On 8/8/05, Ben Cooper <[EMAIL PROTECTED]> wrote: > I thought it would be interesting to get these > running on Mac OS X/ > PPC since there's no Darwin makefile included in the > release.
Hey, this is really great. > So the tests compiled and ran, but as soon as I got > to the fork I got > resource temporarily unavailable messages (as in the > snippet below). > Now I would've expected this to be because of the > low(ish) maxproc > and maxprocperuid limit in OS X, so I bumped them up > to > > kern.maxproc=2048 > kern.maxprocperuid=512 > > with still no luck. Can someone tell me what this > setting is in > Solaris? I should post this on the Darwin mailing > lists but I thought > I'd share my experience with you guys first. <a href="http://cvs.opensolaris.org/source/xref/usr/src/uts/common/conf/param.c#param_init"; >param_init</a> and <a href="http://cvs.opensolaris.org/source/xref/usr/src/uts/common/conf/param.c#param_calc"; >param_calc</a> figure the global and per user limit on the max number of processes. As you can see, it depends on a few things like the max number of users, amount of memory, etc. On an x86 test machine i'm seeing: > v::print v_proc v_maxup v_proc = 0x207a v_maxup = 0x2075 > 0x207a=D 8314 > 0x2075=D 8309 So around 8300 for both. The "c_" tests deal with some number of processes/threads that are handing locks around in a ring. The "_10" tests should be creating 10 processes, the "_200" test should be creating 200, etc...so it's interesting that you're hitting what seems like should be a limit of 512 when the test is creating at most 200 processes.... I could see this happening if libMicro wasen't waiting for the processes from one test to terminate before starting the new test, or if there was some delay in the OS between the time a process exited and the time that it's resources are returned to the system... _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
