Status: In-Development

See https://github.com/angular/benchpress for information about benchpress.

This project allows automated execution of already-built benchpress benchmarks through Karma.

See example/karma-experiment.conf.js for reference configuration and see [example/jasmine-spec-experiment.spec.js] for reference spec.

bpSuite

This plugin comes with an adapter that provides a global function called bpSuite that allows imperative execution of a benchmark with custom configuration. bpSuite returns a promise that will resolve with the result object

it('should be within acceptable limits', function(done) {
  bpSuite({url: 'base/largetable/index-auto.html', variable: 'ngBind', numSamples: 15, iterations: 20}).
    then(function(result) {
      expect(result.$apply.testTime.avg.mean).toBeLessThan(15);
      expect(result.create.testTime.avg.mean).toBeLessThan(1500);
      done();
    }, function(reason) {
      console.error('failed because', reason.message);
    }).then(null, function(e) { console.error('something went wrong', e); done()});
});

The bpSuite options object can have the following properties:

| Property | Required | Description | +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ | url | yes | Url of the benchmark to be executed (ie. base/largetable/index-auto.html) | | variable | no | Benchpress variable with which to run the benchmarks | | numSamples | no | How many samples to collect. Will use benchpress default (currently 20) if not specified | | iterations | no | How many iterations to run (should be greater than samples). Will use benchpress default (currently 25) if not specified |

The adapter will serialize all non-reserved properties into query parameters in the benchmark's url, so any supported benchpress paramaters may be used here.

Result Object

The result object has the following structure:

{
  stepName1 : ...,
  stepName2: {
    testTime: {
      avg: {
        mean: 5.0, //milliseconds of how long the step took to execute
        stdDev: 0.1, //Standard deviation of sample
        coefficientOfVariation: 0.02 //stdDev as percentage of mean
      },
      min: 4,
      max: 6,
      history: [
        4.0,
        5.0,
        6.0
      ]
    },
    //All other measured characteristics have the same structure, though they may represent values other than time
    gcTime: ..., //time spent collecting garbage after this step
    garbageCount: ..., //how much garbage was generated during this step in KB
    retainedCount: ..., //how much memory is retained by the step
  }
}

Gotchas & FAQs

• jasmine.DEFAULT_TIMEOUT_INTERVAL should be set to a number high enough to run all iterations of the benchmark. This would be best inside of a beforeEach/afterEach which will set the interval back to its original value
• A custom launcher based on Chrome Canary should be used to provide the best memory profiling characteristics. An example launcher configuration can be found in the example karma config.
• This plugin and the bpSuite function have no dependency on Jasmine, though Jasmine is the only framework with which it has been tested.