Skip to the content.

Engineering diagram performance evidence

NeqSim records a small, synthetic performance baseline for the representative multi-area engineering-diagram delivery path. The benchmark is regression evidence: it detects unexpectedly slow or nondeterministic changes in review, but it is not a capacity claim, standards-conformance assessment, engineering approval, or fitness-for-construction determination.

Covered operations

EngineeringDiagramPerformanceBenchmarkTest measures the median and maximum wall-clock time from three recorded samples after one warm-up run for:

Every sample also produces a semantic output fingerprint. The validator rejects an operation when the fingerprints differ within a run, when the operation set drifts from the controlled budget, or when the recorded median exceeds its budget. The reference fixture includes multiple areas and explicit material, energy, and information connections.

Package intake measures exact-content inspection and reassessment. It does not reconstruct a ProcessModel. Rendering measures the repository’s native SVG/PDF projection; it does not measure Graphviz, external DEXPI viewers, commercial CAE tools, or accountable drawing review.

Run locally

./mvnw -B -ntp \
  -Dtest=EngineeringDiagramPerformanceBenchmarkTest \
  -Dgroups=benchmark \
  -DexcludedTestGroups= \
  -Djacoco.skip=true \
  test

python devtools/validate_engineering_diagram_performance.py \
  --report target/engineering-diagram-performance.json \
  --budget devtools/baselines/engineering_diagram_performance.json

python -m unittest devtools.test_validate_engineering_diagram_performance

The benchmark writes target/engineering-diagram-performance.json. Its fixed boundary fields are engineeringStatus: PERFORMANCE_REGRESSION_EVIDENCE_ONLY, approvalStatus: REVIEW_REQUIRED, and fitnessForConstruction: false.

Budget policy

Budgets in devtools/baselines/engineering_diagram_performance.json are intentionally conservative hosted-runner medians. A single slow sample is retained in the artifact as maximum but does not fail the gate. Change budgets only with repeated evidence from comparable runners and explain the reason in review; never relax them to hide a deterministic functional regression.

This gate does not establish throughput, scalability, production sizing, or release completion on its own. Final completion still requires the independent current-master audit and the applicable validation, interoperability, licensing, and accountable-engineering decisions documented in the DEXPI and P&ID current-master audit.