Workpackage : WP10
Processor test beds in large scale
mathematical calculations -
collaboration and exchange visits.
Start date or start event : 12 month
Collaboration and exchange of researches participating in the large scale mathematical computations:
- Exchange of experience in application of the available microprocessor to the CPU intensive
mathematical double-precision calculations.
- Planning the performance and accuracy tests of the OO/C++ new real world-wide
packages: GEANT4 tool-kit, Athena/Gaudi framework of HEP.
- Dissemination of the research knowledge towards Jagellonian University undergraduate
and graduate students.
The revolution in the High Energy Physics (HEP) software, the process of world-wide changing from the Fortran to OO/C++ programming technologies requires CPU and human intensive testing of the applicability of the available computing platforms as well as compilers and mathematical libraries to such massive computations.
The merit of the work is to exchange experience in using the real-world complicated software of the HEP experiments and organization of the massive computations with large data flows. Such software typically requires simulations of hundred thousands of events with simulation time 20-30/min per event on the modern HP/Intel processors. The typical size of code is several thousands of lines, requires double precision floating-point operations, loading and correctly executing several external libraries. Stability of such very complicated software system, with multi-language support, both in performance and accuracy is a big challenge for the modern OO/C++ technologies. Organization of the massive production and bookkeeping using conditional databases and other modern technologies (eg. WWW) will be also an important ingredient of the activity within this work package.
The researchers of the DACM have already had the experience with such tests of the large scale computation with the HP/Intel microprocessor with explicit parallelism. One of them participated in the two-years joint project of CERN and Hewlett Packard Company (1998-2000).
Number of visits :
Partners involved :
- 1yr - 1 visit (Barcelona, Valencia),
- 2 yr - 1 visit (Milan).
Universidad de Valencia (Spain),
Sezione Particelle Elementari, Dipartimento di Fisica, Universita degli Studi di Milano,
Instituto de Fisica de Altas Energias (IFAE), Universidad Autonoma de Barcelona, Spain.
- Report on the status of the collaboration and obtained results. (D19)
- Lectures for undergraduate and graduate students: "Massive computations and
their organisation in large scale physics experiments".
Milestones and expected results :
- Active collaboration and the exchange of experience through short-term exchange visits.
- Development of modern technologies for the organisation of massive computations with
large data-flows and intensive floating-point calculations.