Skip to content

Adding SIMD instruction for some loops.

Nicolas Tancogne-Dejean requested to merge improve_simd_support into develop

Description

Improved SIMD support. In the case of nested loops, the OpenMP without SIMD prevents vectorization, which limits performances for single thread runs. This is at least the case for Intel compiler, and the optimization reports reveal that adding the SIMD statement changes how the Intel compiler optimize the loops.

There are many more potential routines, but these ones are important for performances and show a clear gain by ~10% when the SIMD statement is added.

News snippet

Improved OpenMP SIMD support.

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