From the description of CSVQUERY, performance implications:
If you specify an address as a search argument for a module in the PLPA, the
search might take longer than if you specify a name because the PLPA is
organized by name. You can obtain the best performance on a CSVQUERY request by
specifying an entry point token.
Does this mean that CSVQUERY will be faster than scanning the LLE list? Does
anybody have practical experience with CSVQUERY?
-------- Original-Nachricht --------
Betreff: searching the LLE list
Datum: Fri, 19 Feb 2010 12:49:34 +0100
Von: Bernd Oppolzer <email@example.com>
An: IBM Mainframe Assembler List <ASSEMBLER-LIST@LISTSERV.UGA.EDU>
I have some problems with a vendors software that I'm not allowed to change.
This software walkes through the LLE list searching for a specific module and
does this very often, consuming lot of CPU, because in the specific situation,
the LLE list is very long (thousands of modules).
I got some suggestions from IBM-Main, but one question remains open:
is there a more efficient way to look up a specific entry in the LLE list? From
looking at various dumps, I had the impression that the LLE entries are sorted
by module name !!!
If during LOAD new LLE entries were inserted at the proper place, this would be
very expensive. So I wonder, if the sequential organized LLE list might be
super-imposed by another more sophisticated structure, which allows fast
insertion at the right place. If so, there should be new z/OS services to find
LLE entries by name faster using there new structures (maybe balanced trees or
anything like that). This would make perfect sense to me - keeping the
sequential list for compatibility, but improving performance by adding another
I have not enough time to scan all the z/OS manuals. Maybe someone of you could
give me a hint, if there is such a structure and if new services exist to avoid
scanning the LLE list sequentially?