search

edoddridge / aronnax

An idealised isopycnal model that can be run either with n+1/2 layers, or with n layers and variable bathymetry.
http://aronnax.readthedocs.io/en/latest/
MIT License
23 stars 5 forks source link

Benchmarks up to 64 cores on MIT cluster #216

Closed edoddridge closed 5 years ago

edoddridge commented 5 years ago

Ran some more serious benchmarks on the local cluster, using up to 64 cores for a 480x480 simulation.

The 1.5 layer simulation scales atrociously. There are far too many exchange calls, and we see very little benefit from the extra cores.

The 2 layer simulation scales acceptably - this is because the runtime is dominated by the matrix inversion, which scales very well. For larger numbers of CPUs the inefficiencies in the rest of the code become an increasingly large contributor to the runtime.

codecov[bot] commented 5 years ago

Codecov Report

Merging #216 into master will not change coverage. The diff coverage is n/a.

Impacted file tree graph

@@           Coverage Diff           @@
##           master     #216   +/-   ##
=======================================
  Coverage   91.82%   91.82%           
=======================================
  Files          20       20           
  Lines        1822     1822           
  Branches       96       96           
=======================================
  Hits         1673     1673           
  Misses        124      124           
  Partials       25       25

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 796acce...0ebd96b. Read the comment docs.