All pull requests

Pull requests

• asahi: probe HIGH admission and report effective EGL priority

  #31·satchlj opened Nov 3, 2026·

  ·
  7

• asahi: admit T8122 HAL200 with native sampler state

  #30·satchlj opened Nov 3, 2026·

  ·
  4

• asahi: fix failed-context cleanup and Zink query test initialization

  #29·satchlj opened Nov 3, 2026·

  ·
  1

• agx: flatten guarded read-only loads in compute shaders

  #28·techmore opened Nov 2, 2026·

  ·
  5

• asahi/hk: opt-in 8x8 cooperative matrices on AGX3B

  #27·techmore opened Nov 2, 2026·

  ·
  2

• agx: encode the AGX3B 8x8 matrix multiply-accumulate from G16 witnesses

  #26·techmore opened Nov 2, 2026·

  ·
  3

• hk: flush with memory scope for a G17P alias dependency recorded before any work

  #23·DjDeveloperr opened Oct 31, 2026Collaborator·

  ·

• Mirror Aurora Mesa #103, #105–#108: virtio guest, cube samples, render area, UBO record loads, cached variants

  #22·DjDeveloperr opened Oct 29, 2026Collaborator·

  ·
  8

• agx: let G17P ray queries commit opaque triangle hits in hardware

  #21·DjDeveloperr opened Oct 29, 2026Collaborator·

  ·
  4

• honeykrisp: validate the existing Aurora capture/replay heap

  #18·joshuaswarren opened Oct 26, 2026·

  ·
  27

• agx: exact division, log and sin/cos for compute; graphics keeps the base math

  #14·joshuaswarren opened Oct 25, 2026·

  ·
  15

• honeykrisp: fold workgroup memory address deltas on G14 (stacked on #6)

  #11·joshuaswarren opened Oct 15, 2026·

  ·
  5

• honeykrisp: map placed memory at the allocation size

  #8·joshuaswarren opened Oct 12, 2026·

  ·
  35

• asahi: prepare clean M3 foundation and attachment compression

  #7·chriskearney opened Oct 12, 2026·

  ·
  2

• honeykrisp: M1 compute speedups and the deferred CDM barrier on G13G (stacked on #3, #4, #5)

  #6·joshuaswarren opened Oct 1, 2026·

  ·
  49