Date: Fri, 25 Jan 2002 22:42:49 -0700 From: Nate Williams <nate@yogotech.com> To: Daniel Eischen <eischen@pcnet1.pcnet.com> Cc: Nate Williams <nate@yogotech.com>, Terry Lambert <tlambert2@mindspring.com>, Dan Eischen <eischen@vigrid.com>, k Macy <kip_macy@yahoo.com>, Peter Wemm <peter@wemm.org>, Julian Elischer <julian@vicor-nb.com>, arch@FreeBSD.ORG Subject: Re: KSE question Message-ID: <15442.16857.446446.208426@caddis.yogotech.com> In-Reply-To: <Pine.SUN.3.91.1020125183209.10999A-100000@pcnet1.pcnet.com> References: <15441.59822.481182.325298@caddis.yogotech.com> <Pine.SUN.3.91.1020125183209.10999A-100000@pcnet1.pcnet.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> > > > > The kernel knows if the FPU has been used and it also knows > > > > > the format (x87 vs SSE/XMM). As long as the FPU context > > > > > comes from the kernel, then it can also tell us whether > > > > > it is valid and it's format. > > > > > > > > Right, but this has a huge effect on the userlands threads scheduler, > > > > since multiple threads can be active during one time-slice, so the > > > > userland scheduler will have no way of knowing which thread used the > > > > FPU. (At least, not w/out making a system call, defeating most of the > > > > advantages of having userland threads...) > > > > > > I think it only matters when threads are preempted or trap. > > > > With KSE's, won't threads be pre-empted? (I guess they won't unless you > > get an upcall from the kernel, so the flag could get set.) > > I'm not sure, but if you resume a preempted thread without > running any other threads or signal handlers in between, then > you don't have to restore the FPU state anyways. Possibly. > It'll still > be in the pcb and the next trap will load it into the FPU. > Unless the thread is allowed to migrate between KSEs... Eggsactly. :) > > > Can't we treat normal system calls as we would library routines (if > > > you call a function then shouldn't the compiler assume the FPU state > > > could be trashed and be forced to save and restore FPU state that it > > > needs?). > > > > This is much less effecient, since the kernel normally doesn't touch the > > FPU state. (FPU operations in the kernel are illegal currently.) > > A system call looks the same way to an application as any other > library routine doesn't it? Usually, but not always. But, it causes the application to give up it's current scheduling slot, which isn't optimum. > The compiler doesn't know that sigprocmask() is different than > scanf(). Right, but scanf doesn't require saving/restoring the FPU context, while system calls might. :( Nate To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?15442.16857.446446.208426>