Article

 

Superlinear Speedup Phenomenon in Parallel 3D Discrete Element Method (DEM) Simulations of Complex-shaped Particles Public Deposited

Downloadable Content

Download PDF
https://scholar.colorado.edu/concern/articles/1r66j236j
Abstract
  • Strong superlinear speedup has been discovered in large scale simulationsof parallel 3D DEM for complex-shaped particles, which is based on an al-gorithm of spatial domain decomposition, and exhibits the “high-CPU-low-memory” characteristics. The interpretation of this phenomenon requires acareful examination of the speedup theory and practice in the field of parallelcomputing. The superlinear speedup is investigated from three perspectives:(i) memory footprint per process, (ii) cache miss rates of L1, L2 and L3 levelcaches, and (iii) uniprocessor performance, using a wide range of problemsize (across five orders of magnitude of simulation scale regarding number ofparticles) and number of compute nodes (1 to 2,048 nodes) on DoD super-computers. The Performance-API (PAPI) is employed in the source code tomeasure cache miss rate and FLOPS. The strong scaling measurements showthat cache miss rate is sensitive to the memory consumption shrinkage perprocessor, and the last level cache (LLC) contributes most significantly tothe strong superlinear speedup among all of the three cache levels, and this isalso revealed in the weak scaling measurements. The findings are associatedwith the inherently perfect scalability of 3D DEM: its memory scalabilityfunction is a nonlinearly decreasing function of the number of processors. Inaddition, a constant (non-increasing) uniprocessor FLOPS performance w.r.tproblem size can also contribute to the superlinear speedup.

    The superlinear speedup is a common phenomenon for large scale 3D DEM simulations of complex-shaped particles, and the larger the scale, thestronger is the superlinear speedup. DEM researchers should take advantageof this effect to speedup their parallel simulations.

Creator
Date Issued
  • 2018-07
Academic Affiliation
Journal Title
Journal Volume
  • 75
Last Modified
  • 2021-07-19
Resource Type
Rights Statement
DOI
ISSN
  • 1872-7336
  • 0167-8191
Language

Relationships

Items