linux-mips-fnet
[Top] [All Lists]

Re: Booting Linux on Vadem Clio (really Vr41xx identity)

To: "Gleb O. Raiko" <raiko@niisi.msk.ru>
Subject: Re: Booting Linux on Vadem Clio (really Vr41xx identity)
From: Ralf Baechle <ralf@uni-koblenz.de>
Date: Tue, 20 Jul 1999 00:13:05 +0200
Cc: Dominic Sweetman <dom@algor.co.uk>, "Bradley D. LaRonde" <brad@ltc.com>, linuxce-devel@linuxce.org, linux-mips@fnet.fr
In-reply-to: <37931E61.794228D4@niisi.msk.ru>; from Gleb O. Raiko on Mon, Jul 19, 1999 at 04:47:29PM +0400
References: <017301bed191$bcdb3be0$b1119526@tecra.ltc.com> <199907191119.MAA29895@mudchute.algor.co.uk> <37931E61.794228D4@niisi.msk.ru>
On Mon, Jul 19, 1999 at 04:47:29PM +0400, Gleb O. Raiko wrote:

> Dominic Sweetman wrote:
> > I know it fails to provide the (core MIPS-3) instructions LL/SC,
> > because it's designers erroneously thought they were only useful for
> > multiprocessor systems.  Otherwise, it should be pretty compatible.
> > 
> 
> Fortunately, Harald Koerfgen has implemented ll/sc for r3k. Thus, for
> the user mode, they are already handled (iirelevant from CPU type, I
> guess). In the kernel, you need to choose r3k logic for ops on
> semaphores.

That's last week's state of the art.  A newer trick is to rely on the fact
that the exception handlers will leave k0 set to a non-zero value which
allows an implementation that is fast on all MIPS systems.  This will
of course only work for uniprocessors.  The stage beyond that is to
combine this with code patching at runtime.

  Ralf

<Prev in Thread] Current Thread [Next in Thread>