[Bug tree-optimization/92655] Suboptimal vectorization of variable shift

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92655

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
"Fixed" source (MULWU), side-stepping GENERIC folding/narrowing of the
subtract:

inline uint64_t MULWU(uint64_t k, uint64_t m)
{
  unsigned long xyz = BITS-m;
  return (( (k)<<(m) & M61) + ( (k) >> (xyz))  )  ;
}