Adversarial Fuzz Torture Benchmark (120 Degenerate Polygon Pairs × 4 Ops = 480 Boolean Ops)
Tests random self-intersecting bowties, near-collinear edges, coincident vertices (10-10 jitter),
and verifies exact Boolean Inclusion-Exclusion Area Identities:
Area(A ∪ B) + Area(A ∩ B) = Area(A) + Area(B) and
Area(A \ B) + Area(A ∩ B) = Area(A).
| Library | Throughput | Total Time | Area Identity Pass Rate | Crashes | Silent Area Failures | Robustness Verdict |
|---|
Workload-by-Workload Performance & Numerical Accuracy (11 Workloads Across 3 Categories)
Each sample executes all 4 Boolean operations (union, intersect,
difference, xor) and reports per-operation throughput, latency percentiles, and
maximum inclusion-exclusion area error.
| Library | Ops / sec | Relative Speed | Mean (ms) | p50 (ms) | p95 (ms) | Area Identity Err | Status |
|---|