Re^2: Fisher-Yates Shuffle

Random Walk:

Update: I bungled it. The results were suspiciously fast. When copying a block, I left in one of the --$i statements, so I was skipping every other element. Updated code & results:

        'ROBO' => sub {
            my $i = @array;
            do {
                ++$cRO;
                my $j = int rand $i;
                @array[$i,$j] = @array[$j,$i];
            } while ($i--);
        },

roboticus@sparky:~$ perl 1088227.pl 
Benchmark: timing 500 iterations of BRANCH, IFIDDLE, ROBO, WITHOUT...
    BRANCH: 17 wallclock secs (17.05 usr +  0.00 sys = 17.05 CPU) @ 29
+.33/s (n=500)
   IFIDDLE: 14 wallclock secs (14.29 usr +  0.00 sys = 14.29 CPU) @ 34
+.99/s (n=500)
      ROBO: 19 wallclock secs (18.59 usr +  0.00 sys = 18.59 CPU) @ 26
+.90/s (n=500)
   WITHOUT: 15 wallclock secs (15.69 usr +  0.00 sys = 15.69 CPU) @ 31
+.87/s (n=500)
5000000, 5250000, 5000000, 5125250
[download]

My code is still executing the wrong number of loops (the last number on the last line), but not so far out of line. Also, WITHOUT (second) also has an unexpected number of loops. Original node below:

You could also move the comparison to the end of the loop to avoid that +1:

        'ROBO' => sub {
            my $i = @array;
            do {
                my $j = int rand $i--;
                @array[$i,$j] = @array[$j,$i];
            } while (--$i);
        },
[download]

Doing so gave me:

roboticus@sparky:~$ perl 1088227.pl 
Benchmark: timing 500 iterations of BRANCH, IFIDDLE, ROBO, WITHOUT...
    BRANCH: 17 wallclock secs (16.77 usr +  0.00 sys = 16.77 CPU) @ 29
+.82/s (n=500)
   IFIDDLE: 14 wallclock secs (14.02 usr +  0.00 sys = 14.02 CPU) @ 35
+.66/s (n=500)
      ROBO:  8 wallclock secs ( 7.85 usr +  0.00 sys =  7.85 CPU) @ 63
+.69/s (n=500)
   WITHOUT: 15 wallclock secs (15.33 usr +  0.00 sys = 15.33 CPU) @ 32
+.62/s (n=500)
roboticus@sparky:~$
[download]

It seems to help (unless I bungled it).

...roboticus

When your only tool is a hammer, all problems look like your thumb.

Comment on Re^2: Fisher-Yates Shuffle Select or Download Code

Replies are listed 'Best First'.
Re^3: Fisher-Yates Shuffle by RichardK (Parson) on Jun 02, 2014 at 13:33 UTC
That doesn't look right, now you're decrementing $i twice in each ROBO loop. Isn't that only doing half the work , hence taking half the time?	[reply]
Re^4: Fisher-Yates Shuffle by roboticus (Chancellor) on Jun 02, 2014 at 14:20 UTC
RichardK: You're right. My code is all screwed up at the moment. I'll try to fix it correctly when I get off work today. <sigh...> ...roboticus When your only tool is a hammer, all problems look like your thumb.	[reply]


laziness, impatience, and hubris
	PerlMonks