Well, when thinking about it again I'm not sure it will be that simple. All boards including several armv5 ones have different kernel default configs so apparently there are some compile time differences. I'm not sure if the instruction set is somehow part of the configuration. We'll simply have to play around and see how it goes (if we get to that point). Andreas -----Original Message----- From: Brian Smith [mailto:bls2129 at columbia.edu] Sent: den 15 april 2009 18:40 To: Andreas Nilsson Cc: android-virt at lists.cs.columbia.edu Subject: Re: armv5 vs armv6 Andreas, That's true, I think we would also need to do something in terms of tricking the guest OS into thinking it's on an armv5 CPU even when its on an armv6. This would prevent anyone who dynamically checks what they can do from thinking they have more capabilities than we implemented. That seems simple enough, we just need to change what is returned by CP15_r0. I don't know if Linux only relies on compile time decisions or if it does stuff at runtime to determine what it can use, for example what layout of page table to use. I think it's all done at compile time, so this is may not be an issue for at least Linux. Brian Andreas Nilsson wrote: > Hi, > > I just read the "The ARM Architecture Version 6 (ARMv6)" whitepaper. It > seems all new v6 features can be toggled on/off so my guess is that from > an ISA point of view a working emulator solution should work on the > physical device. However there are several performance improvements > (such as ASID) to be used on armv6. > > Quoting the paper: > > "ARMv6 maintains 100% backward compatibility at the binary level for > operating > systems and applications." > > /Andreas > > _______________________________________________ > Android-virt mailing list > Android-virt at lists.cs.columbia.edu > https://lists.cs.columbia.edu/cucslists/listinfo/android-virt > >