Re: Processor requirements

[ness]

ness wrote:
> Here is my position:
> For the beginning, it is OK to only support i686+. Running coyotos in 
> qemu is a need.
> But i think, at least in future, coyotos should run on early hardware. 
> This hasn't to be absolutely efficient. Who wants a high speed system 
> will not try old hardware. But there are people (like me) that don't 
> have much money (I'm working on a 600€ box here, and it's the best I 
> ever had. I'm not going to experiment with it. I got a really old box 
> for doing so. btw, I bought new mouse, too, but my monitors and 
> non-primary graphic cards are 10+ years old.).

I missed sth. here. What I actually wanted to say is that that coyotos 
wouldn't've run on the box I had a year ago.

> And often old hardware is 
> still sufficient for simple jobs, like a local dhcp server.
>
> Jonathan S. Shapiro wrote:
>
> > I'm currently settling the cross-build tools for Coyotos, and I'ld like
> > some input on something.
> >
> > In EROS, I decided early *not* to support the i386. EROS required i486
> > or better. This is because the i386 does not honor the write-protect bit
> > in supervisor mode, and we need this for efficiency in the IPC
> > implementation.
> >
> > Given that it is now 2005, does anybody see a need today to support
> > processors earlier than i686 in desktop PCs?
> >
> > The specific features I am looking for are the "page size
> > extensions" (4M pages) and the per-page global bit. I *think* that both
> > of these were actually present in the appendix H extensions for the
> > Pentium, but I'm not certain about PSE. I have appendix H in a box
> > somewhere, but I'm damned if I can locate it at the moment.
> >
> > I am aware that embedded processors like Geode don't all do PSE. We need
> > to support those, but the question I'm trying to ask right now has to do
> > with the minimum level of PC motherboard that the Hurd group wants to
> > support.
> >
> > shap

--
-ness-