https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125666
--- Comment #1 from ktkachov at gcc dot gnu.org --- Created attachment 64673 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64673&action=edit Reproducer for failed inlining Attaching a synthetic reproducer that I think showcases the limitation in various C++ benchmarks in SPEC2026. LLVM inlines all instances of h() that enables a big speedup through vectorisation, but GCC refuses to do so even at -O3 due to inline-unit-growth
