https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91796
--- Comment #7 from Marc Glisse <glisse at gcc dot gnu.org> --- (In reply to Maxim Egorushkin from comment #3) > It seems to me that register allocation has been a weak spot in gcc for > years. Most such testcases show issues with arguments/return in very small functions, where indeed the allocation is problematic. But one usual comment is that if this function is performance-critical, it should be inlined, which makes function boundaries disappear, and thus the issue as well. It isn't that this part couldn't be improved, but more that it isn't a priority because it should not affect the running time of most applications. (I am just repeating what is written in other PRs, not contributing anything, not even my opinion)