benchmarking

[Eric Wilhelm]

# from Andy Armstrong
# on Sunday 16 September 2007 11:40:

>   554:  1.384 # More accessor shaving
>trunk:  1.392 # -

That should probably say *what* is trunk, eh?

prove:  1.000 # (by definition) -- 27.899s
  270:  1.726 # pre speedy branch
  309:  1.929 # pre merge
  310:  1.664 # speedy merge
  461:  1.785 # console output (broke Parallel)
  463:  1.669 # undo console output
  494:  1.834 # major console output phase 1
  534:  2.004 # lots of formatting (phase 2)
  535:  1.982 # grammar streamline
  537:  1.903 # remove _trim() redundancy
  543:  1.592 # conditional utf8
  545:  1.599 # More minor performance gains
  546:  1.474 # Another small speed hike
  547:  1.422 # More minor speed-ups
  548:  1.441 # Banish accessor bloat
trunk:  1.251 # -

I've just updated my main workstation to debian etch, so 5.8.8 from 
5.8.4, plus the kernel maybe makes a difference.

Maybe we don't have the combination of test_files, num_lines and 
num_runs right, but you can easily lose a few days messing with that.

Of course, Benchmark.pm runs your subs in an unchecked eval {}, so maybe 
it's just dieing and isn't actually faster :-D

--Eric
-- 
Minimum wage help gives you minimum service.
--David Schomer
---------------------------------------------------
    http://scratchcomputing.com
---------------------------------------------------