On Fri, 5 Nov 2021, Sven Schreiber wrote:
Am 05.11.2021 um 17:37 schrieb Cottrell, Allin:
> On Fri, Nov 5, 2021 at 11:57 AM Sven Schreiber <svetosch(a)gmx.net> wrote:
>> Hm, even trying Listing 1 from the gretl + MPI guide (wrapped inside an
>> mpi block) fails for me on the latest snapshot, whereas it runs OK
>> (only) if I comment out the following two lines at the top:
>>
>> set use_dcmt off # or not?
>> set seed 12337
>>
>> Is there a difference between an mpi block and directly running this
>> stuff in the gretlmpi program?
> Yes, "use_dcmt" is specific to gretlmpi. If you want to turn off dcmt
> for an MPI block, use the --single-rng option.
Alright. Follow-up question: If I run a slight variation of Listing 1,
namely inside an mpi block, and without the use_dcmt setting (see below
for the pasted code), I get the output OK, but then gretl also tells me
that the "external command failed" (translated). What's that supposed to
mean? Oh, and this is happening quite often, but not in every attempt,
so there seems to be some kind of flakiness involved, too.
I can sort of see why that might be. It might be better to give up
on trying to relay gretlmpi output to the GUI in real time, on
Windows. If we just put up a message such as "This may take a
while..." and let MPI run to completion before trying to pick up
output, there would be less chance of getting spurious "error"
conditions.
Allin