Finaly the winner is the vec function,but is more expencive on space that pack w.. ..

use strict;
use Benchmark qw( timethese cmpthese );

my $df=1000000;my $tf=3;
our $packedW ='';
our $packedVec ='';
my $nr=0;
foreach(0 .. $df)
{

    $packedW .= pack 'w*', $_;
    $packedW .= pack 'w*', $tf;
    
   vec ($packedVec, $nr++, 32) = $_; # DOC ID......
   vec ($packedVec, $nr++, 32) = $tf;  # TF......

}


cmpthese 10, {
    
    unpackW   => q[ my $odd=0;my %pack;my $doc=0; 
                    foreach (unpack 'w*', $packedW)
                    {
                        if($odd%2!=0)  
                        {
                                $pack{$doc}=$_;$odd++;next;
                        }
                        $doc=$_;
                        $odd++;
                    } 
],

    unVec     => q[ 
                    my $docID=0;
                    my $tf=0;
                    my $index=0;
                                        my %vec;
                    foreach (0 .. 1000000)
                    {
                            $docID = vec ($packedVec, $index++, 32);
                            $tf = vec ($packedVec, $index++, 32);
                            $vec{$docID}=$tf;                    
                    }
        ],
        
};


print "$_: ", length( do{ no strict; ${$_} } ) 
    for qw[ packedVec packedW ];
__END__

            s/iter    unpackW unVecNoGaP
unpackW      21.7         --       -60%
unVec        8.69       149%         --
packedVec: 8000008
packedW: 3983492
[download]

If you have any sugestion how can distinguish eficient the two difirent numbers(docID and TF) with the pack function will be great...

Thanks for your help...