[daip] AIPS Y2K problem

[Eric Greisen]

Joseph Lazio writes:

 > -> uname -a
 > Linux hood.nrl.navy.mil 2.4.22-1.2199.nptlsmp #1 SMP Wed Aug 4 11:48:29 EDT 2004 i686 i686 i386 GNU/Linux
 > 
 > -> gcc --version
 > 2.95.3
 > 
 > -> g77 --version
 > GNU Fortran 0.5.25 20010315 (release)
 > 
 > 
 > I've run the Y2K LARGE and HUGE tests (in TMODE='M' mode).  In all
 > cases, the RMS number of correct bits is larger than the MAX number of
 > correct bits.  However, the number of correct MAX bits is sometimes
 > quite small.
 > 
 > Where do I start looking for the source of the problem?  Should I
 > upgrade the gcc and g77 versions to 3.2.x, is it something in the
 > optimization, ...?  Thanks in advance for any help.
 > 
 > 
 > Here's the number of correct bits for the LARGE test
 > 
 > -> grep BITS Y2K_LARGE.log 
 > AIPS 1: 'NUMBER CORRECT BITS MAX:'      12.7283573
 > AIPS 1: 'NUMBER CORRECT BITS RMS:'      15.5307283
 > AIPS 1: 'NUMBER CORRECT BITS MAX:'      14.2665386
 > AIPS 1: 'NUMBER CORRECT BITS RMS:'      16.5649376
 > AIPS 1: 'NUMBER CORRECT BITS MAX:'      12.2058344
 > AIPS 1: 'NUMBER CORRECT BITS RMS:'      16.5839443
 > AIPS 1: 'NUMBER CORRECT BITS MAX:'      15.3931131
 > AIPS 1: 'NUMBER CORRECT BITS RMS:'      22.8904324
 > AIPS 1: 'NUMBER CORRECT BITS MAX:'      10.8945446
 > AIPS 1: 'NUMBER CORRECT BITS RMS:'      16.4354191
 > AIPS 1: 'NUMBER CORRECT BITS MAX:'       3.5376263
 > AIPS 1: 'NUMBER CORRECT BITS RMS:'      11.5676575
 > AIPS 1: 'NUMBER CORRECT BITS MAX:'       3.4324644
 > AIPS 1: 'NUMBER CORRECT BITS RMS:'      11.1726274

    This suggests that VTESS is the problem.  Note that VTESS changed
- so I changed the Y2K procedure and the master data.  The new master
data is available at the ftp site

     ftp://ftp.aoc.nrao.edu/pub/software/aips/FITS/31DEC04

and you should use the latest Y2KLOAD.  The current HUGE data are
there too - all previous HUGEs are obsolete and never really worked.


 > 
 > and here it is for the HUGE test
 > 
 > -> grep BITS Y2K_HUGE.log
 > AIPS 2:    TYPE 'NUMBER CORRECT BITS MAX: ', MAPDIF(1,JJCNT)
 > AIPS 2:    TYPE 'NUMBER CORRECT BITS RMS: ', MAPDIF(2,JJCNT)
 > AIPS 2: 'NUMBER CORRECT BITS MAX:'      13.1464624
 > AIPS 2: 'NUMBER CORRECT BITS RMS:'      17.0951157
 > AIPS 2: 'NUMBER CORRECT BITS MAX:'      15.2415934
 > AIPS 2: 'NUMBER CORRECT BITS RMS:'      18.2353897
 > AIPS 2: 'NUMBER CORRECT BITS MAX:'       0.8875204
 > AIPS 2: 'NUMBER CORRECT BITS RMS:'       7.1682196
 > AIPS 2: 'NUMBER CORRECT BITS MAX:'       2.7311242
 > AIPS 2: 'NUMBER CORRECT BITS RMS:'       7.5666523
 > AIPS 2: 'NUMBER CORRECT BITS MAX:'       2.1505158
 > AIPS 2: 'NUMBER CORRECT BITS RMS:'       7.1010847
 > AIPS 2: 'NUMBER CORRECT BITS MAX:'       6.807672
 > AIPS 2: 'NUMBER CORRECT BITS RMS:'      15.635848

The low ones here are residual images and so often do not have the
same number of bits.  The last 6.8 is more of a worry.  Check your
data and procedure first however.

Eric Greisen


 > 
 > 
 > -- Joe
 > 
 > -- 
 > T. Joseph W. Lazio, Ph.D.        voice: +1 202 404 6329
 > Remote Sensing Division          fax:   +1 202 404 8894
 > Naval Research Lab, Code 7213    Joseph.Lazio at nrl.navy.mil
 > Washington, DC 20375-5351 USA    http://rsd-www.nrl.navy.mil/7213/lazio/
 > 
 > _______________________________________________
 > Daip mailing list
 > Daip at listmgr.cv.nrao.edu
 > http://listmgr.cv.nrao.edu/mailman/listinfo/daip