Subject: | |
From: | |
Reply To: | |
Date: | Fri, 5 Feb 1999 14:55:33 -0800 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Paul asks an excellent question:
> >A great performance hint: OCTCOMP and ALLOCATE fcopy!
>
> I've always been curious. If 'octcomping' a CM run module is supposed to
> improve performance why doesn't HP ship 'octcomped' versions of their utilities
> that are still in CM.
I don't know the answer, but I've been recommending to vendors for years
that they OCTcomp their CM programs because it's important for *them* to
do the testing! Sure, the program files are a lot bigger (8 to 10 times!),
but DDS tapes and disk drives are relatively cheap.
Why have the vendor test? As I pointed out before, the OCTCOMP is a
*compiler*. Compilers have been known to have bugs, even OCTCOMP. I'd
rather have the vendor use *their* validation suites to test the
OCT'd program than have *my* data be the test bed :)
BTW, we always do:
:newgroup cm
:copy fcopy.pub, fcopy.cm
:octcomp fcopy.cm, fcopy.pub
(or similar)
Note: I don't recommend OCT'ing the CMSTORE program...the last time
I did that (over 10 years ago), the result corrupted file labels
when I ran it.
--
Stan Sieler [log in to unmask]
http://www.allegro.com/sieler.html
|
|
|