NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/bit/listserv/ibmmain/2604

Text File | 1992-11-09 | 1.8 KB | 32 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!stanford.edu!unixhub!fnnews.fnal.gov!overload.lbl.gov!zeus.ieee.org!europa.asd.contel.com!paladin.american.edu!auvm!AWIIMC12.IMC.UNIVIE.AC.AT!CHRISTIAN.J.REICHETZEDER
Organization: Institut fuer Medizinische Computerwissenschaften
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.ibm-main
Date:         Mon, 9 Nov 1992 22:54:21 CET
Sender:       IBM Mainframe Discussion list <[e-mail]>
From:         [e-mail]
Subject:      Re: Forcing page boundaries in high level code
In-Reply-To:  Message of Mon, 9 Nov 92 16:25:00 EST from <JURASCHEK@GMUVAX>
Lines: 19

On Mon, 9 Nov 92 16:25:00 EST David F. Juraschek said:
>For example, is there some way to make various groups of paragraphs to be
>aligned at an even 2K or 4K (or whatever) page.  The object here is to
>allow portions of a COBOL program, for example to be loaded into the same
>physical page of memory [....]
In my experience the primary reason for paging is due to
jumping through the data sections (arrays of structures vs. structures of
arrays and the like).
Admitted - for large  programs code too can cause heavy  paging. As Larry said
overlays can do  some tricks to keep  certain parts of code  together but with
sloppy coding you not only get paging but also I/O due to fetching of modules.
Keep in mind that  HLL's often use library calls so you  always have a certain
minimum working  set. If you  want to bundle  parts together (given  your flow
already  avoids unnecessary  jumps through  the code)  compile the  program in
small chunks.  You can then  direct the loader to  load the parts  at specific
locations. Or you  could write dummy assembler routines which  punch SLC cards
and contain references to the COBOL entry points.

Christian