[Paper Review] First Results with HIJING++ on High-energy Heavy Ion Collisions

Gábor Papp, G. G. Barnaföldi+7 more|Repository of the Academy's Library (Library of the Hungarian Academy of Sciences)|May 7, 2018

High-Energy Particle Collisions Research3 citations

TL;DR

This paper presents HIJING++ 3.1.1, a fully C++-based, modular, and parallelized event generator for high-energy heavy-ion collisions, built on PYTHIA8 and enhanced with updated nuclear effects. It achieves ~24x speedup via thread-level parallelization (4.2% non-parallelizable fraction) and shows fair agreement with ALICE data for charged hadron spectra and nuclear modification factors in pp and pPb collisions at √sNN = 5.02 TeV, with room for improvement in jet quenching modeling.

ABSTRACT

We present preliminary results with HIJING++ (3.1.1) for identified hadron production in high-energy heavy ion collisions at LHC energies. The recently developed HIJING++ version is based on the latest version of PYTHIA8 and contains all the nuclear effects that have been included in the HIJING2.552, which will be improved by a new version of the shadowing parametrization and jet quenching module. Here, we summarize the structure and the speed gain due to parallelization of the new program code, also presenting some comparison between experimental data.

Motivation & Objective

• To modernize the legacy FORTRAN-based HIJING event generator into a portable, modular, and parallel C++ framework compatible with modern HEP software like AliRoot.
• To implement updated nuclear effects—such as $Q^2$-dependent shadowing and jet quenching—using the latest PYTHIA8 and LHAPDF6 libraries.
• To achieve significant computational speedup through multithreaded parallelization while maintaining physical accuracy comparable to the original HIJING 2.552.
• To enable future integration into large-scale detector simulation frameworks and support for GPU-accelerated random number generation.
• To provide a foundation for next-generation simulations at future colliders with improved phenomenological modeling and extensibility.

Proposed method

• Rewrote the original FORTRAN HIJING 2.552 code in object-oriented C++ using PYTHIA8's class structure as a base.
• Integrated nuclear effects including shadowing (with $Q^2$ dependence), Cronin effect, jet quenching, and multiple parton interactions via modular classes.
• Implemented thread-based parallelization using HijQueue and HijManager classes to distribute events across CPU cores.
• Enabled user-defined analysis via the HijAnalysis class with configurable histogram collection modules.
• Optimized performance by replacing the default PRNG with a GPU-compatible version in preliminary tests.
• Applied Amdahl’s law to quantify speedup and identify the 4.2% non-parallelizable portion (mainly queue management).

Experimental results

Research questions

• RQ1 Can a modern, C++-based, and parallelized event generator be developed from the legacy HIJING FORTRAN code while preserving its core physics?
• RQ2 To what extent does the new HIJING++ model reproduce experimental data on charged hadron spectra in pp and pPb collisions at √sNN = 5.02 TeV?
• RQ3 How effective is thread-level parallelization in reducing computational time, and what is the theoretical speedup limit?
• RQ4 Why does the current jet quenching model lead to over-suppression in the nuclear modification factor R_{pPb}, and how can it be improved?
• RQ5 Can the modular C++ design enable seamless integration into large-scale HEP simulation frameworks like AliRoot?

Key findings

• HIJING++ 3.1.1 achieves a maximum theoretical speedup of ~24x on a 20-thread system, with only 4.2% of the code being non-parallelizable, as confirmed by Amdahl’s law fitting.
• The charged hadron yield in pp collisions at √s = 5.02 TeV is within 50% of ALICE data, with HIJING++ outperforming PYTHIA8 at mid- and high-pT regions depending on tune.
• In pPb collisions at √sNN = 5.02 TeV, HIJING++ results agree with ALICE data within 30% for hadron spectra and nuclear modification factors.
• The nuclear modification factor R_{pPb} is generally suppressed in HIJING++, indicating an overestimation of energy loss, likely due to incorrect treatment of N_bin scaling relative to N_part.
• The code structure supports future integration of advanced jet quenching models such as GLV and GPU-accelerated random number generation, with preliminary tests showing minor speed gains.
• The modular C++ design enables easier maintenance, extensibility, and compatibility with modern HEP software stacks like AliRoot, reducing code bloat and cross-linking.