To Re-compile or Not to Re-complie



Support for OS/VS COBOL, VS COBOL II, COBOL for OS/390 & VM and Enterprise COBOL for z/OS

To Re-compile or Not to Re-complie

Postby The Butcher » Sat Mar 19, 2011 1:35 am

Hello,

My site at work has a newer compiler since 2007
I was checking the DB2 packages of some heavily used subroutines that the language type column of these packages
shows a mix bags of '2' and '3'.

When the value is '2' its description is "VS COBOL II or IBM COBOL REL1"
When the value is '3' its description is "IBM COBOL REL2+"

Is there a benefit or performance gain to re-compile Cobol source from VS Cobol II to Cobol REL2 ?
Some of those source modules has last precompile timestamp of 1998.
Any danger also to use the OPT(FULL) when re-compile these old programs>

Thanks in advance for any suggestions.
Best Regards,
The Butcher
The Butcher
 
Posts: 3
Joined: Thu Mar 17, 2011 6:52 pm
Has thanked: 0 time
Been thanked: 0 time

Re: To Re-compile or Not to Re-complie

 

Re: To Re-compile or Not to Re-complie

Postby BillyBoyo » Sun Mar 20, 2011 11:27 pm

The Butcher wrote:Is there a benefit or performance gain to re-compile Cobol source from VS Cobol II to Cobol REL2 ?


One thing I can suggest is to look at the Release Notes, carefully. These would summarise all changes and point to anything with potential performance improvements. You can also read up exactly on your option and then search in details on the Internet. If there are any problems I'm sure you'll find texts dealing with them

This is far from the end of the story. I guess you know that you can't just recompile everything and stick it in production? It is one of those situations where the testing effort far outweighs the development effort.

Unless you are all twiddling your thumbs, I don't think management would buy the change-it-all-at-once approach. If performance is a problem, pick on the program you think is likely to be the worst culprit, and test it against the original. Will give you some idea of what you are looking at as far as the recompile is concerned. "If it ain't broke, don't fix it" is often not a bad start when looking at such things. So, is it broke (performance)?
BillyBoyo
Global moderator
 
Posts: 3804
Joined: Tue Jan 25, 2011 12:02 am
Has thanked: 22 times
Been thanked: 264 times

Re: To Re-compile or Not to Re-complie

Postby The Butcher » Mon Mar 21, 2011 6:47 am

Hello BillyBoyo,

There is an objective in the horizon to lower MIPS that the company is paying each month. These sub programs get called around 60,000 to 70,000 times each per week. No it is not broken but if I can remove a single perform statement I believe, sorry I hope that it may improve things even though if it may seems not worthy. May be I am barking up the wrong tree. Your comments are welcome and appreciated.
The Butcher
 
Posts: 3
Joined: Thu Mar 17, 2011 6:52 pm
Has thanked: 0 time
Been thanked: 0 time

Re: To Re-compile or Not to Re-complie

Postby BillyBoyo » Mon Mar 21, 2011 1:09 pm

So pick on the one likely to be the most greedy and try it out. Make a test program which calls the original sub-routine (I know you're not changing it, but calls the old non-recompiled one) some number of times (like 70,000). Then clone the test program and do it with the re-compile and whatever else you are considering. Run it the same number of times and compare the CPU. Will give you an idea of the size of the savings.

Do all the background work already mentioned as well.

I have to say that 70,000 usage per week does not sound like a lot. Even when multiplied by the number of sub-routines. Try it out and see. Let us know the results, it sounds interesting.
BillyBoyo
Global moderator
 
Posts: 3804
Joined: Tue Jan 25, 2011 12:02 am
Has thanked: 22 times
Been thanked: 264 times


Return to IBM Cobol

 


  • Related topics
    Replies
    Views
    Last post