Skip to content

drm/asahi: enable GPU completion inputs on J716c - #2207

Open
erikwb wants to merge 1 commit into
integration/aurorafrom
j716c/gpu-completion-inputs
Open

erikwb wants to merge 1 commit into
integration/aurorafrom
j716c/gpu-completion-inputs

Conversation

@erikwb

@erikwb erikwb commented Nov 18, 2026

Copy link
Copy Markdown
Contributor

J716c makes the host wait for compute work whose producer can already signal completion to the GPU. Enable the existing counted submission path for this board. The driver accepts an eligible producer in the same VM with a visible completion. Older userspace and other dependencies retain the existing host wait.

In Cyberpunk 2077 at 3840×2160, High with FSR2 Balanced, the two control runs rendered 18.26 and 17.75 FPS. The changed kernel rendered 22.52 and 22.61 FPS (+25.3% across the means). Frame generation and VSync were off. Both kernels ran on the same device with the same Mesa build, ARM64EC tool and single-queue setting. Runs were grouped by kernel across a reboot. Both kernels passed 64 of 65 synchronization CTS cases, with one unsupported and zero failures.

The flag applies directly to integration/aurora. To boot this J716c hardware configuration, the test kernel also included the board enablement stack in #1173, #1650 and #2185.

Two game launches on kernels with completion inputs enabled lost the Vulkan device while a render queue was waiting for its completion stamps. One used stock vkd3d-proton. The fault is intermittent, and a matched stability comparison with completion inputs disabled is still outstanding. Its relationship to this change is unresolved.

Reviewed with sol-loop: 1 round with GPT-6.2 Sol. Sol verdict: approved.

Created with Codex.

Allow eligible same-VM compute dependencies to use the counted submission form on G17C. Producers without visible completion continue through the host wait path.

Developed with Claude Code.

Signed-off-by: Erik Bourget <erik@ebourget.net>
@RyanTheTide

Copy link
Copy Markdown
Contributor

Thanks for the Cyberpunk numbers; +25% is a big gain.

We try to select behaviour by a hardware or firmware property rather than by the board it was measured on. This flips the per-board flag for one more machine (drivers/gpu/drm/asahi/g17/fw/m5/initdata.rs:2470, completion_inputs: true on J716C), while J714 has it and J775D and J813 do not, and nothing in the code says why they differ.

Whether explicit waits can take the firmware's completion class is a property of the firmware interface, not of the board. Could you make it depend on that (the firmware build/layout that supports it), or turn it on for every M5 board and remove the field? If some board must keep it off, a comment with the reason would be the way. Before that, the two lost Vulkan devices you mention need the matched comparison with completion inputs disabled. We can help with rows on other machines.

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.

2 participants