Terminal minimizer relations (SYN-02)

Descriptive single-adapter measurements using the public TerminalOptimizer with HiGHS LP and convex providers.

Source report syn02-terminal-20261013-c234a1a0-35de886a7fb4; measured algorithm commit c234a1a04b2f90e34776a2bfa349a91a1d30257d.

Scope: Terminal optimizer on resolved objectives and a supplied terminal H-slice; not system simulation or complete control synthesis. No comparator, speedup, or application-scale claim is made.

View solve latency ranges by fixture

Median and min–max solve time; values appear only for fully expected fixture groups.

Fixture	Expected outcome	Expected termination	Matched/requested	Non-success	Solve time
linear-face	optimal	completed	7/7	0	0.226 ms (214,000–242,600 ns)
constant-linear	optimal	completed	7/7	0	0.172 ms (162,000–184,300 ns)
psd-null-direction	optimal	completed	7/7	0	0.362 ms (335,200–373,700 ns)
rectangular-residual	optimal	completed	7/7	0	0.345 ms (337,100–379,700 ns)
shifted-residual-degenerate	undetermined	numerical_failure	7/7	7	Unavailable

The shifted-residual fixture retains undetermined with numerical_failure and no constructed relation in all 7 measured trials. These expected numerical failures remain visible and are not filtered from the results.

Evidence limits

binary64 literal fixture checks, not exact or validated arithmetic
relation samples and H representation checks do not certify full set equality or full minimizer-family completeness
global optimum and relation support remain Unestablished
the shifted residual case returns undetermined with numerical_failure and no constructed relation
tiny analytic fixtures only; no comparator, speedup, application-scale, throughput, or scalability claim
limiter unknown because no call-stack profile evidence was collected

Profile limiter: unknown. Outcome support and relation support remain Unestablished.

Saved evidence

Versioned envelope · Publication manifest