GPU Compilation on Nova
Steps
Load the following modules:
module load git openmpi cudaCheck out the Alamo repository:
git clone https://github.com/solidsgroup/alamo.gitFor this workflow, check out the
gpubranch:git checkout gpuIn the Alamo directory, configure with:
./configure --get-eigen --cudaThe Nova
eigenmodule will not work for this workflow.Compile the currently working example:
make bin/heatOptionally start an interactive compilation session to speed up the build:
srun --partition=interactive --pty bashThe modules must be reloaded inside the interactive session.
You may need to run
makemore than once; the old wiki noted odd issues when compiling in parallel.With Alamo compiled, start a new interactive session:
srun --partition=interactive --gres=gpu:a100:1 --pty bashLoad CUDA:
module load cudaRun:
./bin/heat-3d-cuda80-g++ tests/HeatConduction/inputThe old wiki reported that this did not work because
libcuda.so.1could not be found. Symlinking to the/stublibrary was attempted, but the stub library could not be used.
Last updated
Was this helpful?