论文信息 - Adding shared memory parallelism to FLASH for many-core architectures

Adding shared memory parallelism to FLASH for many-core architectures

In this paper we discuss evolutionary changes to FLASH to enable enhanced applications to run efficiently on both the current generation BG/P and the next generation BG/Q. We motivate the need for change by discussing current FLASH applications and the challenges they are facing on today’s architectures. Our solution to current challenges with a view to the next generation is mixed-mode MPI+OpenMP FLASH applications. We show some preliminary results and discuss next steps.

[1] B. Fryxell,et al. FLASH: An Adaptive Mesh Hydrodynamics Code for Modeling Astrophysical Thermonuclear Flashes , 2000 .

[2] Wei-keng Liao,et al. A case study for scientific I/O: improving the FLASH astrophysics code , 2012 .

[3] Michael E. Papka,et al. In situ data analysis and I / O acceleration of FLASH astrophysics simulation on leadership-class system using GLEAN , 2011 .

[4] Leonid Oliker,et al. Modeling the Office of Science ten year facilities plan : The PERI Architecture Tiger Team , 2009 .

[5] Andrew Siegel,et al. Extensible component-based architecture for FLASH, a massively parallel, multiphysics simulation code , 2009, Parallel Comput..

[6] Peter MacNeice,et al. Paramesh: A Parallel Adaptive Mesh Refinement Community Toolkit , 2013 .

[7] Caltech,et al. Flame Evolution During Type Ia Supernovae and the Deflagration Phase in the Gravitationally Confined Detonation Scenario , 2007, 0706.1094.