Magenheimer, Dan (HP Labs Fort Collins)
2006-Apr-28 16:26 UTC
[Xen-devel] Different instruction stream for PV OS on VT/Pacifica?
I was asked a question and I think I know the answer for Xen/ia64 but am not sure I know the answer for Xen/x86 -- or if there is a different answer for VT vs Pacifica. In the three scenarios: A. Paravirtualized (PV) OS running on a pre-HVM processor using a virtualization-capable Xen B. PV OS running on a VT (or Pacifica) processor using a virtualization-capable Xen but NOT running any fully virtualized (FV) OS C. PV OS running on a VT (or Pacifica) processor using a virtualization-capable Xen which is sharing the processor with a FV OS is the sequence of instructions executed by the processor on behalf of the PV OS -- including instructions in firmware and in Xen -- identical or different/faster/slower? (Ignoring irrelevant factors such as domain switch, non-virtualization- related processor differences, etc.) On Xen/ia64, a PV OS should always execute the same sequence of instructions, whether or not it is on a VT-capable processor. Can experts for the others (VT-x and Pacifica) comment? Thanks, Dan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Petersson, Mats
2006-Apr-28 16:49 UTC
RE: [Xen-devel] Different instruction stream for PV OS on VT/Pacifica?
> -----Original Message----- > From: xen-devel-bounces@lists.xensource.com > [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of > Magenheimer, Dan (HP Labs Fort Collins) > Sent: 28 April 2006 17:27 > To: xen-devel > Subject: [Xen-devel] Different instruction stream for PV OS > on VT/Pacifica? > > I was asked a question and I think I know the answer for > Xen/ia64 but am not sure I know the answer for Xen/x86 -- or > if there is a different answer for VT vs Pacifica. > > In the three scenarios: > A. Paravirtualized (PV) OS running on a pre-HVM processor > using a virtualization-capable Xen > B. PV OS running on a VT (or Pacifica) processor using a > virtualization-capable Xen but NOT running any fully > virtualized (FV) OS > C. PV OS running on a VT (or Pacifica) processor using a > virtualization-capable Xen which is sharing the processor > with a FV OS > > is the sequence of instructions executed by the processor on > behalf of the PV OS -- including instructions in firmware and > in Xen -- identical or different/faster/slower? (Ignoring > irrelevant factors such as domain switch, non-virtualization- > related processor differences, etc.) > > On Xen/ia64, a PV OS should always execute the same sequence > of instructions, whether or not it is on a VT-capable processor. > > Can experts for the others (VT-x and Pacifica) comment?I''d say that this answer applies to AMD processors (and as far as I know it applies to Intel too - but I''m not really a spokesperson for AMD, and even less so for Intel.) There''s been some thinking that we could use HVM (VT/SVM (pacifica)) to enhance PV guest too, but so far nothing has been implemented [and I guess once it does get implemented, IA64 will follow shortly afterwards, assuming it''s generally useful and not specific to x86]. -- Mats> > Thanks, > Dan > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel > >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel