Dataset v1.0.0

1D Cutting Benchmark Suite

Thirteen fixed, synthetic cutting-stock cases make material accounting inspectable. The suite measures the production Rust engine, then independently rebuilds every length balance and placement assertion before a result can pass.

Suite 1.0.0Methodology 1.0.013 casesCC BY 4.0 data

Dataset

13 synthetic cases

Versioned inputs and constraints are frozen for suite v1.0.0.

Validation

Independent accounting

Placement, inventory, kerf, trim, remnant and summary invariants are reconstructed.

Timing

Raw runs plus p50/p95

The execution target and timing boundary must be named with every result.

Publication

Reference run pending

No dirty-worktree diagnostic timing is presented as a public release.

Material accounting

Waste is not one number

Every used stock length must balance, within a 0.001 base-unit tolerance, into finished parts, kerf loss, trim loss, reusable remnants and scrap leftover.

Public metric	Definition
Finished-part yield	Finished required length divided by the full length of every used stock piece.
Process loss	Modeled kerf plus start and end trim on used stock.
Reusable remnant	Leftover meeting the case's declared reusable-offcut threshold.
Disposable material	Process loss plus leftover below the reusable threshold.
Recovery-adjusted efficiency	Finished parts plus reusable remnants divided by stock input.

Balance invariant: stock input = finished parts + kerf + trim + reusable remnants + scrap leftover.

Public case catalog

All suite cases stay visible

Failure-boundary cases are intentionally included. A useful benchmark cannot silently exclude oversized or inventory-exhausted inputs.

Case	Category	Expected placement	Allowed status	Coverage
1d-exact-fit	exact	All parts placed	optimal	exact-fitoptimal-proof
1d-kerf-exact-fit	kerf-trim	All parts placed	optimal	kerf-sensitiveexact-fitoptimal-proof
1d-asymmetric-trim	kerf-trim	All parts placed	optimal	trimkerfleftover
1d-offcut-threshold	remnant-quality	All parts placed	heuristic	reusable-remnantscrapthreshold
1d-wood-furniture	wood	All parts placed	heuristic	woodfurnituresaw-kerfremnant
1d-steel-fabrication	steel	All parts placed	optimal	steelmixed-stockoptimal-proof
1d-cost-tradeoff	mixed-stock	All parts placed	optimal	costmixed-stockobjective
1d-mixed-finite-stock	mixed-stock	All parts placed	heuristic	finite-inventorymixed-stocksteel
1d-pipe-batch	pipe	All parts placed	heuristic	pipebatchtrimkerf
1d-aluminium-profile	aluminium	All parts placed	heuristic	aluminiumprofilerepeated-cuts
1d-oversized-cut	failure-boundary	1 unplaced	heuristic	unplacedoversized
1d-stock-exhausted	failure-boundary	1 unplaced	heuristic	unplacedfinite-inventory
1d-high-piece-count	scale	All parts placed	heuristic	largeunlimited-stockrepeated-cuts

Reproducibility boundary

Results will point to an immutable dataset and engine

The public repository scaffold and dataset are ready locally. Result publication waits for a confirmed GitHub remote, an immutable dataset tag, a clean engine commit and a documented reference-machine run. That prevents local diagnostic timings from becoming permanent marketing claims.

Try the production 1D optimizer Review import templates