This code is an excercise to evaluate Julia for HEP application. The antikt hadronic jet clustering is implemented using the fastjet tile N² algorithm [1].

We started with translating the C++ code extracted from the fastjet package. Then the code was profiled to improve the performance. We had little experience with the Julia programming language when translating the code.

It should be noted that the fastjet code has been written to run very fast, the code in addition to the algorithm itself has been highly optimized. The code use heavily pointers.

Timing has been peformed using 100 multijet events from proton-proton collisions at sqt{s} = 13TeV generated with Pythia8 [2].

Hardware: laptop with an Intel(R) Core(TM) i7-8550U CPU, 8 GB of RAM.

Reference timing from C++ code: 263 ± 9 μs

Time with the latest julia implementation: 315 ± 7 μs, 1.2x C++ implementation.

JIT compiling is excluded from the measurement: function called once before the timing to trigger the compilation.

Optimisation history:

The HepMC3 is used by the c++ version of the code to read the data file. Install it in a directory you will call externals at the root directory of this project or set the directory where it is installed in the Makefile.def file. If you have access to LCG cvmfs, it can be found in /cvmfs/sft.cern.ch/lcg/releases/hepmc3/<release>/<architecture>

A script to run the benchmark can be found in the directory benchmark.

• ./run_benchmarks will create the events.hepmc3 input file and runs the benchmark on the C++ and Julia code.

Instead of using the run_benchmarks script the test codes can be executed directly. For this, you need first to:

1. Create a events.hepmc3 or uncompress the one from data/events.hepmc3;
2. Either install the AntiKt and HepMC3 packages located in the directory antikjl and hepmc3jl using the dev command of the Julia package manager or add these two directories in the JULIA_LOAD_PATH environment variable (as done in the run_benchmark script).

The Julia benchmark code is benchmark/run_antikt_jl and the c++ one is antiktcxx/antikt. The script benchmark/run_antikt_cxx can be used to ensure the c++ code is built and run it. Example of execution:

• ./run_antikit_jl -m 100 events.hepmc3
• ./run_antikit_jl -m 100 events.hepmc3

A compressed data file is located under the path data/events.hepmc3.gz. The fata file can also be generated using the code from the genevts directory.

[1] Pileup subtraction using jet areas, M. Cacciari and G. P. Salam, Phys. Lett. B 641 (2006) 57
[2] A comprehensive guide to the physics and usage of PYTHIA 8.3, C. Bierlich et al., arxXiv: 2203.11601