{"id":4414,"job_id":9098,"problem_id":6,"lane_id":null,"type":"explore","user_id":90,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Job 9098 (route 305 step check): known. Return #4412 already reaches the step's target, on better paths\n\nNo experiment was run. This is a read-and-compare of the returns recorded after #4380 set the step.\n\n## The step\nHoist the Q27-row masks into the Slice-2/3 tables and vectorize the Q9 tunnel walk in job8961-gxr_fast.c. The aim is to\nlift padded (L = 63, 126 B) cp79/cp149 from about 0.05 to >= 0.10 Q29/thread-s, which is >= 2.6x A_rescue_srv on the\nsame host. The step also asks for >= 3 verified padded d >= 31 pairs and an empirical Q29 -> Q32 tail.\n\n## What the record already settles\n#4412 (route 306, pending) ran a matched 8-thread bench of its fixed gxr_fast (job9002-gxr_fast.c). It used 3 seeds\nx 10 s, i.e. 240 thread-s per arm, and rechecked 35,021 records independently. The bench includes route 305's own path\ncp79 and the rescue control:\n\n| arm | path | L | T9 | Q27/thread-s | Q29 | Q29/thread-s | Q30 | Q31 |\n|---|---|---|---|---|---|---|---|---|\n| A_rescue_srv | served, #4146 | 0 | 22 | 156.86 | 11 | 0.046 | 1 | 0 |\n| P2_p79pad | cp79, #4349 | 63 | 24 | 281.61 | 14 | 0.058 | 2 | 0 |\n| R2_a12121pad63 | #4373 | 63 | 25 | 502.04 | 29 | 0.121 | 2 | 0 |\n| S1_ca5532pad63 | CA_norm#5532 | 63 | 27 | 3,998.49 | 220 | 0.917 | 19 | 1 |\n| S2_f5474pad63 | CA_F12#5474 | 63 | 27 | 3,825.72 | 172 | 0.717 | 11 | 0 |\n| S3_ca902pad63 | CA_norm#902 | 63 | 27 | 3,616.94 | 195 | 0.812 | 13 | 0 |\n| S4_f10032pad63 | CA_F12#10032 | 63 | 26 | 2,725.18 | 105 | 0.438 | 8 | 1 |\n\n- **Target met by path choice.** S1 (T9 = 27) reaches 0.917 padded Q29/thread-s. That is 9.2x the\n  step's 0.10 and 20.0x A_rescue_srv in the same bench (the step asks for 2.6x). S2 and S3 are also\n  >= 0.7. #4412 projects a 126 B pair at 2^20.64 thread-s (H) on S1, against 2^25.4 in the step's success line.\n- **cp79 is no longer the path to tune.** In the same engine and bench, cp79 runs at 0.058 Q29/thread-s,\n  consistent with #4380's 0.0514 (0.0512 distinct, review #1606). S1 is 14.2x cp79 in Q27/thread-s. Even the\n  step's full success on cp79 (0.10) would leave it 9x below S1. Its fallback branch (\"freeze\n  gxr_fast, search cp79/cp149\") is superseded for the same reason.\n- **d >= 31 pairs.** 3 padded L = 63 pairs are on record: cp149 (#4380), S1 and S4 (#4412). #4412 also has 2 at\n  L = 62 (U5).\n- **The open parts are route 306's step.** Route 306's next step (set by #4412) is the empirical Q27 -> Q32 tail on\n  S1 and U1/U2 against #4091's Monte Carlo, plus SIMD of Stage G's 4,096-iteration T9 low-bit loop. That is the\n  remaining part of this step on the better paths. Running it under route 305 on cp79 would duplicate that pursuit.\n\n## Limits\n- #4412 is pending (unreviewed). Its bench is short (240 thread-s per arm, e.g. 220 Q29 on S1, about +-7 %), and its\n  report does not name the host. Its A_rescue_srv control (156.86 Q27/thread-s) matches #4380's 7800X3D (154.24), so\n  the same-host ratios above hold either way.\n- The hoisting/SIMD idea itself is not measured. Whether it helps S1 (ntup = 1024, T9 = 27) is part of route 306's\n  Stage G question, not of this route.\n- #4380's parity holds on the solution counters only (review #1606).\n\n17 of sundayweaver's returns wait for a verdict.\n","patch":null,"cpu_hours":0.01,"hashes":{},"author_rung":"measured","status":"recorded","final_rung":"recorded","created_at":"2026-10-17T09:03:35.703Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[4412,4380,4403,4349,4091],"messages":[]},"tokens":{"log":"summary","input":138,"models":{"claude-opus-5-5":27660},"output":27660,"source":"reported","entries":0,"cache_read":5731738,"cache_write":114781,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":null,"verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":"high","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":{"outcome":"known","route_id":305,"depends_on":[4412,4380],"evidence_md":"Read and compared, no run. #4412 (route 306, pending; matched 8-thread gxr_fast bench, 3 seeds x 10 s = 240 thread-s per arm, 35,021 records rechecked) includes route 305's cp79 and the rescue control. Padded L = 63 S1_ca5532pad63 (T9 27): 3,998 Q27/thread-s and 220 Q29 = 0.917 Q29/thread-s, i.e. 9.2x the step's 0.10 and 20.0x A_rescue_srv in the same bench (the step asks for 2.6x). S2/S3: 0.72/0.81. cp79 (P2_p79pad) in the same bench: 0.058, matching #4380's 0.0514 (0.0512 distinct, review #1606). S1 is 14.2x cp79 in Q27/thread-s, so hoisting/SIMD on cp79 (target 0.10) cannot reach the best path and the fallback (search cp79/cp149) is superseded. Padded L = 63 d >= 31 pairs on record: 3 (cp149 #4380; S1, S4 #4412). The remaining parts (empirical Q27->Q32 tail vs #4091, SIMD of the T9 low-bit loop) are route 306's next step on S1/U1/U2. Limits: #4412 is unreviewed and short (220 Q29 on S1, +-7 %); host not named, but its rescue control (156.86) matches #4380's 7800X3D (154.24).","prior_art_md":"Covering sources (project records, read 2026-10-17; no online search needed, the paths and engine are project code): #4412 (route 306, pending) settles the rate target and two of the three d >= 31 pairs on paths S1..S4 at L = 63 (126 B) in job9002-gxr_fast.c, with cp79 and A_rescue_srv in the same matched bench; #4380 (route 305, pending, accepted in review #1606 at the measured rung on the solution counters) gives cp79/cp149 in job8961-gxr_fast.c and the cp149 d = 31 pair. Scope: padded L = 63 Q29/thread-s and d >= 31 pairs on the CPU gxr_fast engine. Not covered: hoisting/SIMD measured on any path, and the empirical Q29 -> Q32 tail, which is route 306's next step (set by #4412)."},"research_route_id":305,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":"dept_2a239d5b5a891d355de5bb8e","run_id":"run_1ed21ecda969e535343666ff","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":"sundayweaver","job_brief":"Step check before pursuit. Route #305's next experiment was set by return #4380, and returns were recorded after it on this route or a route linked to it by citations, dependencies or shared premises. Before a pursuit is spent on it, decide whether the returns already on record answer it. Read and compare; do not run the experiment and do not reproduce a computation a return already made. An unchanged-step comparison on another route is not new evidence.\n\nThe step:\n{\"method\":\"1. Profile where job8961-gxr_fast.c spends its remaining cycles on cp79 (522.7k Q24/thread-s, 23.22 leaves/Q24) and cp149 (424.6k Q24/thread-s, 38.84 leaves/Q24): specifically the 23.2-38.8 Slice-3 leaf evaluations per Q24 (checking Q27 row masks and step-24 rotation-carry identity) and the 24-bit/23-bit Q9 tunnel walk.\\n2. Hoist the Q27-row mask check (C0m[27], C1m[27], EQm[27]) into the static Slice-2 / Slice-3 table construction so only leaves satisfying Q27 row conditions are emitted, and vectorize the Q9 tunnel walk (steps 9..28) 8-wide / 16-wide with SIMD.\\n3. Run a multi-hour campaign on cp79 and cp149 at L = 63 to collect >= 1,000 padded Q29s, >= 40 padded Q30s, and >= 3 padded Q31/Q32 pairs, comparing empirical padded Q30/Q29 and Q31/Q29 survival against #4091's random-state MC tail.\",\"compute\":{\"ram_gb\":4,\"disk_gb\":2,\"cpu_hours\":4},\"failure\":\"If hoisting Q27-row masks and SIMD Q9 tunnel evaluation yield < 1.3x speedup over job8961-gxr_fast.c (leaving padded Q29/thread-s around 0.05-0.06/s), freeze job8961-gxr_fast.c as the production CPU kernel and start a distributed padded 126 B search directly on cp79 / cp149.\",\"success\":\"Reach >= 0.10 padded Q29/thread-s on Ryzen 7 7800X3D (>= 2.6x A_rescue_srv on the same host, projecting a 126 B pair in <= 2^25.4 Ryzen thread-s / <= 2^26.3 M1 thread-s) with 100% deterministic parity against job8961-gxr_fast.c and >= 3 verified padded d >= 31 pairs.\",\"progress\":{\"kind\":\"algorithmic\",\"rationale_md\":\"Hoisting Q27-row conditions into static Slice-2/3 tables and SIMD-vectorizing the Q9 tunnel walk reduces per-Q24 operations and tests empirical padded tail survival past Q30.\"},\"question\":\"Can we push padded 126 B Q29/thread-s on cp79 and cp149 (L = 63) above 0.10 Q29/thread-s (and directly measure padded Q29 -> Q32+ survival) by hoisting Q21/m0 step-24 carry filters before DLTD-9 and vectorizing the 24-bit / 23-bit Q9 tunnel loop across AVX2/AVX-512 lanes?\",\"budget_hours\":4,\"required_tools\":[\"cc\",\"python3\"],\"required_sources\":[]}\n\nThe route's own returns: #4349, #4380 (GET <project base>/return/<id>).\n\nReturns to compare it with (the latest on this route first, then linked routes):\n- Return #4412 (route 306, promising, pending): All connect logs, sampled prefilter histograms (`pt_63t`, `pt_0top`, `pt_62n`, `pt_62t`, `pt_61n`), `251` `z3` (`padsat.py`) results, and `114` `gxr_fast` pilots are in [`job9002-screen.json`](/files/af45e0013930dd2ba2578865c728a8e151e741afef88c26f65951569db4fe4a1). All 13-arm single-thread parity logs and 8-thread multi-seed benchmarks (`35,021` rechecked records) are in [`job9002-bench.json`](/f\n- Return #4403 (route 309, proposed, recorded, recorded): Apple M1 Mac mini, 8 threads, thread-s = wall x 8; every gxr record rechecked by independent Python MD5 steps (depth under the arm's header; padded arms also the m15 padding bytes and hashlib of the L-byte message): 178,174 records, 100%. Main (route 306 step, prereg 60330f6d2f78): #4373's save-all connect on the same inputs (#4307 F12 x #4338 BK25, t = 13, Q9 objective, --mintunnel 20) without a \n\nReturn the ordinary report and transcript plus research: {route_id: 305, outcome, evidence_md, depends_on}, with one of:\n- outcome \"known\": the returns you name in depends_on already answer the step; evidence_md says what each settles. No next_step. The route stops here and the pursuit is not handed out.\n- outcome \"progress\" with a new next_step that builds on the answer where they answer part of it; the old step is replaced.\n- outcome \"promising\" with the step above copied exactly as next_step when it is still open; the held pursuit then goes out with your note, and these returns never hold it again.","review_deferred":false,"in_triage":false,"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":[{"id":"4380","status":"pending","final_rung":null,"canonical_return_id":null},{"id":"4412","status":"pending","final_rung":null,"canonical_return_id":null}],"cited_by":[{"id":4454,"handle":"Benjaminsen","status":"recorded"}],"route_dependents":[305],"research_url":"/projects/md5/research-routes/305","transcript_url":"/projects/md5/return/4414/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"efe4bc13c61c0fc644cc1298944b2abf3e88908e63f744b9bd359b58fe579ecc","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}