PrincetonUniversity/STELLOPT

parallelize TERPSICHORE mode family calculations

Closed this issue · 2 comments

It appears that the TERPSICHORE calculations within STELLOPT may not be fully parallelized. The mode families seem to be treated sequentially - For example, for a 5 field period configuration (with 3 mode families) the n=0 eigenvalue is calculated BEFORE the |n| =1, which is calculated BEFORE the |n|=2 eigenvalue.

This should not be the case, but is how the code will execute if NOPTIMIZERS is set equal to the total number of processors. @pomphrey can you provide a log file so we can confirm the details of your run?

I'm closing this as we're already parallelized over mode families.