| |
|
 |
|
 |
|
|
|
| |
|
|
xen-users
Re: [Xen-users] Xen 3.1 - 32 vs 64 bit hypervisor
On Fri, 2007-06-08 at 17:38 +0100, Mark Williamson wrote:
> 32-bit PV guests potentially are actually faster on 64-bit than 64-bit PV
> guests. Due to <more arcane x86isms> I believe the system call interface can
> be faster for a 32-bit PV guest than a 64-bit guest. So conceivably it's
> more efficient to run a 32-bit dom0 (maybe it won't *actually* matter, I
> don't know).
I'm now thoroughly confused. You mean, if you could run a 64 bit guest /
32 bit hypervisor, right?
> [1] Actually, it's very simple as long as you don't expect it to sound
> remotely reasonable. It *is* x86 after all :-) Xen wants to be in the top
> part of the address space. 32-bit glibc wants to mess with the top of the
> address space in order to implement TLS. On 32-bit hosts this conflicts. On
> a 64-bit host, Xen stashes itself at the top of the virtual address space;
> because it's a 64-bit host this is waaaaaaaaaaaay higher than a 32-bit guest
> is capable of addressing. Therefore, it's fine to let glibc play with the
> top of the 32-bit space - it just can't get at Xen :-)
That's why I'm wondering what you meant above, or does this completely
change if you run a 32 bit hypervisor with a 64 bit guest?
So if a 64 bit tree fell in 32 bit address space ... , oh sod it never
mind. Just tell us when it works and we'll say "Thank you."
Best,
--Tim
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
|
|
| |
|
Home