I want to sort an array whose elements are release versions in a.b.c format. e.g. my @versions = (2.4.74, 3.2.5, 1.14.56, 1.45.2, 3.14.75)

I have put together the following solution ..

#!/usr/bin/perl -w
#
#  Sort version numbers

use strict;
use warnings;

{
    my @versions = qw/2.4.74 3.2.5 1.14.56 1.45.2 3.14.75/;

    print "Original version string:\n" . join( "\n", @versions ) . "\n
+";
    print "Sorted version strings:\n"
      . join( "\n", sort compareVersionStrings @versions ) . "\n";
}

#  Compare two version strings.
#
#  Among other assumptions, this code assumes that the version strings
+ have the
#  same number of elements in them. Thus, comparing 3 against 2 will w
+ork; as
#  will 3.1 against 2.7 and 3.1.4 against 2.7.18. Comparing 2.0 agains
+t 4 will
#  fail, as will 2.0 against 4.5.6.

sub compareVersionStrings {

    my ( @a, @b );

    my @left  = split( /\./, $a );
    my @right = split( /\./, $b );

    return ( $left[0] <=> $right[0]
          || $left[1] <=> $right[1]
          || $left[2] <=> $right[2] );
}
[download]

It seems to work, although there are some limitations, as I've explained in the comment.

However, as others have already pointed out, this sounds like premature optimization.

There are many ways to achieve it but I need a very optimized way.

You haven't explained why an optimized method is required. Are you expecting to have to do this thousands of times a second?

If you're sincere in your request to speed it up, you'll need to hash the version numbers into something that's very easily sorted -- a trivial solution would be to multiply major by 1,000,000, minor by 1,000 and add the three values together, giving you a single number that should sort correctly.