src/connectivity/network/netstack3/bench/README.md

# Netstack3 Benchmarks

Netstack3 currently has microbenchmarks exercising the internals of Core,
but in the future this may be expanded to include both the Bindings and the
public API surface. Netstack3 benchmarks are based on the
[Criterion](https://docs.rs/criterion/latest/criterion/) benchmark
infrastructure. In Core, benchmark functions are declared with the
[`bench!` macro](https://cs.opensource.google/search?q=f:netstack3%20%22macro_rules!%20bench%22&ss=fuchsia)
 and can be run either as standard unit tests (picked up both by
`cargo test` and `fx test`), or to profile the performance of Netstack3. As
unit tests, the benchmarks may perform debug assertions to verify the
correctness of Netstack3. When run as performance benchmarks, the build is
optimized and all debug assertions are stripped, allowing the benchmark to
measure Netstack3's performance in a production-like environment.

## Running the Benchmarks

1. Add the benchmark test target to your `fx set` line, and configure your
   build for release (optimized). Note that the benchmarks rely on `SL4F`
   which is currently only available on terminal and workstation builds:

    ```
    fx set terminal.x64 --with //src/tests/end_to_end/perf:test --release
    ```

2. Build Fuchsia

    ```
    fx build
    ```

3. Start the Fuchsia emulator, using TAP networking.

    ```
    ffx emu start --headless --net tap
    ```

4. In a separate terminal, serve Fuchsia packages

    ```
    fx serve -v
    ```

5. Run the tests

    ```
    fx test --e2e netstack3_benchmarks
    ```

After completing, the tests will print the name of the
[catapult_json](https://github.com/catapult-project/catapult/blob/main/docs/histogram-set-json-format.md)
output file containing the benchmark results.