On Mon, 25 Mar 2013 11:19:47 -0500, John Gilmore wrote: >If, as seems likely, you are going to read a significant number of >members from the same PDS[E], then the DESERV macro is likely to yield >better performance (and a simpler program) than BLDL either above or >under the covers. This would also be the case if you had occasion to >read the same member more than once. > But be careful. D.R. is likely designing a library of callable services. While I counseled the prudent use of STOW DISC to avoid dangling references, "Using Data Sets" also states that after STOW DISC another BLDL/DESERV/FIND is necessary to re-open the member. This may tax the intended audience's awareness of z/OS internals. In fact, his first posting implied that D.R. planned to do BLDL or find singly, not en masse.
>I also have some measurements that suggest (without really >establishing in a definitive way) that DESERV is better at dealing >with a significant concatenation of a number of libraries than BLDL; >but I would be more enthusiastic about this result if I had some >inkling of why it is the case. > Again, the apparent deficiency of DESERV with respect to mixed concatenations (U.D.S mentions only PDS and PDSE), may be a show stopper. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
