Based on stats I have had, command chaining or similar seems to be what they
do. Did that years ago in a bunch of programs for tape spooling and
subsequent processing.
-----Original Message-----
From: Tom Brennan
Sent: Sunday, September 14, 2014 5:25 PM Newsgroups: bit.listserv.ibm-main
To: [email protected]
Subject: Re: SORT JCL
I remember I needed some quick changes to a very large PS dataset, and
not wanting to struggle with DFSORT syntax I wrote a quick assembler
program (QSAM with a large block size and extra buffers). Nothing could
be faster than an assembled program, right? Later I figured out how to
do the same work with DFSORT or ICETOOL, and was surprised when the SORT
job ran in about 25% less clock time than my own simple program. All I
could guess is that DFSORT has some DASD tricks to speed things up, like
you say.
Martin Packer wrote:
Without wishing to appear a "die hard defender of DFSORT" :-) I would
expect DFSORT's I/O speed to be better than that of a program (even with
decent Sequential File tuning). But that quite possibly DOESN'T matter.
Cheers, Martin
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
