[Hallc] Sent on Behalf of Thia Keppel

[Erica Jones]

ACCELERATOR SEMINAR

GPT Best Practices

Dr. Bas van der Geer
Author, General Particle Tracer Software

Over the past two decades the General Particle Tracer (GPT) code has
become a well-established simulation platform for the design of charged
particle accelerators and beamlines. However, a significant increase in
hardware complexity, in addition to ever increasing demands from the
accelerator community, make it more and more challenging to maintain a
generic code that performs well over a broad range of applications. On
the one hand we are proud that GPT is being used for both the simulation
of keVelectron microscopes and GeV class accelerators, but on the other
hand we see an increasing number of performance related problems where
simple changes in beam line modelling and numerical settings can
severely impact the performance of the code. This presentation focuses
on the design philosophy of GPT, explaining why certain approaches work
better than others. Topics include modelling external fields, stochastic
Coulomb interactions, the PIC space-charge model, running the
multi-objective optimiser, load-balancing on MPI clusters, and any
application of GPT proposed by the audience.


Tuesday, November 7, 2017

10:30 a.m.

CEBAF Center, Room F113