On 2016-09-02 18:25, Paul Gilmartin wrote: > o Strange. It avoids SORTIN and SORTOUT, using the E15 and E35 > exits instead. > > o Even stranger. The user must offset key positions by 4 bytes to > account for RDWs. I'd expect DFSORT to handle this internally. > Does it just regard the RDW as data? > > I'm trying to imagine all the ways a delaying stage might cause > a stall. Can one run two DFSORT stages in parallel?
You seem to do fine answering your own questions, and most of this is from before I could write, there are days that I don't use DFSORT/CMS, but my 0.02 on this... When DFSORT is using CMS QSAM for its input and output, the data must be where CMS supports it. I'm not aware of ways to use data in memory, so we would end up using temporary disk files. Something you could do already. CMS Pipelines does not parse the parameter string for DFSORT; that would require it to keep track of all options. So we can't tell which fields to tweak to compensate for the offset. Don't think you can have multiple dfsort stages active at the same time. might imply that you could have multiple dfsort stages active. Sir Rob the Plumber
