RE: Re: Algorithm Efficiency vs. Specialized Hardware?

I did a test where I did this, and increasing the %url population from 1000 to 100_000 caused each algorithm to take over 100 times longer.

the change is:

old:

my %url = ( monday => { @{[map(($_,1), (1..1000))]} } );
[download]

new:

my %url = ( monday => { @{[map(($_,1), (1..100_000))]} } );
[download]

I turned the iterations down for the large dataset, because otherwise I would be posting tomorrow instead of today. ;)
results are:

Large dataset

Benchmark: timing 100 iterations of Grep, Max, Ternary...
      Grep: 83 wallclock secs (80.90 usr +  0.02 sys = 80.92 CPU) @  1.24/s (n=100)
       Max: 74 wallclock secs (71.38 usr +  0.00 sys = 71.38 CPU) @  1.40/s (n=100)
   Ternary: 77 wallclock secs (74.15 usr +  0.00 sys = 74.15 CPU) @  1.35/s (n=100)

Small dataset

Benchmark: timing 10000 iterations of Grep, Max, Ternary...
      Grep: 73 wallclock secs (68.50 usr +  0.05 sys = 68.55 CPU) @ 145.88/s (n=10000)
       Max: 61 wallclock secs (57.77 usr +  0.05 sys = 57.82 CPU) @ 172.95/s (n=10000)
   Ternary: 64 wallclock secs (61.54 usr +  0.02 sys = 61.56 CPU) @ 162.44/s (n=10000)

Can you post the rest of the Benchmark output?

Paris Sinclair    |    4a75737420416e6f74686572
pariss@efn.org    |    205065726c204861636b6572
I wear my Geek Code on my finger.

Comment on RE: Re: Algorithm Efficiency vs. Specialized Hardware? Select or Download Code

Replies are listed 'Best First'.

RE: RE: Re: Algorithm Efficiency vs. Specialized Hardware?
by lhoward (Vicar) on Jun 20, 2000 at 23:34 UTC

   my @size=(100,1000,10000,100000);
   foreach(@size){
     my $size=$_;
     print "size=$size\n";
     my $now = 8;
     my %url = ( monday => { @{[map(($_,1), (1..$size))]} } );
     timethese(0, {
       Grep => q{ $now = (sort grep {$_ <= $now} keys %{$url{'monday'}
+})[-1]; },
       Ternary => q{ $now = ($now < $_ && $_ < 8 ? $_ : $now) for keys
+ %{$url{'monday'}}; },
       Max => q{ foreach ( keys %{$url{'monday'}} ) { $now = $_ if $_ 
+> $now }; }
     });
   }
[download]

I think I am onto something... I don't think the code posted in teh original question really benchmarks what the user thinks it does... I'm crafing a new reply to the original question that should be up soon.

[reply]
[d/l]


Keep It Simple, Stupid
	PerlMonks