igneum-labs
|
b0af074fb4
|
Latency ladder (Horizon finding 4, 0.3.17): docs/design/latency-ladder.md (where N lives, the ladder as a genesis list, the 90 percent seven-window step rule on the class signal's machinery, the verifier bound and its measured table, the X9 chip anchor, the hostile review, consequences per tier); igneum-pow: v4_class_at, v4_rung_reps, v4_counted_ops, generate_from_seed_bytes_program_class_shadow and Epoch::chain_program_shadow (rung 0 is V4_CLASS byte for byte; a rung above carries its shadow size in the id through program_id_class), era draws 8 and 9 consumed and not used, --shadow-reps on the CLI, the known-failed test first (a changed N hashed another program under one id); tools/ladder/verify-bench-remote.sh (the bench per rung on core 40 alone and with core 88 loaded under the box's measure hold; the first run's sibling finished early and is recorded as the script's failed case) with both runs' raw lines under docs/design/latency-ladder-bench; infra/fast-time/latency-ladder.mjs (step, no-step and the known-failed case) and the 60x file's three ladder fields plus the two 0.3.16-lane fields it lacked; ledger M34; the litepaper's hash-class paragraph and the corrected RandomX precedent (Bitmain's Antminer X9, approximate)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-06 22:18:48 +00:00 |
|