Computing strategies, Algorithms and Code Optimization for existing and new architectures.

The goal of this WP is to help the theoretical community to use as eciently as possible the new computational building blocks that are becoming more and more relevant in high-performance systems.
We plan to select a number of relevant large size computing projects directly relevant to the INFN theoretical community, in areas such as LGT, uid dynamics, complex systems, quantitative-biology, numerical relativity, and explore at all appropriate levels - from programming strategies to algorithms to coding style - the best ways to use new architectures and networks. This work will be supported by several experimental machines either directly available at INFN or made available by collaborating institutions. We also plan to identify computational problems studied in collaborating institutions and work together with their researchers to help optimize their algorithms and programs.



Upgrade of the TheoPhys Cluster.

The goal of this WP - to be activated in the early phases of the project - is a substantial upgrade of the TheoPhys cluster currently operational at Sezione di Pisa. This machine, currently used by approximately 100 users, is rapidly becoming obsolete; we plan a complete overhaul of the system, replacing the proces sors and the network infrastructure, and improving access to all users. A key element here is the integration of the new cluster with a similar system to be installed soon at SISSA, so that INFN and SISSA users share both machines.



Assembly of a large prototype of a state-of-the-art parallel system.

This WP builds on the assessment of the results made by the WPs in the first year of work, and selects and procure a reasonably large experimental system that may be considered as an intermediate step towards Exascale level machines. This system should combine at best commercially available building blocks, as well as any element directly or indirectly stemming from the development activities made by INFN in recent years. The large prototype will be a crucial resource tool for all physics groups (in LQCD and elsewhere) to conceive and test new
computing strategies for new probes which are particularly well suite for numerical computation and are also easily connected with interesting physical observables. We consider it very important that this procurement involves an industrial vendor and that the system be installed at a mayor national computer center and be made available for tests also to external collaborators, in order to provide the greatest possible visibility to the results of our project.



Advanced developments.

This WP provides continued support to the computer development activities that have been carried out for years
at INFN, aiming to develop and test fresh ideas that may evolve into key building blocks for scienti c computing at the Exascale level.