https://github.com/IceCubeOpenSource/offline_production
with cmake3.17.0 or newer: env CC=gcc-8 CXX=g++-8 cmake ..
in build: ./env-shell.sh export I3_TESTDATA= PATH_TO_ICECUBECUDA/test-data
python clsim/resources/runSmallSim.py (needs variable I3_TESTDATA to test-data folder too) (other one: python ./clsim/resources/scripts/benchmark.py --gcd-file=$I3_TESTDATA/GCD/GeoCalibDetectorStatus_2016.57531_V0.i3.gz --numevents 1000 , has not yet been adapted to run both..)
Slim version of IceCube's physics production software, including only the projects required to run up to photon generation.
Full documentation of all projects are available at https://docs.icecube.aq/trunk .
- cmake
- python3
- numpy
- boost
- zmq
- gsl
- OpenCL
apt-get install build-essential cmake libbz2-dev libgsl0-dev libcfitsio-dev
libboost-system-dev libboost-thread-dev libboost-date-time-dev libzmq5-dev
libboost-python-dev libboost-serialization-dev libboost-filesystem-dev
libboost-program-options-dev libboost-regex-dev libboost-iostreams-dev
opencl-dev python3-numpy sprng2-dev libarchive-dev
mkdir build
cd build
export I3_TESTDATA=<path_to_source>/test-data
cmake <path_to_source>
make
./env-shell.sh
./clsim/resources/scripts/benchmark.py --gcd-file=$I3_TESTDATA/GCD/GeoCalibDetectorStatus_2016.57531_V0.i3.gz --numevents 100
This should produce output similar to this:
WARN (clsim): Propagating muons and photons in the same process. This may starve your GPU. (I3CLSimMakePhotons.py:315 in I3CLSimMakePhotons)
# these numbers are performance figures for the GPU:
time per photon (GPU): 17.113565755277392 ns
photons per second (GPU): 58433176.01369108 photons per second
# these numbers include the host utilization and are probably not meaningful for --numevents=1 (the default). You need more events to even out the startup/setup time.
(avg) time per photon (actual, including under-utilization): 17.276601912161702 ns
(avg) photons per second (actual, including under-utilization): 57881752.73611296 photons per second
(total) host time: 12.2 s
(total) waiting time: 91.9 µs (0.001%)
(total) number of kernel calls: 4
wallclock time: 12.9 s
(avg) device utilization: 99.05368708262631 %