Back to square one I guess. This

 sub popnum3 {
    my ( $x, $y, $z ) = @_;  # $_[3] is $zazb
    if ( $y == 0 ) {
        $aob[$x][0] = $initial * ( 1 + $z );
    }
    else {
        my @cPosns;
        push @cPosns, $-[0] while $_[3] =~ m{c}g;
        if (@cPosns) {
            my $offset = $cPosns[rand @cPosns];
            $aob[$x][$y] = $aob[$offset][ $y - 1 ] * ( 1 + $z );
        }
    }
    return $aob[$x][$y];
}
[download]

only seems to work relatively short strings, but not when the string gets much more than 500 characters. Then it throws 'uninitialized value in $offset' errors which I'm guessing means the algorithm didn't find the one 'c' that had to be there.

I admit I'm at my wit's end here. This looks like it should work, but it doesn't. There is a 'c' in every row, so that's not the problem.

THIS works

sub popnum3 {
    my ( $x, $y, $z, $zazb ) = @_;
    if ( $y == 0 ) {
        $aob[$x][0] = $initial * ( 1 + $z );
    }
    else {
        while (1) {
            my $xda = int rand( $total + 1 );
            
            if ( substr( $zazb, $xda, 1 ) eq 'c' ) {
                $aob[$x][$y] = $aob[$xda][ $y - 1 ] * ( 1 + $z );
                last;
            }
        }
    }
    return $aob[$x][$y];
}
[download]

But takes about 5 seconds per row on the full array and since there's 8400 rows in the full run, that's about 12 hours.

Getting the run time down would be very helpful.

Thanks