search

drwells / fiddle

4 stars 3 forks source link

Improve ghost updates in IFEDMethod::computeLagrangianForce() #143

Closed drwells closed 1 year ago

drwells commented 1 year ago

In the Turek-Hron benchmark we spend 2% (via callgrind) of total runtime waiting for ghost updates here. There are two possible improvements:

  1. Examine if we actually have ghost values present at this point. I don't think we do but it's worth double-checking.
  2. If not: only conditionally update the velocity (e.g., the beam doesn't use drag forces so we don't need updated ghost values)
  3. If not: interleave communication to get better load balancing in the multi-part case.