Awesome

Recursive Fibonacci Benchmark using top languages on Github

Top 10: JavaScript, Python, Java, TypeScript, C#, Php, C++, C, Shell, Ruby reference

Others: Go, Rust, Swift, Crystal, Pony, Ada, Pascal, Fortran, Kotlin, Clojure, Scala, Mono, R, Dart, Julia, D, Nim, Cython, Python3, PyPy, Ruby jit, OCaml, Lisp, Haskell, Erlang, Elixir, Escript, Dart, Scheme, Lua, Perl, Perl6, Bash, Emoji

The code performs a recursive fibonacci to the 47th position with the result of 2,971,215,073.

Fibonacci can be written many different ways. The goal of this project is to compare how each language handles the exact same code.

Here is the Ruby version:

def fib(n)
  return n if n <= 1
  fib(n - 1) + fib(n - 2)
end

puts fib(47)

Here is the Crystal version:

def fib(n)
  return n if n <= 1
  fib(n - 1) + fib(n - 2)
end

puts fib(47_u64)

Too keep a level playing field, only common "release" flags are used in the compilation step. This allows for compiler optimizations like inlining and constant propogation but removes anything considered dangerous i.e. bypassing out of bounds checks.

All tests are run on:

Hetzner CCX23 - 4 vCPU
Processor: Intel Xeon 3.1Ghz
Memory: 16 GiB
OS: Ubuntu 24.04
Docker Base Image: ubuntu:24.04

How to run them

You can run the tests using Docker: docker run -it drujensen/fib

By default, it will compile and run all languages 5 times. Totals are calculated by adding the average compile and run times.

To only run a subset of the languages, provide a list of extensions and optionally the count:

docker run -it drujensen/fib ./run.sh s,c,cpp,go,rs,swift 5

To run in the background using screen:

screen
docker run drujensen/fib > results.txt 2>&1

Detach from the session by pressing Ctrl+A followed by D.
You can reattach to the session later with:

screen -r

NOTE: Please see issues with benchmarks like this. Its original goal was to compare at a marco level how much faster Crystal is to Ruby. Any language faster than Assembly is performing unrolling type optimizations. Modern languages like Go, Swift and Crystal have bounds checking which have safety built-in, but also have a cost associated with runtime performance.

Results

Last benchmark was ran on December 05, 2024

Natively compiled, statically typed

Language	Total	Compile	Time	Run	Time	Ext
C	4.835	gcc -O3 -o fib fib.c	0.112	./fib	4.723	c
C++	4.888	g++ -O3 -o fib fib.cpp	0.139	./fib	4.750	cpp
Fortran	6.334	gfortran -O3 -o fib fib.f03	0.130	./fib	6.204	f03
Ada	6.811	gnat make -O3 -gnatp -o fib fib.adb	0.220	./fib	6.591	adb
Rust	8.108	rustc -C opt-level=3 fib.rs	0.358	./fib	7.750	rs
Odin	8.838	odin build fib.odin -file -0:speed	0.087	./fib	8.751	odin
Mojo	8.947	mojo build fib.mojo	0.241	./fib	8.706	mojo
V	9.168	v -prod -o fib fib.v	4.450	./fib	4.718	v
Assembly	9.367	gcc -no-pie -O3 -o fib fib.s	0.025	./fib	9.342	s
Pascal	10.460	fpc -O3 -Si ./fib.pas	0.041	./fib	10.419	pas
Pony	11.106	ponyc -s -b fib -p ./fib.pony	0.879	./fib	10.226	pony
OCaml	16.018	ocamlopt -O3 -o fib fib.ml	0.187	./fib	15.831	ml
Zig	16.181	zig build-exe -OReleaseFast ./fib.zig	6.882	./fib	9.299	zig
Haskell	18.095	rm ./fib.o && ghc -O3 -o fib fib.hs	0.001	./fib	18.094	hs
D	18.193	dmd -release -of=fib fib.d	0.342	./fib	17.851	d
Swift	18.801	swiftc -O fib.swift	1.411	./fib	17.390	swift
Go	18.948	go build fib.go	1.107	./fib	17.842	go
Crystal	20.525	crystal build --release fib.cr	3.106	./fib	17.420	cr
Lisp	25.726	sbcl --load fib.lisp	0.979	./fib	24.747	lisp
Dart Compiled	31.707	dart compile exe -o fib ./fib.dart	1.558	./fib	30.149	dartc
Cobol	4380.728	cobc -x -O3 -o fib ./fib.cbl	0.133	./fib	4380.596	cbl

VM compiled bytecode, statically typed

Language	Total	Compile	Time	Run	Time	Ext
C#	12.760	dotnet build -c Release -o ./bin	2.015	dotnet ./bin/fib.dll	10.744	cs
Java	13.040	javac Fib.java	0.733	java Fib	12.307	java
Scala	15.836	scalac Fib.scala	2.682	scala Fib	13.153	scala
Kotlin	16.286	kotlinc Fib.kt	4.074	java FibKt	12.212	kt
Erlang	28.378	erlc +native +'{hipe,[o3]}' fib.erl	0.402	erl -noinput -noshell -s fib	27.976	erl
Groovy	70.123	groovyc Fib.groovy	1.519	groovy Fib	68.604	groovy

VM compiled before execution, mixed/dynamically typed

Language	Time	Run	Ext
Clojure	17.815	clojure -M fib.cljc	cljc
Julia	18.000	julia -O3 fib.jl	jl
Bun	21.312	bun fib.js	bun
Dart	29.984	dart fib.dart	dart
Node	34.736	node fib.js	js
Elixir	35.243	ERL_COMPILER_OPTIONS='[native,{hipe, [o3]}]' elixir Fib.exs	exs
Lua Jit	37.837	luajit fib.lua	luajit
Python (PyPy)	54.078	pypy fib.py	pypy
Ruby (jit)	81.454	ruby --jit fib.rb	rbjit

Interpreted, dynamically typed

Language	Time	Run	Ext
Escript	28.380	escript fib.es	es
Scheme	102.887	guile fib.scm	scm
Php	157.312	php fib.php	php
Lua	203.702	lua fib.lua	lua
Ruby	393.625	ruby fib.rb	rb
Python	423.427	python fib.py	py
Janet	479.663	janet ./fib.janet	janet
Perl	1490.416	perl fib.pl	pl
Raku	1672.015	rakudo fib.raku	raku
Tcl	2230.883	tclsh fib.tcl	tcl
R	2575.249	R -f fib.r	r

Versions

All compilers are installed using apt or asdf on Ubuntu 24.04 docker image:

Language	Version
ada	13.2.0
assembly	13.2.0
bash	5.2.21
bun	1.1.38
crystal	1.14.0
clojure	1.12.0.1488
dart	3.5.4
dmd	2.109.1
dotnet	9.0.101
elixir	1.17.3
elm	0.19.1
erlang	27.1.2
fortran	13.2.0
g++	13.2.0
gcc	13.2.0
gnucobol	3.2.0
golang	1.23.2
groovy	4.0.24
guile	3.0.10
haskell	9.8.3
janet	1.36.0
java	openjdk-23
julia	1.11.1
K	3.6
kotlin	2.1.0
ldc	1.39.0
lua	5.4.7
luajit	2.1.1
mojo	24.5.0
nim	2.2.0
nodejs	23.3.0
ocaml	5.2.1
odin	dev-2024-11
pascal	3.2.2
perl	5.40.0
php	8.4.1
pony	0.58.7
powershell-core	7.4.6
python	3.12.0
pypy	7.3.17
r	4.4.2
rakudo	2024.10
ruby	3.3.6
rust	1.83.0
sbcl	2.4.11
scala	3.3.4
swift	6.0.2
tcl	9.0.0
v	0.4.8
zig	0.13.0