note davido <p>I'm hesitant to post this because I haven't had the time to turn an idea into a viable solution (and won't have time tonight). I haven't tested this against your code or even on a large vector. It's just an idea that hit me. Then as I was looking for other ideas on algorithms I came across [http://en.wikipedia.org/wiki/Hamming_weight|Hamming Weight], which mentions this:</p> <blockquote><em><p>With unlimited memory, we could simply create a large lookup table of the Hamming weight of every 64 bit integer.</p></em></blockquote> <p>Well, obviously we don't have unlimited memory, but using an array as a lookup for 16-bit integers would allow you to count 16 bits at a time rather than a single bit at a time. So this is a quick rough-draft. It doesn't take into consideration bit strings that are of lengths that aren't evenly divisible by 16. It's just an incomplete proof of concept before I go to bed... there's work to be done on it before it's a solution, and of course the work needs to be followed by benchmarks. :)</p> <c> my @lookup; for( 0 .. 65535 ) { my $bits = sprintf "%b", $_; my $count = $bits =~ tr/1/1/; push @lookup, $count; } my $bitstring = ''; vec( $bitstring, 0, 32 ) = 1234567891; my $count = 0; for( 0 .. length( $bitstring ) / 2 - 1 ) { $count += $lookup[ vec( $bitstring, $_, 16 ) ]; } print $count, "\n"; </c> <div class="pmsig"><div class="pmsig-281137"> <br /><p>Dave</p> </div></div> 1015564 1015564