more-similar less-idiosyncratic

When programs are idiomatic someone will say they are not comparable.

Are they similar-enough to be comparable for your purposes? Will someone then say the programs are not idiomatic.

If only more-idiomatic was not less-comparable!

Here are a few naive un-optimised single-thread #8 programs transliterated lowest-common-denominator style into different programming languages from the same original.

Also let's repeat the calculations a dozen times in-process to amortize startup and JIT warmup costs even more —

elapsed secs / fastest
1×		12× in-process
×	source	×
1.0	Intel C	1.0	Intel C	spectral-norm
1.0	Swift	1.0	Swift	fannkuch-redux
1.0	C gcc	1.0	C gcc	n-body
1.0	C clang	1.0	C clang	n-body
1.0	C clang	1.0	C clang	fannkuch-redux
1.1	Intel C	1.1	Intel C	n-body
1.1	C gcc	1.1	C gcc	fannkuch-redux
1.2	Swift	1.3	Swift	n-body
1.2	Intel C	1.2	Intel C	fannkuch-redux
1.3	C# aot	1.3	C# aot	n-body
1.3	C# aot	1.3	C# aot	fannkuch-redux
1.3	Java	1.6	Java	fannkuch-redux
1.3	Go	1.3	Go	fannkuch-redux
1.4	Dart	1.3	Dart	n-body
1.4	Java	1.4	Java	n-body
1.7	Node.js	1.9	Node.js	fannkuch-redux
1.7	Node.js	1.8	Node.js	n-body
1.7	Go	1.7	Go	n-body
2.0	C# aot	2.0	C# aot	spectral-norm
2.0	Swift	2.0	Swift	spectral-norm
2.0	Go	2.0	Go	spectral-norm
2.0	Dart	2.0	Dart	fannkuch-redux
2.0	C gcc	2.0	C gcc	spectral-norm
2.0	C clang	2.0	C clang	spectral-norm
2.0	Java	2.1	Java	spectral-norm
2.4	Node.js	2.4	Node.js	spectral-norm
4.1	Dart	4.1	Dart	spectral-norm
20	Java -Xint			fannkuch-redux
29	Java -Xint	29	Java -Xint	n-body
30	Java -Xint	31	Java -Xint	spectral-norm
44	PHP	44	PHP	n-body
49	PHP			fannkuch-redux
51	PHP	51	PHP	spectral-norm
55	Ruby yjit			fannkuch-redux
57	Ruby yjit	58	Ruby yjit	n-body
63	Ruby yjit	66	Ruby yjit	spectral-norm
64	Python 3			fannkuch-redux
94	Python 3			n-body
142	Python 3			spectral-norm
586	Matz's Ruby			n-body
626	Matz's Ruby			spectral-norm

Seemingly similar programs behave differently when the language implementations behave differently. Do they have the same default stdio buffering? Do the programs need to be written differently and run differently to make them behave the same.

To be less similar so they are more comparable!

elapsed secs / fastest		significant io
×	source
1.0	Intel C	mandelbrot
1.0	C# aot	fasta
1.0	Go	k-nucleotide
1.0	C gcc	mandelbrot
1.1	C clang	mandelbrot
1.1	C# aot	mandelbrot
1.1	Swift	mandelbrot
1.1	Go	mandelbrot
1.2	Java	mandelbrot
1.2	C# aot	k-nucleotide
1.2	Intel C	fasta
1.3	PHP	k-nucleotide
1.3	C clang	fasta
1.3	C gcc	fasta
1.3	Java	k-nucleotide
1.6	Java	fasta
1.8	Go	fasta
2.3	Node.js	k-nucleotide
2.9	Swift	k-nucleotide
3.0	Node.js	fasta
3.0	Node.js	mandelbrot
3.0	Swift	fasta
3.9	Python 3	k-nucleotide
5.8	Dart	fasta
7.9	PHP	fasta
8.4	Dart	mandelbrot
8.6	Ruby yjit	k-nucleotide
9.0	Dart	k-nucleotide
18	PHP	mandelbrot
25	Python 3	fasta
26	Java -Xint	mandelbrot
31	Ruby yjit	fasta
39	Ruby yjit	mandelbrot
53	Java -Xint	fasta
64	Python 3	mandelbrot
126	Matz's Ruby	fasta

Such comparisons raise a nest of issues all their own: for example, is it really the same program when written in the two compared languages? (Will Partain, 1992.)

 

Throw the ratios together and give them a shake!

elapsed secs / fastest		Geometric Mean
1.1	Intel C
1.3	C clang
1.3	C gcc
1.3	C# aot
1.5	Java
1.5	Go
1.6	Swift
2.2	Node.js
3.4	Dart
22	PHP
30	Java -Xint
41	Ruby yjit
42	Python 3
359	Matz's Ruby