Cosmological simulations on many-core architectures
|Research Area||Astro Sciences|
|Principal Investigator(s)||Dr. Simon Portegies Zwart|
In astronomy large cold dark matter simulations are used to study the formation of large-scale structures in the universe. Such simulations help us understand details of the universe, such as how galaxies form.
Since we want to study these problems in more detail, we plan to optimize our simulation code to allow runs of 81923 particles on more than 10^5 cores. Simulations of such size require a highly scalable code that is able to use large amounts of processing power (i.e. CPU cores) and computer memory. Current bottlenecks in our code therefore need to be removed. Our objectives are to study and improve the scalability of the code, to determine the performance on different architectures and if possible to conduct some experiments with larger scale meta-computing (current experience with 3 machines).