The overhead compared to the EXECIO would be greater, but it would likely be less than the EXECIO plus the REXX loop if you can do all of the processing in the pipe.
Regards, Richard Schuh > -----Original Message----- > From: The IBM z/VM Operating System > [mailto:[EMAIL PROTECTED] On Behalf Of Wakser, David > Sent: Thursday, August 07, 2008 11:51 AM > To: [email protected] > Subject: Re: VM Read problem > > It would seem to me that that you are correct in that Pipes > would make the code a lot nicer - but I would assume the > overhead has to be greater. So what is it buying in this instance? > > I'll look at the EXECIO and see what I can find. > > David Wakser > > -----Original Message----- > From: The IBM z/VM Operating System > [mailto:[EMAIL PROTECTED] On Behalf Of Bruce Hayden > Sent: Thursday, August 07, 2008 2:47 PM > To: [email protected] > Subject: Re: VM Read problem > > It would go on your EXECIO command - that is the one getting > the output from CP. I don't know if this is really your > problem or not, since I didn't try to make sense of your > code, but it is a "gotcha" > anyway. The reason I mention CMS Pipelines is because it > will automatically reissue a CP query command when the buffer > fills up and use a large enough buffer. The best way to use > Pipelines is to redesign your exec and put the logic in your > Do loop into the pipeline. But, a small first step would be > to replace the EXECIO command with "PIPE CP Q RDR * ALL | stack" > the next improvement is to drop the header in the pipe: > "PIPE CP Q RDR * ALL | drop 1 | stack" > and so forth until you've eliminated the loop and done all > the work in the pipe.. > > On Thu, Aug 7, 2008 at 2:34 PM, Wakser, David > <[EMAIL PROTECTED]> wrote: > > Bruce: > > > > RECEIVE has no BUFFER option - so I am confused as > to what you > > mean. > > > > David Wakser > > > > -----Original Message----- > > From: The IBM z/VM Operating System > [mailto:[EMAIL PROTECTED] > > On Behalf Of Bruce Hayden > > Sent: Thursday, August 07, 2008 1:15 PM > > To: [email protected] > > Subject: Re: VM Read problem > > > > Well, 100 reader files at about 80 chars per line plus the > header, is > > 8080 bytes, just short of the default 8192 default buffer for > > returning the output from a CP command. Maybe you need to > specify a > > bigger buffer using the BUFFER option? And, of course, CMS > Pipelines > > would be a much better solution... > > > > > > -- > Bruce Hayden > Linux on System z Advanced Technical Support IBM, Endicott, NY >
