Re: [Xen-ia64-devel] Time for hybrid virtualization?

[Isaku Yamahata <yamahata@xxxxxxxxxxxxx>]

On Wed, Jan 16, 2008 at 05:02:07PM -0700, Alex Williamson wrote:
>    Is it possible?  Is it a good idea?  What are some of the issues?  We
> would lose support for non-VT capable processors (pre-Montecito), but is
> that so bad?  Is it a "fast track" to upstream Linux Xen/ia64 support?
> Let me know your thoughts.  Thanks,

I think hybrid virtualization is quite possible.
Some (related or unrelated) random thoughts:

- In x86 case, paravirt_ops has already developped. It was the key.
  I'm not sure about ia64 case, especially hybrid-virtualization is
  "fast track".
  Presumably we need more experiance in this area. I guess hybrid
  virtualization may require efforts. I'm not sure how big the effort is.

- I guess the overhead difference from x86 case roots from shadow page
  table or shadow TLB. Does the figure ia64 shadow TLB works quite well?
  Does the figure mean that ia64 HVM case is quite well or IA64 PV case
  can be better? (or both?)

- I'm expecting that PV case overhead can be reduced by using fast hypercall.
  It can eliminate several cycles from hyperprivops path.
  I don't have any figures how much improved by fast hyperprivops,
  but the VT-i assisted way would be faster than PV case in principle.
  (Otherwise HW engeneers try to make it faster.)
  Anyway the question here is 'is it worth while or is
  hybrid-virtualization better?'

- Some kind of hybridness is inevitable in the future, I think.

-- 
yamahata

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel