[Qemu-devel] Agenda items for today's MTTCG call

[Qemu-devel] Agenda items for today's MTTCG call

[Alex Bennée]

User-agent: mu4e 0.9.12; emacs 24.5.50.4
Hi,

It's been a while since we last synced up via phone call so could I add
the following items for discussion:

1. Status of Alvise and Fred's trees
2. Any cross-pollination ideas between Alvise's LL/SC work and Emilio's
series

Emilio's series has a number of interesting features to it, not least
that it works with softmmu and linux-user. However I do prefer the TCG
op based approach of Alvise's tree. Is there any scope for getting the
best of both worlds?

3. Memory barriers

I now have working barrier tests which fail on real hardware when
barriers don't exist. I'll be sending a new series soon. Again Emilio's
tree introduces some barrier TCG ops.

4. Other outstanding work

I've had a couple of people prod me with offers of help so it would be
nice to see what other tasks are currently pending to avoid too much
duplication of work.

Does anyone else have anything then want to bring up?

-- 
Alex Bennée

Re: [Qemu-devel] Agenda items for today's MTTCG call

[Claudio Fontana]

On 28.09.2015 11:19, Alex Bennée wrote:
> User-agent: mu4e 0.9.12; emacs 24.5.50.4
> Hi,
> 
> It's been a while since we last synced up via phone call so could I add
> the following items for discussion:
> 
> 1. Status of Alvise and Fred's trees
> 2. Any cross-pollination ideas between Alvise's LL/SC work and Emilio's
> series
> 
> Emilio's series has a number of interesting features to it, not least
> that it works with softmmu and linux-user. However I do prefer the TCG
> op based approach of Alvise's tree. Is there any scope for getting the
> best of both worlds?
> 
> 3. Memory barriers
> 
> I now have working barrier tests which fail on real hardware when
> barriers don't exist. I'll be sending a new series soon. Again Emilio's
> tree introduces some barrier TCG ops.
> 
> 4. Other outstanding work
> 
> I've had a couple of people prod me with offers of help so it would be
> nice to see what other tasks are currently pending to avoid too much
> duplication of work.
> 
> Does anyone else have anything then want to bring up?
> 

Hi Alex, thanks for this and we'll hear in a few :-)

In general the way I see this is that we should start upstreaming one small thing at a time, rather than wait for a perfect solution which starts to include more and more patches and diverge more and more from upstream qemu.
I don't think we want to maintain a separate "MT" branch right?

Ciao,

Claudio

Re: [Qemu-devel] Agenda items for today's MTTCG call

[Claudio Fontana]

Hi Alex and all,

today I wanted to test the whole MTTCG thing to see if I can speed up a specific use case (running a build process multiple times involving QEMU),
do we have some clear instructions about how to set the whole thing up using the latest publicly available code? (Which patches to apply, in which order etc)?

Btw these instructions could go in the dedicated MTTCG wiki page of QEMU as well I guess..

Ciao

Claudio

Re: [Qemu-devel] Agenda items for today's MTTCG call

[Alexander Spyridakis]

> do we have some clear instructions about how to set the whole thing up using the latest publicly available code? (Which patches to apply, in which order etc)?
>
> Btw these instructions could go in the dedicated MTTCG wiki page of QEMU as well I guess..

Hi Claudio,

You can find mttcg + atomic handling here:
https://git.virtualopensystems.com/dev/qemu-mt.git branch
slowpath-for-atomic-v5-mttcg

git clone https://git.virtualopensystems.com/dev/qemu-mt.git -b
slowpath-for-atomic-v5-mttcg
./configure --target-list=arm-softmmu
make -j8

Then you can use the following command line to start an SMP linux guest:

./qemu-system-arm \
    -smp 8 \
    -nographic \
    -kernel zImage \
    -m 1024 -M virt -cpu cortex-a15 \
    -drive if=none,file=disk.img,id=fs,format=raw \
    -device virtio-blk-device,drive=fs \
    -netdev type=user,id=net0 \
    -device virtio-net-device,netdev=net0,mac="52:54:00:12:34:56" \
    -append "console=ttyAMA0 root=/dev/vda rw ip=dhcp"

Regards.