From: Rick DeNatale <rick.denatale@...>
Date: 2007-03-03T06:37:17+09:00
Subject: Re: default passed block for method

On 3/2/07, Robert Dober <robert.dober@gmail.com> wrote:
> On 3/2/07, Rick DeNatale <rick.denatale@gmail.com> wrote:
> > Just for giggles, I combined Robert and Joel's benchmarks and ran them
> > on both ruby 1.8 and a quite recent 1.9.  Perhaps the results are of
> > interest.
> >
> > rick@frodo:/public/rubyscripts$ ruby1.8 blkgiven.rb
> > Rehearsal -----------------------------------------------------
> > rd - block_given?   6.070000   1.620000   7.690000 ( 15.323920)
> > jv - block_given?   3.530000   0.530000   4.060000 (  9.952521)
> > rd - &blk          29.620000   2.090000  31.710000 ( 55.018634)
> > jv - &blk           8.580000   1.400000   9.980000 ( 13.505903)
> > ------------------------------------------- total: 53.440000sec
> >
> >                         user     system      total        real
> > rd - block_given?   6.430000   1.550000   7.980000 ( 12.312925)
> > jv - block_given?   3.470000   0.630000   4.100000 (  7.956797)
> > rd - &blk          28.920000   2.040000  30.960000 ( 55.456023)
> > jv - &blk           8.860000   1.350000  10.210000 ( 12.264554)
> >
> >
> > rick@frodo:/public/rubyscripts$ ruby1.9 blkgiven.rb
> > Rehearsal -----------------------------------------------------
> > rd - block_given?   1.950000   0.000000   1.950000 (  2.503581)
> > jv - block_given?   1.930000   0.010000   1.940000 (  3.029174)
> > rd - &blk          10.870000   0.020000  10.890000 ( 21.230573)
> > jv - &blk           2.430000   0.000000   2.430000 (  6.296742)
> > ------------------------------------------- total: 17.210000sec
> >
> >                         user     system      total        real
> > rd - block_given?   2.040000   0.010000   2.050000 (  3.862906)
> > jv - block_given?   1.770000   0.000000   1.770000 (  2.239562)
> > rd - &blk          10.520000   0.050000  10.570000 ( 17.941764)
> > jv - &blk           2.450000   0.010000   2.460000 (  3.961397)
> > rick@frodo:/public/rubyscripts$
> > ---
> > Rick DeNatale
> >
> > My blog on Ruby
> > http://talklikeaduck.denhaven2.com/
> >
> >
>
> Rick I think as a matter of fact that both benchmarks are important.
> I was too humble, but that will not happen again ;)
>
> Let us look at Jo�l's first, it applies to the following pattern:
>
> def method &blk
>     blk ||= proc{true}
>    MANY.times do
>         ...
>        blk.call
>        ...
>   end
> end
>
> from the performance POV you can get away with it.
> ==================================================
> my naive benchmark was benchmarking the following pattern - which
> exits quite often too.
>
> def method &blk
>     blk ||= proc{ true }
>    ...
>    blk.call
>    ...
> end
>
> Now of course the proc is a killer as we have as many proces as #call :(
>
> Hopefully this kind of a reasonable conclusion about the performance
> cost of the whole stuff.
>

I was suggesting that an analysis of how the relative numbers are
changing between 1.8 and 1.9 might be interesting.

Besides the fact that YARV seems quite a bit faster overall, I think
that some of the rules of thumb based on benchmarking on the 'classic'
Ruby implementation are going to change for the new Ruby.

-- 
Rick DeNatale

My blog on Ruby
http://talklikeaduck.denhaven2.com/