_x86_64.a
Enter $ (CUDA_SDK)/c/common, execute make, build libcutil_x86_64.a
(b) Compiling the SDK sample program
Sample program source code in the $ (CUDA_SDK)/C/SRC directory, compiled the successful executable file in the $ (CUDA_SDK)/c/bin directory.
Take Devicequery as an example:
Enter $ (CUDA_SDK)/c/src/devicequery directory, execute make compilation
Enter $ (CUDA_SDK)/c/bin/linux/release directory, execute./devicequery Run
To recompile, you can perform make clean; Make