QCD simulations with staggered fermions on GPUs

Bonati, Claudio; Cossu, Guido; D’Elia, Massimo; Incardona, Pietro

doi:10.1016/j.cpc.2011.12.011

Cited by 28 publications

(26 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Once again, we believe that this is due to some immaturity of the compiler, for which we expect will be resolved in future versions. Fig.4 addresses the question of the efficiency costs (if any) of our architecture-portable code; it compares the execution time for a full Monte Carlo step (in double precision) of the OpenACC code and a previously developed CUDA implementation [9], optimized for NVIDIA GPUs. Although the two codes are not exactly in a one to one correspondence, the implementations are similar enough to make the test quantitatively meaningful.…”

Section: Performance Analysismentioning

confidence: 99%

“…Lattice QCD simulations is a typical and well known HPC grand challenge, where physics results are strongly limited by available computational resources [3,4]; over the years, several generations of parallel machines, optimized for LQCD, have been developed [5,6], while the development of LQCD codes running on many core architectures, in particular GPUs, has seen large efforts in the last decade [7][8][9]. Our target is to have a single code able to run on several processors without any major code change while looking for an acceptable trade-off between portability and efficiency [10].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Portable LQCD Monte Carlo code using OpenACC

et al. 2018

View full text Add to dashboard Cite

Abstract. Varying from multi-core CPU processors to many-core GPUs, the present scenario of HPC architectures is extremely heterogeneous. In this context, code portability is increasingly important for easy maintainability of applications; this is relevant in scientific computing where code changes are numerous and frequent. In this talk we present the design and optimization of a state-of-the-art production level LQCD Monte Carlo application, using the OpenACC directives model. OpenACC aims to abstract parallel programming to a descriptive level, where programmers do not need to specify the mapping of the code on the target machine. We describe the OpenACC implementation and show that the same code is able to target different architectures, including state-of-the-art CPUs and GPUs.

show abstract

Section: Performance Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Portable LQCD Monte Carlo code using OpenACC

et al. 2018

View full text Add to dashboard Cite

show abstract

“…All numerical simulations have been performed using an Rational Hybrid Monte-Carlo (RHMC) algorithm running on Graphics Processing Units (GPUs) [86,87].…”

Section: Simulation Detailsmentioning

confidence: 99%

Gauge-invariant screening masses and static quark free energies in Nf=2+1 QCD at nonzero baryon density

et al. 2018

View full text Add to dashboard Cite

We discuss the extension of gauge-invariant electric and magnetic screening masses in the quark-gluon plasma to the case of a finite baryon density, defining them in terms of a matrix of Polyakov loop correlators. We present lattice results for N f ¼ 2 þ 1 QCD with physical quark masses, obtained using the imaginary chemical potential approach, which indicate that the screening masses increase as a function of μ B . A separate analysis is carried out for the theoretically interesting case μ B =T ¼ 3iπ, where charge conjugation is not explicitly broken and the usual definition of the screening masses can be used for temperatures below the Roberge-Weiss transition. Finally, we investigate the dependence of the static quark free energy on the baryon chemical potential, showing that it is a decreasing function of μ B , which displays a peculiar behavior as the pseudocritical transition temperature at μ B ¼ 0 is approached.

show abstract

“…Consequently, there is an on-going softwareand algorithm development in order to incorporate GPUs effectively into lattice simulations. See for example [1,2,3,4]. These applications are developed and carried out predominantly on NVIDIA hardware, consistently using the NVIDIA exclusive CUDA language [5] for the interaction with the GPU.…”

Section: Figure 1: Loewe-cscmentioning

confidence: 99%

“…The first lattice simulations in OpenCL were performed in [1] with staggered fermions. On NVIDIA hardware, a significantly lower performance (25% on C1060 and 60% on S2050) of OpenCL was reported compared to CUDA for Hybrid Monte Carlo (HMC) updates.…”

Section: Figure 1: Loewe-cscmentioning

confidence: 99%