Awesome

Toy Language Benchmark

I dabble with toy scripting languages, having implemented several including:

• BASIC
• FORTH
• Lisp
• Monkey
• TCL
• evalfilter

This repository is designed to host a quick, unscientific, benchmark of those implementations.

Benchmark

The idea is to calculate 12! (i.e. factorial of 12), and see how they compare.

To run the benchmark clone this repository then run:

go test -run=Bench -bench=.

Or

go test -run=Bench -bench=. -benchtime=30s

Results

As of today the results, on my desktop system, look like this:

$ go test -run=Bench -bench=.
goos: linux
goarch: amd64
pkg: github.com/skx/toy-language-benchmarks
cpu: AMD A10-6800K APU with Radeon(tm) HD Graphics
BenchmarkGoFactorial-4           35637698        30 ns/op
BenchmarkEvalFilterFactorial-4      61542     17458 ns/op
BenchmarkFothFactorial-4            44751     26275 ns/op
BenchmarkBASICFactorial-4           36735     32090 ns/op
BenchmarkMonkeyFactorial-4          14446     85061 ns/op
BenchmarkYALFactorial-4              2607    456757 ns/op
BenchmarkTCLFactorial-4               292   4085301 ns/op
PASS
ok  	github.com/skx/toy-language-benchmarks	21.479s

Which means in terms of speed:

• Native Go
  • Obviously this would be best, being compiled.
• Evalfilter
  • Expected as this uses a bytecode virtual machine, which is pretty fast.
• FOTH
  • Expected as this is essentially a bytecode virtual machine, which is pretty fast.
• BASIC
  • This is a real surprise, although the code is nice and simple so maybe it shouldn't be?
• Monkey
  • This was a surprise, as there's no particular optimisation or cleverness here.
• Lisp
  • Expected as TCO makes this a reasonably fast benchmark.
• TCL
  • The constant string-conversion kills performance of maths.