[PATCH] D129396: [InstCombine] Add support for div, ldiv, lldiv, and imaxdiv folding

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a comment.

There really isn't any reason to preserve the call to div() in the IR, even if we can't fold it.  If you use the normal LLVM IR div/rem instructions, usually we can optimize them to a more efficient sequence. Worst-case, the backend will just generate a call to __divmodsi4(), which is basically the same thing as div().

Given that, the lowering would be a lot simpler to implement in clang because we wouldn't need to worry about the ABI stuff.  (And there shouldn't be code in languages other than C using div().)


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D129396/new/

https://reviews.llvm.org/D129396