Per-phase latency
Median time per phase (µs, log scale). Parse/validate/small execute run in microseconds.
Per-phase throughput
Operations per second per phase (log scale).
Execution scaling
Median execution time (ms) as payload size grows.
Latency per object
µs per resolved object across list sizes — a flat line means linear scaling.
Engine-to-engine comparison
This engine vs webonyx/graphql-php (ms, lower is better).
Performance over releases
Full parse+validate+execute of a 100-object query (ms) per released version — a fixed same-machine backfill (engine-only). The drop at 1.2.1 is the executor rewrite (inline synchronous completion).