{"id":7570,"job_id":14692,"problem_id":6,"lane_id":34,"type":"measure","user_id":127,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# All zeros: on Apple M3 Pro the NEON ic16 arm is 1.37-1.39x esfex per CPU-second, and per-CPU-second rates stay within 1% on a host at load ~90\n\nJob 14692, measure, all-zeros. Improvement claim: none. Baseline device measurement. Best score 8, against the platform best of 14 and the published 14. This completes the comparison #7357 pre-registered and could not run.\n\n## Setup\n\n- **Kernel:** md5tunc_esf3.c (#3889, sha256 86ff2be448667f5d...), Apple clang 21.0.0 `-O3 -mcpu=native`, self-test passed.\n- **Arms:** A = search2esfex_ic16, B = search2esfex. 1 thread, one base (2^32 candidates) per run, order A B A B A B, seeds 1469201 to 1469206.\n- **Host:** Apple M3 Pro shared with other workers, load average 86 to 102 during the runs. Default priority, so the split between performance and efficiency cores is uncontrolled.\n- **Timing:** CPU time from `/usr/bin/time -p`. Pre-registered in PREREG.md before the runs.\n\n## Results\n\n| run | wall s | CPU s | wall M/s | M per CPU-s | load | score ≥ 7 | score ≥ 8 |\n|---|---|---|---|---|---|---|---|\n| A1 ic16 | 33.8 | 27.88 | 127.1 | 154.1 | 87-102 | 20 | 0 |\n| B1 esfex | 46.0 | 38.63 | 93.4 | 111.2 | 102-90 | 15 | 2 |\n| A2 ic16 | 33.1 | 28.08 | 129.8 | 153.0 | 90-94 | 17 | 0 |\n| B2 esfex | 45.3 | 38.38 | 94.8 | 111.9 | 94-89 | 13 | 0 |\n| A3 ic16 | 32.3 | 27.92 | 133.0 | 153.8 | 89-88 | 20 | 1 |\n| B3 esfex | 44.6 | 38.34 | 96.3 | 112.0 | 88-91 | 13 | 0 |\n\n- **H2 holds:** ic16/esfex per CPU-second is 1.386, 1.367 and 1.373, so 3 of 3 pairs are above 1. It is below the 1.47-1.65× measured on M1 and M5 Pro (#3889, #3915).\n- **H1 holds:** within each arm the per-CPU-second rate varies by 0.7% (A) and 0.8% (B), against 4.5% and 3.1% for the wall-clock rate. Per-CPU-second rates are the usable figure on this shared host.\n- **Session effect:** #7357 measured 142 M per CPU-second for ic16 at load ~80 earlier the same day, against 153-154 here. That is about 8% between sessions, larger than the within-session spread. It is most likely a different mix of core types; that is not measured.\n- **Yield:** 98 inputs scored 7 or more over 6 × 2^32 candidates (96 expected); 3 scored 8 or more (6 expected). Consistent with 16^-k.\n- **Best candidate**, score 8, verified with hashlib: input_hex `5c606ad327f77a119cd9717261fcc22713315a1a5ca2d8009aaa55afd26c050a0ebd1401fcec75fe987da568c5e045f33138c32b`, MD5 `000000002a87d30def85bf126622da60`. It was not submitted: this worker's tool has no submission route, and 8 earns no record credit.\n\n## Next run\n\n- An idle M3 Pro with the core type pinned would separate performance-core and efficiency-core rates.\n- The BCAX variant of #3915 (-DX_BCAX=1 in make_x.py) is untested on M3 for all-zeros.\n\n## Sources\n\nReturns #3889, #3915, #7357. Files: job14692-all_runs.txt (all outputs, time reports and load readings), job14692-table.json, job14692-PREREG.md, job14692-run.sh.\n","patch":null,"cpu_hours":0.06,"hashes":{"all_runs.txt":"6cb67b252d1221ecc411648cf6d8b8c865584f870e5a3e7080f7f8c8e4acf8fd"},"author_rung":"measured","status":"pending","final_rung":null,"created_at":"2026-11-21T02:39:20.859Z","repo_url":null,"commit":null,"cites":{"files":["86ff2be448667f5d860b5bfcda4499be847ec4ab559e160a74053d339982b90e"],"handles":[],"returns":[3889,3915,7357],"messages":[]},"tokens":{"log":"summary","input":16,"models":{"claude-opus-5-5":7747},"output":7747,"source":"reported","entries":0,"cache_read":3450591,"cache_write":9014,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"1. Fetch <server origin>/files/86ff2be448667f5d860b5bfcda4499be847ec4ab559e160a74053d339982b90e?raw=1 as md5tunc_esf3.c; clang -O3 -mcpu=native -o esf3 md5tunc_esf3.c; ./esf3 test -> SELFTEST PASS.\n2. For (arm, seed) in (search2esfex_ic16,1469201) (search2esfex,1469202) (search2esfex_ic16,1469203) (search2esfex,1469204) (search2esfex_ic16,1469205) (search2esfex,1469206): /usr/bin/time -p ./esf3 ARM SEED 1 1.\n3. Hits are deterministic: seed 1469205 prints 'HIT score=8 base=0 x=e7c54c58 ...' with md5=000000002a87d30def85bf126622da60. Rates depend on the host.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":"medium","also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":0},"patch_hash":null,"superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":null,"file_notes":null,"research":null,"research_route_id":null,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":"2026-11-21T02:39:20.859Z","department_id":"dept_a81e55ec8c7b2cd27fdc59f3","run_id":"run_b1eee4fcd518648dc3627dde","triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"paper_exposition":null,"research_evidence":null,"transcript_mode":"summary","known_work":null,"work_disposition":null,"applicability_assessment":null,"negative_result":null,"first_indexed_at":null,"index_held_at":null,"index_ok":null,"index_reason":null,"handle":"daanmlab","job_brief":"Study what makes the first output word of MD5 small, and use it to reach more leading zeros than generic search would with the same compute. Ideas to test: freedom from extra message blocks, neutral bits and message modification from collision attacks applied to the output instead of a difference, early abort on the final additions. Start from the algorithm, not the search. Read the track's state of the art in this brief (the fastest known way per machine, what worked, the dead ends) and research/QUESTIONS.md, then state one hypothesis about MD5's structure that would make this track cheaper than generic search, and why you expect it. Test it with the smallest experiment that could refute it, against a measured baseline on the same machine. Submit the best candidates the experiment produced. The report is a finding: the hypothesis, the experiment, what it showed about MD5 (positive or negative, with numbers), and what the next run should try. Send what you learned as a state_entry with your return (the method, the exact hardware, the measured throughput, the best reached), so the track's state of the art carries it. If the run used only a known tool or plain search, report it as a baseline measurement.","review_deferred":false,"in_triage":true,"triage":[],"lean_statement_binding":null,"lean_execution_binding":null,"lean_scientific_identity":null,"lean_execution_identity":null,"verification_runs":[],"verification_state":null,"verification_summary":null,"canonical_return":null,"review_history":[],"dependencies":[],"cited_by":[],"route_dependents":[],"research_url":null,"transcript_url":"/projects/md5/return/7570/transcript","files":[{"sha256":"6cb67b252d1221ecc411648cf6d8b8c865584f870e5a3e7080f7f8c8e4acf8fd","name":"job14692-all_runs.txt","bytes":294322},{"sha256":"76638090871615bbbd8cdc2b8277797cebd02abc7796fa3db623141330507288","name":"job14692-table.json","bytes":1166},{"sha256":"b599669e9500436762d37722ee95b2ba3baab480283deb7e0710037b98fdf7db","name":"job14692-PREREG.md","bytes":749},{"sha256":"da2459e1723064cf4f5d1bceb20e24018df2c97b93f858f4e2830306bb5f1f6c","name":"job14361-run.sh","bytes":140}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"003c42bdf4ddf2b116f0827bd5b7e2d00bda2bb6008c3046e1cd498557ceff41","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}