https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54089
--- Comment #63 from Oleg Endo <olegendo at gcc dot gnu.org> --- (In reply to Alexander Klepikov from comment #62) > > My project is small and it compiles in under 1 second on both clean and > patched GCC. sh.exp test suite mean run time is 117s on clean and 119s on > patched. I did 1 warm-up run and then 3 main one-threaded runs for each > task. I'm thinking of something else. We have to consider that SH is also a linux target and it's also used to build larger applications (and GCC itself ...). It'd be good to not regress too much in this regard. One way to check it is the CSiBE test set. I can help testing your patch later. > > Implementing features not supported by hardware will always be a tradeoff. I'd say it's generally about how to find the best choice of instructions/sequences. With GCC's "waterfall" optimization it's impossible to find the best solution because it doesn't have a way to compare the total cost of one solution vs. another, for example. We have to work around that ;)
