Experiment: Optional minification

[moreati]

Controlled by MITOGEN_MINIFY environment variable on controller.

Alternative to #942

[moreati]

import mitogen.master

def foo():
    return 42

if __name__ == '__main__':
    broker = mitogen.master.Broker()
    router = mitogen.master.Router(broker)

    context = router.local()

    result = context.call(foo)
    print(result)

➜  mitogen git:(minify-optional) ✗ hyperfine --warmup 1 'python2 hello_local.py' 'MITOGEN_MINIFY=0 python2 hello_local.py' 'python3.14 hello_local.py' 'MITOGEN_MINIFY=0 python3.14 hello_local.py'    
Benchmark 1: python2 hello_local.py
  Time (mean ± σ):     878.6 ms ±  14.7 ms    [User: 670.5 ms, System: 134.8 ms]
  Range (min … max):   857.2 ms … 908.0 ms    10 runs
 
Benchmark 2: MITOGEN_MINIFY=0 python2 hello_local.py
  Time (mean ± σ):     725.4 ms ±   5.4 ms    [User: 523.6 ms, System: 129.0 ms]
  Range (min … max):   717.5 ms … 734.9 ms    10 runs
 
Benchmark 3: python3.14 hello_local.py
  Time (mean ± σ):     276.4 ms ±   7.6 ms    [User: 198.9 ms, System: 27.6 ms]
  Range (min … max):   265.8 ms … 294.1 ms    11 runs
 
Benchmark 4: MITOGEN_MINIFY=0 python3.14 hello_local.py
  Time (mean ± σ):     261.8 ms ±   4.2 ms    [User: 185.8 ms, System: 26.6 ms]
  Range (min … max):   256.7 ms … 271.3 ms    10 runs
 
Summary
  MITOGEN_MINIFY=0 python3.14 hello_local.py ran
    1.06 ± 0.03 times faster than python3.14 hello_local.py
    2.77 ± 0.05 times faster than MITOGEN_MINIFY=0 python2 hello_local.py
    3.36 ± 0.08 times faster than python2 hello_local.py

[moreati]

import mitogen

def foo():
    return 42

@mitogen.main()
def main(router):
    context = router.local()
    result = context.call(foo)
    print(result)

➜  mitogen git:(minify-optional) ✗ hyperfine --warmup 1 'python2 hello_local2.py' 'MITOGEN_MINIFY=0 python2 hello_local2.py' 'python3.14 hello_local2.py' 'MITOGEN_MINIFY=0 python3.14 hello_local2.py'
Benchmark 1: python2 hello_local2.py
  Time (mean ± σ):     727.8 ms ±   3.8 ms    [User: 524.1 ms, System: 128.5 ms]
  Range (min … max):   721.5 ms … 733.6 ms    10 runs
 
Benchmark 2: MITOGEN_MINIFY=0 python2 hello_local2.py
  Time (mean ± σ):     666.5 ms ±  10.2 ms    [User: 463.7 ms, System: 126.5 ms]
  Range (min … max):   658.2 ms … 693.7 ms    10 runs
 
Benchmark 3: python3.14 hello_local2.py
  Time (mean ± σ):     195.8 ms ±   1.6 ms    [User: 123.3 ms, System: 22.2 ms]
  Range (min … max):   192.7 ms … 199.0 ms    15 runs
 
Benchmark 4: MITOGEN_MINIFY=0 python3.14 hello_local2.py
  Time (mean ± σ):     200.1 ms ±   7.6 ms    [User: 126.3 ms, System: 22.4 ms]
  Range (min … max):   194.5 ms … 226.9 ms    15 runs
 
  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs. It might help to use the '--warmup' or '--prepare' options.
 
Summary
  python3.14 hello_local2.py ran
    1.02 ± 0.04 times faster than MITOGEN_MINIFY=0 python3.14 hello_local2.py
    3.40 ± 0.06 times faster than MITOGEN_MINIFY=0 python2 hello_local2.py
    3.72 ± 0.04 times faster than python2 hello_local2.py

[moreati]

fork benchmarks are too noisy to judge by

import mitogen

def foo():
    return 42


@mitogen.main()
def main(router):
    context = router.fork()
    result = context.call(foo)
    print(result)

➜  mitogen git:(minify-optional) ✗ hyperfine --warmup 1 --min-runs 40 'python2 hello_fork.py' 'MITOGEN_MINIFY=0 python2 hello_fork.py' 'python3.14 hello_fork.py' 'MITOGEN_MINIFY=0 python3.14 hello_fork.py'
Benchmark 1: python2 hello_fork.py
  Time (mean ± σ):     125.0 ms ±  20.2 ms    [User: 66.7 ms, System: 39.1 ms]
  Range (min … max):   111.9 ms … 171.2 ms    40 runs
 
  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs. It might help to use the '--warmup' or '--prepare' options.
 
Benchmark 2: MITOGEN_MINIFY=0 python2 hello_fork.py
  Time (mean ± σ):     126.0 ms ±  19.9 ms    [User: 67.1 ms, System: 39.2 ms]
  Range (min … max):   112.2 ms … 169.1 ms    40 runs
 
  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs. It might help to use the '--warmup' or '--prepare' options.
 
Benchmark 3: python3.14 hello_fork.py
  Time (mean ± σ):      87.8 ms ±  24.6 ms    [User: 42.2 ms, System: 11.2 ms]
  Range (min … max):    49.6 ms … 107.7 ms    55 runs
 
  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs. It might help to use the '--warmup' or '--prepare' options.
 
Benchmark 4: MITOGEN_MINIFY=0 python3.14 hello_fork.py
  Time (mean ± σ):      90.1 ms ±  23.7 ms    [User: 42.5 ms, System: 11.3 ms]
  Range (min … max):    51.0 ms … 107.8 ms    54 runs
 
  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs. It might help to use the '--warmup' or '--prepare' options.
 
Summary
  python3.14 hello_fork.py ran
    1.03 ± 0.39 times faster than MITOGEN_MINIFY=0 python3.14 hello_fork.py
    1.42 ± 0.46 times faster than python2 hello_fork.py
    1.43 ± 0.46 times faster than MITOGEN_MINIFY=0 python2 hello_fork.py