[daip] AIPS 31DEC09- IMAGR

Eric Greisen egreisen at nrao.edu
Thu Mar 5 10:50:25 EST 2009


Amy Mioduszewski wrote:
> Hi Gabi,
> 
> I'm sorry our IMAGR expert is away so you'll have to bear with me.  That is very 
> strange.  The 08 and 09 versions of IMAGR are identical.  Do you have an "old" 
> version of 08?  or the most updated one?
> 
> OVERLAP seems to control what fields are reimaged.  What do you have it set to?
> 
> Amy
> gcastell at iafe.uba.ar wrote:
>> Dear AIPS group,
>>
>> I am imaging low radio frequency data using AIPS 31DEC09 version.
>> Surpringly the task IMARG runs more slow than in the older AIPS 31DEC08
>> version. Now,
>> I noted that after each major cycle IMAGR makes images of some fields (I am
>> working with 31 facets each of one of 4096X4096 px, cell=1".5).
>>
>> Assuming that this routine slows IMAGR, I would like to know what inputs I
>> should modify in order to avoid imaging fields each time.
>>
>> In addition, I do not understand the way in which IMAGR selects the set of
>> fields for imaging after each iteration. In my case IMAGR only selects
>> fields 1 2 3 5 7 14 to make the intermediate image.
>>
>> I would be very grateful if you can give me some information about these
>> points.

I changed IMAGR to try to take advantage of more memory to reduce the 
I/O.  There are a variety of places in which that is now done.  The most 
uncertain area for this was the decision to make current images of 
several facets each time we are seeking a new facet to Clean.  Otherwise 
it often makes one image, decides it is not good enough, makes another 
image, decides that is not good enough, etc with a full read of the data 
with each pass.  This happens with some source structures a great deal 
and with other source distributions, no where near as often.  The 
default is to examine the size of the images, the size of the allowed 
max pseudo AP, and the size of the uv work file and to then make a guess 
as to the maximum number of facets imaged each time.  I have added to 
the 31DEC09 IMAGR another IMAGRPRM option (18) so that you may limit the 
maximum number of facets re-imaged each time.  A MNJ on this release 
will update your IMAGR to have that option.

Eric Greisen




More information about the Daip mailing list