[daip] Re: 31DEC04: would you mind "vetting" this Core adjustment checklist?

[Eric Greisen]

Samuel Conner writes:
 >  
 > 1. Edit $INC/PKSZ.INC  to adjust
 > KBLOCK ( X 256 = dimension of K, C, CH)
 > LBLOCK ( X 256 = dimension of LISTF and the size of the Core "source"
 > segment)
 > KPOINT = size of Core "program" segment
 >  
 >  
 > Comments:  Core variable segment size = KBLOCK X 256 - KPOINT - 11
 >            Core program segment size = KPOINT
 >  

      I think that's right

 > 2. a. COMPRL all $AIPSUB routines which include DCON.INC
 >    b. COMLNK AIPS, POPSGN, AIPSB, AIPSC

     use AIGREP:
primate<156>$ cd $TST/AIPS
/home/primate2/AIPS/31DEC05/AIPS
primate<157>$ AIGREP PKSZ $PWD
/home/primate2/AIPS/31DEC05/AIPS/PGM/FILAI2.FOR
/home/primate2/AIPS/31DEC05/AIPS/PGM/FILAIP.FOR
/home/primate2/AIPS/31DEC05/AIPS/PGM/FILINI.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/AU2A.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/SGLOCA.FOR

primate<158>$ AIGREP DCON.INC $PWD
/home/primate2/AIPS/31DEC05/AIPS/PGM/AIPS.FOR
/home/primate2/AIPS/31DEC05/AIPS/PGM/AIPSB.FOR
/home/primate2/AIPS/31DEC05/AIPS/PGM/AIPSC.FOR
/home/primate2/AIPS/31DEC05/AIPS/PGM/DAIP.FOR
/home/primate2/AIPS/31DEC05/AIPS/PGM/POPSGN.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/ADVERB.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/ADVRBS.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/ASSGN.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/AU1A.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/AU2.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/AU2A.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/AU5E.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/AU8.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/AUT.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/COMPIL.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/CONCAT.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/EDITOR.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/EQUIV.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/GETFLD.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/HELPS.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/INIT.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/K2TEXT.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/KWICK.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/LTSTOR.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/MASSGN.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/PASSGN.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/POLISH.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/PSEUDO.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/SGLAST.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/STORES.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/SUBS.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/SYMBOL.FOR
/home/primate2/AIPS/31DEC05/AIPS/SUB/VERSTR.FOR

 >  
 > 3. RUN FILINI to generate ME files.  FILINI asks for the size of the ME
 >    file in 1024 byte blocks.
 >  
 >    The dimension of the ME file must be 4 * 256 * (KBLOCK + LBLOCK)

      No - go delete the $TSTMEM/ME* file, then run FILAIP.

 >   
 > 4. RUN POPSGN to populate the $DA00/ME files
 >  
 >   
 > ------------------
 > 
 > Question: I have 1GB of ram, of which 80MB has been allocated to the AP.
 > These POPS arrays are tiny by comparison. 
 > 
 > * Is there any good reason not to make them much larger, say KBLOCK =
 > 2000 (it's currently 388)?
 > 
 > * I see that all the array dimension issues are now absorbed into the
 > KBLOCK, LBLOCK and KPOINT parameters in PKSZ.INC . Are there any
 > conventions that I need to respect in choosing these parameters (aside
 > from KPOINT < KBLOCK*256)?

The reason not to is that there are many copies of the file that is
the size of the K and LISTF arrays.  As many ME as interactive + batch
POPS numbers allowed plus all SAVE/GET files, etc.  A factor of a few
increase would not hurt on modern disks but really large factors would
hurt.  POPS is too slow to do any real computing with.

Eric Greisen