Re: Population count in SSE2, again

I built my business on software for bit counting, running since 1976.
I thought I understood the objective of the posts about "popcount" and
"wrap", but I see I don't.

If one wants to count "1" or "0" bits in strings of bytes or words,
you have to access memory anyway for the bytes. So there are a variety
of very simple, very fast ways of doing this using the continous sum
of table-indexed words or bytes, whose index is the words or byte to
count bits in. There is no need for extenal fixed/floating point
chips; just four or five machine language intructions found in the
most basic cpu chips. So you call one pre-builtt function from Fortran
which returns the answer for a vast string.