Hi Peter!

[..]
 PF> So,  for example,  creating a named pipe server program with 500
 PF> blocking mode threads (assuming most of them are not used most of the
 PF> time) adds no overhead to the scheduler.  
I found this problem: when 568 (or around) threads block on a single
semaphore, obviously the semaphore managing mechanism runs out of memory
and traps (ring 3). Afterwards, the process can't complete it's
exitlist-processing and becomes a zombie. Is there an APAR for that ?
     __
     |_)
cu/2 |atrick
[Team  OS/2]

---