Re: [LSF/MM TOPIC] NUMA, memory hierarchy and device memory

[Jerome Glisse <jglisse@redhat.com>]

I see that the schedule is not full yet for the mm track and i would
really like to be able to have a discussion on this topic

Schedule:
https://docs.google.com/spreadsheets/d/1Z1pDL-XeUT1ZwMWrBL8T8q3vtSqZpLPgF3Bzu_jejfk/edit#gid=0


On Fri, Jan 18, 2019 at 12:45:13PM -0500, Jerome Glisse wrote:
> Hi, i would like to discuss about NUMA API and its short comings when
> it comes to memory hierarchy (from fast HBM, to slower persistent
> memory through regular memory) and also device memory (which can have
> its own hierarchy).
> 
> I have proposed a patch to add a new memory topology model to the
> kernel for application to be able to get that informations, it
> also included a set of new API to bind/migrate process range [1].
> Note that this model also support device memory.
> 
> So far device memory support is achieve through device specific ioctl
> and this forbid some scenario like device memory interleaving accross
> multiple devices for a range. It also make the whole userspace more
> complex as program have to mix and match multiple device specific API
> on top of NUMA API.
> 
> While memory hierarchy can be more or less expose through the existing
> NUMA API by creating node for non-regular memory [2], i do not see this
> as a satisfying solution. Moreover such scheme does not work for device
> memory that might not even be accessible by CPUs.
> 
> 
> Hence i would like to discuss few points:
>     - What proof people wants to see this as problem we need to solve ?
>     - How to build concensus to move forward on this ?
>     - What kind of syscall API people would like to see ?
> 
> People to discuss this topic:
>     Dan Williams <dan.j.williams@intel.com>
>     Dave Hansen <dave.hansen@intel.com>
>     Felix Kuehling <Felix.Kuehling@amd.com>
>     John Hubbard <jhubbard@nvidia.com>
>     Jonathan Cameron <jonathan.cameron@huawei.com>
>     Keith Busch <keith.busch@intel.com>
>     Mel Gorman <mgorman@techsingularity.net>
>     Michal Hocko <mhocko@kernel.org>
>     Paul Blinzer <Paul.Blinzer@amd.com>
> 
> Probably others, sorry if i miss anyone from previous discussions.
> 
> Cheers,
> Jérôme
> 
> [1] https://lkml.org/lkml/2018/12/3/1072
> [2] https://lkml.org/lkml/2018/12/10/1112
>