Tutorial for the K2I 2015 summer institute on accelerators
You can download these instructions from here: link
Running on davinci
On davinci login node :
- Grab a copy of OCCA:
git clone https://github.com/tcew/OCCA2 - Tell OCCA where it lives by editing ~/.bashrc
export OCCA_DIR=$HOME/OCCA2
export PATH=$PATH:$OCCA_DIR/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$OCCA_DIR/lib - Grab a copy of the examples:
git clone https://github.com/tcew/HPC15.git - Request a GPU node on davinci:
qsub -I -l nodes=1:ppn=12:gpu,walltime=00:30:00 -W group_list=k2i -q classroom
On GPU node:
- Set up terminal session for CUDA and OpenCL (sets path and library path):
module load cuda - Build OCCA (on the GPU node):
module load cuda
cd OCCA
make -j - Test OCCA (should report all compute devices on all thread models detected):
occainfo - . Build a simple OCCA example:
cd ~/HPC15/examples/occa/simple
make - Run OCCA example:
./main