14th Geant4 Users and Collaboration Workshop

Context of session - and proposed agenda

19 Oct 2009, 14:10

10m

Room "Azzurra" (Laboratori Nazionali del Sud - INFN)

Parallel Session I - Particle tables and processes

Description

** Focus ** Proposed as a working session - focusing on: * understanding the current use cases for Physics tables (in sequential Geant4 applications) * understanding the characteristics (and limitations) when used in a multi-process (copy on write) application; * understanding the characteristics and requirements in the use case of a prototype multi-threaded Geant4 application; * clarifying the requirements for Physics table(s) coming from multi-process and multi-core extensions; * discussing potential ways to overcome these current weaknesses * summarising our considerations, as a starting point for a deeper examination and potential revision of the design of the Physics tables to cope with the new requirements.

Summary

Potential agenda
* Brief introduction to Physics Tables
( focusing on use cases, requirements ) - talk of 3-4 slides, ( ~ 5 minutes )
- Memory cost of Physics tables
*Considerations and challenges of multi-core ( ~ 30 minutes )
Mini-presentation: Maximum 10 slides [ 15 minutes ]
- Why share physics tables - either as pages of memory or as objects ?
- Current difficulties in multi-core (mixing of read-only and read-write addresses)
- How are physics tables (and client classes) treated in Xin's Multi-Threaded prototype
Discussion [ ~ 15 minutes ]
- Challenge of multi-threading (explanation of context in M/T prototype)
- Are there potential benefits for ordinary G4 (sequential) from revising the PhysicsTable implementation?
* Clarification of use case(s) - with focus on multi-core/multi-threading ( ~ 20 minutes )
* Distillation of requirements, potential constraints ( ~ 20 minutes )
* Potential solutions ( ~ 10 minutes )