[Qemu-devel] Re: using mmap?

[Anthony Liguori <aliguori@linux.vnet.ibm.com>]

Mark Williamson wrote:
>> I'm also doubtful how much benefit it gave in practice. I'm sure it would
>> be good for synthetic CPU benchmarks. However using mmap significantly
>> increases the overhead of context switches/tlb misses.
>>
>> To get good overall performance I suspect you're going to need closer
>> cooperation with the kernel than mmap gives you. If you really want to make
>> cross-emulation go fast I suggest working with the xen and/or kvm people to
>> integrate qemu dynamic translation into those products. In theory I'd guess
>> you should be able to plug it in as an alternative to the HVM code. I've no
>> idea how close that is to being practical.
> 
> http://wiki.xensource.com/xenwiki/HVM/V2E
> 
> The v2e stuff allows execution state to be extracted from the real CPU, 
> plugged into QEmu for a bit for emulation, then transferred back to the real 
> CPU again.  This is initially to be used for supporting Big Real Mode 
> emulation on HVM platforms.  Later on it's planned to be used to accelerate 
> IO emulation further.
> 
> Eventually this may provide a means to use QEmu's translator to execute kernel 
> code whilst running user mode code under Xen.  It may be that this isn't as 
> fast as other approaches, but it'd be a useful feature for Xen to have IMO.

I'd absolute like to get there.  The current Xen HVM code is a bit of a 
mess at the moment.  There are far too many assumptions about having 
VT/SVM hardware present.  However, I'd really like to get to a point 
where Xen could run unmodified guests on non VT/SVM hardware using qemu 
with user virtualization.

KVM is having similar trouble ATM with big real mode.  It would be very 
nice to also add a V2E like interface to KVM.

I'd also like to see the translator even used for paravirtual guests. 
This way we could still have a proper boot loader run without having to 
deal with paravirtualizing it.  It would be nice to use emulation to 
avoid paravirtualizing the non-performance critical bits.

Regards,

Anthony Liguori

> Cheers,
> Mark
> 
>>> Wouldn't this be a *significant*
>>> performance enhancement for this setup which I'm sure is a common one?
>>> Maybe this can be implemented for regular processes on the guest and
>>> only use the softmmu for the kernel?  Would someone point me in the
>>> right direction for technical information?  I have had to switch to
>>> vmware free player until Qemu+KQEMU reaches a point of similar
>>> performance, but I would really rather see Qemu advance further.
>> If you're using an accelerator (eg. kqemu or kvm) this is all irelevant as
>> most code isn't run by qemu, it's virtualized by the accelerator. qemu just
>> does the IO emulation.
>>
>> Paul
>>
>>
>> _______________________________________________
>> Qemu-devel mailing list
>> Qemu-devel@nongnu.org
>> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>