https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68981
--- Comment #3 from Zdenek Sojka <zsojka at seznam dot cz> --- (In reply to Richard Biener from comment #2) > I don't see any vectorization happening and on the GCC 5 branch (where I can > reproduce only) the segfault also happens with -fno-tree-vectorize. Note > that -O3 already includes -ftree-vectorize. -ftree-vectorize was needed for reproducing. I am not trying to explain why it was so. > So to me this looks like a dup of PR66616, fixed on the branches as well > after > that backport. Indeed.