Skip to content

J714c (M5 Max): sync aurora-wip and two T6050 patches #35

Open

Description

@edimoldovan

What I did

The J714c (M5 Max, T6050) runs its desktop on aurora-silicon/mesa aurora-wip 8ac88a2 plus two patches of my own:

  • GL through zink on the T6050, because gallium asahi's M5 render commands don't match the kernel UAPI;
  • the HAL300 path on by default on the T6050, with ASAHI_M5_BRINGUP=0 to turn it off.

The M5 Mesa work is by Jonathan Goyvaerts, DjDeveloperr, Thomas Glanzmann, Ryan Murray, Acelogic, Artyom Astafurov, Satya Benson and others.

The issue

integration/aurora doesn't accept the M5 Max: its M5 bring-up takes T8142 and G17S 0x6050, but not G17C with 4 clusters, which is what the kernel reports. Its kmsro also recognizes only the "asahi" DRM name. Merging 8ac88a2 into integration/aurora conflicts in 7 files (asahi_clc.c, agx_chip.c/.h, agx_device.c, test-chip-admission.c, hk_cmd_dispatch.c, hk_device_memory.c). asahi-native-descriptor and three agx_tests pack-binary tests also fail on plain aurora-wip.

Proposed solution

I propose a sync of aurora-wip 8ac88a2 into integration/aurora, like #19. I've resolved the 7 conflicts locally, keeping both sides. The two T6050 patches would follow as separate PRs; both apply cleanly after the sync.

I can open the sync PR myself, or wait if you'd rather run the aurora-wip sync yourselves.

What would you advise?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions