felixr / janet-benchmarks

janet-benchmarks

A suite of benchmarks targeted at exposing performance problems.

How to read the output

Each benchmark+param is run 3 times and the min/max time spent is printed. Example when run with version 1.12.2:

ver:     1.12.2
build:   b91fe8be

noop
  param=0 ...           min=0.007 max=0.007
hexagon_tuples
  param=50 ...          min=2.677 max=2.717
table_with_tuple_keys
  param="0 200" ...             min=2.500 max=2.543
  param="1000 200" ...          min=0.268 max=0.279
  param="10000 200" ...         min=0.044 max=0.045
  param="100000 200" ...        min=0.016 max=0.017
  param="1000000 200" ...       min=0.059 max=0.061
  param="10000000 200" ...      min=5.503 max=5.647
dict_words
  param=1000000 ...             min=0.387 max=0.403

This output shows that the table_with_tuple_keys takes much long if the numbers used are in the range of 0 to 200 or 10,000,000 to 10,000,200.

Benchmarks under a heading (eg. table_with_tuple_keys) should all do the same amount of work and take the same amount of time.

Results

Non dev versions are releases downloaded from Github. Names consist of janet/version and janet/build. All benchmarks are run on Intel(R) Core(TM) i7-1165G7 @ 2.80GHz.