BrianP,

Maybe this helps?

    use strict;
    use warnings;

#    my $file = '249465.pprgb-srgb.absv.4000.cs.jpg';    #    Original
+ File downloaded
    my $file = '249465.pprgb-srgb.absv.4000.cs.jpg2';    #    New File
+ multiple copies
    my $compraw = -s $file; my $pixels = int ( $compraw / 6 );
    my @UINT48;
    $UINT48[$pixels] = '';        ## Allocate size of array ( faster )
    open ( my $in, "<", "./$file") or die "$!\n";
    if ( 1 )
    {    my $loc = -1;
        my $size = sysread( $in, my $buffer, $compraw );    #    No bu
+ffering
        if ( $size ne $compraw ) { die "Disk error: $!\n" }
        while( $buffer )
        {    $UINT48[$loc++] = substr( $buffer, 0, 6, '' );
        }
    }
    close $in;                    ## $buffer is gone!
                                ## @UINT48 is populated with 6 byte RG
+B strings
    my $uint48 = scalar @UINT48;

    print "\nImage has $uint48 pixels\n";

##  How to get the values back!

    my ( $R,$G,$B ) = unpack( "SSS", $UINT48[0] );
    print "\t$R,$G,$B\n";
    ( $R,$G,$B ) = unpack( "SSS", $UINT48[$pixels] );
    print "\t$R,$G,$B\n";

__END__

> time pyrperl uniqcolors.plx

Image has 41473711 pixels
    17994,17993,256
    55551,57599,4096

real    0m13.745s
user    0m12.905s
sys     0m0.840s
>
[download]

For you earlier question about reading from disk, Linux reads/writes in 4096 byte blocks. So I multiply the pagesize * 6 (since 4096 is not exactly divisible by 6). Use you system page size and multiply by a factor of 6 ( 6,12,...).

If you want '$buffer' around after the populate cycle, just remove the "if ( 1 ){ }" or define '$buffer' before the loop,

Regards...Ed