Skip to content

Weak scaling

DMI-Cavelan Aurélien requested to merge weakScaling into gpu-hackathlon

This branch is based on the gpu-hack-noallocs branch.

  • Reduce memory footprint of the tree (no longer the bottleneck)
  • Remove unnecessary data (ro_0 and p_0)
  • Remove initFluidDensityAtRest call
  • Fix work allocation and load balancing for both small and high number of ranks
  • Fix a debug macros and add some more
  • Fix Global Tree configuration for running at scale

Merge request reports