tree: a96d6ebb5d2007e0505d00663fe19fadb2ad8655 [path history] [tgz]
  1. expected_metric_names/
  2. test/
  3. BUILD.gn
  4. OWNERS
  5. pubspec.yaml
  6. README.md
src/tests/end_to_end/perf/README.md

Performance tests

This directory contains performance tests. In many cases, the code in this directory is just an SL4F wrapper (that is, Dart code using the SL4F framework) for a test that is located elsewhere in the Fuchsia source tree.

Check the instructions to run these test locally.

This directory contains the following performance tests written in Dart:

  • audio_mixer_profiler_test - SL4F wrapper for profiling the different steps in the audio mixing process.

  • fidl_microbenchmarks_test - SL4F wrapper for running FIDL microbenchmarks which test FIDL bindings for multiple languages.

  • flatland_benchmarks_test - Tests the performance of Flatland's end-to-end present latency, measured by tracing flow events, for minimal Flatland client (flatland-view-provider).

  • input_latency_benchmarks_test - Tests the performance of end-to-end input latency, measured by tracing flow events, for minimal Scenic clients (simplest_app).

  • kernel_boot_stats_test - SL4F wrapper for a test that records the time taken by different parts of the kernel boot process.

  • microbenchmarks_test - SL4F wrapper for fuchsia_microbenchmarks. The microbenchmarks this contains tend to be for low-level operations such as Zircon syscalls.

  • netstack_benchmarks_test - SL4F wrapper for benchmarks of TCP, UDP, and ICMP echo sockets. These benchmarks run against Netstack2, Netstack2 with Fast UDP enabled, and a “fake netstack” that does minimal work in order to isolate the API overhead.

  • netstack_iperf_test - Tests network stack performance using the benchmarking tool iperf3.

  • rust_inspect_benchmarks_test - Tests the performance of various Rust Inspect operations, such as creating and deleting nodes and updating properties.

This directory contains the following performance test written in Python using Lacewing:

  • py_flatland_benchmark - Tests the performance of Flatland's end-to-end present latency, measured by tracing flow events, for minimal Flatland client (flatland-view-provider).

  • py_input_latency_benchmark - Tests the performance of end-to-end input latency, measured by tracing flow events, for minimal Scenic clients (simplest-app-flatland-session).

As well as the following examples:

  • perf_publish_example - Simple example test that publishes a performance metric.

You can view the test results from CI builds in Chromeperf.