[Date Prev][Date Next][Thread Prev][Thread Next]
[Author Index]
[Date Index]
[Thread Index]
[SQR-USERS Info]
[SQRUG Home Page]
Re: more info - performance with local procedures...
- Subject: Re: more info - performance with local procedures...
- From: "Rao, Sam" <Sam_Rao@AFCC.COM>
- Date: Mon, 30 Aug 1999 13:36:58 -0500
Clark,
Based on your sample programs: blob1.sqr contains an additional assignment
statement ( let #count_in = #count_out) that gets executed once for every
row (i.e. 109135 times!). Could the difference in execution times be partly
attributed to this assignment statement that gets executed so many times? To
properly benchmark this kind of a scenario you should select a scenario that
does not require additional statements like this (between the local and
global procedures) which could give a misleading result!
-Sam Rao
> -----Original Message-----
> From: the dragon [SMTP:ceprn@HOTMAIL.COM]
> Sent: Monday, August 30, 1999 12:30 PM
> To: Multiple recipients of list SQR-USERS
> Subject: more info - performance with local procedures...
>
> Dear gurus, guruettes and lurkers,
>
> Thanks for the feedback I received last week about the performance of
> local
> procedures vs global procedures. There seem to be some strong feelings
> about whether they are good to use of not, with a lot of what I heard
> being
> that local procedures *should* run faster since they are in memory, or
> something like that; however, no one cited any real evidence of whether
> that
> was true or not. So, while I didn't take a lot of time to test this
> theory
> out, I created a couple of simple programs, blob1.sqr (with a local
> procedure) and blob2.sqr (with no local procedure), compiled them and ran
> them. The programs being basically identical, except one procedure is
> either local or global, produced the following results:
>
> for 109135 records on an idle RS6000 running Oracle 8,
> blob1 (local) 15 seconds
> blob2 (global) 13 seconds.
>
> Based upon this very limited data (no significant validity) it appears
> global procedures are faster and easier to work with.
>
> I welcome any additional evidence that either supports or debunks this
> theory. please attache code and results in postings - either here or
> directly to me.
>
> long life and continued success
>
> clark
> dragon enterprises
>
> ______________________________________________________
> Get Your Private, Free Email at http://www.hotmail.com << File: blob1.log
> >> << File: blob1.sqr >> << File: blob2.log >> << File: blob2.sqr >>