https://gcc.gnu.org/bugzilla/show_bug.cgi?id=49263
--- Comment #40 from Oleg Endo <olegendo at gcc dot gnu.org> --- (In reply to Alexander Klepikov from comment #39) > > I'm sorry, but .md lang is too complicated for me. Yeah, it looks alien at first sight. But it's where a lot of things happen w.r.t. instruction selection. > It looks like with optimization enabled it converts bitwise AND to right > shift and then optimizes again. But SH4 has 'shad' and 'shad' can be > optimized to 'tst'. And SH2E has libcall instead of dynamic shift and libcll > cannot be converted. It seems that very first optimization spoils things. > > But when we have numerous 'shar' instructions, optimization joins the game > again and converts them to 'tst'. Yes, something like this is what I remember happening there. I'll try to look into the issue with your test cases and see if it's possible to add some patterns to catch those. BTW, have you tried it on a more recent GCC? There have also been some optimizations in the middle-end (a bit more backend independent) for this kind of thing. > You are absolutely right, the code will be larger when we do right shifts. > But there's situations when you can't use library. For exmple, SH7055 engine > control unit can barely contain the program. The library just won't fit. Have you tried to use whole program optimization via -flto and -ffunction-sections? It should be able to strip out all unnecessary library functions.
