#performance

One method:

const t0 = performance.now();
await doWork();
console.log(`took ${(performance.now() - t0).toFixed(2)}ms`);

#performance.now()

console.log(performance.now());   // 0        (origin latched here)
console.log(performance.now());   // 0.041…

It reads the monotonic clock (.awake — CLOCK_MONOTONIC / UPTIME_RAW), latches the origin on first use, and divides nanoseconds by 1e6. Only performance.now exists — there are no mark, measure, timeOrigin, or navigation members.

#Timing patterns

#Measure a block

const t0 = performance.now();
const rows = db.rows("SELECT * FROM todos");
const ms = performance.now() - t0;
console.detail(`query: ${ms.toFixed(1)}ms (${rows.length} rows)`);

#Time a request end to end

http.serve({ port: 3000 }, async (url) => {
  const t0 = performance.now();
  const res = await fetch("https://example.com/");
  console.log(`${url} → ${res.status} in ${(performance.now() - t0).toFixed(1)}ms`);
  return new Response(await res.text(), { headers: { "content-type": "text/html" } });
});

#Benchmark loop

const N = 100_000;
const t0 = performance.now();
for (let i = 0; i < N; i++) JSON.stringify({ i });
const per = (performance.now() - t0) / N;
console.log(`${(per * 1e6).toFixed(0)}ns/op`);

#vs console.time

Use performance.now() for measurements; use console.time for human-visible timings in logs.

#Gotchas

• The origin is the first call, not process start — anything before your first performance.now() is invisible. Call it once at boot if you need a stable baseline.
• It's a float: compare and print with rounding (toFixed) to avoid 0.041000000000000002 noise.
• Sub-millisecond values are real but coarse — single samples are noise; average over iterations.
• In Workers each runtime has its own origin.

#See also

• Logging — timing in logs
• Timers — scheduling, not measuring
• Event loop