On Sat, 10 May 2014 15:03:11 -0500, John McKown wrote:
>
>Go Debian!
>It installs from source.
>
Thereby validating it.  VM (used to be?) that way.

On Sat, 10 May 2014 22:38:01 -0500, Ed Gould wrote:
>
>I had a few occurrences of looking at code that way. After looking at
>a IFCEREP1 module I found a really bad code sequence (causing cpu
>usage to sky rocket when running a specific report) The damn module
>did a sequential  table look up that was absolutely terrible. It
>could have been handled with very few instructions with a load and a
>shift left double. For each device it looped through a table.
>Something a rookie would have done. I tried to apar it and they
>refused to take it as a performance issue. Fast forward 20+ years and
>the damn module still sucks up cpu  time. No one cares.
> 
I once did likewise, wondering why CLIST was so slow.  A colleague's
execution profiling utility located the culprit: the character fetch
routine in lexical analysis (JWG knows the principle well):
BALR; STM; GETMAIN; chain; IC; STC; unchain; LM; FREEMAIN;
BR14 (roughly, from memory) for each character!  An egregious
misuse of coding standards.

>On May 10, 2014, at 6:00 PM, Shane Ginnane wrote:
>
>> I do however recall sitting down at a (very old) fiche reader in
>> the computer room at one time looking through the source for
>> IEBCOPY. Hard way to read code.
>> Can't remember if we had an actual problem, of it was just a "I
>> wonder" moment.

-- gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to