Benchmarks

Test System

• 15-inch Macbook Air M2, 2023 (my personal laptop)
• rust-parallel uses 8 parallel jobs by default on this machine
• MacOS 13.6

Test Setup

• Run echo on 1000 strings 1 to 1000
• Benchmarks run using hyperfine testing tool
• Link to benchmark script

Summary

• rust-parallel is 1.43 times faster when the path cache is enabled (this is enabled by default)
• rust-parallel is 1.65 times faster than xargs with 8 processes. Note that this use of xargs does not prevent output interleaving.
• rust-parallel is 13.34 times faster than GNU parallel

Versions

$ rust-parallel -V
rust-parallel 1.10.3

$ parallel -V
GNU parallel 20230922

Results

$ hyperfine --warmup 3 \
  'seq 1 1000 | rust-parallel echo' \
  'seq 1 1000 | rust-parallel --disable-path-cache echo' \
  'seq 1 1000 | xargs -P8 -L1 echo' \
  'seq 1 1000 | parallel echo'

Benchmark 1: seq 1 1000 | rust-parallel echo
  Time (mean ± σ):     147.5 ms ±   1.4 ms    [User: 225.0 ms, System: 737.1 ms]
  Range (min … max):   145.3 ms … 150.8 ms    19 runs

Benchmark 2: seq 1 1000 | rust-parallel --disable-path-cache echo
  Time (mean ± σ):     210.3 ms ±   1.5 ms    [User: 253.3 ms, System: 981.8 ms]
  Range (min … max):   208.3 ms … 212.7 ms    14 runs

Benchmark 3: seq 1 1000 | xargs -P8 -L1 echo
  Time (mean ± σ):     243.0 ms ±   1.8 ms    [User: 145.9 ms, System: 542.7 ms]
  Range (min … max):   240.5 ms … 247.4 ms    12 runs

Benchmark 4: seq 1 1000 | parallel echo
  Time (mean ± σ):      1.968 s ±  0.012 s    [User: 2.436 s, System: 3.561 s]
  Range (min … max):    1.951 s …  1.989 s    10 runs

Summary
  seq 1 1000 | rust-parallel echo ran
    1.43 ± 0.02 times faster than seq 1 1000 | rust-parallel --disable-path-cache echo
    1.65 ± 0.02 times faster than seq 1 1000 | xargs -P8 -L1 echo
   13.34 ± 0.15 times faster than seq 1 1000 | parallel echo