Warning, /jana2/src/examples/SubeventCUDAExample/README.md is written in an unsupported language. File is not indexed.
0001 # Run on JLab ifarm GPUs
0002
0003 This is an example that can run on a single GPU.
0004
0005 1. On the `ifarm` node, ask for a GPU compute node with 8000 MB memory.
0006
0007 ```bash
0008 module load cuda
0009
0010 srun --gres gpu:${gpu_type}:1 -p gpu --mem-per-cpu=8000 --pty bash
0011 ```
0012 ${gpu_type} can be `T4`, `TitanRTX` or `A100`.
0013
0014 2. On the compute node, build the whole application with cmake option "USE_CUDA=On".
0015
0016 ```bash
0017 # bash-4.2$ rm -rf build/*
0018 bash-4.2$ cmake -DUSE_CUDA=On -S . -B build # add the cmake option to config
0019 #### output
0020 #-- Found CUDA: /apps/cuda/11.4.2 (found version "11.4")
0021 bash-4.2$ cmake --build build -j 32 --target install # build and install as normal
0022 ```
0023 Launch the application.
0024 ```bash
0025 bash-4.2$ ./install/bin/SubeventCUDAExample
0026 ```
0027
