[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Memory allocation jump after running for a while with a largenumber of threads



Ulrich Drepper wrote:
Hui Huang wrote:


They aren't from JVM, as we mmap with rwx- (for heap, generated code,
etc) and/or ---p (for guard pages; no, we don't use 992KB guards).
Chances are they either came from kernel or libc.


Just to follow up on this, I cannot see such anomalies when writing a C
program and limiting the stack to a hew kB.  I get the expected stack
and guard size.


As mentioned in Hrvoje Niksic's email, the problem shows after days into the run. That looks to me like a timing related bug and I doubt a simple test (either C or Java) can reproduce the problem in just a few runs.

What are the attributed the JVM code uses when creating threads?


pthread_attr_setdetachstate(PTHREAD_CREATE_DETACHED) and pthread_attr_setstacksize(128K), that's it.

thanks,
-hui





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]