This repository was archived by the owner on Nov 12, 2026. It is now read-only.

Repository navigation

Package MLX and Apple Silicon eGPU updates - #705

Merged

techmore merged 16 commits into

devfrom

codex/warren-upstream-audit-20261030

Oct 30, 2026

Merged

Package MLX and Apple Silicon eGPU updates#705
techmore merged 16 commits into
devfrom
codex/warren-upstream-audit-20261030

Commits on Oct 30, 2026
- Record Warren AI and eGPU upstream qualification
  techmore
  committed
  4a56763
- Prepare separate Aurora eGPU kernel candidate
  techmore
  committed
  ad58f2d
- Record M1 Pro batched MLX candidate measurements
  techmore
  committed
  336935e
- Verify batch dispatch and live USB4 device tree
  techmore
  committed
  d30a580
- Prepare isolated eGPU boot tree candidate
  techmore
  committed
  a6ba1cc
- Prepare separate ConnectX Ethernet kernel profile
  techmore
  committed
  04cb970
- Record ConnectX candidate build run
  techmore
  committed
  1b76561
- Record eGPU firmware and ConnectX compile checks
  techmore
  committed
  29e7407
- Qualify updated TensorFold Vulkan draft on M1 Pro
  techmore
  committed
  326dd10
- Record TensorFold long-generation comparison
  techmore
  committed
  b58868b
- Measure TensorFold teacher-forced CPU reference fidelity
  techmore
  committed
  5be1a84
- Verify Aurora eGPU kernel artifact
  techmore
  committed
  adaf081
- Investigate long TensorFold divergence against CPU
  techmore
  committed
  259da19
- Deploy MLX release and stage eGPU kernel
  techmore
  committed
  06a2661
- Package eGPU and ConnectX support in one kernel
  techmore
  committed
  f70f5cc
- Record combined kernel build
  techmore
  committed
  476b558