Doesn't the chipset prevent SMP cpu's from accessing MMIO's at the same
time?
I mean it's a single headed environment.. ofcourse latency is an issue
though.
--- Jim
[All those programs will be lost.... like bits in the stream...time to
crash.]
[ --- Modernization of Rutger Haurs last speech in BladeRunner
]
> -----Original Message-----
> From: Niklas H�glund [SMTP:[EMAIL PROTECTED]]
> Sent: Monday, October 04, 1999 5:46 PM
> To: [EMAIL PROTECTED]
> Subject: Re: Synchronized MMIO access on SMP
>
> On Mon, Oct 04, 1999 at 10:26:29AM -0400, James Simmons wrote:
> > Good point. Have to look at a way to handle this. I have started to
> > rewrite the schedular to handle this. I realized this is not a problem
> for
> > just video cards but really any device that has MMIO. On a SMP machine
> you
> > could have two processors access the MMIO at the same time causing the
> > card to go into a indetermined state. Also with some cards they might
> have
> > more than one MMIO area and each MMIO area thats accessed can effect the
> > card in such a way that the results you expected from another MMIO
> > region might not be what you want. Well I going to take this to the MM
> > guys to figure out away to handle this.
>
> I guess most other hardware are only accessed from kernel device
> drivers that have internal locking.
>
> --
> Niklas
