https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69564
Richard Biener <rguenth at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |amker at gcc dot gnu.org --- Comment #31 from Richard Biener <rguenth at gcc dot gnu.org> --- I can't see any difference in the vectorizer dumps, one notable thing is that we (again) do peeling for alignment which is hardly useful on x86_64 these days, at least if you only have a single store stream. The code after the vectorizer looks 1:1 the same, so do the alias checks. IVOPTs does different decisions for the outer loop (LU.c:87) that is because with LTO we have more context (inlined the function) and thus "expansion" brings up bigger expressions (well, that's my guess).