Efficient Computing for High Energy Physics

Lead Research Organisation: University of Warwick
Department Name: Physics

Abstract

Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.

Planned Impact

See main proposal summary

Publications

10 25 50
 
Description The work revealed that the software package 'Sherpa' is spending a major fraction of its CPU time in calls to a subsystem 'LHAPDF'. This unexpected wasteful behaviour has triggered subsequent work which has sped up the code a factor 10, leading to a paper "Accelerating LHC event generation with simplified pilot runs and fast PDFs" explicitly highlighting this work in the title. The worldwide grid, hundreds of thousands of CPUs, devotes a significant fraction of its resources to this activity, so this order of magnitude saving results in lower power consumption, better physics, or both.
Exploitation Route The worldwide grid, hundreds of thousands of CPUs, devotes a significant fraction of its resources to this activity, so this order of magnitude saving results in lower power consumption, better physics, or both.
Sectors Other
URL https://arxiv.org/abs/2209.00843
 
Title Sherpa 
Description Sherpa is a multipurpose simulation programme for high energy event generation. The grant did _not_ lead to its creation, but to performance improvements, which will, for example, save many hundreds of CPU years/year for the ATLAS experiment at LHC alone. 
Type Of Technology Grid Application 
Year Produced 2020 
Open Source License? Yes  
Impact A fix was made providing a lookup for a function accessed frequently, saving 5% CPU. It was also discovered that half the Sherpa CPU usage was in PDF access and potential other improvements due to caching in the associated LHAPDF package were identified, but a fix was not completed during the award period.