Jon Perryman wrote:

<begin extract>
Rarely will radical changes such as changing optimization algorithms
be implemented thru a PTF, In general these types of changes occur at
FMID level (Although there are exceptions).
</end extract>

I think I understand and agree with the thrust of what he is saying,
but 'optimization algorithms' that are implemented incorrectly must be
and are corrected in PTFs.  Moreover, such PTFs are, at least in my
experience, common.

Wholly new optimization schemes are certainly not likely to be
implemented in PTFs.

Changing focus now, my recent experience suggests that for
statement-level languages values of ARCH(<level>) can be the locus of
serious problems: bad code may be generated for some but not all of
them; and for this reason they too should be captured, as I do for my
own code using the binder IDENTIFY statements, which generate IDRs.

John Gilmore, Ashland, MA 01721 - USA

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to