Re: [Perlunit-devel] commits

[Ian R. <ia...@ez...>]

>>>>> " AS" == Adam Spiers <ad...@sp...> writes:

  AS> Ian Robertson (ia...@ez...) wrote:
 IBR> FWIW, on my box, the shift method seems to be a touch faster if there
 IBR> are one or two arguments (including $self); the @_ method starts to
 IBR> win at three.  However, the difference is in the fraction of a
 IBR> microsecond range, so it's probably not a deciding factor :).

  AS> That's odd, my tests show that the more shifts you use, the slower it
  AS> gets.  Try the attached benchmark.  It may be something to do with a
  AS> difference in Perl versions, of course.  Mine's 5.6.0.

That's the same experience I had - for two arguments or less
(including $self), using shift is faster; after that, @_ is faster.

Running 5.005_03 (and changing cmpthese to timethese for my
(presumably older) version of Benchmark), I get these results:

Benchmark: timing 500000 iterations of two_all, two_none, two_one...
   two_all:  4 wallclock secs ( 2.79 usr +  0.00 sys =  2.79 CPU)
  two_none:  3 wallclock secs ( 2.82 usr +  0.00 sys =  2.82 CPU)
   two_one:  2 wallclock secs ( 3.05 usr +  0.00 sys =  3.05 CPU)
Benchmark: timing 500000 iterations of three_all, three_none, three_one...
 three_all:  2 wallclock secs ( 3.29 usr +  0.00 sys =  3.29 CPU)
three_none:  2 wallclock secs ( 3.03 usr +  0.00 sys =  3.03 CPU)
 three_one:  3 wallclock secs ( 3.38 usr +  0.00 sys =  3.38 CPU)
Benchmark: timing 500000 iterations of four_all, four_none, four_one...
  four_all:  7 wallclock secs ( 6.39 usr +  0.00 sys =  6.39 CPU)
 four_none:  7 wallclock secs ( 6.03 usr +  0.00 sys =  6.03 CPU)
  four_one:  7 wallclock secs ( 6.42 usr +  0.00 sys =  6.42 CPU)
Benchmark: timing 500000 iterations of four_all, four_none, four_one...
  four_all:  6 wallclock secs ( 6.44 usr +  0.00 sys =  6.44 CPU)
 four_none:  7 wallclock secs ( 6.09 usr +  0.00 sys =  6.09 CPU)
  four_one:  7 wallclock secs ( 6.34 usr +  0.01 sys =  6.35 CPU)
Benchmark: timing 500000 iterations of five_all, five_none, five_one...
  five_all:  7 wallclock secs ( 6.84 usr +  0.00 sys =  6.84 CPU)
 five_none:  6 wallclock secs ( 6.19 usr +  0.00 sys =  6.19 CPU)
  five_one:  7 wallclock secs ( 6.50 usr +  0.00 sys =  6.50 CPU)


  - Ian