Skip to content

Apple A8 (T7000) GPU support — iPod touch 6, testers wanted on iPad mini 4 and iPhone 6 - #196

Open
brkosebay wants to merge 7 commits into
OrchardLinux:mainfrom
brkosebay:a8-gpu
Open

brkosebay wants to merge 7 commits into
OrchardLinux:mainfrom
brkosebay:a8-gpu

Conversation

@brkosebay

@brkosebay brkosebay commented Nov 21, 2026 •

Copy link
Copy Markdown
Contributor

This PR adds the GPU of the A8 (T7000, PowerVR GX6450) to powervr's Apple glue, with the iPod touch 6 (N102) as the tested board. The userspace side is OrchardLinux/mesa#18.

Results (iPod touch 6 only)

  • powervr's A8 variant probes the GPU (BVNC 3.10.4.5), powers it up in Apple's order and boots Apple's GPU firmware. It runs compute and TA/3D render jobs.
  • Vulkan 1.2 through Mesa's powervr driver. Every pixel-checked Vulkan test in the ladder passes.
  • OpenGL ES 3.1 through unmodified Zink.
  • vkmark: 842 at 256x256 (910 with a8_joint=1), 332 at 750x1334.
  • glmark2 (OpenGL ES through Zink): 611 at 256x256, 311 at 750x1334.
  • A soak of 514k jobs ran with no fault. No job is refused.
  • This exact series (built as the installed system's kernel) runs an installed Omarchy A desktop from the iPod's internal storage: labwc composites on the GPU (wlroots GLES2 renderer through Zink on renderD128, shown on the boot framebuffer), Orchard's gpu-probe passes, and two-process dmabuf and implicit-sync checks (wlroots-style partial redraws, 60 frames) pass with no fault and no refused job.

Kernel patches (kernel/patches/drivers/gpu/)

Patch What
0063 The S8003 (A9) chip file describes its chip in a variant table. A9 values and functions unchanged.
0064 A chip may read its own GPU ID and give its own heap table. A9 and A10 unchanged.
0065 dt-bindings: apple,t7000-gpu.
0066 powervr: a GPU node without power-domains; a failed probe names its step.
0067 A8 firmware boot, power up and down in Apple's order.
0068 A8 compute jobs.
0069 A8 render jobs: parameter manager and its pool (a8_pm_blocks, a8_pm_grow), the region-header ring kept in the HWRT data set.
0070 A8 fault and hang reports and recovery.
0071 A8 completion in the firmware interrupt's thread, the re-check after an early interrupt, a publish that leaves nothing behind when refused, Apple's doorbell.
0072 A fragment job submitted behind its geometry job, with two alternating 3D records (a8_joint).
0073 Statistics and the job timeline in debugfs apple_jobs.

CONFIG_DRM_POWERVR_APPLE goes from unset to =y on apple-4k, as on apple-16k.

The bring-up debug code is gone (probe crumbs, a8_stop, submit_debug, a8_trace). The final tree is the one tested on the device minus that debug code; this branch's kernel itself has been built but not yet flown. drivers/gpu/drm/imagination builds at every patch (apple-4k), and the full series builds on apple-16k.

Other commits

  • devices/n102 and devices/t7000/0011: the iPod touch 6 and the GPU node.
  • devices/t7000/0012: the GPU enabled on the iPad mini 4 (J96, J97) and the iPhone 6 and 6 Plus (N61, N56). Same SoC and GPU, but not run there yet.
  • examples/ipod-gpu-drm: a RAM test payload for any A8 board (probe, DRM and Vulkan tests, the Vulkan ladder sources). The GPU firmware is extracted from Apple's public IPSW; no Apple bytes are in the repository.

Experimental

  • a8_joint is off by default. It is faster (vkmark 910 against 842) and ran 118k jobs with no fault, but it has had one test run.
  • 0073 is statistics, last in the series. It also points the firmware's TA and 3D time words at the channel's times; the series without it has not run on the device.
  • Every patch but 0065 and 0068 also touches code the A9 and A10 run (pvr_apple.c, pvr_apple_chip.h, the S8003 file, generic powervr files). Their paths are meant to be unchanged; the series builds on apple-16k, but it was not retested on an A9 or A10.

Testers wanted

The driver has run only on the iPod touch 6. If you have an iPad mini 4 or an iPhone 6 / 6 Plus, please boot examples/ipod-gpu-drm (see its README) and report:

  • the board;
  • the output of a8-gpu-drm-test probe with its kernel lines (/run/a8-gpu-drm/probe.txt, or the serial log if the device froze);
  • the ladder's RESULT lines.

On an image that autoloads modules, powervr probes the GPU at boot on these untested boards. The test payload does not autoload it.

🤖 Generated with Claude Code

The A8's GX6450 runs the same family of firmware as the A9's, so it
becomes a variant of the S8003 chip file rather than a new driver
(drivers/gpu/0063-0073):

- 0063-0064: the S8003 chip described in a variant table, and chip
  hooks for its own GPU ID read and heap table. These touch the A9 and
  A10 paths; their values and functions are unchanged.
- 0065: dt-bindings for apple,t7000-gpu.
- 0066: a GPU node without power domains, and the failing probe step
  named (shared powervr code; nodes with power domains are unchanged).
- 0067-0069: the A8's firmware boot and power up and down in Apple's
  order, compute jobs, and TA/3D render jobs with the parameter-manager
  pool (a8_pm_blocks, a8_pm_grow) and the region-header ring kept in
  the HWRT data set.
- 0070: fault and hang reports and recovery.
- 0071: completion checks in the firmware interrupt's thread, the
  re-check after an early interrupt, a publish that cannot leave a
  failed job in the firmware's ring, Apple's doorbell.
- 0072: a fragment job submitted behind its geometry job, with two
  alternating 3D records (a8_joint, off by default).
- 0073: statistics and the A8's job timeline in debugfs apple_jobs,
  last. It also points the firmware's TA and 3D time words at the
  channel's times; the series without it has not run on the device.

Config: CONFIG_DRM_POWERVR_APPLE goes from unset to =y on apple-4k, as
on apple-16k, so the A8 gets powervr's Apple glue. It builds into the
powervr module on every A7, A8 and A8X kernel; only boards whose device
tree enables an apple,t7000-gpu node bind it.

Assisted-by: LLM
The iPod touch 6 RAM-boots through checkm8, PongoOS and m1n1 to the USB
serial shell. Its descriptor, boot assets and firmware recipe pin the
public iOS 12.5.8 IPSW.

devices/t7000/0011 describes the A8's GPU in t7000.dtsi (the Rogue core,
Fender, the power manager and GPU bridge registers, five interrupts,
the firmware name) and enables it on the N102, where powervr's A8
variant runs Vulkan and OpenGL ES (through Zink) jobs.

Orchard's feature list has no GPU entry and the chip's [gpu] table
records dEQP results, which the A8 does not have yet, so the GPU state
is in devices/n102/README.md. docs/compatibility.md is regenerated.

Assisted-by: LLM
The iPad mini 4 (J96, J97) and the iPhone 6 and 6 Plus (N61, N56) have
the iPod touch 6's A8 GPU. devices/t7000/0012 enables the GPU node in
their shared board files (t7000-mini4.dtsi, t7000-6.dtsi) so people with
these boards can test it; the driver has run only on the iPod touch 6.
Each board's README asks for the probe and ladder results. The A8X
(T7001) has a different GPU and is unchanged.

Assisted-by: LLM
minimal-boot with a constant USB serial number, the kernel's modules,
the GPU firmware and a8-gpu-drm-test, which loads powervr and checks the
probe, Mesa's Vulkan path and two DRM-level replays, each step under a
hardware watchdog with its kernel lines saved. It builds for every A8
board (n102, j96, j97, n61, n56).

tools/extract-fw.py writes the GPU firmware and its border colour table
from Apple's public iPod touch 6 iOS 12.4 IPSW, read over HTTP ranges
with orchard's own IPSW and kernelcache code, or from another IPSW or
kernelcache. Nothing from Apple is in the repository, and everything
build.sh stages is git-ignored.

Assisted-by: LLM
a8-vk-smoke (Vulkan probe, compute, a triangle) and a8-vk-ladder, a set
of Vulkan tests from one clear to vkmark's scenes with every pixel or
value checked on the CPU, both MIT. vk/make-bundle.sh builds them with a
Mesa build of the A8 branch into the self-contained /opt/a8vk bundle
that build.sh's A8_VK_DIR stages, with vkmark when it is installed.

Assisted-by: LLM
@brkosebay
brkosebay marked this pull request as ready for review November 21, 2026 22:14
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.

1 participant