I, myself, have never been against a good clean vga emulation..., I was just
(more than anything else) thinking about how we could _just_get_it_to_work_
with one instance only--and then start doing the more daunting task of
emulation/virtualization of a video device after we know that the thing
seems to work.
I agree with Kevin.., get the important CPU stuff out of the way before
moving to the important VGA stuff.... I must say though that I like some of
the ideas being tossed around here. I would just like to throw out the
general reminder to everyone: "The best code its that which runs the fastest
while using the simplest solution model possible and the fewest resources."
I think that this has been best expressed by the push for the most basic vga
possible, but let's wait until we know what we are writing it to run upon.
Drew Northup, N1XIM
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf
> Of Kenneth C. Arnold
> Sent: Tuesday, October 31, 2000 9:19 PM
> To: [EMAIL PROTECTED]
> Subject: Re: plex86 video proposal
>
>
> On Tue, Oct 31, 2000 at 07:52:14PM -0600, Nelson Rush wrote:
> > Speaking of X Windows, here is an interesting idea...
> >
> > Instead of trying to map the video card to the guest operating system or
> > hack at it some other way you could just run X Windows and xhost the
> > localhost on the guest. Install an X server on the host
> operating system and
> > then you trap the guest X Window messages over a loopback and
> display it on
> > the host. Surely this would be a better method, and much
> faster. So, to run
> > MS Windows as the guest you'd need to install an X server on
> it. Maybe that
> > defeats the purpose, I don't know. But, IMHO Plex86 shouldn't
> have to worry
> > about the video card. Of course, there is also VNC which is
> based off of the
> > X Windows protocol and I think it gets better mileage.
> >
> > http://www.uk.research.att.com/vnc/
> >
> > If you're going to decide to map the video card memory instead, there is
> > going to be very noticeable latency. And as you pointed out there are
> > disadvantages to virtualizing the VGA access.
>
> VNC and X Windows are indeed good ideas for optimized video access
> (though AFAIK SDL has more hardware-acceleration support), but the user
> will (probably) want to see things that go on before the special driver
> is loaded. Think about how VMWare does it. Video _always_ works;
> the VMWare
> special display drivers just make it faster by direct pipes similar to
> what you describe.
>
> Speaking of VMWare, has anyone realized that all the kernel-level code
> for VMWare is (and has to be) open-source? And it's quite well documented
> internally, also ... could any of the real hackers here get ideas from
> that code?
>
> [snip quoted message]
>
> --
> Kenneth Arnold <[EMAIL PROTECTED]> / kcarnold / Linux user #180115
> http://arnoldnet.net/~kcarnold/
>
>
