[Ocaml-lib-devel] Re: benchmarking enum approaches

[Alan P. <ap...@re...>]

In article <Pin...@ea...>, Brian Hurt wrote:
> 
> That depends upon the program.  I'd be most worried about 10+ maps on
> 10_000_000 elements myself.  With non-trivial map functions.  As that's
> where Enum really starts to shine- as you do wavefront processing.  And
> where tail recursion stops being an optimization and starts being required
> for correctness.  This is the problem with microbenchmarks.
> 
> I'll see if I can't whip something up.

Yes, real-world applications are better than benchmarks.  I suggest,
however, that benchmarks are better than speculation.