search

sa2c / thirring-rhmc

For Simon Hands' 3D Thirring Model LFT simulation code.
0 stars 1 forks source link

mpi=no seems broken? #2

Open chillenzer opened 1 year ago

chillenzer commented 1 year ago

I just tried to compile with mpi=no in MkFlags in order to "keep things simple". It turned out to do the opposite. Could it be that this case is broken? comms4.F90 will get compiled either way and that one explicitly asks for use mpi. Have I overlooked something?

mmesiti commented 1 year ago

You most probably have not. Most of the relevant cases (and bugs) concerned the mpi version, so I have exclusively worked on the mpi=yes case. It was not clear to me whether thempi=no -related parts of the code should be kept or not, so I kept them on a "best effort" basis, even if this made the code very hard to read (it's quite a c-preprocessor hell).

In hindsight the best ways to maintain both the mpi=yes and mpi=no version would have been to have some CI pipelines set up, but I was not so well versed in the CI-fu at the time, and there is always the risk of making the code harder and harder to maintain and to use by throwing yet more tools at it.

chillenzer commented 1 year ago

Thanks for the answer! This particular issue might be readily fixed with another #ifdef MPI, so I might come back to that later. But as you said, development-focus will probably be on the MPI case, anyways.