mirror of
https://github.com/flutter/flutter.git
synced 2025-06-03 00:51:18 +00:00

1. Allow for selective benchmarks to run locally (see README.md) 2. Shuffle the tests; the seed being rotated daily or specified (see README.md)
31 lines
960 B
Markdown
31 lines
960 B
Markdown
# microbenchmarks
|
|
|
|
To run these benchmarks on a device, first run `flutter logs` in one
|
|
window to see the device logs, then, in a different window, run:
|
|
|
|
```sh
|
|
flutter run -d $DEVICE_ID --profile lib/benchmark_collection.dart
|
|
```
|
|
|
|
To run a subset of tests:
|
|
|
|
```shell
|
|
flutter run -d $DEVICE_ID --profile lib/benchmark_collection.dart --dart-define=tests=foundation/change_notifier_bench.dart,language/sync_star_bench.dart
|
|
```
|
|
|
|
To specify a seed value for shuffling tests:
|
|
|
|
```shell
|
|
flutter run -d $DEVICE_ID --profile lib/benchmark_collection.dart --dart-define=seed=12345
|
|
```
|
|
|
|
|
|
The results should be in the device logs.
|
|
|
|
## Avoid changing names of the benchmarks
|
|
|
|
Each microbenchmark is identified by a name, for example,
|
|
"catmullrom_transform_iteration". Changing the name passed to `BenchmarkResultPrinter.addResult`
|
|
will effectively remove the old benchmark and create a new one,
|
|
losing the historical data associated with the old benchmark in the process.
|