Optimizing Barnes-Hut simulations for many-core super computers using the scalable parallel runtime

Barnes-Hut;Scalable Parallel Runtime (SPR);Qthreads;Parallel Programming Language (PPL);Many-Core;Exascale

Nusbaum, Kurtis ; Snir ; Marc

关键词: Barnes-Hut;    Scalable Parallel Runtime (SPR);    Qthreads;    Parallel Programming Language (PPL);    Many-Core;    Exascale;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/44186/Kurtis_Nusbaum.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

It is now generally accepted that the road to Exascale Super Computing will no doubt include increasing core counts on individual compute nodes. The Scalable Parallel Runtime (SPR) is a project being developed at Sandia National Laboratories which attempts to address the challenge of efficiently utilizing the power of these new many-core systems. SPR is a combination of the Qthreads Library, Portals, and MPI. In this paper, we investigate optimizing the Barnes-Hut simulation by using the PPL Runtime which we implement using SPR. PPL provides a PGAS runtime with one-sided com- munication facilities, ideal for irregular applications with dynamic commu- nication patterns like that of Barnes-Hut.