Skip to content

Fix a performance regression for Lanczos.

Nicolas Tancogne-Dejeanrequested to merge
fix_perf_lanczos into main

Description

Fix a performance regression for Lanczos. For bulk Si, these initializations where making the code twice slower.

This is revealed by VTune image

and the impact of the initialization can be understood as the two arrays have in 164k which is 4 times more point than a batch for a typical Si calculation. In this example, the setting to zero are the dominant cost of the calculation...

As these initializations are useless anyway, I am simply removing them.

News snippet

Fix a performance regression for Lanczos.

Checklist

  • I have checked that my code follows the Octopus coding standards
  • I have added tests for all the new features added in this request.

Merge request reports