----- Original Message ----- From: "Daniel Reinhold" <danielr@xxxxxxxxxxxxx> To: <openbeos@xxxxxxxxxxxxx> Sent: Friday, August 16, 2002 12:20 PM Subject: [openbeos] Re: undocumented syscalls -- NewOS scheduler doc > Here's another piece of useful info I forgot about. > > A gentleman named Brian Mastenbrook did an analysis of the NewOS > scheduler earlier this year and posted his notes online: > > http://acm.roosevelt.edu/~chandler/articles/newos/NewOS_Scheduler.html > > This is very good reading and will help you get a handle on the basic > workings of the scheduler. But beware that it is slightly out of date! :) The code in that page handled RT and TS threads in the same fashion, this is no longer the case. manuel, > > > >The scheduler quantum is the maximum time slice each thread in the > list > >of running threads is allowed. That is, the scheduler will allow each > >to run up to XXX ms before switching to the next. They don't *have* to > >use that much time -- a thread could reschedule itself earlier. > >Undoubtedly, that's what microsleep() does. > > > >The amount of time alloted to each running thread does not represent > >the timing granularity of the system at all. I think for the BeOS, the > >smallest time unit that is guaranteed to be honored is 1 us (one > >microsecond). But I'm not totally sure about that -- I'm just > recalling > >from memory. > >