BITs extraction from a byte

[frktons]

i get 1 clock cycle on my P4 - lol
(it's probably 2 cycles)
that's with no call/ret overhead

That was the target. We got there. Thanks master for your inspiration.  :t

[frktons]

I tried it on my Core Duo, because Kaspersky is giving me problems on
my PIV, and I cannot measure the test, it is always 0.
I need to test it with some more data to have an idea of the performance.

Let's do some modification to the test...

[frktons]

For 12 bytes, adding the overhead of the unrolled cycles
I get these results on Core Duo PC.

96 cycles
113 cycles
45 cycles

Still a couple of times faster than other solutions anyway.

[dedndave]

prescott w/htt

139 cycles
164 cycles
62 cycles

163 cycles
143 cycles
55 cycles

163 cycles
142 cycles
61 cycles

[hool]

deleted