OPM vs ILE Performance

[Markjstev]

Does anyone know the following or where I can find the following

Performance difference between OPM programming vs ILE programming performance?

What does ILE programs do to the system (esp a CISC box)? does it hurt it?
need more memory?

TIA

Mark

[IraFrosch]

ILE programs can be faster in many circumstances. The initial startup
time is longer than for an OPM program, due to the activation or
"linking" of all programs to be used upon the initial call. After that,
all program to program calls are generally faster. There are lots of
caveats to the above, depending on your usage and how activation groups
are used, but generally, ILE should be faster and provide a lot more
functionality built in.

Memory is always subjective, but yes, if you run a lot more ILE, you
will need a larger working size for each client; i.e. more memory. The
translator (from MI to IMPI in your case) is much improved, and
equivalent programs should be faster in ILE. This is especially true
for ILE/C to EPM/C. Haven't played much with the ILE/RPG, but I would
imagine similar results.

Ira

[danh...@millcomm.com]

In <19970911031...@ladder02.news.aol.com>, mark...@aol.com (Markjstev) writes:
>Does anyone know the following or where I can find the following
>
>Performance difference between OPM programming vs ILE programming performance?
>
>What does ILE programs do to the system (esp a CISC box)? does it hurt it?
>need more memory?

If you make an apples-to-apples comparison you will find that ILE programs
are generally faster than OPM programs (though the difference is more
pronounced on a RISC box). However, ILE programs tend to require more
storage and will usually take longer for startup.

Old MI (OPM) is well-suited to RPG and pretty good for COBOL, but it is a
lousy target for C or C++.

Dan Hicks
Hey!! My advice is free -- take it for what it's worth!
http://www.millcomm.com/~danhicks

[@bigfoot.com NKP]

IraFrosch <ifr...@ibm.net> skrev i artiklen <34184C49...@ibm.net>...

> ILE programs can be faster in many circumstances. The initial startup
> time is longer than for an OPM program, due to the activation or
> "linking" of all programs to be used upon the initial call. After that,
> all program to program calls are generally faster. There are lots of
> caveats to the above, depending on your usage and how activation groups
> are used, but generally, ILE should be faster and provide a lot more
> functionality built in.
>
> Memory is always subjective, but yes, if you run a lot more ILE, you
> will need a larger working size for each client; i.e. more memory. The
> translator (from MI to IMPI in your case) is much improved, and
> equivalent programs should be faster in ILE. This is especially true
> for ILE/C to EPM/C. Haven't played much with the ILE/RPG, but I would
> imagine similar results.

At my company we have a huge order entry program, which during compilation
produces a 900+ page spoolfile. When the spoolfile is generated, it takes a
while before the compilation is finished. I assume that is the actual
program building. In OPM that takes approx. 10 minutes, but if I convert
the program to ILE, the compilation takes about the same time, but the
building of the program takes a couple of hours!!! Why is that?

--
Nils Kilden ~ Pedersen

"Your suffering will be legendary. Even in Hell." - Pinhead

[danh...@millcomm.com]

In <01bcc038$f8b8e660$LocalHost@nkp>, "NKP" <nkp @ bigfoot . com> writes:
>At my company we have a huge order entry program, which during compilation
>produces a 900+ page spoolfile. When the spoolfile is generated, it takes a
>while before the compilation is finished. I assume that is the actual
>program building. In OPM that takes approx. 10 minutes, but if I convert
>the program to ILE, the compilation takes about the same time, but the
>building of the program takes a couple of hours!!! Why is that?

The translator that backs up the ILE compiler is much more sophisticated
(and produces much faster-running code) than the one that backs up the OPM
compiler, especially if you compile for high optimization levels. Most of
the time you are observing is translator (ie, compiler backend) time.

[Martin Neugebauer]

I suspect, that you give the binder a hard time to resolve to the
modules you want to link. Did you specify something like '*ALL/*LIBL' on
the MODULE or BNDSRVPGM paramters? If so, the binder has to search thru
many places for unresolved labels. Try to be specific. Use a binding
directory.

Reagrds, Martin.

NKP wrote:

> IraFrosch <ifr...@ibm.net> skrev i artiklen
> <34184C49...@ibm.net>...
> > ILE programs can be faster in many circumstances. The initial
> startup
> > time is longer than for an OPM program, due to the activation or
> > "linking" of all programs to be used upon the initial call. After
> that,
> > all program to program calls are generally faster. There are lots
> of
> > caveats to the above, depending on your usage and how activation
> groups
> > are used, but generally, ILE should be faster and provide a lot more
>
> > functionality built in.
> >
> > Memory is always subjective, but yes, if you run a lot more ILE, you
>
> > will need a larger working size for each client; i.e. more memory.
> The
> > translator (from MI to IMPI in your case) is much improved, and
> > equivalent programs should be faster in ILE. This is especially
> true
> > for ILE/C to EPM/C. Haven't played much with the ILE/RPG, but I
> would
> > imagine similar results.
>

> At my company we have a huge order entry program, which during
> compilation
> produces a 900+ page spoolfile. When the spoolfile is generated, it
> takes a
> while before the compilation is finished. I assume that is the actual
> program building. In OPM that takes approx. 10 minutes, but if I
> convert
> the program to ILE, the compilation takes about the same time, but the
>
> building of the program takes a couple of hours!!! Why is that?
>