README.md

Benchmarks

The benchmark suite separates algorithm cost from text tokenization:

• diffs.rs is a Criterion suite covering small text fixtures, large sparse and disjoint sequences, dense repeated inputs at multiple sizes, and end-to-end line tokenization.
• allocations.rs uses a tracking global allocator to report allocation calls, cumulative requested bytes, peak live requested bytes, and bytes retained by the result. Allocator metadata and fragmentation are intentionally excluded, which makes the output deterministic and useful for before/after comparisons.

Run the timing suite with:

cargo bench --bench diffs

Run the allocation report with:

cargo bench --bench allocations

Criterion filters can narrow a run, for example:

cargo bench --bench diffs -- dense_repeated_scaling

Input generation and fixture loading happen outside timed iterations. Every iteration includes construction and destruction of the resulting diff so that allocation and cleanup costs are represented.