 |
 |
On Tue, 20 Dec 2011, Malcolm J. Currie wrote: > There are fixed-sized arrays declared in the routines. Note CCDPACK was > first written in the VAX era. Peter Draper would have a better idea how > much effort it would take it to change these to dynamic arrays. Or not, it has been quite a while since I wrote this all. As Malcolm says these limits have crept up over the years as people have requested it. The main driver behind keeping these values low was the requirement for handling the variance arrays (that is N^3 for image sized arrays), rather than the list processing. > It is a little strange that the maximum number of NDFs is 1000, yet the > maximum number of input files for FINDOFF is 400. It's not the 400 that's > large, but the requirements for the intercomparisons is that goes as > 0.5*N*N-1. So if we increase N from 400 to 1000 that means declaring arrays > over six times larger. The storage for these arrays in FINDOFF > is effectively 26*N*N. So if we increased this storage to 1000 files, over > 20MB more storage would be needed to run FINDOFF. > > If you have the source, look at file applications/ccpdack/CCD1_PAR and > increase Parameter CCD1__MXLIS. Alternatively divide and conquer: form > overlapping subsets, combine those, and then register those subsets to form > the final mosaic. > > It looks as if 400 was a compromise. The original version had a 40-file > limit, that's gradually been increased over the years as demands and memory > have grown. > > Malcolm > > ---- > Starlink User Support list > For list configuration, including subscribing to and unsubscribing from the > list, see > https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK > -- Peter W. Draper, http://astro.dur.ac.uk/~pdraper ---- Starlink User Support list For list configuration, including subscribing to and unsubscribing from the list, see https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK