Quick Start

To get the source code for esi-g4ox, download or clone the repository from GitHub. If you have the code in your $HOME directory and prefer not to install all external dependencies manually, you can use esi-shell for development. The following command mounts your $HOME inside the container:

esi-shell -- -v $HOME -e HOME=$HOME -w $HOME -u $(id -u ${USER}):$(id -g ${USER})

Once inside the container, run the following commands:

cd $HOME
cmake -S esi-g4ox -B build
cmake --build build

To run a test that generates a NumPy file (out/photons.npy) with the simulated photons, execute:

./build/src/simtox

Similarly, to start a Geant4-based simulation of optical photons, run:

./build/src/simg4ox

Before runnning simg4ox it is recommended to set the following environment variables:

export TMP=/tmp/myname
export GEOM=mygeom
export OPTICKS_EVENT_MODE=DebugLite

Visualization

Plot any volume serialized

scripts/plot-csg.py ../out/csg/CSGFoundry/SSim/scene/meshmerge/0/

GPU code profiling

nsys software can be used to profile GPU code. By adding

nsys profile

An examples as follows:

nsys profile ./build/src/simg4ox -g esi-g4ox/geom/opticks_raindrop.gdml -m esi-g4ox/run.mac

The resulting file has nsys-rep extension. This can be opened either by nsys-ui or can be exported with "nsys stats" command into an sqlite database.

In order to achieve a more detailed profiling one can use the nsight compute framework. The user needs to download nsight compute, accept the license agreement. If one is using docker the following argument needs to be added:

--cap-add=SYS_ADMIN

An example of running the nsight compute profiling command:

/usr/local/NVIDIA-Nsight-Compute/ncu --set full --target-processes all -o report --import-source on ./build/src/simg4ox -g  esi-g4ox/geom/pfrich_min_added_parameters.gdml -m esi-g4ox/run.mac

Afterwards the resulting profile file can opened in the following GUI:

/usr/local/NVIDIA-Nsight-Compute/ncu-ui