Clone this repo:
  1. 5936c53 Update leakIframes URL param when toggling state in dev menu by Camillo Bruni · 7 weeks ago main-custom
  2. f227c26 Add params.leakyIframes to main-custom by Camillo Bruni · 8 weeks ago
  3. a4d98e2 Push the latest upstream main changes to main-custom by Camillo Bruni · 8 weeks ago
  4. 68d8f57 News Site Next: simplify message popup integration (#471) by Thorsten Kober · 3 months ago
  5. a3b0e71 News site Next: run build command (#474) by Thorsten Kober · 3 months ago

What is Speedometer?

Speedometer is a benchmark for web browsers that measures Web application responsiveness by timing simulated user interactions on various workloads. Our primary goal is to make it reflect the real-world Web as much as possible. When a browser improves its score on the benchmark, actual users should benefit. In order to achieve this, it should:

  • Test end-to-end user journeys instead of testing specific features in a tight loop. Each test should exercise the full set of what’s needed from the engine in order for a user to accomplish a task.
  • Evolve over time, adapting to the present Web on a regular basis. This should be informed by current usage data, and by consensus about features which are important for engines to optimize to provide a consistent experience for users and site authors.
  • Be accessible to the public and useful to browser engineers. It should run in every modern browser by visiting a normal web page. It should run relatively quickly, while providing enough test coverage to be reflective of the real-world Web.

Score and Test Measurement

Each test can contain several steps who contributed to the test duration. Note that the prepare step of a test is unmeasured and thus does not contribute to the score. All following test steps are measured and summed up in the test step time.

Each step consists of the following phases:

Score

The final benchmark score is calculated based of the inverse of geomean of all tests. We average the score over multiple iterations with the arithmetic mean. By using the geomean to combine the test durations we maintain the invariant that relative improvements are favoured equally amongst all tests, even though they can have vastly different durations.