Read below two papers on the results of work from the previous project, HiDALGO.
The first one published in Future Generation Computer Systems is entitled: “Profiling and optimization of Python-based social sciences applications on HPC systems by means of task and data parallelism” and authored by Łukasz Szustak, Marcin Lawenda, Sebastian Arming, Gregor Bankhamer, Christoph Schweimer and Robert Elsaesser. The article presents optimisation techniques for two Python-based large-scale social sciences applications: SN (Social Network) Simulator and KPM (Kernel Polynomial Method). These applications use MPI technology to transfer data between computing processes, which in the regular implementation leads to load imbalance and performance degradation. To avoid this effect, we propose a 2-stage optimization (tasks scheduling and division) and facilitate the use of multiple NUMA domains.
The second paper “Large-Scale Parallelization of Human Migration Simulation” published in IEEE Transactions on Computational Social Systems is authored by Derek Groen, Nikela Papadopoulou, Petros Anastasiadis, Marcin Lawenda, Lukasz Szustak, Sergiy Gogolenko, Hamid Arabnejad, and Alireza Jahani. This work focuses on an agent-based modelling tool (the Flee simulation code) that can forecast population displacements in civil war settings. However, performing accurate simulations requires non-negligible computational capacity. We present an approach parallelisation for fast execution on multicore platforms and discuss the algorithm’s computational complexity and implementation.
The findings of those two contain general conclusions on optimisation and co-design techniques that can also be used in the current work of HiDALGO2.