On 03/13/2015 01:43 PM, Charles Mills wrote:
The classic approach is TRT, which is a little bit of a PITA (R2 anyone?) but
does the job. It is a single machine instruction but that does not mean it's
fast!
There is a new instruction (new relative to a lot of us LOL) "search string"
(as I recall) that is more straightforward and might be faster. Again, fundamentally a
loop whether you code it or the machine does it under the covers, so not as fast as you
might guess. Faster than a CLI/JE loop I would guess.
Watch out for the "no second qualifier" special case. FOOBAR is a valid dataset
name.
Charles
I'm an old ALC kinda guy, and I'm use to TRT.
I would build a TRT table that would hit on C'.', c' ', and c','
for processing working with qualifiers and DDNames. I might even
go so far as to hit on non-valid characters...
And I would set each of those to have a unique content so I could
use it to drive logic (branch table style) (c'.', C' ', C',', and
then all the invalids to use the same value).
TRT is an instruction whose use precludes the use of R2 as a base
register. But if you have gone to the Relative Branch (Jxx) type,
you probably don't care. However, if you are using R2 for data
access, beware!!
Regards,
Steve Thompson
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
