Hello Racketeers,
it didn't take long before I hit another strange behavior when
extensively using futures. The setup is relatively simple: I am
processing a large fxvector/flvector and I am processing it on a
line-by-line basis (the vectors themselves are width*height in nature).
To avoid races I setup a (make-vector height #f) to serve as a vector of
locks. Locking is simple:
(let lloop ()
(when (not (vector-cas! locks y #f #t))
(lloop)))
.. do something ...
(vector-set! locks y #f)
Of course, the code in question is in 8 futures on my CPU and they are
scheduled in parallel pretty consistently.
Most of the time it works fine, but sometimes it just hangs with
multiple CPU cores being utilized at 100% (I assume they are in the
busy-wait "lloop").
The problem is I cannot isolate a minimal working example from that. The
code in question is about 3000 LoC and the issue expresses itself only
in about 1 in 30 times just after the program has started. The GUI
thread is running and there are lots of flonum operations and lots of
thrown away flonum results (meaning GC gets triggered quite often).
Also I cannot trigger the same issue with unsafe-vector*-cas! - I'll
look into the differences in the source.
I see this with latest Racket 3m.
Any suggestions how to approach finding the source of the problem are -
as always - very welcome.
Cheers,
Dominik
--
You received this message because you are subscribed to the Google Groups
"Racket Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to racket-users+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/racket-users/0e6de88c-1d39-38fa-ea69-2043c4f4071b%40trustica.cz.
