Qlgyro4tgyro

[bpatel2107]

I think this the interface between QLGYRO and TGYRO in Fortran is now complete and can be merged into unstable!

[jcandy]

@bpatel2107 there is a conflict with cgyro_write_restart.F90 (which no longer exists). Can you have a look?

[bpatel2107]

@bpatel2107 there is a conflict with cgyro_write_restart.F90 (which no longer exists). Can you have a look?

I've merged master into this branch and removed that files now so should be good now. The issue I fixed in that file was fixed in master anyway .

[bpatel2107]

There is a small issue that I haven't quite figured out yet but it is not always a problem. Currently QLGYRO runs CGYRO for multiple KY/PX0 and runs them asynchronously. We split the total number of MPI processes over the number of requested parallel runs (set by N_PARALLEL) in the input file. This is generally more efficient as different runs can take different times so it is best to split your MPI processes such that many individual CGYRO runs can be done whilst the longer CGYRO runs take place (usually the case with slow growing/stable points).

The issue I am facing is that sometimes 2 CGYRO runs will finish at the same time and both will attempt to look at the next (ky,px0) to run by looking at the out.qlgyro.status file. Usually that MPI process will modify to the file to "claim" that (ky,px0). However if they happens to access the file at the same time then both MPI processes (usually a group of them) will run the same (ky,px0) at the same time which leads to problems. This is particularly bad when restarting a QLGYRO run.

Does anyone know of a way to prevent this. Maybe there is a way to lock the out.qlgyro.status file so only one process can read it at a time.

[gmstaebler]

I approve this