Skip to content

Use shape matrix cache in LiquidFlow process to reduce memory consumption

Tom Fischer requested to merge TomFischer/ogs:ShapeMatrixCacheForLF into master

In the actual master the shape function is stored in integration point data for each element. Since the shape function is constant per mesh element type, it is sufficient to store the shape function once per element type in a cache. Such a cache was/is already implemented and is used in this MR.

The memory consumption was measured using a unit cube domain with 10^6 hexahedra. The peak memory consumption reduced from 5.1 GB (master) to 4.5 GB (this branch).

  1. Feature description was added to the changelog
  2. Tests covering your feature were added? - all available benchmarks pass
  3. Any new feature or behaviour change was documented? - internal change - no user documentation necessary

Merge request reports