Fibonacci Benchmark

Fibonenchmark

# Run complete. Total time: 00:34:14

Benchmark                                                   Mode  Samples     Score    Error  Units
o.g.m.f.FibonacciMicroBenchmark.groovy_primitive_cs         avgt       30     1.961 ±  0.055  ms/op
o.g.m.f.FibonacciMicroBenchmark.baseline_java_30            avgt       30     2.004 ±  0.084  ms/op
o.g.m.f.FibonacciMicroBenchmark.baseline_java_boxing_30     avgt       30     5.504 ±  0.036  ms/op
o.g.m.f.FibonacciMicroBenchmark.groovy_primitive_30         avgt       30     5.733 ±  0.024  ms/op
o.g.m.f.FibonacciMicroBenchmark.groovy_indy_30              avgt       30     5.903 ±  0.033  ms/op
o.g.m.f.FibonacciMicroBenchmark.golo_30                     avgt       30     6.545 ±  0.060  ms/op
o.g.m.f.FibonacciMicroBenchmark.nashorn_30                  avgt       30     7.812 ±  0.061  ms/op
o.g.m.f.FibonacciMicroBenchmark.groovy_indy_primitive_30    avgt       30     8.033 ±  0.130  ms/op
o.g.m.f.FibonacciMicroBenchmark.clojure_30                  avgt       30     9.277 ±  0.161  ms/op
o.g.m.f.FibonacciMicroBenchmark.groovy_30                   avgt       30    23.930 ±  0.228  ms/op
o.g.m.f.FibonacciMicroBenchmark.jruby_30                    avgt       30    25.868 ±  0.663  ms/op

o.g.m.f.FibonacciMicroBenchmark.groovy_primitive_cs_40      avgt       30   238.734 ±  0.994  ms/op
o.g.m.f.FibonacciMicroBenchmark.baseline_java_40            avgt       30   238.913 ±  2.652  ms/op
o.g.m.f.FibonacciMicroBenchmark.baseline_java_boxing_40     avgt       30   703.751 ± 15.198  ms/op
o.g.m.f.FibonacciMicroBenchmark.groovy_primitive_40         avgt       30   710.078 ±  6.669  ms/op
o.g.m.f.FibonacciMicroBenchmark.groovy_indy_40              avgt       30   726.623 ±  7.280  ms/op
o.g.m.f.FibonacciMicroBenchmark.golo_40                     avgt       30   800.427 ± 10.386  ms/op
o.g.m.f.FibonacciMicroBenchmark.nashorn_40                  avgt       30   961.904 ± 10.431  ms/op
o.g.m.f.FibonacciMicroBenchmark.groovy_indy_primitive_40    avgt       30   992.273 ± 23.452  ms/op
o.g.m.f.FibonacciMicroBenchmark.clojure_40                  avgt       30  1131.375 ±  7.095  ms/op
o.g.m.f.FibonacciMicroBenchmark.groovy_40                   avgt       30  2937.296 ± 21.417  ms/op
o.g.m.f.FibonacciMicroBenchmark.jruby_40                    avgt       30  3153.625 ± 46.717  ms/op

The numbers above are comparing the performance of the infamous Fibonacci benchmark for various dynamic languages. It computes fib(30) or fib(40).

Tests done with Groovy 2.4.0-SNAPSHOT (results may be updated).

VM:

java version "1.8.0_40-ea"
Java(TM) SE Runtime Environment (build 1.8.0_40-ea-b15)
Java HotSpot(TM) 64-Bit Server VM (build 25.40-b18, mixed mode)

Why so many Groovy versions of the benchmark ? Basically we tested:

• untyped Groovy: fib(n)
• typed Groovy: long fib(long n)
• invokedynamic versions of the above (_indy, Groovy has 2 runtimes: legacy and indy)
• statically compiled Groovy (_cs, using @CompileStatic)

see melix/golo-lang@a2148dc

Thanks, but what a thing happen with groovy indy results?

why groovy_indy_primitive is much slower than groovy_primitive_40 ?