Only yesterday I solved a similar problem with calling a cobol module from rexx. The cobol routine was called some 100,000 times with a 5 to 10 cs overhead (elapsed). That really hurt.
In my case the cobol was only a veneer to some existing extensive assembler code. So I rewrote the veneer in assembler (rather clumsily, I don't do much). The job now ran in under 8 minutes rather than being cancelled after 4+ hours (I gave up on the job). I profiled my rexx btw, it does rather a lot of work and the slowdown could have been anywhere. I write a lot of code in rexx and try to concentrate on good modular design, and only worry about performance when problems arise (known really poor practice avoided of course). Usually even surprisingly rexx is not the root cause, as was the case here. Anyway the I am nonetheless very interested in ceepipi. Writing assembler veneers is not a generic solution. I'd like to investigate calling ceepipi from rexx so I can consider calling COBOL from rexx without concern. Could you (Jan Moyersons in particular, re stored procedure) share your code for ideas? Has anybody else been down this route? Thanks -- Julian Levens Sent from my Android phone with K-9 Mail. Please excuse my brevity. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: INFO IBM-MAIN
