# Browser Startup Benchmark

This benchmark measures the time to the first non-empty paint (`InitialWebUI.Startup.ReloadButton.FirstContentfulPaint`) for the browser startup sequence. It compares a **baseline** configuration against one with the **InitialWebUI** feature enabled.

## Running the Benchmark

This configuration uses the `loading` subcommand. To run the benchmark locally and compare the variants using Poetry, first install the required dependencies:

```bash
poetry install
```

Then run the benchmark:

```bash
poetry run cb loading \
  --browser-config config/benchmark/browser_startup/browser.config.hjson \
  --probe-config config/benchmark/browser_startup/probe.hjson \
  --page-config config/benchmark/browser_startup/story.hjson \
  --env-validation=warn \
  --repeat=30
```

> [!NOTE]
> If you are in a Chromium development environment with `depot_tools` (`vpython3`), you can run `./cb.py loading ...` directly without `poetry install`.

## Interpreting Results

After the run finishes, Crossbench outputs a path to the results directory.
Within the `trace_processor/` sub-directory, look for `reload_button_fcp.csv` (or `.json`). This file contains the calculated `reload_button_fcp_ms` metric for each run, which you can use to directly compare the performance of each variant.