[daip] Imagr limits

[Eric Greisen]

Hans-Rainer Kloeckner wrote:
> Dear Eric,
> 
> Sorry to have another request. I'm currently running into the following 
> limitation in imagr. Would there be a way to increase this limit please?
> 
> cheers,
> 
> Hans
> 
> 
> IMAGR1: Field   24 min =   -7.4 MilliJy,max =    7.9 MilliJy
> IMAGR1: Field   50 min =   -6.5 MilliJy,max =    7.1 MilliJy
> IMAGR1: Field   50 min algorithm flux=   -4.054 MilliJy iter=     1167
> IMAGR1: Field   50 Clean flux density=  372.065 MilliJy     1167 comps
> IMAGR1: Total Cleaned flux density     =   11.615      Jy   174461 comps
> IMAGR1: Field   24 min =   -7.3 MilliJy,max =    7.9 MilliJy
> IMAGR1: Field   87 min =  -11.9 MilliJy,max =   10.9 MilliJy
> IMAGR1: Field   87 Iter limit, Max resid=   4.104 MilliJy iter=     2212
> IMAGR1: Field   87 Clean flux density=   31.019 MilliJy     2212 comps
> IMAGR1: Total Cleaned flux density     =   11.616      Jy   174468 comps
> IMAGR1: I Polarization model processed
> IMAGR1: Total Clean components   174468 reaches limit  174468
> IMAGR1: Merging the Clean components files
> IMAGR1: ZABORS: signal 7 received
> IMAGR1: ABORT!

There is no such limit in the software.  The maximum iteration number is 
set by the adverb NITER and is changed only with the same adverb in a 
TELL operation.  The message indicates that CLNLIM (adverb NITER) had 
the value 174468 for some reason.   I suppose there is some chance that 
an array overflowed its bounds and altered CLNLIM but the messages that 
precede this all look too normal for that.

What is rather more concerning is the BUS error that follows.  That 
suggests that something has gone rather wrong in the memory addressing 
presumably in the merging of the CC files.  IMAGR is used a lot and for 
significant numbers of facets and NITER values way more than 175 
thousand.  What sort of computer are you using and if Linux do you know 
whether you are using the 32-bit ($ARCH=LINUX in AIPSpeak) or 64-bit 
($ARCH=LNX64) version?

i assume that this is some sort of massive data set with BOXFILEs etc to 
do the imaging, making it hard for me to try to debug it...

Eric Greisen

We did have a user with wierd problems that cleared up when he reset the 
size of the AP allowed.  I recently changed the code to double precision 
from single and this led to issues about what the max AP size means.
Try the verb SETMAXAP with no argument and then re-issue
         SETMAXAP n
where n is the number of Megabytes you want.  I don't think that will 
fix anything here but it can't hurt to do this.