Research in HPC

Intel Parallel Computing Centre (Intel PCC) at LRZ/TUM

The Intel Parallel Computing Center at Leibniz Supercomputing Centre and Technische Universität München is a project financially supported by Intel to modernise community codes for the newest generations of Intel architectures. In this framework, our team at LRZ is active on the optimisation of performance of P-Gadget3, a widely used community code for computational astrophysics, on multi and many-core computer architectures, including the new Intel Xeon Phi processor of second generation, codenamed Knights Landing (KNL). The author is proposer and co-investigator of the project, funded by Intel for two years with a yearly grant of 110 000 USD. The figure below shows the improvement in the execution time of a Gadget kernel on different architectures, indicated in the legend, for the original and optimised versions.

Bar plot from Baruffa, Iapichino et al. 2017

For the latest updates on the IPCC project please consult our local webpage.

Collaboration with astrophysicists: AstroLab projects

The Astro and Plasma Physics Application Lab (AstroLab) is a LRZ initiative for high-level support to users of the LRZ facilities from the astrophysics community.
In this framework, I collaborated in 2016 to the project The world's largest turbulence simulations (with Christoph Federrath, Ralf Klessen and Nicolay Hammer), for the memory optimisation of the FLASH code and the preparation of a large-scale computing project (44 million core-hours, granted by the Gauss Center for Supercomputing). The figure below shows the weak scaling of the optimised version of the code, performing very well up to the whole SuperMUC Phase 2 system.

Weak scaling, Federrath et al.

The analysis of the scientific results is ongoing, the technical description can be found here.

Currently I am collaborating with Daniel Ceverino (ITA Heidelberg) in the project FirstLight: simulations of the first galaxies in the early Universe. During this project we are optimising the node-level performance of the ART code. More results are coming soon!  
All content: © Luigi Iapichino, last modification: 24-Jul-2017