Files
chip_ignite/experiments/2026-05-12-archive-v6-v11/.gitignore
cah ebec884751
Some checks failed
CI / rtl-lint (push) Has been cancelled
fix(submission): archive v6-v11 wrapper experiments, add LVS waiver
The May 7-11 series of wrapper-hardening attempts (cf_wrapper_v6
through v11) all failed to eliminate the 208 cosmetic LVS pin-match
errors. The root cause is a Magic SPICE-extraction limitation: it
collapses constant-tied output nets into shared global power/ground
nets, losing per-pin labels. Per-pin conb_1 cells (v8) and placement
tweaks don't change this behavior; v9/v10 regressed to 1300-1800
routing DRC errors; v11 was interrupted.

This commit:
- Restores the working tree to the May 7 PDN-fix wrapper (HEAD before
  experiments started).
- Archives the v6-v11 experiment files under experiments/ with a
  per-file README explaining what was tried and why it failed. Large
  binary debris (mag, multicorner spef, uncompressed wrapper GDS) is
  preserved locally but excluded from git via a local .gitignore.
- Adds docs/LVS_WAIVER.md formally requesting the 208 cosmetic LVS
  errors be accepted as Magic-extraction artifacts. Cites netgen's own
  "Device classes equivalent" / "Cell pin lists altered to match"
  lines, the v8 disproof that per-pin tieoffs cannot fix this, and
  the 17/19 mpw_precheck PASS at cf_wrapper_v5 (May 1, 2026).
- README.md now links the waiver from the Precheck Results section.
- .gitignore adds /precheck_logs/ for local snoke run logs.

The submission ships with the May 7 cf_wrapper_v5 + PDN-fix wrapper
(commit 8cc8414). All 5 GLS cocotb tests pass (GPIO=0xAB). Magic DRC
clean. 17/19 mpw_precheck. The cosmetic LVS pin-match issue is
documented in the waiver and not fixable by wrapper-only hardening.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-12 23:16:18 -06:00

9 lines
398 B
Plaintext

# Binary debris from failed v6-v11 wrapper hardening experiments.
# The text files (manual_tieoffs.vh, manual_placements.json, README.md)
# in this directory capture all the lessons. The binaries below are
# regenerable by re-running `cf harden user_project_wrapper` with the
# v8 config and have no future analytical value, so they are not tracked.
mag/
spef_multicorner/
user_project_wrapper.gds