IPAC2016 - List of Keywords (GPU)

Paper	Title	Other Keywords	Page
MOPMW037	FEL Simulation Using Distributed Computing	simulation, FEL, electron, distributed	483
	J. Einstein, S. Biedron, H. Freund, S.V. Milton, P.J.M. van der Slot CSU, Fort Collins, Colorado, USA G. Bernabeu Altayo Fermi National Accelerator Laboratory, Batavia, Illinois, USA S. Biedron University of Ljubljana, Faculty of Electrical Engineering, Ljubljana, Slovenia J. Einstein Fermilab, Batavia, Illinois, USA P.J.M. van der Slot Twente University, Laser Physics and Non-Linear Optics Group, Enschede, The Netherlands
	While simulation tools are available and have been used regularly for simulating light sources, the increasing availability and lower cost of GPU-based processing opens up new opportunities. This poster highlights a method of how accelerating and parallelizing code processing through the use of COTS software interfaces.
DOI •	reference for this paper ※ DOI:10.18429/JACoW-IPAC2016-MOPMW037
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEPMY043	Parallel Particle Movement Simulation Algorithm Based on Heterogeneous Computing	hardware, simulation, controls, framework	2654
	L.G. Zhang, L. Cao, K. Fan, J. Huang, K.F. Liu, W. Qi, J. Yang HUST, Wuhan, People's Republic of China
	Particle in cell (PIC) algorithm studies the self-consistent motion of multi-particle system by solving equations of particle dynamics, this algorithm is widely used to evaluate the nonlinear space charge effect of the high intensity or low energy beam. In order to reduce the random noise in the simulation, a huge number of particles should be traced, the process expends many computer hardware resources and a lot of computing time. Heterogeneous computing can greatly improve the efficiency of large quantities of the particle tracking by making full use of different types of computing resources. In this paper we give the algorithm which uses both CPU and GPU to trace the particles in electromagnetic field. The results show that the given algorithm increases the efficiency significantly.
DOI •	reference for this paper ※ DOI:10.18429/JACoW-IPAC2016-WEPMY043
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEPOY044	Review of CPU and GPU Faddeeva Implementations	timing, interface, simulation, space-charge	3090
	A. Oeftiger, R. De Maria, L. Deniau, K.S.B. Li, E. McIntosh, L. Moneta CERN, Geneva, Switzerland A. Aviral BITS Pilani, Pilani, India S. Hegglin ETH, Zurich, Switzerland A. Oeftiger EPFL, Lausanne, Switzerland
	Funding: CERN, Doctoral Studentship EPFL, Doctorate The Faddeeva error function is frequently used when computing electric fields generated by two-dimensional Gaussian charge distributions. Numeric evaluation of the Faddeeva function is particularly challenging since there is no single expansion that converges rapidly over the whole complex domain. Various algorithms exist, even in the recent literature there have been new proposals. The many different implementations in computer codes offer different trade-offs between speed and accuracy. We present an extensive benchmark of selected algorithms and implementations for accuracy, speed and memory footprint, both for CPU and GPU architectures.
DOI •	reference for this paper ※ DOI:10.18429/JACoW-IPAC2016-WEPOY044
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEPOY051	Performance Optimization of Multi-particle Beam Dynamics Code IMPACT-Z on NVidia GPGPU	operation, lattice, linac, diagnostics	3110
	Z.Q. He, G. Shen, Y. Yamazaki FRIB, East Lansing, Michigan, USA X. Wang ICER, MSU, East Lansing, USA
	Funding: The work is supported by the U.S. National Science Foundation , the U.S. Department of Energy Office of Science, the Institute for Cyber-Enabled Research, MSU. Facility for Rare Isotope Beams is designed using a multiparticle tracking code IMPACT-Z. IMPACT-Z is originally for the purpose of accelerator design, so it is precise, however, quite time consuming, therefore usually not suitable for on-line beam tuning applications. IMPACT-Z is originally boosted using Message Passing Interface (MPI) technology. For single node mode, performance of IMPACT-Z is usually bounded by CPU performance, and for multimode mode, communication between MPI processes would become bottleneck. However, new emerging High Performance Computing (HPC) technology, like general-purpose graphics processing unit (GPGPU), brings new possibility in accelerating IMPACT-Z, so that the speed of IMPACT-Z satisfies for on-line beam tuning applications. This paper presents the efforts in exploring the capability of Nvidia GPGPU and the results of speed up of IMPACT-Z.
DOI •	reference for this paper ※ DOI:10.18429/JACoW-IPAC2016-WEPOY051
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

Paper

Title

Other Keywords

Page

MOPMW037

FEL Simulation Using Distributed Computing

simulation, FEL, electron, distributed

483

J. Einstein, S. Biedron, H. Freund, S.V. Milton, P.J.M. van der Slot
CSU, Fort Collins, Colorado, USA
G. Bernabeu Altayo
Fermi National Accelerator Laboratory, Batavia, Illinois, USA
S. Biedron
University of Ljubljana, Faculty of Electrical Engineering, Ljubljana, Slovenia
J. Einstein
Fermilab, Batavia, Illinois, USA
P.J.M. van der Slot
Twente University, Laser Physics and Non-Linear Optics Group, Enschede, The Netherlands

While simulation tools are available and have been used regularly for simulating light sources, the increasing availability and lower cost of GPU-based processing opens up new opportunities. This poster highlights a method of how accelerating and parallelizing code processing through the use of COTS software interfaces.

DOI •

reference for this paper ※ DOI:10.18429/JACoW-IPAC2016-MOPMW037

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEPMY043

Parallel Particle Movement Simulation Algorithm Based on Heterogeneous Computing

hardware, simulation, controls, framework

2654

L.G. Zhang, L. Cao, K. Fan, J. Huang, K.F. Liu, W. Qi, J. Yang
HUST, Wuhan, People's Republic of China

Particle in cell (PIC) algorithm studies the self-consistent motion of multi-particle system by solving equations of particle dynamics, this algorithm is widely used to evaluate the nonlinear space charge effect of the high intensity or low energy beam. In order to reduce the random noise in the simulation, a huge number of particles should be traced, the process expends many computer hardware resources and a lot of computing time. Heterogeneous computing can greatly improve the efficiency of large quantities of the particle tracking by making full use of different types of computing resources. In this paper we give the algorithm which uses both CPU and GPU to trace the particles in electromagnetic field. The results show that the given algorithm increases the efficiency significantly.

DOI •

reference for this paper ※ DOI:10.18429/JACoW-IPAC2016-WEPMY043

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEPOY044

Review of CPU and GPU Faddeeva Implementations

timing, interface, simulation, space-charge

3090

A. Oeftiger, R. De Maria, L. Deniau, K.S.B. Li, E. McIntosh, L. Moneta
CERN, Geneva, Switzerland
A. Aviral
BITS Pilani, Pilani, India
S. Hegglin
ETH, Zurich, Switzerland
A. Oeftiger
EPFL, Lausanne, Switzerland

Funding: CERN, Doctoral Studentship EPFL, Doctorate
The Faddeeva error function is frequently used when computing electric fields generated by two-dimensional Gaussian charge distributions. Numeric evaluation of the Faddeeva function is particularly challenging since there is no single expansion that converges rapidly over the whole complex domain. Various algorithms exist, even in the recent literature there have been new proposals. The many different implementations in computer codes offer different trade-offs between speed and accuracy. We present an extensive benchmark of selected algorithms and implementations for accuracy, speed and memory footprint, both for CPU and GPU architectures.

DOI •

reference for this paper ※ DOI:10.18429/JACoW-IPAC2016-WEPOY044

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

WEPOY051

Performance Optimization of Multi-particle Beam Dynamics Code IMPACT-Z on NVidia GPGPU

operation, lattice, linac, diagnostics

3110

Z.Q. He, G. Shen, Y. Yamazaki
FRIB, East Lansing, Michigan, USA
X. Wang
ICER, MSU, East Lansing, USA

Funding: The work is supported by the U.S. National Science Foundation , the U.S. Department of Energy Office of Science, the Institute for Cyber-Enabled Research, MSU.
Facility for Rare Isotope Beams is designed using a multiparticle tracking code IMPACT-Z. IMPACT-Z is originally for the purpose of accelerator design, so it is precise, however, quite time consuming, therefore usually not suitable for on-line beam tuning applications. IMPACT-Z is originally boosted using Message Passing Interface (MPI) technology. For single node mode, performance of IMPACT-Z is usually bounded by CPU performance, and for multimode mode, communication between MPI processes would become bottleneck. However, new emerging High Performance Computing (HPC) technology, like general-purpose graphics processing unit (GPGPU), brings new possibility in accelerating IMPACT-Z, so that the speed of IMPACT-Z satisfies for on-line beam tuning applications. This paper presents the efforts in exploring the capability of Nvidia GPGPU and the results of speed up of IMPACT-Z.

DOI •

reference for this paper ※ DOI:10.18429/JACoW-IPAC2016-WEPOY051

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)