version française rss feed
HAL : in2p3-00585625, version 1

Fiche détaillée  Récupérer au format
Conference on Computing in High Energy and Nuclear Physics 2010 (CHEP 2010), Taipei : Taïwan (2010)
ng: What Next-Generation Languages Can Teach Us About HENP Frameworks in the Manycore Era
S. Binet1

Current HENP frameworks were written before multicore systems became widely deployed. A 'single-thread' execution model naturally emerged from that environment, however, this no longer fits into the processing model on the dawn of the manycores era. Although previous work focused on minimizing the changes to be applied to the LHC frameworks (because of the data taking phase) while still trying to reap the benefits of the parallel-enhanced CPU architectures, this paper explores what new languages could bring to the design of the next- generation frameworks. Parallel programming is still in an intensive phase of R&D and no silver bullet exists despite the 30+ years of literature on the sub ject. Yet, several parallel programming styles have emerged: actors, message passing, communicating sequential processes, task-based programming, data flow programming, . . . to name a few. We present the work of the prototyping of a next-generation framework in new and expressive languages (python [4] and Go [5]) to investigate how code clarity and robustness are affected and what are the downsides of using languages younger than Fortran/C/C++.
1 :  LAL - Laboratoire de l'Accélérateur Linéaire
Informatique/Langage de programmation
Liste des fichiers attachés à ce document : 
binet-chep-2010-id-114.pdf(329.7 KB)