LambdaMicrobenchmarking: microbenchmarking tool for C# and F#.

LambdaMicrobenchmarking is a library for microbenchmarking orchestration in C# and F#. The programer can measure the execution performance of thunks (lambdas with no arguments).

Func<long> sumLinq       = () => v.Sum();
Func<long> sumSqLinq     = () => v.Select(x => x * x).Sum();
Func<long> sumSqEvenLinq = () => v.Where(x => x % 2 == 0).Select(x => x * x).Sum();
Func<long> cartLinq      = () => (from x in vHi
                                  from y in vLow
                                  select x * y).Sum();
Script.Of("sumLinq", sumLinq)
      .Of("sumSqLinq", sumSqLinq)
      .Of("sumSqEvensLinq", sumSqEvenLinq)
      .Of("cartLinq", cartLinq)
      .WithHead()
      .RunAll();

The corresponding script for F#:

Script.Of("sumLinq", Func<int64> sumLinq)
      .Of("sumOfSquaresLinq", Func<int64> sumSqLinq)
      .Of("sumOfSquaresEvenLinq", Func<int64> sumSqEvenLinq)
      .Of("cartLinq", Func<int64> cartLinq)
      .WithHead()
      .RunAll |> ignore

Sample output:

Benchmark          Mean      Mean-Error  Sdev  Unit
sumLinq          98.324           0.543 0.359 ms/op
sumSqLinq       224.831           1.804 1.193 ms/op
sumSqEvenLinq    82,789           5,729 3,789 ms/op
cartLinq        200,775           6,477 4,284 ms/op

* the statistics part is inspired by JMH (AbstractStatistics, ListStatistics).

References

• Clash of the Lambdas
• Microbenchmarks in Java and C#
• JMH

Contributing

Sending PRs and participating in the discussion for improving LM is highly encouraged. Our goal is to create a tool for .NET that produces valid results, giving real control over the intented experiments.