---

Speed comparison of programming languages

This projects tries to compare the speed of different programming languages. In this project we don't really care about getting a precise calculation of pi. We only want to see how fast are the programming languages doing.

It uses an implementation of the Leibniz formula for π to do the comparison.

Here is a video which explains how it works: Calculating π by hand

Disclaimer

I'm no expert in all these languages, so take my results with a grain of salt.
Also the findings just show how good the languages can handle floating point operations, which is only one aspect of a programming language.

You are also welcome to contribute and help me make fix my possible misuse of some languages. 😄

Languages used in this comparison

• Julia - JIT
• Python - interpreted (CPython)
• Ruby - interpreted
• Rust - compiled
• Javascript using Node.js - interpreted, JIT
• Lua - interpreted
• Nim - compiled
• PHP - interpreted
• C++ - compiled
• Crystal - compiled
• Go - compiled
• C - compiled

Results

======= Comparison =======
Iterations: 1000000

> Julia
Version: julia version 0.6.2
Speed (all): 747ms, 723ms, 715ms, 730ms, 724ms, 701ms, 718ms, 726ms, 710ms, 725ms
Speed (best): 701ms
Speed (worst): 747ms
Speed (median): 723.5ms
Result: 3.1415916535917745
Accuracy: 66.67%

> Python (CPython)
Version: Python 3.6.4
Speed (all): 245ms, 259ms, 223ms, 235ms, 247ms, 255ms, 260ms, 242ms, 226ms, 228ms
Speed (best): 223ms
Speed (worst): 260ms
Speed (median): 243.5ms
Result: 3.1415936535887745
Accuracy: 72.22%

> Ruby
Version: ruby 2.5.0p0 (2017-12-25 revision 61468) [x86_64-linux]
Speed (all): 139ms, 133ms, 125ms, 120ms, 124ms, 134ms, 127ms, 123ms, 133ms, 136ms
Speed (best): 120ms
Speed (worst): 139ms
Speed (median): 130.0ms
Result: 3.1415916535917745
Accuracy: 66.67%

> Java
Version: 8.u144-1
Speed (all): 72ms, 76ms, 73ms, 83ms, 75ms, 68ms, 77ms, 79ms, 76ms, 77ms
Speed (best): 68ms
Speed (worst): 83ms
Speed (median): 76.0ms
Result: 3.1415936535887745
Accuracy: 72.22%

> JS (node)
Version: v9.6.1
Speed (all): 69ms, 65ms, 67ms, 66ms, 66ms, 66ms, 66ms, 70ms, 72ms, 65ms
Speed (best): 65ms
Speed (worst): 72ms
Speed (median): 66.0ms
Result: 3.1415936535887745
Accuracy: 72.22%

> Python (pypy)
Version: 5.10.0-5
Speed (all): 112ms, 40ms, 41ms, 43ms, 42ms, 41ms, 40ms, 41ms, 41ms, 43ms
Speed (best): 40ms
Speed (worst): 112ms
Speed (median): 41.0ms
Result: 3.1415936535887745
Accuracy: 72.22%

> Lua
Version: Lua 5.3.4  Copyright (C) 1994-2017 Lua.org, PUC-Rio
Speed (all): 74ms, 67ms, 72ms, 70ms, 72ms, 72ms, 72ms, 68ms, 75ms, 76ms
Speed (best): 67ms
Speed (worst): 76ms
Speed (median): 72.0ms
Result: 3.1415916535917745
Accuracy: 66.67%

> Rust
Version: rustc 1.24.0
Speed (all): 75ms, 74ms, 73ms, 71ms, 74ms, 71ms, 71ms, 67ms, 68ms, 70ms
Speed (best): 67ms
Speed (worst): 75ms
Speed (median): 71.0ms
Result: 3.1415936535887745
Accuracy: 72.22%

> PHP
Version: PHP 7.2.2 (cli) (built: Jan 30 2018 19:18:38) ( NTS )
Speed (all): 60ms, 58ms, 57ms, 59ms, 58ms, 58ms, 58ms, 56ms, 55ms, 60ms
Speed (best): 55ms
Speed (worst): 60ms
Speed (median): 58.0ms
Result: 3.1415936535887745
Accuracy: 72.22%

> Nim
Version: 0.17.2-2
Speed (all): 43ms, 39ms, 45ms, 41ms, 40ms, 39ms, 40ms, 40ms, 38ms, 39ms
Speed (best): 38ms
Speed (worst): 45ms
Speed (median): 40.0ms
Result: 3.1415916535917745
Accuracy: 66.67%

> C++
Version: g++ (GCC) 7.3.0
Speed (all): 11ms, 10ms, 10ms, 9ms, 10ms, 9ms, 10ms, 8ms, 8ms, 8ms
Speed (best): 8ms
Speed (worst): 11ms
Speed (median): 9.5ms
Result: 3.1415936535887745
Accuracy: 72.22%

> Crystal
Version: Crystal 0.24.1 (2017-12-20)
Speed (all): 12ms, 10ms, 9ms, 9ms, 9ms, 9ms, 9ms, 9ms, 9ms, 10ms
Speed (best): 9ms
Speed (worst): 12ms
Speed (median): 9.0ms
Result: 3.1415916535917745
Accuracy: 66.67%

> Go
Version: go version go1.10 linux/amd64
Speed (all): 7ms, 6ms, 6ms, 6ms, 5ms, 5ms, 6ms, 5ms, 6ms, 6ms
Speed (best): 5ms
Speed (worst): 7ms
Speed (median): 6.0ms
Result: 3.1415936535887745
Accuracy: 72.22%

> C
Version: gcc (GCC) 7.3.0
Speed (all): 9ms, 7ms, 6ms, 7ms, 7ms, 6ms, 6ms, 6ms, 6ms, 6ms
Speed (best): 6ms
Speed (worst): 9ms
Speed (median): 6.0ms
Result: 3.1415936535887745
Accuracy: 72.22%

Run it yourself

Everything is run by a Docker container and a bash script which envokes the programs.

To measure the execution time a python package is used.

Requirements

• Docker
• Makefile support

Run measurement

Just run: make

Print plot

make plot

Run cli

make cli

FAQ

Why do you also count reading a file and printing the output?

Because I think this is a more realistic scenario to compare speeds.

Are the compile times included in the measurements?

No they are not included, because when running the program in the real world this would also be done before.

TODO

• Add C++
• Add run pypy for Python
• Produce graphic with results
• Add Java
• Add Kotlin
• Add Swift
• Add Ada
• Add Haskell
• Add Elixir

Thanks

This projects takes inspiration from Thomas who did a similar comparison on his blog.