GEO-04 measurement allocation: clean MSVC Release executable from algorithm source 9579f9e936d85d66c59003b2a04157c153df6b63 and harness b83b8dc415dbe5811371b8446931ae4c6f3d2e28.
Host: Gigabyte H610M K DDR4, Windows 11 Home 10.0.22631, Intel Core i5-12400F, 6 physical cores / 12 logical processors, Balanced power plan. Five total-processor-load samples from 2026-10-12T22:49:23Z through 2026-10-12T22:49:27Z: 2.17%, 1.22%, 0.00%, 0.73%, 0.54% (mean 0.93%); free physical memory was about 15.92 GiB.
This is an ordinary interactive workstation, not an isolated benchmark host. This session launched no competing local benchmark or build during the measurement window; remote hosted CI did not run locally during the observation. Background OS services and user activity cannot be ruled out. Requested algorithm resources: one outer worker, one thread per worker, one total thread; HiGHS internal thread count one. OS-observed per-trial native thread counts and memory are unavailable through the public API.
Trials use fresh scenario inputs/workspaces/providers, one excluded warmup per scenario, seven measured repetitions, and alternating forward/reverse scenario order. The host is shared, so results are descriptive microfixture timings, not production-scale or scaling claims.
