https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54089
--- Comment #105 from Oleg Endo <olegendo at gcc dot gnu.org> --- (In reply to Alexander Klepikov from comment #104) > I've been thinking about something. I suspect that this patch could take > work away from other patches. I'm sorry, I don't know how to express myself > properly. I mean there's several patches that corrects shift patterns and > 'tst' instruction generation (most of them are written by you, by the way). > I suspect that some of them might not run anymore because this patch looks > more general and should cover more cases, including yet unknown cases, I > hope. And, in the end, dead code may appear because of it. I hope I was able > to make my point clearly. Yes, I understand what you're saying. As other parts of the compiler evolve, the RTL input that the backend code has to work with changes. It's a normal thing that happens during the course of development. Some patterns might stop working (especially those combine patterns are prone to that). And sometimes things magically start working because something got fixed somewhere else. I've tried to add SH specific test cases to try and keep it in check. Ideally we'd have to go through all of the specific SH quirks in the backend periodically, try to remove them one by one and run tests to see if the patterns are still working/needed or whether they can be removed. Let me know if you have more test cases (that work or don't work).
