[Gretl-devel] Re: MPI bootstrap failures depending on number of iterations
Am 05.11.2021 um 19:06 schrieb Sven Schreiber:
I have GUI-gretl open and work in there (so gretlw32.exe or
whatever),
with mpi blocks, and then in parallel (no pun intended) I execute
gretlcli.exe in batch mode from time to time. (In fact, in
package-building mode --makepkg.) Now, with gretlcli there's no mpi
involved, but it seems that somehow it causes subsequent trouble with
the mpi blocks in GUI-gretl. Maybe affecting the dotdir with some side
effects?
That strange interaction effect above still stands. So now I'm avoiding
this mode of working, but I believe the issue should be further
addressed, because it feels totally legit to run a CLI instance of gretl
in parallel to a GUI instance, no?
Maybe that already gives you some clue - I will keep investigating and
debugging.
The rest is still weird. I did see an error from MPI again, but then I
can run the same script with the same settings and data later and
there's no error.
(I am using --single-rng right now and set seed inside the mpi block, so
I hope the numbers used should also be identical across runs.)
Perhaps this is somehow due to gretl's attempt to spit out MPI output in
real time on Windows, as Allin mentioned. So we can hope that this
problem would be fixed as well, as soon as that attempt is abandoned.
Still haven't had the time to test on Linux.
thanks
sven