Abstract

Poster - Splinter Computation   (virtual room B)

Fast and efficient data analysis of extensive hydro simulations with the Julia language

M. Behrendt
USM/LMU - Max-Planck-Institut für extraterrestrische Physik

I present my work on developing a fast and efficient post-processing tool for the next generation of large-scale hydro simulations. I discuss typical bottlenecks and different concepts used in existing libraries. I show how it is solved in MERA, followed by benchmarks. Currently, it works with RAMSES but can be easily extended to other grid and N-body codes. Key goals are: - performance of statically-typed languages like Fortran or C/C++ - efficient use of the memory - simple syntax like in high-level languages (e.g. Python, Matlab) - many functionionalities, customizable to the user's problem - avoid overdoing high-level abstraction (black boxes) - interactivity - easy to extend - east to install and update - reproducibility - runs on multiple platforms Find the current public version at https://github.com/ManuelBehrendt/Mera.jl