Skip to content

Conversation

@mckees
Copy link

@mckees mckees commented Nov 17, 2025

Description

Previous review here:
https://code.launchpad.net/~mckeesh/sru-docs/+git/sru-docs/+merge/491986

This PR proposes a process for allowing full-stack upgrades for Intel GPU enablement in LTS releases. This is needed because aside from Mesa and the kernel, our LTS releases do not have support for new Intel hardware. This leaves important cases such as compute, media, and raytracing completely broken for products released over a year ago in Noble.


Related issue


Checklist

  • I have read and followed the Ubuntu Project contributing guide
  • My pull request is linked to an existing issue (if applicable)
  • I have tested my changes, and they work as expected

Additional notes (optional)

Changes made since previous upload:

  • Add platform testing info and update versioning
  • Remove hard 75% threshold for test devices
    • Change made after a discussion with @RAOF
  • Fix spelling mistakes

 Changes made since previous upload:
 - Add platform testing info and update versioning
 - Remove hard 75% threshold for test devices
  - Change made after a discussion with @RAOF
 - Fix spelling mistakes
@mckees
Copy link
Author

mckees commented Nov 17, 2025

Just a quick note on the "linked issue (if applicable)" checklist item: I think it should always be applicable, given that the contributing guide requires us to create an issue if one does not exist. Though it is functionally an honesty/due diligence check for your first checklist item, so you may prefer to keep it as is :)

@basak
Copy link
Collaborator

basak commented Nov 17, 2025

Thank you for moving this across.

As well as https://code.launchpad.net/~mckeesh/sru-docs/+git/sru-docs/+merge/491986 that you linked there's also https://lists.ubuntu.com/archives/ubuntu-release/2025-February/006315.html and discussion over subsequent months, and https://wiki.ubuntu.com/IntelGraphicsUpdates. All should be checked for anything outstanding before landing this.

this group:

#. intel-gmmlib
#. libva
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Aargh, sorry I didn't catch this before - libva is used by AMD video acceleration, as well, and any regression there is not likely to be caught by Intel internal testing (I presume?).

libva looks mostly un-scary, but we should probably at least do a smoke test; even something as simple as running vainfo on a couple of AMD GPUs.

Copy link
Author

@mckees mckees Nov 20, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, good point. Libva is quite reliable, but testing the integration makes sense. I've gone a step further and found that our ffmpeg testing works on an AMD machine, so we should run that.

Gstreamer testing currently appears to be Intel specific, but the goal of this testing is to be more broadly applicable (especially as I start merging media testing to checkbox). Since the testing relies heavily on libva tracing, I suspect AMD support will be a small gap to fill in over time.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Included an AMD test machine with ffmpeg regression testing

@mckees mckees requested a review from RAOF November 21, 2025 07:29
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Migrate Intel Graphics SRU exception proposal to this repo

3 participants