Hi,
2009-06-15 20:44 UTC+0200 Viktor Szakats (harbour.01 syenar.hu)
* include/hbextern.ch
* source/vm/thread.c
+ Added HB_MUTEXWAITERSCOUNT() to core.
(slightly rewritten using local functions)
Such function by definition cannot return valid result so it should
be marked as helper function with intentional race condition inside.
I know that thread count could be changed between HB_MUTEXWAITERSCOUNT()
and actual HB_THREADSTART(), but I do not care a lot about it. But it
would be bad if this race condition can cause return garbage on some
systems. In this case it need to be protected by mutex. It could be
renamed to __MUTEX*() of couse, but I feel, that sometimes it could be
useful to have mutex information function. I found two things useful:
waiters count and queue length. Maybe we can create some more general
mutex protected function like:
HB_MUTEXQUEUEINFO( hMutex, [ @nWaitersCount ], [ @nQueueLength ] )
or another name similar name.
Przemek, I think your opinion is most valuable here. Or maybe you
propose a completely different problem solution. The problem is worker
threads count management: if too many workers are free, it would be good
to decrease workers count; if job queue length is too big, it would be
good to start new worker threads.
Regards,
Mindaugas
_______________________________________________
Harbour mailing list
Harbour@harbour-project.org
http://lists.harbour-project.org/mailman/listinfo/harbour
