Docs: give the local onboard sweep the same marker shape as CI

[ChaoWao]

Summary

Everywhere a doc told the reader to run the full a2a3 scene-test sweep on real hardware, it gave a bare pytest examples tests/st --platform a2a3 — the exact invocation testing/SKILL.md warns "is not what CI runs and will report failures that CI never sees": it mixes SDMA cases with the sweep's aicore_op_timeout fault-injection test, the collision the dedicated SDMA step exists to prevent (#1425, ~306 s fault teardown on an SDMA-provisioned device).

Nine sites now carry -m "not sdma" and name the marker:

File	What
README.md	hardware scan line
docs/testing.md	both hardware scan sites; first gains the separate -m sdma pass with the reason
docs/user/reference/cli.md	hardware scan line
examples/README.md	hardware scan line
docs/troubleshooting/a2a3-507899-aicpu-shared-so-fault.md	repro scan
SUBMIT_BY_CLUSTER.md ×3 (a2a3/a5 × tmr, a2a3 hbg)	final-validation scans

Untouched, deliberately: sim lines (no SDMA cases), single-directory invocations (marker irrelevant), and all a5 scans (st-onboard-a5 has no marker filter).

Also fixed, the deferred ci.yml comments — the two that claimed the SDMA step runs "on its own dedicated device(s)", flagged in #1617 and left until ci.yml was next touched for a substantive reason. Device disjointness holds only on aarch64 (task-submit --device auto); the x86_64 branch shares ${DEVICE_RANGE} and ordering is the separation. This change touches ci.yml, so the full matrix runs — which was the reason it had been deferred.

Testing

• Swept every pytest examples tests/st --platform a2a3 / pytest examples --platform a2a3 site: none remain unfiltered (except the SKILL.md warning sentence, which is not a command)
• a5 scan lines verified unchanged (no marker filter in st-onboard-a5)
• YAML parses after the comment edits
• ci.yml in the diff → full matrix expected

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 4f530903-587a-4450-8e66-e4cf2a367492

📥 Commits

Reviewing files that changed from the base of the PR and between 266b123 and eba81a8.

📒 Files selected for processing (9)

• .github/workflows/ci.yml
• README.md
• docs/testing.md
• docs/troubleshooting/a2a3-507899-aicpu-shared-so-fault.md
• docs/user/reference/cli.md
• examples/README.md
• src/a2a3/runtime/host_build_graph/docs/SUBMIT_BY_CLUSTER.md
• src/a2a3/runtime/tensormap_and_ringbuffer/docs/SUBMIT_BY_CLUSTER.md
• src/a5/runtime/tensormap_and_ringbuffer/docs/SUBMIT_BY_CLUSTER.md

---

📝 Walkthrough

Walkthrough

The change documents SDMA test isolation in CI and updates hardware test commands to exclude SDMA cases by default. Separate commands now run SDMA cases with the sdma marker.

Changes

SDMA test isolation

Layer / File(s)	Summary
CI ordering and device isolation .github/workflows/ci.yml	The CI comments document disjoint-device handling on aarch64, ordering on x86_64, and separate SDMA execution.
Default test selection and SDMA commands README.md, docs/testing.md, docs/troubleshooting/..., docs/user/reference/cli.md, examples/README.md, src/a2a3/runtime/*/docs/SUBMIT_BY_CLUSTER.md, src/a5/runtime/tensormap_and_ringbuffer/docs/SUBMIT_BY_CLUSTER.md	Hardware test commands exclude sdma cases by default. Documentation provides separate -m sdma commands and dedicated device guidance.
Estimated code review effort: 2 (Simple)	~10 minutes

Possibly related PRs

• hw-native-sys/simpler#1620: Updates SDMA quarantine guidance and marker-based execution.
• hw-native-sys/simpler#1617: Documents SDMA isolation and dedicated CI execution.
• hw-native-sys/simpler#1609: Updates SDMA test isolation and pytest selection.

Poem

I’m a rabbit guarding the SDMA lane,
General tests hop on the main train.
With not sdma, the sweep stays bright,
Then -m sdma runs in its own light.
Separate devices keep order just right.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the documentation change that aligns local onboard sweeps with CI marker filtering.
Description check	✅ Passed	The description directly explains the documentation updates, CI comment corrections, scope, and testing performed.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}