[Bug tree-optimization/120647] [X86] Sub optimal code generated for counting the number matches between two array elements

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120647

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Like we have reduc_sbool_{and,ior,xor}_scal optabs we could add a
popcount_sbool
optab for the case vector masks do not have distinct modes like V4BI.  For
integer mode vector masks we could use regular popcount as well.

This still involves quite some special-casing as the accumulator will be
an integer and not a vector of integers.  Possibly multiple integers with
high enough VF.