On Wed, Jun 25, 2008 at 5:41 PM, SPITZ, HOBART CTR DFAS <[EMAIL PROTECTED]> wrote:
> Has anyone recently quantified the performance benefits of ADDRESS > COMMAND vs. ADDRESS CMS, or the COMMAND stage vs. the CMS stage? We I'd be surprised to see big potential there. The cost of running CMS commands would be in what that does, not how expensive it is to invoke it (if the command is so trivial that the cost of invocation is relevant, I would question the modularity of the program design). If you talk about compiled REXX programs that get invoked, then there may be a lot of low hanging fruit in rethinking those. The REXX applications that I have worked with in the past would gain a factor when being compiled, and order of magnitude when done in proper plumbing. :xmp. Back then we had a REXX program that would scan our RSCS CONFIG for a particular LUNAME (to find whether we needed RSCS changes when workstations were changed). It had a loop with EXECIO plus some REXX logic and was pretty slow. My quick & dirty pipeline with a few locates was so much faster that the original author was convinced immediately. When the amount of change grew and we sometimes had 100 names to be checked, that needed to be automated too. Clearly he used the fast program - in a REXX loop to invoke it once for each name to be checked...:-( The replacement (reading two files and using lookup) was so fast that people did not believe it actually did what it should do. :exmp. -Rob
