InlineBenchmarker is an easy to use inline benchmarking system.
The goal of inline benchmarking is to help you find bottle-necks in your application, rather than to benchmark abstract units.
Therefor Dart's benchmark harness library would probably be better suited.
Inline benchmarks can be used in projects that are otherwise hard to benchmark (perhaps because they are too much of a spaghetti), or to monitor alghorithms on how they respond with production processes and data.
void loginMethod()
{
Benchmark loginBenchmark = Benchmarker.generate("loginBenchmark");
// Perform login logic
loginBenchmark.stop();
Benchmarker.report();
}
loginBenchmark
is the name of the benchmarkBenchmarker.report();
will print out the results of the benchmark
[DEFAULT] loginBenchmark: Runtime: 1000 us
[DEFAULT]
stands for the group the benchmark belongs to1000 us
is how long the benchmark took to run
µs is an abbreviation for microseconds.
void calculateMethod()
{
Benchmark calculationBenchmark = Benchmarker.generate("calculationBenchmark", autoRun: false);
for (int i = 0; i < 100 ; i++)
{
calculationBenchmark.start();
// Perform calculation logic
calculationBenchmark.stop();
}
Benchmarker.report();
}
- Setting
autoRun
to false will prevent the benchmark from running instantly. - Benchmarks can be started manually using the
start()
method
[DEFAULT] calculationBenchmark:
› (1/100) Runtime: 10 us
› (2/100) Runtime: 10 us
› (3/100) Runtime: 10 us
...
› (100/100) Runtime: 10 us
(1/100)
is which iteration was benchmarked
For more complex applications, which combine nested methods and loops, it might be useful to format the output text using the group
and indentation
parameters.
void calculateMethod()
{
Benchmark calculateMethodBenchmark = Benchmarker.generate("Calculation Benchmark", group: 'CALCULATION');
Benchmark calculateLoopBenchmark = Benchmarker.generate("Calculation Loop Benchmark", group: 'CALCULATION', indentation: 1, autoStart: false);
for (int i = 0; i < 100 ; i++)
{
calculateLoopBenchmark.start();
// Perform calculation logic
calculateOtherStuffMethod();
calculateLoopBenchmark.stop();
}
calculateMethodBenchmark.stop();
Benchmarker.report(group: 'CALCULATION');
}
void calculateOtherStuffMethod()
{
Benchmark calculateOtherStuffMethodBenchmark = Benchmarker.generate("Calculate Other Stuff Benchmark", group: 'CALCULATION', indentation: 2);
// Perform other calculation logic
calculateOtherStuffMethodBenchmark.stop();
}
- The
group
parameter will group benchmarks together in the output console. - When no
group
is specified, a 'DEFAULT' group will be created - The
indentation
parameter will indent the output with the number of tabs specified Benchmarker.report(group: 'CALCULATION');
will only report the benchmark specified by thegroup
parameter
[CALCULATION] Calculation Benchmark: Runtime: 10000 us
[CALCULATION] Calculation Loop Benchmark:
› (1/100) Runtime: 10 us
[CALCULATION] Calculate Other Stuff Benchmark: Runtime: 10 us
› (2/100) Runtime: 10 us
[CALCULATION] Calculate Other Stuff Benchmark: Runtime: 10 us
› (3/100) Runtime: 10 us
[CALCULATION] Calculate Other Stuff Benchmark: Runtime: 10 us
...
› (100/100) Runtime: 10 us
[CALCULATION] Calculate Other Stuff Benchmark: Runtime: 10 us
When you are running in production mode, you probably don't want the benchmarks running constantly. A better option would be to give your application the ability to enable/disable benchmarks (perhaps by passing a URL parameter, in the case of a web application).
Disabling benchmarks is very easy:
Benchmarker.disable();
Enable benchmarks to produce a report
For more info about this project, contact: