[daip] Re: A question about hacking PAPC.INC in and older version of AIPS

Eric Greisen egreisen at cv3.cv.nrao.edu
Sat Oct 20 23:31:18 EDT 2001


Samuel Conner writes:

 >    I am running 15JUL94 AIPS on an old SPARC5. I have plenty of
 > RAM and would like to increase the array processor size beyond the
 > largest option in PAPC.INC (which is 1.25 Mword or 5 MByte) to

        The code in this old version makes less good use of larger
memory than the code you could install in less than a day from our web
site - and IMAGR is very much better than MX.  The robustness
parameter really helps avoid uncleanable stripes.

 > speed the FFT step on some large area maps (2048 square) I am
 > making.  I recognize that this is a very obsolete revision of
 > classic AIPS that I am running and if you don't want to bother
 > with this question, I understand fully. If you are willing, here it
 > is:
 > 
 >     Is there any problem that I would cause by increasing the
 > array processor size to a much larger value, say 30 MByte? I
 > see nothing in PAPC.INC (which is a very small include) to
 > preclude an arbitrary size array processor, but I am reluctant
 > to get myself "lost" searching in the guts of the array processor routines 
 > for any potential landmines. Are there any that you are
 > aware of?

       We run with PAPC set to a larger size here:

C                                       Include for AP memory sizes
      INTEGER   APSIZE, PKPWRD, PKPWD2, PKPWDC
C                                       20.25 Megaword = 81 Megabyte size
C     PARAMETER (APSIZE=21233664)
C                                       12.5 Megaword = 50 Megabyte size
C     PARAMETER (APSIZE=13107200)
C                                       5 Megaword = 20 Megabyte size
      PARAMETER (APSIZE=5242880)
C                                       1.25 Megaword = 5 Megabyte size
C     PARAMETER (APSIZE=1310720)
C                                       1 Megaword = 4 Megabyte size
C     PARAMETER (APSIZE=1048576)
C                                       2 Megabyte size
C     PARAMETER (APSIZE=524288)
C                                       1 Megabyte size
C     PARAMETER (APSIZE=262144)
C                                       256 Kbyte size
C     PARAMETER (APSIZE=65536)
C                                       PKPWRD=primary AP size
      PARAMETER (PKPWRD=64)
C                                       PKPWD2=secondary memory
      PARAMETER (PKPWD2 = (APSIZE / 1024 - PKPWRD))
C                                       PKPWDC = clean memory
C                                       non-vector = all
      PARAMETER (PKPWDC = APSIZE)
C                                                          End PAPC.

And it works fine.  You have to rebuild a lot (in our current
systems).  I do not know if these are entirely correct instructions
for such an antique:

1. Edit the PAPC.INC

2. COMRPL $APLGEN/ZDCHIN
   COMRPL $QPSAP/*.FOR
   COMRPL $QOOP/*.FOR
   COMLNK $AIPNOT/SETPAR $AIPPGM/{SETSP,FILAI*}.FOR
   COMLNK $QPGM/*.FOR $QPGNOT/*.FOR $QPGOOP/*.FOR
   COMLNK $QYPGM/*.FOR $QYPGNOT/*.FOR

3. RUN SETSP
   29          (the type number for pseudo ap memory)
   999999      (much too much - it should then tell you how much it
                will allow)
   enter the right amount  ours is 5056.

4 restart all aips's

 > 
 >     I thank you for your time. You may not recall, but we corresponded
 > in the early '90s about increasing the CORE size in my local customized
 > version of AIPS. Your instructions worked perfectly then and again
 > recently when I did the same thing to the version I am now running.
 > 
 > Respectfully yours,
 > 
 > Sam Conner

I vaguely remember the name.  BTW - we ship with much larger CORE
these days.

Eric Greisen



More information about the Daip mailing list