https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126466
--- Comment #2 from Reshma Roy <Reshma.Roy at amd dot com> --- Thank you for catching this. I can confirm this is a regression introduced by my commit 8ca1e887847e2f. The original pattern matched a generic 32-bit population count but failed to validate the final bitwise AND mask (@7). Hence any arbitrary mask value was accepted.
