Turning off the pesky MPI_INSTRUMENTATION flag since it spits out a ton of files into the run directory (one for each MPI processor) upon completion. Also document it and how to turn it back on in comments and in docs.
UPDATE1: updated to remove the #define MPI_INSTRUMENTATION from the source, and documented how to enable it through the docs (consistent with the comments I received here).
I think it should be specified in the MACH_FLAGS instead of a compile-time option because I see Chronos phasing out this old timing module.
So do you think I should remove the commented form from macros_and_parameters.h? and just leave the information in the docs on how one can manually do this by editing his/her Makefile at compile time?
+1 on this rather than adding yet another compile time flag.
It probably should be, yes, but I honestly didn't think we even needed it at all. Sam and my performance tools cover most of this stuff, and I was just removing it from default (and documenting it) in hopes that we'd remove it entirely in a newer version of enzo.
@ngoldbaum makes a very valid point in the IRC channel, in that why don't we remove this module entirely? There are lots of "#ifdef MPI_INSTRUMENTATION" commands littering the code, and if no one is using this functionality (largely because of the reasons I address in the documentation in the PR), then why not just rip it out instead of leaving it around?
If there is functionality in these tools that isn't in the stuff you and Sam wrote, you should leave it in. When the functionality is totally covered, then I would vote for pulling it out. Just because it isn't used by people who are on IRC this evening doesn't mean it isn't useful
+1 on BWO
OK, I will pull out the comment in the source, and just leave the docs with how to enable it through editing one's makefile. We can look at fully reproducing this functionality in a future version of chronos.