Benchmarking example

[rokicki]

Here's a benchmarking example if you want it. It takes about five hours with Monoid Automata Factory, and about five minutes (if you give the -t 2000000 option) with kbmag. I'd be curious how this Julia implementation performs.

If you don't give a -t option to kbmag, it will run for days; I've posted an issue and suggested fix in the kbmag repo.

#aaa,ab'b'b'a'b'a'b
_RWS := rec
(
  isRWS := true,
  ordering := "shortlex",
  generatorOrder := [a,A,b,B],
  inverses := [A,a,B,b],
  equations := 
  [
   [a*a*a,IdWord],
   [a*B*B*B*A*B*A*b,IdWord]
  ]
);

[kalmarek]

I'd be curious how this Julia implementation performs.

why don't you tell me? ;)

But honestly, this implementation is not really tuned for performance. If it does perform well here, that would be by sheer coincidence.

1. How large is the confluent system?
2. Does it experience one (or more) collapses i.e. when a large portion of the rules is rendered unnecessary?
3. Does it run into the problem of cubic (w.r.t. number of rules) complexity? Then it's likely the case we're reducing too often.
4. Does it finish abruptly (due to backtrack confluence check)?

There are many parameters that will influence the performance of Knuth-Bendix completion ;)