This versioned report records seven alternating measurements per registered strategy on the same exact two-stage integrator problem and optimal-family request. It is a single tiny exact-profile experiment.
Conclusion: inconclusive; no winner is claimed. No separate limiter profile was performed, so the latency limiter is unknown. The observed timing differences do not establish a general strategy ranking, speedup, application-scale result, or scalability result.
Problem identity csl.problem.v1.sha256:56871be36e6d8b0db2dde142a02722e82684a101171bd2975da3f4f708cb1d30; scenario SHA-256 f9f067beb75288884c79133d2c81e4578f9de445f03eca7f6b697c5486681c04. The effective family request is ts01_exact_integrator, target exact_optimal_relation, H representation, numerical assurance, absolute/relative tolerance 1e−8, with no materialization limit.
Both strategies received the same immutable resolved problem and provider registries. Each fresh plan/workspace ran through the public StrategyPort path; the report verifies the exact-profile complete-family result, selects two distinct witnesses, and checks both against the original problem. Original-witness checks establish bindings, feasibility, and objective only; they do not establish optimality or family completeness independently.
Selector request copied from pinned measured source b7f1ac0a9800326c825b6d8e35a8f23656496682 / examples/registered_synthesis_comparison.cpp SHA-256 eeaa8ddadceb3d30b219486b81375f977b108369f999cb39a05e0fb3a2067c3c: count 2, distinctness tolerance 0, seed 31, directions [[1,-1],[-1,1]]. Objective/problem identity is retained in each family identity.
One excluded A/B warmup; seven measured pairs alternate A/B, B/A, A/B, B/A, A/B, B/A, A/B. Per-strategy interval: fresh StrategyRegistry::prepare through exclusive workspace creation, public StrategyPort family construction and exact-profile verifier, selector preparation/query, and original-problem verification of both selected histories. Shared scenario/provider setup is reported separately. Process start, JSON/file serialization, build, plots and report rendering are excluded. Times are nanoseconds; CTest duration is not used.
| Strategy | Median (ns) | Min–max (ns) | Range width (ns) |
|---|---|---|---|
| csl.synthesis.reach-back | 4,992,900 | 4,626,500–5,207,700 | 581,200 |
| csl.synthesis.strategy-b | 1,373,600 | 1,318,500–1,440,000 | 121,500 |
The rows report descriptive sample summaries only. Seven repetitions on this fixture do not identify a performance winner. Internal LP/convex, projection, affine-sum, conversion, hull-membership, verifier subphase, CPU-time, and per-call memory counters are unavailable.
| Repetition | Kind | Strategy | Total (ns) | Family construction (ns) | Witness query (ns) | Original checks (ns) | Status |
|---|---|---|---|---|---|---|---|
| 0 | warmup | csl.synthesis.reach-back | 6,466,500 | 5,846,500 | 476,800 | 27,900 | passed |
| 0 | warmup | csl.synthesis.strategy-b | 1,682,400 | 1,037,500 | 585,600 | 11,900 | passed |
| 1 | measured | csl.synthesis.reach-back | 5,123,800 | 4,627,700 | 428,800 | 12,000 | passed |
| 1 | measured | csl.synthesis.strategy-b | 1,318,500 | 852,300 | 419,300 | 13,100 | passed |
| 2 | measured | csl.synthesis.strategy-b | 1,345,000 | 883,900 | 416,200 | 11,500 | passed |
| 2 | measured | csl.synthesis.reach-back | 4,866,800 | 4,364,400 | 442,300 | 12,700 | passed |
| 3 | measured | csl.synthesis.reach-back | 4,992,900 | 4,481,400 | 446,300 | 15,300 | passed |
| 3 | measured | csl.synthesis.strategy-b | 1,440,000 | 863,200 | 526,100 | 16,700 | passed |
| 4 | measured | csl.synthesis.strategy-b | 1,373,600 | 907,100 | 421,200 | 11,400 | passed |
| 4 | measured | csl.synthesis.reach-back | 5,153,000 | 4,625,700 | 446,800 | 29,400 | passed |
| 5 | measured | csl.synthesis.reach-back | 5,207,700 | 4,627,100 | 515,900 | 13,100 | passed |
| 5 | measured | csl.synthesis.strategy-b | 1,354,000 | 886,100 | 423,000 | 11,400 | passed |
| 6 | measured | csl.synthesis.strategy-b | 1,429,100 | 959,000 | 426,200 | 11,200 | passed |
| 6 | measured | csl.synthesis.reach-back | 4,967,700 | 4,409,400 | 493,500 | 13,100 | passed |
| 7 | measured | csl.synthesis.reach-back | 4,626,500 | 4,094,500 | 469,400 | 12,300 | passed |
| 7 | measured | csl.synthesis.strategy-b | 1,434,000 | 879,500 | 503,400 | 17,300 | passed |
All 16 strategy trials passed (2 excluded warmup trials and 14 measured trials); failures: 0. Every result was registry-verified for this exact profile and returned two distinct selected histories. Each selected history passed numerical original-problem binding, feasibility, and objective checks with maximum reported residual 0 under 1e−8 tolerances. The numerical evidence label does not mean validated arithmetic.
Measured source commit b7f1ac0a9800326c825b6d8e35a8f23656496682; Release executable SHA-256 78707cf76bec303a1bf28839a803b6232d99141c0fe78aa23d938737607eec02; compiler MSVC 19.44.35215.0; requested resources: {"memory_bytes":null,"outer_workers":1,"threads_per_worker":1,"total_threads":1}. Host: 12th Gen Intel(R) Core(TM) i5-12400F, 6 physical / 12 logical cores, Windows build 10.0.22631. Pre-run system CPU samples: 0% total processor utilization in all five 1-second pre-run samples.
The algorithm process returned 0 and emitted all 17 records. The collection wrapper returned 1 because it compared Windows CRLF stdout bytes to the LF binary JSONL sidecar. Raw stdout and JSONL were preserved; they match byte-for-byte after CRLF-to-LF normalization. This collection defect is documented and did not alter algorithm return status or event data.
Source strategy implementations are not modified by this report. The result covers only this exact discrete T-S01 fixture and the declared request, with numerical assurance. It does not serialize the live complete-family proof payload and does not close SYN-03/SYN-04 or OBS-04/#83.