Add ParallelBenchmark

[manufacturist]

As requested by @vasilmkd, moved the ParallelBenchmark to a new PR. The reason for the 6 benchmarks is to compare the performance of parTraverse vs traverse under different CPU load.

Posted the benchmarks for the series/3.x:

[info] Benchmark                                    (size)   Mode  Cnt    Score   Error  Units
[info] ParallelBenchmark.parTraverseCpuTokens100     10000  thrpt   20   52.561 ± 1.691  ops/s
[info] ParallelBenchmark.parTraverseCpuTokens1000    10000  thrpt   20   44.141 ± 1.213  ops/s
[info] ParallelBenchmark.parTraverseCpuTokens10000   10000  thrpt   20   18.174 ± 0.254  ops/s
[info] ParallelBenchmark.traverseCpuTokens100        10000  thrpt   20  741.288 ± 9.223  ops/s
[info] ParallelBenchmark.traverseCpuTokens1000       10000  thrpt   20  134.859 ± 5.774  ops/s
[info] ParallelBenchmark.traverseCpuTokens10000      10000  thrpt   20   15.097 ± 0.458  ops/s

LE:

Executing Machine Details

Processor Name:         Quad-Core Intel Core i7
Processor Speed:        2,2 GHz
Number of Processors:   1
Total Number of Cores:  4

[djspiewak]

Outstanding! If I may ask, how many physical threads does your machine have? (Runtime.getRuntime().availableProcessors())

Edit: Oh, reading. :-D

[manufacturist]

Outstanding! If I may ask, how many physical threads does your machine have? (Runtime.getRuntime().availableProcessors())

^ Just updated the machine details

[vasilmkd]

I'm running the benchmark too, I will post base results.

[vasilmkd]

Benchmark                                    (size)   Mode  Cnt    Score   Error  Units
ParallelBenchmark.parTraverseCpuTokens100     10000  thrpt   20   95.980 ± 0.942  ops/s
ParallelBenchmark.parTraverseCpuTokens1000    10000  thrpt   20   67.336 ± 8.235  ops/s
ParallelBenchmark.parTraverseCpuTokens10000   10000  thrpt   20   26.116 ± 0.104  ops/s
ParallelBenchmark.traverseCpuTokens100        10000  thrpt   20  361.982 ± 7.369  ops/s
ParallelBenchmark.traverseCpuTokens1000       10000  thrpt   20   54.546 ± 0.286  ops/s
ParallelBenchmark.traverseCpuTokens10000      10000  thrpt   20    5.711 ± 0.016  ops/s

[djspiewak]

Hold on merging this. Some changes @manufacturist and I discussed directly

[vasilmkd]

Waiting for discussed changes then.

[manufacturist]

Ran the benchmark with cpuTokens values 100 & 100000 as well to get more data & pushed the discussed changes from yesterday:

• Scale the cpuTokens from 0.1k - 10k to 1k & 100k Scale cpuTokens to 0.1k - 1kk
• Parametrize size as well

[info] Benchmark                      (cpuTokens)  (size)   Mode  Cnt      Score     Error  Units
[info] ParallelBenchmark.parTraverse          100     100  thrpt   20   6487.985 ± 251.364  ops/s
[info] ParallelBenchmark.parTraverse          100    1000  thrpt   20    791.526 ±   6.699  ops/s
[info] ParallelBenchmark.parTraverse          100   10000  thrpt   20     75.203 ±   0.372  ops/s
[info] ParallelBenchmark.parTraverse         1000     100  thrpt   20   4505.063 ± 170.821  ops/s
[info] ParallelBenchmark.parTraverse         1000    1000  thrpt   20    545.267 ±   3.686  ops/s
[info] ParallelBenchmark.parTraverse         1000   10000  thrpt   20     58.031 ±   4.109  ops/s
[info] ParallelBenchmark.parTraverse        10000     100  thrpt   20   2294.218 ±   7.110  ops/s
[info] ParallelBenchmark.parTraverse        10000    1000  thrpt   20    237.894 ±   1.132  ops/s
[info] ParallelBenchmark.parTraverse        10000   10000  thrpt   20     22.876 ±   0.288  ops/s
[info] ParallelBenchmark.parTraverse       100000     100  thrpt   20    320.065 ±   0.294  ops/s
[info] ParallelBenchmark.parTraverse       100000    1000  thrpt   20     32.039 ±   0.173  ops/s
[info] ParallelBenchmark.parTraverse       100000   10000  thrpt   20      3.186 ±   0.040  ops/s
[info] ParallelBenchmark.parTraverse      1000000     100  thrpt   20     33.128 ±   0.042  ops/s
[info] ParallelBenchmark.parTraverse      1000000    1000  thrpt   20      3.318 ±   0.008  ops/s
[info] ParallelBenchmark.parTraverse      1000000   10000  thrpt   20      0.351 ±   0.035  ops/s
[info] ParallelBenchmark.traverse             100     100  thrpt   20  68114.475 ± 750.640  ops/s
[info] ParallelBenchmark.traverse             100    1000  thrpt   20   8024.076 ± 104.133  ops/s
[info] ParallelBenchmark.traverse             100   10000  thrpt   20    918.410 ±   4.294  ops/s
[info] ParallelBenchmark.traverse            1000     100  thrpt   20  15431.432 ±  32.394  ops/s
[info] ParallelBenchmark.traverse            1000    1000  thrpt   20   1615.529 ±  19.228  ops/s
[info] ParallelBenchmark.traverse            1000   10000  thrpt   20    166.345 ±   0.537  ops/s
[info] ParallelBenchmark.traverse           10000     100  thrpt   20   1766.005 ±   7.986  ops/s
[info] ParallelBenchmark.traverse           10000    1000  thrpt   20    178.010 ±   0.157  ops/s
[info] ParallelBenchmark.traverse           10000   10000  thrpt   20     17.718 ±   0.106  ops/s
[info] ParallelBenchmark.traverse          100000     100  thrpt   20    179.566 ±   0.178  ops/s
[info] ParallelBenchmark.traverse          100000    1000  thrpt   20     17.847 ±   0.110  ops/s
[info] ParallelBenchmark.traverse          100000   10000  thrpt   20      1.728 ±   0.061  ops/s
[info] ParallelBenchmark.traverse         1000000     100  thrpt   20     17.867 ±   0.094  ops/s
[info] ParallelBenchmark.traverse         1000000    1000  thrpt   20      1.717 ±   0.084  ops/s
[info] ParallelBenchmark.traverse         1000000   10000  thrpt   20      0.178 ±   0.006  ops/s

[vasilmkd]

Excellent.

[djspiewak]

Outstanding! That's actually more data points than I was expecting. 🙂 This is going to be super useful for some other changes as well, thank you!

[vasilmkd]

Thank you @manufacturist! Great work.