Probably it is a little too late and my brain doesn't work as it is
supposed to do anymore, but I don't understand! From the numbers Charles
posted doesn't result at all that "OPM is faster", on the contrary! For
example, over 5M iterations, Bound Procedures are six times faster (1s
vs. 6s), over 50M iterations is over 15 times faster (4s vs.
62s)....(???)
Florin Todor
-----Original Message-----
From: rpg400-l-bounces@xxxxxxxxxxxx
[mailto:rpg400-l-bounces@xxxxxxxxxxxx] On Behalf Of Scott Klement
Sent: Thursday, March 18, 2010 3:32 PM
To: RPG programming on the IBM i / System i
Subject: Re: Timestamp vs Date/time Decimal fields,Overhead in RPG
programs...
These numbers seem reasonable to me. No surprises. However, I'd
caution against posting something like this because I have no doubt that
someone is going to make a bad decision because "OPM is faster" (which
is ridiculous -- the difference in performance here is something you'll
never notice!)
On 3/18/2010 7:37 AM, Charles Wilt wrote:
#Iterations
5M 50M
OPM calling OPM 6s 62s
Dynamic Program Call 12s 122s
Bound Procedure 1s 4s
Bear in mind that ILE programs have a PEP and a UEP. So when you call
an ILE program, it creates a call stack entry for the PEP (which is
generated by the OS) and then a call stack entry for the UEP (which is
your main procedure -- your code so to speak.)
By contrast, OPM doesn't have a PEP... you call the main procedure
directly.
Therefore, it makes perfect sense for the OPM calls to be twice as fast,
all else being equal. One routine on the call stack vs. two routines in
the call stack.
But I hate to see someone post something like this. The performance you
posted will make ZERO difference to anyone. Really. 6 seconds over
FIFTY MILLION iterations? Who the hell cares? "I converted to ILE and
my 8-hour batch job now takes 6 seconds longer". Give me a break.
midrange.com
