Benchmarks

zig build bench
zig build bench -Doptimize=ReleaseFast

benchmarks/parse_bench.zig generates a 20,000-operand expression corpus (~80KB, 100,000 nodes) and prints: source bytes, initial parse, repeated parse, worst-case leading-edge incremental parse with reused_nodes, traversal visits, and query matches — timed on the std.Io monotonic clock. Reference numbers live in Performance. When optimizing, profile the ReleaseFast numbers, change one thing, and keep the Debug suite green; per-node allocation or per-token allocation regressions show up immediately in the incremental line.

Related: Performance, Incremental Runtime.