silicon: NUCLEO-G474RE anchor protocol scaffolding (board-prep)

[avrabe]

Summary

Scaffolding for the silicon-anchor protocol so that when the
NUCLEO-G474RE arrives this week, taking a capture is a single
`bash silicon/capture.sh --board nucleo_g474re --variant baseline --sweep long`
away. No firmware code touched; no CI changes. Manual-flow only,
recorded directly into the repo.

Companion to PR #36 (overhead compensation + SCOPE.md). Composes
cleanly on top — when #36 lands, silicon captures will already
emit the `overhead_cycles,` metadata and be apples-to-apples
with the post-compensation Renode CI numbers.

Architecture

• CI = Renode, deterministic, parallel-safe, every PR.
• Silicon = manual, periodic, single board shared across the
  team, recorded as immutable evidence in `silicon/runs//`.
  Citeable from any blog post via stable git URL.

The relationship `silicon_median / renode_median` per RPM step is
what the anchor establishes. Once consistent across multiple
captures it can be cited as the Renode-silicon multiplier for that
bench/board combination.

What's in this PR

Path	What
`silicon/README.md`	Protocol doc — why we anchor, the recorded-run-in-git convention, capture procedure for NUCLEO-G474RE, comparison workflow against Renode CI, anchor cadence, don't-do-this list (overwriting dated dirs, mixing pre/post-overhead-compensation captures, claiming WCET)
`silicon/capture.sh`	Build + flash + capture + tag + manifest, one invocation. Auto-detects the USB serial port on macOS / Linux. Refuses to overwrite an existing dated dir.
`silicon/capture.py`	Cross-platform pyserial UART capture. Reads until `=== END ===` or wall-clock timeout.
`silicon/boards/nucleo_g474re/`	Board overlay (currently empty — Zephyr defaults are right) + board notes (clock, UART, programming)
`silicon/runs/.gitkeep`	Placeholder; first dated capture goes here
`README.md`	Methodology section now points at `silicon/`

Each captured run will commit:

• `output.csv` — raw firmware UART
• `events.csv` — tagged through `tag_events.py`
• `firmware.elf` + `firmware.elf.sha256`
• `manifest.txt` — board, MCU, gale_sha, rustc, west, zephyr_sha, ELF sha256, capture timestamp, port, timeout

CSV row counts are small (~50–500 KB per long-sweep run). At one
capture per board per major bench-relevant commit, repo growth is
modest.

Test plan

• `bash -n capture.sh` syntax-clean
• `python3 -c 'import ast; ast.parse(open(...))'` clean
• First end-to-end capture on the NUCLEO-G474RE when it
  arrives — that's the validation that matters; this PR is
  preparation only.

Out of scope

• ESP32-C3-DevKit-RUST-1 board: separate work, will need its own
  Renode equivalent built first (RISC-V port of engine_control)
  before silicon makes sense. Tracked separately.
• Analyzer `--silicon-anchor` flag for single-call Renode-vs-silicon
  side-by-side rendering: deferred until the first capture exists
  to test against.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!