vec's no good because it only does powers of 2-bits.

Just out of interest (and to try to get my head around how it works) I thought I'd see if I could come up with a vec solution although I didn't expect it to challenge in the speed stakes. By combining 4-bit and 2-bit vecs with 2- or 4-bit shifts you can obtain your 6-bit values. As expected, it trails the field. (I didn't test the C solutions as I don't have Inline installed).

sub johngg {
   map
   {
       my $offsetVec4 = $_ * 6;
       my $offsetVec2 = $_ * 12;
       my @vals = (
          ( ( vec( $_[0], $offsetVec4 +  1, 4 ) << 2 ) +
              vec( $_[0], $offsetVec2 +  1, 2 ) ),
          ( ( vec( $_[0], $offsetVec2,      2 ) << 4 ) +
              vec( $_[0], $offsetVec4 +  3, 4 ) ),
          ( ( vec( $_[0], $offsetVec4 +  2, 4 ) << 2 ) +
              vec( $_[0], $offsetVec2 + 11, 2 ) ),
          ( ( vec( $_[0], $offsetVec2 + 10, 2 ) << 4 ) +
              vec( $_[0], $offsetVec4 +  4, 4 ) ),
          );
   }
   0 .. 7;
}
[download]

00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 2
+3 24 25 26 27 28 29 30 31
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 2
+3 24 25 26 27 28 29 30 31
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 2
+3 24 25 26 27 28 29 30 31
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 2
+3 24 25 26 27 28 29 30 31
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 2
+3 24 25 26 27 28 29 30 31
             Rate    johngg       buk    jethro jmcnamara     salva
johngg    22191/s        --       -3%      -29%      -42%      -51%
buk       22974/s        4%        --      -27%      -39%      -49%
jethro    31438/s       42%       37%        --      -17%      -30%
jmcnamara 37959/s       71%       65%       21%        --      -16%
salva     45081/s      103%       96%       43%       19%        --
[download]

I hope this is of interest.