Hi! On Tue, Dec 04, 2018 at 10:12:51AM -0600, Peter Bergner wrote: > We talked about replacing rs6000'c regno_or_subregno() with the generic > reg_or_subregno() function from jump.c. I agree the geberic version is > better because it has an assert that ensures we have a REG. There were > also a couple of places that could use reg_or_subregno() where we weren't. > I made those changes and tested it and ran into no regressions. > > Nearby the changes above, there were cases where could could have been using > the rtl convenience macros like REG_P, etc. to simplify the code so I did > that too...and kept going and just ended up converting all the cases I could > find that should have been using them, to use them. So the patch is large, > but pretty straight forward.
In the future, please do separate patches? Especially mechanical ones could be separated (and then be much easier to review -- you also post the script, and then I don't have to check it really is mechanical, instead of just boring and you could have made mistakes). > (rs6000_init_hard_regno_mode_ok, direct_move_p): Use > HARD_REGISTER_NUM_P. This line is too long. And that is the worst I see in that whole patch! Please apply it. Thanks! Segher
