Skip to content

Benchmarks

Aaron Riekenberg edited this page Apr 27, 2024 · 102 revisions

Test System

  • 15-inch Macbook Air M2, 2023 (my personal laptop)
  • rust-parallel uses 8 parallel jobs by default on this machine
  • MacOS 14.4.1

Test Setup

Summary

  • rust-parallel is 1.38 times faster when the path cache is enabled (this is enabled by default)
  • rust-parallel is 1.67 times faster than xargs with 8 processes. Note that this use of xargs does not prevent output interleaving.
  • rust-parallel is 14.07 times faster than GNU parallel

Versions

$ rust-parallel -V
rust-parallel 1.18.0

$ parallel -V
GNU parallel 20240422

Results

Benchmark 1: seq 1 1000 | rust-parallel echo
  Time (mean ± σ):     139.5 ms ±   1.6 ms    [User: 236.5 ms, System: 577.3 ms]
  Range (min … max):   137.8 ms … 145.4 ms    21 runs

Benchmark 2: seq 1 1000 | rust-parallel --disable-path-cache echo
  Time (mean ± σ):     191.8 ms ±   0.9 ms    [User: 245.9 ms, System: 786.4 ms]
  Range (min … max):   190.5 ms … 194.3 ms    15 runs

Benchmark 3: seq 1 1000 | xargs -P8 -L1 echo
  Time (mean ± σ):     233.7 ms ±   1.4 ms    [User: 151.0 ms, System: 451.8 ms]
  Range (min … max):   230.8 ms … 236.6 ms    12 runs

Benchmark 4: seq 1 1000 | parallel echo
  Time (mean ± σ):      1.963 s ±  0.042 s    [User: 2.678 s, System: 3.094 s]
  Range (min … max):    1.916 s …  2.072 s    10 runs

Summary
  seq 1 1000 | rust-parallel echo ran
    1.38 ± 0.02 times faster than seq 1 1000 | rust-parallel --disable-path-cache echo
    1.67 ± 0.02 times faster than seq 1 1000 | xargs -P8 -L1 echo
   14.07 ± 0.34 times faster than seq 1 1000 | parallel echo
Clone this wiki locally