jojo added a comment. > While I agree with Bradley that the repetition is not pretty, I think it will > expose all issues to make a class design simple and straightforward, once we > get all the sharp edges out. But we need to know what are the difficulties on > Clang, llc and the back-ends, and make sure that the architectural part of > the change is correct, before we that.
A class based design *will* change how every TargetParser method is being called now, and will touch a large number of files with mechanical changes, including in the ARM side, unrelated to the addition of AArch64. Yes.It will be a big project really,especially for me.I need to have a considerable understanding of the front-end, back-end, and the arm architecture, if I'm going to do it. > After my failed attempt at getting a class design across, I'd rather > introduce functionality first, then design better, than the other way round. > But I think they should be separated commits based on their intents and > merits alone. Totally agree.And that's exactly what I think.I think we can take class design as a low priority task.Maybe we could put it in our backlog list. Repository: rL LLVM http://reviews.llvm.org/D20089 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits