https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127975
--- Comment #4 from GCC Commits <cvs-commit at gcc dot gnu.org> --- The master branch has been updated by Richard Biener <[email protected]>: https://gcc.gnu.org/g:f4b64250a70482fbb51d6bb97fae481f90aa9649 commit r17-5452-gf4b64250a70482fbb51d6bb97fae481f90aa9649 Author: Richard Biener <[email protected]> Date: Tue Oct 20 14:17:01 2026 +0200 tree-optimization/127975 - mixed AVX512 / AVX2 style masks with loop masking loop masking with AVX512 assumes integer type masks throughout, but when only a limited subset of AVX512 is enabled we can end up with mixed style masks which we do not support. Verify this doesn't happen and disable loop masking in that case instead of ICEing at transform time. PR tree-optimization/127975 * tree-vect-loop.cc (vect_verify_full_masking_avx512): Require all loop masks to be integer mode masks. * gcc.dg/vect/vect-simd-22.c: New testcase.
