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

RE: [Xen-users] 3D acceleration question (again)



 

> -----Original Message-----
> From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx 
> [mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of 
> Andre.Weidemann@xxxxxx
> Sent: 08 December 2006 14:43
> To: xen-users@xxxxxxxxxxxxxxxxxxx
> Subject: [Xen-users] 3D acceleration question (again)
> 
> Hi,
> I know this subject has been discussed several times here on the list 
> before, but I'd like to bring it up again.
> 
>  From what I have read in the Wiki's FAQ and the list here, an IOMMU 
> would cure the problem of translating the IO-Addresses for 
> accessing the 
> GFX card directly.
> 
>  From what I have read, the Pacifica architecture inside the AMD 
> Athlon64s respectively the CPU's northbridge provides an IOMMU.
> Is this correct?

Not quite. There is a possibility to use the GART (Graphics Aperture 
Redirection Table(??)) to translate one PCI (including PCI, PCI-x, PCI-e and 
AGP[248]x) address to another physical memory address - this is used in the 
Linux kernel to do IOMMU for memory buffers above 4GB to hardware that isn't 
64-bit addressing. 

I think Mark Langsdorf at AMD was working on a xenified version of this, but I 
don't know what happened to that work. Unforutnately, the GART only translates 
for ONE GUEST AT A TIME, so if you have a multiprocessor (or even multicore) 
system, you can't have two guests use the GART translation (no problem in this 
application, you'd probably only _WANT_ give the graphics card to ONE guest 
anyways) which limits it's usefulness. 

There are plans for a FULL IOMMU (actually, I'd expect this to be more than a 
plan at this stage, but I don't know what the hardware guys are doing on this, 
and _IF_ I did know, I couldn't tell anyone anyways... ;-) )


> Could this be used for the previously discussed address translations?
> If so, are there any plans yet on using this unit and could one get 
> direct access to the GFX card so full hardware acceleration 
> was possible?

As I said, there's been plans for it, but as of right now I don't know if it's 
realistic to expect this to happen. 

--
Mats
> 
> 
> Thank you in advance.
>   André
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-users
> 
> 
> 



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.