Future of Many-Core Computing in Science

Europe/Budapest
2nd Floor, Meeting Room (Wigner Datacenter)

2nd Floor, Meeting Room

Wigner Datacenter

29-33 Konkoly-Thege Miklós Str. H-1121 Budapest, Hungary
Dániel Berényi (Wigner RCP), Gergely Gábor Barnaföldi (Wigner RCP RMI of the Hungarian Academy of Sciences), No name No surname (No institute), No name No surname (No institute)
    • 08:30 09:00
      Registration - enter to the Campus 30m
      Slides
    • 09:00 09:15
      Opening -- Greetings from the General Director 15m
      Speaker: Peter Levai (WIGNER RCP)
    • 09:15 10:00
      Portable HPC for High-Performance Simulation 45m
      High Energy Physics code has been known for making limited use of high performance computing architectures. Efforts in optimising HEP code on vector and RISC architectures have yield limited results and recent studies have shown that, on modern architectures, it achieves a performance between 10% and 50% of the theoretical one. Although several successful attempts have been made to port selected codes on GPUs, no major HEP code suite has a "High Performance” implementation for accelerators. With LHC undergoing a major upgrade and a number of challenging experiments on the drawing board, HEP has to try making the best usage of the new hardware. This activity is one of the foci of the SFT group at CERN, which hosts, among others, the Root and Geant4 projects. One important initiative is centred on the development of a high-performance prototype for particle transport, the GeantV project. A good concurrency level has been achieved by Geant4 with the parallelisation at event level. However, apart the sharing of data structures, this does not increase the number of events per second produced on a loaded system, which requires an efficient use of the low level instruction parallelism and pipelining of modern processors. In GeantV project we have implemented a framework that allows scheduling vectors of particles to an arbitrary number of computing resources in a fine grain parallel approach. Via template code specialisation we are able to recast simple computational kernels do different architectures, minimising code duplication. This approach is already providing very promising results in the treatment of the detector geometry and we plan to extend it soon to the optimisation of the physics code. The talk will review the current satis and the perspectives of the development of a simulation framework able to profit best from the recent technology evolution in computing.
      Speaker: Dr Federico Carminati (CERN)
      Slides
    • 10:00 10:20
      The Future Computing with Silicon Computer Systems 20m
      Speaker: Mr Gábor Lehoczki (Silicon Computers LTD)
      Slides
    • 10:20 10:40
      Coffee Break 20m
    • 10:40 11:50
      AMD Heterogeneous System Architectures : Software and Hardware Ecosystem 1h 10m
      Speakers: Bruno Stefanizzi (Advanced Micro Devices INC.), Dr Dmitry Kozlov (Advanced Micro Devices INC.)
      Slides
    • 11:50 12:10
      Bi-directional particle tracing on the GPU with applications in PET 20m
      Speaker: Prof. László Szirmay-Kalos (Technical University of Budapest)
    • 12:30 13:30
      Lunch Break 1h
    • 13:30 13:50
      QCD on Lattice 20m
      In numerical simulation of quantum chromodynamics the presence of large number of integration variables requires up-to-date hardware technologies, and software developed for the particular hardware. Due to the locality in field theory the main source of performance improvement nowadays is parallelization and simulation on GPU clusters. We present the GPU cluster at the Eotvos Lorand University. We briefly review the Krylov-Schur algorithm (which we use for matrix diagonalization and matrix function computation) and compare the performance of the GPU and parallel CPU implementation.
      Speakers: Mr Ferenc Pittler (Eötvös Univeristy, Budapest), Prof. Sándor Katz
      Slides
    • 13:50 14:10
      GPU Code-generation for Differential Equation Solvers 20m
      Speaker: Mr Dániel Berényi (Wigner RCP of the HAS, Eötvös University)
      Slides
    • 14:10 14:30
      GridRipper >> Lattice Template Library 20m
      Speaker: Mr Máté Ferenc Nagy-Egri (WIgner RCP of the HAS)
      Slides
    • 14:30 14:50
      Accelerated Monte Carlo Particle Generators for the LHC 20m
      Speaker: Dr Gergely Gábor Barnaföldi (Wigner RCP RMI of the Hungarian Academy of Sciences)
      Slides
    • 14:50 15:10
      Coffee Break 20m
    • 15:10 17:00
      Closed Session 1h 50m