[Qemu-devel] qemu-fast target removal ?

[Qemu-devel] qemu-fast target removal ?

[Fabrice Bellard]

Hi,

I plan to remove the 'qemu-fast' target in the next release of QEMU. It 
is too painful to maintain, difficult to port and it needs a patched 
guest OS to work correctly.

This target is replaced by the standard QEMU with soft mmu support. The 
QEMU Kernel Acceleration Layer which will be unveiled very soon will 
give much more performance while working with unpatched guest OSes.

Fabrice.

Re: [Qemu-devel] qemu-fast target removal ?

[Magnus Damm]

Hello,

On Sat, 29 Jan 2005 22:48:24 +0100, Fabrice Bellard <fabrice@bellard.org> wrote:
> I plan to remove the 'qemu-fast' target in the next release of QEMU. It
> is too painful to maintain, difficult to port and it needs a patched
> guest OS to work correctly.

Very good idea IMO.

> This target is replaced by the standard QEMU with soft mmu support. The
> QEMU Kernel Acceleration Layer which will be unveiled very soon will
> give much more performance while working with unpatched guest OSes.

Cool! Will you include an updated bios is the next release?

/ magnus

Re: [Qemu-devel] qemu-fast target removal ?

[Jim C. Brown]

On Sat, Jan 29, 2005 at 10:48:24PM +0100, Fabrice Bellard wrote:
> Hi,
> 
> I plan to remove the 'qemu-fast' target in the next release of QEMU. It 
> is too painful to maintain, difficult to port and it needs a patched 
> guest OS to work correctly.

Glad to hear it.

> 
> This target is replaced by the standard QEMU with soft mmu support. The 
> QEMU Kernel Acceleration Layer which will be unveiled very soon will 
> give much more performance while working with unpatched guest OSes.


I wonder how the internals of that will work...

> 
> Fabrice.
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
> 

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] qemu-fast target removal ?

[Derek Fawcus]

On Sat, Jan 29, 2005 at 10:48:24PM +0100, Fabrice Bellard wrote:
> Hi,
> 
> I plan to remove the 'qemu-fast' target in the next release of QEMU. It 
> is too painful to maintain, difficult to port and it needs a patched 
> guest OS to work correctly.

Well the code is interesting...  So before you remove it,  please tag
the last known working version in the repository.  Then interested
parties can get to it easily.

DF

Re: [Qemu-devel] qemu-fast target removal ?

[Herbert Poetzl]

On Sat, Jan 29, 2005 at 10:48:24PM +0100, Fabrice Bellard wrote:
> Hi,
> 
> I plan to remove the 'qemu-fast' target in the next release of QEMU. It 
> is too painful to maintain, difficult to port and it needs a patched 
> guest OS to work correctly.

hmm, but IMHO it's _a_lot_ faster than the qemu/soft mmu
and I'm using it on a regular basis for kernel development
(the patch is minimal and doesn't change anything relevant)

are you sure that you want to do that?

here is a comparison:
(2.6.11-rc2 kernel testboot, nothing else)

with qemu-fast (0.6.1):

 3.31user 0.33system 0:08.39elapsed 43%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (5638major+1879minor)pagefaults 0swaps

with qemu (softmmu, 0.6.1)

19.25user 0.27system 0:23.25elapsed 83%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (901major+5788minor)pagefaults 0swaps

I would _really_ appreciate not to spend almost
three times the amount of time and a lot more
of cpu power to wait for the kernel to boot ...

> This target is replaced by the standard QEMU with soft mmu support. The 
> QEMU Kernel Acceleration Layer which will be unveiled very soon will 
> give much more performance while working with unpatched guest OSes.

any numbers for that? will it be compareable?
what will it require?

thanks for your work!

best,
Herbert

> Fabrice.
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel

Re: [Qemu-devel] qemu-fast target removal ?

[Fabrice Bellard]

Herbert Poetzl wrote:

> I would _really_ appreciate not to spend almost
> three times the amount of time and a lot more
> of cpu power to wait for the kernel to boot ...

I will just disable its compilation by default, but the code will still 
be there for some time, so you will be able to enable it with the 
--target-list option.

>>This target is replaced by the standard QEMU with soft mmu support. The 
>>QEMU Kernel Acceleration Layer which will be unveiled very soon will 
>>give much more performance while working with unpatched guest OSes.
> 
> 
> any numbers for that? will it be compareable?

All I can say is that for Linux or Windows user code (such as nbench), 
the performance is the same as the native one.

> what will it require?

A small host kernel module is needed. It works with 2.4 and 2.6 Linux 
kernels.

Fabrice.

Re: [Qemu-devel] qemu-fast target removal ?

[zitu]

Quoting Fabrice Bellard <fabrice@bellard.org>:
> > any numbers for that? will it be compareable?
>
> All I can say is that for Linux or Windows user code (such as nbench),
> the performance is the same as the native one.

cool. bye bye vmw****

> > what will it require?
>
> A small host kernel module is needed. It works with 2.4 and 2.6 Linux
> kernels.

ok, means no go for winXP hosts, right ? Will the soft-mmu new std target
support winXP hosts ?

One last question: when will the new kernel module/layer be available as
alpha ?

Thanks Fabrice & co for this *wonder*
Zitu