On Sep 7, 2011, at 15:52, Glenn Knickerbocker wrote:
> On 09/07/2011 01:38 PM, Paul Gilmartin wrote:
>> How might I do this with other stages? I suppose I haven't sufficient
>> stamina.
>
> No other stages, just one huge FANINANY with at least as many input
> streams as you might ever need:
> ...
> Then just connect each new pipeline to the next available output. And,
> actually, when you run out, you can use this exact same routine again to
> add another big FANINANY.
>
It's sneakier than that. The program:
WARP2 REXX M1 V 133 Trunc=133 Size=15 Line=3 Col=1 Alt=0
* * * Top of File * * *
/* Rexx */ signal on novalue; /* =================================== */
trace Err
'addpipe warp FOO | > WARP LOG A'
'addpipe literal +1 | delay | warp FOO' /* ??? */
'callpipe literal +1.0 do other stuff for a while. | delay | console'
do K = 1 until RC<>0
if right( K, 3 )=0 then do
'callpipe literal +0.050000' K '| delay | console'
end
'addpipe literal' K '| warp FOO'; end
say 'exiting with K =' K'; RC =' RC
/*=================================================================== */
* * * End of File * * *
was still going strong after a quarter million iterations, an unreasonable
numbe of connectors to FANINANY. With proper care, WARP seems to be
serially reusable.
But if I omit the CALLPIPE DELAY which seems to give it breathing
room, it suffocates by resource exhaustion after a few thousand.
-- gil
