CUDA Toolkit

Programming model and platform for Nvidia GPU. Find more information on the availability of GPU here.

Tools included:

  • C++ extensions for CUDA programming.
  • CuDNN library for neuron networks.

Preparing the working environment

module load cuda/cuda-8.0

Executables

nvcc
nvidia-smi

Usage instructions for the cluster

Task examples available here:

/opt/exp_soft/user_info/cuda

The following Linux-related graphical tools can be used for programming and debugging: emacs and ddd.

module load cuda/cuda-8.0
emacs hello-world.cu
nvcc -g -G hello-world.cu –o hello-world.out
ddd –debugger cuda-gdb hello-world.out

NVIDIA Nsight, a development environment based on Eiclipse and development by NVIDIA, can be used.

module load cuda/cuda-8.0
nsight

Note. To be able to use GUI tools, X11 forwarding must be performed to provide a remote graphical interface (instruction is available here).