qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] Open source qemu x86 accelerator module.


From: Herbert Poetzl
Subject: Re: [Qemu-devel] Open source qemu x86 accelerator module.
Date: Sun, 3 Apr 2005 20:58:53 +0200
User-agent: Mutt/1.5.6i

On Sun, Apr 03, 2005 at 12:22:16AM +0100, Paul Brook wrote:
> I'd like to announce qvm86; an open source x86 accelerator module for qemu. 
> 
> qvm86 is basically a drop-in replacement for kqemu. However is is an 
> independent replacement written from scratch by myself, and released under 
> the GnNU General Public Licence (GPL).
> 
> The CVS repository and mailing list for qvm86 are hosted on savannah:
> http://www.nongnu.org/qvm86/

excellent work!

> qvm86 is still in the early stages of development, so I provide no guarantees 
> about its reliability. It almost certainly has bugs which could crash your 
> machine or cause data corruption. Any assistance identifying and fixing these 
> bugs is welcomed ;-)

is an ioctl interface really the best choice here?
what about a new syscall instead?

I know that might need some changes to qemu too, but
I guess it might increase performance ...
(i.e. lower overhead)

best,
Herbert

> See the README file in qvm86 CVS for brief instructions how to install. It's 
> basically the same procedure as kqemu.
> 
> qvm86 currently only works on x86-linux hosts, and has only been tested on 
> relatively recent 2.6 kernels with udev installed.
> 
> It should be possible to port qvm86 to other x86 hosts (eg. FreeBSD or even 
> Windows). I don't have immediate plans for doing such ports, but would 
> welcome patches if other people want to do the porting. It may also be 
> possible to x86-64 hosts, but I don't know enough details to say how hard 
> this would be.
> 
> I have successfully booted windows 2000, freebsd and couple of different 
> linux 
> guests. Windows 98 guests don't work, but I think I know what's wrong.
> 
> Performance depends heavily on the guest wokload. It varies from near-native 
> to about the same as normal qemu. For example compiling qemu with gcc is ~5x 
> slower than native (normal qemu is ~17x slower). I believe there is sill 
> quite a bit of scope for improving performance.
> 
> Paul Brook
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/qemu-devel




reply via email to

[Prev in Thread] Current Thread [Next in Thread]