[RFC] Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?

[RFC] Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?

[Viacheslav Dubeyko]

Hello,

Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?
Any comments?

Hi Dan,

I believe we need some talk to summarize:
(1) what we achieved during 2023 year,
(2) what TODOs we have
(3) etc.

What do you think?

Hi Adam,

I assume it will be good to hear what is cooking on hardware vendor’s side
in the CXL memory area. Could we have some summary related to this
(of course, what can be shared with open-source community)? How do you
feel about it?

Hi Vincent,

It will be great to hear what is going on with CXL specification and which activity
has taken place in CXL Consortium. How do you feel to share some 2023 year
summary?

Hi Jonathan,

I think QEMU is important area of CXL memory emulation and testing.
So, I assume we could have interesting topics and discussions here.
Any suggestions for this? What do you think?

Thanks,
Slava.

Re: [RFC] Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?

[Adam Manzanares]

On Fri, Dec 22, 2023 at 06:10:21PM +0300, Viacheslav Dubeyko wrote:
> Hello,
> 
> Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?
> Any comments?

We are starting to spend a bit more time looking beyond the
CXL driver code, specifically interleaving and tiering in
the context of CXL attached memory. It would be great to 
carve out time to get interested parties in the same room.

> 
> Hi Dan,
> 
> I believe we need some talk to summarize:
> (1) what we achieved during 2023 year,
> (2) what TODOs we have
> (3) etc.
> 
> What do you think?
> 
> Hi Adam,
> 
> I assume it will be good to hear what is cooking on hardware vendor’s side
> in the CXL memory area. Could we have some summary related to this
> (of course, what can be shared with open-source community)? How do you
> feel about it?

I would be more than happy to give an update from a device vendor perspective. 
There were several public demonstrations of CXL devices in the past year and
it would be great if we could get representation from other CXL device vendors
to share updates with the community.

> 
> Hi Vincent,
> 
> It will be great to hear what is going on with CXL specification and which activity
> has taken place in CXL Consortium. How do you feel to share some 2023 year
> summary?
> 
> Hi Jonathan,
> 
> I think QEMU is important area of CXL memory emulation and testing.
> So, I assume we could have interesting topics and discussions here.
> Any suggestions for this? What do you think?
> 
> Thanks,
> Slava.
> 
> 
> 

Re: [RFC] Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?

[Viacheslav Dubeyko]

> On Jan 3, 2024, at 7:53 PM, Adam Manzanares <a.manzanares@samsung.com> wrote:
> 
> On Fri, Dec 22, 2023 at 06:10:21PM +0300, Viacheslav Dubeyko wrote:
>> Hello,
>> 
>> Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?
>> Any comments?
> 
> We are starting to spend a bit more time looking beyond the
> CXL driver code, specifically interleaving and tiering in
> the context of CXL attached memory. It would be great to 
> carve out time to get interested parties in the same room.
> 
>> 
>> Hi Dan,
>> 
>> I believe we need some talk to summarize:
>> (1) what we achieved during 2023 year,
>> (2) what TODOs we have
>> (3) etc.
>> 
>> What do you think?
>> 
>> Hi Adam,
>> 
>> I assume it will be good to hear what is cooking on hardware vendor’s side
>> in the CXL memory area. Could we have some summary related to this
>> (of course, what can be shared with open-source community)? How do you
>> feel about it?
> 
> I would be more than happy to give an update from a device vendor perspective. 
> There were several public demonstrations of CXL devices in the past year and
> it would be great if we could get representation from other CXL device vendors
> to share updates with the community.

Maybe, do we need to share this as a special topic? What do you think? Let’s elaborate
some better explanation of suggested discussion.

Thanks,
Slava.

Re: [RFC] Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?

[Adam Manzanares]

On Mon, Jan 15, 2024 at 11:58:03AM +0300, Viacheslav Dubeyko wrote:
> 
> 
> > On Jan 3, 2024, at 7:53 PM, Adam Manzanares <a.manzanares@samsung.com> wrote:
> > 
> > On Fri, Dec 22, 2023 at 06:10:21PM +0300, Viacheslav Dubeyko wrote:
> >> Hello,
> >> 
> >> Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?
> >> Any comments?
> > 
> > We are starting to spend a bit more time looking beyond the
> > CXL driver code, specifically interleaving and tiering in
> > the context of CXL attached memory. It would be great to 
> > carve out time to get interested parties in the same room.
> > 
> >> 
> >> Hi Dan,
> >> 
> >> I believe we need some talk to summarize:
> >> (1) what we achieved during 2023 year,
> >> (2) what TODOs we have
> >> (3) etc.
> >> 
> >> What do you think?
> >> 
> >> Hi Adam,
> >> 
> >> I assume it will be good to hear what is cooking on hardware vendor’s side
> >> in the CXL memory area. Could we have some summary related to this
> >> (of course, what can be shared with open-source community)? How do you
> >> feel about it?
> > 
> > I would be more than happy to give an update from a device vendor perspective. 
> > There were several public demonstrations of CXL devices in the past year and
> > it would be great if we could get representation from other CXL device vendors
> > to share updates with the community.
> 
> Maybe, do we need to share this as a special topic? What do you think? Let’s elaborate
> some better explanation of suggested discussion.
>

See the following thread:
https://lore.kernel.org/all/2b29dd3d-bb2c-6a8c-94d2-d5c2e035516a@google.com/T/

It covers many areas of concern from the mm perspective.


> Thanks,
> Slava.
> 

Re: [RFC] Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?

[Jonathan Cameron]

On Fri, 22 Dec 2023 18:10:21 +0300
Viacheslav Dubeyko <slava@dubeyko.com> wrote:


> 
> Hi Jonathan,
> 
> I think QEMU is important area of CXL memory emulation and testing.
> So, I assume we could have interesting topics and discussions here.
> Any suggestions for this? What do you think?

Currently I'm not seeing anything that would really suit LSF-MM.
There's lots going on it just won't much benefit from the people I'd
expect to talk to in that venue (about QEMU emulation side of things
anyway - even the virtualization topics are going to be too early stage
to really discuss much this year).

Also from person point of view, I'm avoiding travel until the summer.
Might participate remotely.

Jonathan

> 
> Thanks,
> Slava.
> 
> 
> 

RE: [RFC] Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?

[Dan Williams]

Viacheslav Dubeyko wrote:
> Hello,
> 
> Do we need dedicated LSF/MM/BPF 2024 session for CXL memory?

Last year we handled this on the fly in terms of identifying topics that
have the highest level of cross appeal with core-mm folks. My
observation is that some of the sessions were useful for the "mm" track,
like tiering policy and ABI, while other topics were likely more
suitable for the "storage" track, like fabric manager topics.

I expect to do the same this year and not have a dedicated CXL session.
I.e. prefer CXL topics that intersect mm-policy or ABI for the "mm"
track and find a space for the more hardware-centric topics on an as
needed basis.