Scalability/en: Difference between revisions

Scalability/en (view source)

2 bytes added , 6 years ago

Updating to match new version of source page

38,897

edits

@@ Line 98: / Line 98: @@
 |77.6
 |}
-The formula for the efficiency here is quite simple, just the reference run time divided by the run time at <math>n</math> cores then multiplied by a hundred to obtain a percentage. Once again, the goal is run with an efficiency of at least 75%. As is often the case, efficiency remains high up to larger core counts than with strong scaling.
+The formula for the efficiency here is quite simple, just the reference run time divided by the run time at <math>n</math> cores then multiplied by a hundred to obtain a percentage. Once again, the goal is to achieve an efficiency of at least 75%. As is often the case, efficiency remains high up to larger core counts than with strong scaling.
 Weak scaling tends to be especially pertinent for applications that are memory-bound. If the parallel program has been designed to privilege communications between nearest neighbours then the weak scaling is usually good. An application which performs a lot of nonlocal communication (e.g. a fast Fourier transform<ref>Wikipedia, "Fast Fourier transform: https://en.wikipedia.org/wiki/Fast_Fourier_transform</ref>) may exhibit poor performance in a weak scalability analysis.