BoF at LPC: Documenting the Heterogeneous Memory Model Architecture

BoF at LPC: Documenting the Heterogeneous Memory Model Architecture

[Felix Kuehling]

As the programming models for GPU-based high-performance computing 
applications are evolving, HMM is helping us integrate the GPU memory 
management more closely with the kernel's virtual memory management. As 
a result we can provide a shared virtual address space with 
demand-paging and page-based migrations of anonymous pages to/from 
device memory. A patch series by AMD [1, 2] to add support for 
cache-coherent, CPU-accessible device memory has brought up some fairly 
fundamental questions about HMM and its interaction with virtual memory 
management, page cache and file systems. We'd like to use the chance of 
getting together for a BoF [3] at LPC to raise awareness for HMM outside 
the GPU driver code, identify gaps in the architectural documentation 
and clarify our priorities for future development.

Thank you, Daniel, for suggesting the BoF and getting it scheduled. It's 
set for Friday, 10am Pacific, 1pm Eastern, 5pm UTC.

I am registered at LPC. Daniel got a speaker's pass. We're still trying 
to work something out for Alex.

I hope to see many of you on Friday.

Best regards,
   Felix


[1] https://patchwork.freedesktop.org/series/94611/
[2] https://patchwork.freedesktop.org/series/90706/
[3] https://linuxplumbersconf.org/event/11/contributions/1123/

-- 
F e l i x   K u e h l i n g
PMTS Software Development Engineer | Linux Compute Kernel
1 Commerce Valley Dr. East, Markham, ON L3T 7X6 Canada
(O) +1(289)695-1597
     _     _   _   _____   _____
    / \   | \ / | |  _  \  \ _  |
   / A \  | \M/ | | |D) )  /|_| |
  /_/ \_\ |_| |_| |_____/ |__/ \|   facebook.com/AMD | amd.com

Re: BoF at LPC: Documenting the Heterogeneous Memory Model Architecture

[Felix Kuehling]

Change of plan: Instead of a BoF, this is now a session in the 
"GPU/media/AI buffer management and interop MC" micro conference. Thank 
you Daniel Stone for making that happen. 
https://linuxplumbersconf.org/event/11/contributions/1112/

It is scheduled for tomorrow (Friday) 08:40-10:00 Pacific, 11:40-13:00 
Eastern, 15:40-17:00 UTC.

I hope to see you all tomorrow,
   Felix


On 2021-09-21 3:19 p.m., Felix Kuehling wrote:
> As the programming models for GPU-based high-performance computing 
> applications are evolving, HMM is helping us integrate the GPU memory 
> management more closely with the kernel's virtual memory management. 
> As a result we can provide a shared virtual address space with 
> demand-paging and page-based migrations of anonymous pages to/from 
> device memory. A patch series by AMD [1, 2] to add support for 
> cache-coherent, CPU-accessible device memory has brought up some 
> fairly fundamental questions about HMM and its interaction with 
> virtual memory management, page cache and file systems. We'd like to 
> use the chance of getting together for a BoF [3] at LPC to raise 
> awareness for HMM outside the GPU driver code, identify gaps in the 
> architectural documentation and clarify our priorities for future 
> development.
>
> Thank you, Daniel, for suggesting the BoF and getting it scheduled. 
> It's set for Friday, 10am Pacific, 1pm Eastern, 5pm UTC.
>
> I am registered at LPC. Daniel got a speaker's pass. We're still 
> trying to work something out for Alex.
>
> I hope to see many of you on Friday.
>
> Best regards,
>   Felix
>
>
> [1] https://patchwork.freedesktop.org/series/94611/
> [2] https://patchwork.freedesktop.org/series/90706/
> [3] https://linuxplumbersconf.org/event/11/contributions/1123/
>

Re: BoF at LPC: Documenting the Heterogeneous Memory Model Architecture

[Matthew Wilcox]

On Thu, Sep 23, 2021 at 04:25:08PM -0400, Felix Kuehling wrote:
> Change of plan: Instead of a BoF, this is now a session in the "GPU/media/AI
> buffer management and interop MC" micro conference. Thank you Daniel Stone
> for making that happen.
> https://linuxplumbersconf.org/event/11/contributions/1112/
> 
> It is scheduled for tomorrow (Friday) 08:40-10:00 Pacific, 11:40-13:00
> Eastern, 15:40-17:00 UTC.

That's up against:

 Direct map management
Vlastimil Babka, Mike Rapoport, Rick Edgecombe  11:30-12:15.

Seems like a lot of the same people would want to be in both sessions.
Maybe one could be moved?

RE: BoF at LPC: Documenting the Heterogeneous Memory Model Architecture

[Phillips, Daniel]

[AMD Official Use Only]

Hi Matt,

I originally picked the 10 am pdt time slot to avoid those conflicts. But I see the overlap is only partial. I also see Peterz is giving a talk that also overlaps partially. But I think this is survivable. I suggest we just try to make this work as best we can. It's not a presentation, it's primarily intended as a meet and greet and discussion. Whatever starts there will continue on lkml and etc.

Regards,
Daniel

-----Original Message-----
From: Matthew Wilcox <willy@infradead.org> 
Sent: Thursday, September 23, 2021 1:39 PM
To: Kuehling, Felix <Felix.Kuehling@amd.com>
Cc: Linux MM <linux-mm@kvack.org>; linux-ext4@vger.kernel.org; linux-xfs@vger.kernel.org; Maling list - DRI developers <dri-devel@lists.freedesktop.org>; linux-kernel@vger.kernel.org; Christoph Hellwig <hch@lst.de>; Jason Gunthorpe <jgg@mellanox.com>; Jerome Glisse <jglisse@redhat.com>; Ralph Campbell <rcampbell@nvidia.com>; Theodore Ts'o <tytso@mit.edu>; Andrew Morton <akpm@linux-foundation.org>; Mel Gorman <mgorman@techsingularity.net>; Peter Zijlstra <peterz@infradead.org>; Phillips, Daniel <Daniel.Phillips@amd.com>; Sierra Guiza, Alejandro (Alex) <Alex.Sierra@amd.com>; Daniel Stone <daniel@fooishbar.org>
Subject: Re: BoF at LPC: Documenting the Heterogeneous Memory Model Architecture

[CAUTION: External Email]

On Thu, Sep 23, 2021 at 04:25:08PM -0400, Felix Kuehling wrote:
> Change of plan: Instead of a BoF, this is now a session in the 
> "GPU/media/AI buffer management and interop MC" micro conference. 
> Thank you Daniel Stone for making that happen.
> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flinu
> xplumbersconf.org%2Fevent%2F11%2Fcontributions%2F1112%2F&amp;data=04%7
> C01%7CDaniel.Phillips%40amd.com%7Cb5064a1c5649431a90be08d97ed2517b%7C3
> dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637680264672385094%7CUnknown
> %7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJ
> XVCI6Mn0%3D%7C2000&amp;sdata=e%2Bpn4vXZI9yhLAbB8rtqik6OiZsEwOCBMdqk4JZ
> n%2FC8%3D&amp;reserved=0
>
> It is scheduled for tomorrow (Friday) 08:40-10:00 Pacific, 11:40-13:00 
> Eastern, 15:40-17:00 UTC.

That's up against:

 Direct map management
Vlastimil Babka, Mike Rapoport, Rick Edgecombe  11:30-12:15.

Seems like a lot of the same people would want to be in both sessions.
Maybe one could be moved?

Re: BoF at LPC: Documenting the Heterogeneous Memory Model Architecture

[Daniel Stone]

Hi,

On Thu, 23 Sept 2021 at 21:40, Matthew Wilcox <willy@infradead.org> wrote:
> On Thu, Sep 23, 2021 at 04:25:08PM -0400, Felix Kuehling wrote:
> > Change of plan: Instead of a BoF, this is now a session in the "GPU/media/AI
> > buffer management and interop MC" micro conference. Thank you Daniel Stone
> > for making that happen.
> > https://linuxplumbersconf.org/event/11/contributions/1112/
> >
> > It is scheduled for tomorrow (Friday) 08:40-10:00 Pacific, 11:40-13:00
> > Eastern, 15:40-17:00 UTC.
>
> That's up against:
>
>  Direct map management
> Vlastimil Babka, Mike Rapoport, Rick Edgecombe  11:30-12:15.
>
> Seems like a lot of the same people would want to be in both sessions.
> Maybe one could be moved?

Good point, and thanks, but it's hard to keep the longer slot whilst
moving it later; I wonder if we could move direct map management to
the final slot?

Cheers,
Daniel