Executing library benchmark: unicode-segmenter

[cometkim]

Hey, I tried to compile and execute my library unicode-semgneter which is pure JS alternative to Intl.Segmenter API

You can fine the case here: cometkim/unicode-segmenter#61

It compiles and executes well. But the printed output looks like a false positive results:

❯ yarn porf benchmark/porffor/run.out.js

unicode-segmenter/grapheme (small input)
samples: 1000
duration (avg): 0.789

unicode-segmenter/grapheme (medium input)
samples: 1000
duration (avg): 0.793

Here is the result on Hermes for the comparison:

unicode-segmenter/grapheme (small input)
samples: 1000
duration (avg): 0.195

unicode-segmenter/grapheme (medium input)
samples: 1000
medium input (avg): 0.699

And even stranger, if I increase the value of samples 2x (:%s/1e3/2e3/g), the bench emits undefined values.

unicode-segmenter/grapheme (small input)
samples: 2000
duration (avg): 0.7715

unicode-segmenter/grapheme (medium input)
samples: undefined
duration (avg): undefined

And if I increase the number even more, like to 5000, it's takes very very long time and print something like this

unicode-segmenter/grapheme (small input)


0 0 2 undefined   undefined  {} undefined 0 0 2 undefined   undefined  {} undefined undefined undefined undefined undefined undefined
unicode-segmenter/grapheme (medium input)

(exited 0)