Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/michael-tiger-2010/wyvernjs/llms.txt

Use this file to discover all available pages before exploring further.

ShuiHu’s sh.time object is a Proxy that creates benchmark timers on demand — no setup required beyond accessing a property name. Each timer records a start time via performance.now(), up to 500 step measurements stored as flag+timestamp pairs in a Uint32Array (1000 slots, 2 per step), and a stop time. Total overhead is approximately 0.06ms per timer, making it suitable for profiling hot paths without significantly skewing results.

​
API

All benchmark methods are accessed via sh.time.<label>.<method>(). The <label> is any valid JavaScript property name and uniquely identifies the running benchmark.

​
sh.time.<label>.start()

Starts the benchmark timer. Records performance.now() (rounded to one decimal place) as the start time and registers the benchmark under <label> in internal storage. Calls sh.error() and returns null if a benchmark with the same label is already running. Always pair each start() with a stop() or cancel(). 
sh.time.loadData.start();

​
sh.time.<label>.step(flag)

Records a step checkpoint relative to the benchmark’s start time. flag must be a number — it acts as a named identifier in the output (for example, a loop index, phase number, or operation code). Each step stores two values in the Uint32Array: the elapsed time and the flag. Calls sh.error() and returns null if:
  • flag is not a number.
  • The benchmark does not exist (i.e. start() was not called).
  • The step count would exceed 500 (internal index has reached 1000, past the 998 write boundary).
sh.time.renderLoop.step(0);  // flag 0 at this point in time
sh.time.renderLoop.step(42); // flag 42 — e.g. an iteration index

​
sh.time.<label>.stop()

Stops the benchmark, assembles a formatted report string, logs it internally under the 'benchmark' label (level 0), and deletes the benchmark from memory. Calls sh.error() and returns null if the benchmark does not exist. 
sh.time.loadData.stop();

​
sh.time.<label>.cancel()

Removes the benchmark from memory without logging any output. Use this when you want to abort timing without emitting a report. Calls sh.error() and returns null if the benchmark does not exist. 
sh.time.experiment.cancel();

​
Output format

When stop() is called, ShuiHu logs a report with the 'benchmark' label. Each step line is numbered (1-based) and prefixed with f followed by the flag value: 
benchmark: myLabel
Start: 12.3
End: 45.6
Diff: 33.3
Steps:
1 f0: 5.2
2 f1: 12.4
...
  • Start / End — absolute performance.now() values in milliseconds, rounded to one decimal place.
  • Diff — total elapsed time (End - Start).
  • Steps — each step’s elapsed time since Start, paired with its flag value.

​
Examples

// Basic benchmark — no steps
sh.time.loadData.start();
const data = await fetchJson('/api/data');
sh.time.loadData.stop();

// With step flags to trace a render loop
sh.time.renderLoop.start();
for (let i = 0; i < items.length; i++) {
  renderItem(items[i]);
  sh.time.renderLoop.step(i); // flag is the iteration index
}
sh.time.renderLoop.stop();

// Cancel a benchmark when an early exit occurs
sh.time.experiment.start();
if (earlyExit) {
  sh.time.experiment.cancel(); // no output emitted
} else {
  sh.time.experiment.stop();
}
Use step flags as meaningful identifiers — for example, use the index in a loop, a phase number, or an operation code — so you can correlate each step line in the log output to a specific iteration or stage when reading the benchmark report.
Each benchmark supports a maximum of 500 steps. The 500th step writes to positions [998] and [999] of the internal Uint32Array, leaving the index at 1000. Any attempt to record a 501st step finds index > 998 and triggers sh.error() — the step is not recorded. If you need to profile loops longer than 500 iterations, consider sampling — for example, only calling step() every N iterations.

Build docs developers (and LLMs) love

Get started for freeTalk to us