[Difx-users] difxcalc number of antennas

[Mark Kettenis]

> Date: Fri, 11 Jun 2021 10:58:12 +1000
> From: Adam Deller via Difx-users <difx-users at listmgr.nrao.edu>
> 
> Hi difx-users,
> 
> I was trying to abuse difxcalc as a convenient means of generating a model
> for a short-baseline interferometer (by faking up a difx input and calc
> file), but it segfaults when I try to run with 72 antennas.  cmxst11.i says
> that the parameter "Max_stat" was set to 4 1, so I guess that makes sense,
> but when I change that to 80 and recompile, I still get a segfault.
> cmxst11.i says that this parameter should only need to be changed in that
> one place - see
> https://svn.atnf.csiro.au/trac/difx/browser/applications/difxcalc11/trunk/src/cmxst11.i#L7.
> Any suggestions on where else an offending max station limit might be
> lurking?

Hi Adam,

When I looked into generating the model for SFXC with calc10, I ran
into similar issues.  Instead of trying to fix the calc10 code I
decided on a different strategy and have generate_delay_model (SFXC's
exquivalent to difxcalc11) only generate delays for a single station
(i.e. for the baseline of that station to CoE).  It then is very easy
to parallize the model calculation by simply running multiple
processes in parallel (which I implemented using a simple Python
wrapper).  Maybe not the most efficient way to do things (as you will
be calculating the delay for the CoE station multiple times), but in
terms of wall clock time, it makes a huge difference.

Cheers,

Mark