rjmccall added a comment. In D100591#2692978 <https://reviews.llvm.org/D100591#2692978>, @asavonic wrote:
> In D100591#2692599 <https://reviews.llvm.org/D100591#2692599>, @rjmccall > wrote: > >> I think the right thing to do here is to recognize generally that we're >> emitting a mandatory tail call, and so suppress *all* the normal >> transformations on the return value. > > I assume it can be tricky to detect such call. The final decision (tail call > vs normal call) is made before instruction selection, after all LLVM IR > optimization passes. So we can miss tail calls that are not obvious on > non-optimized code, or get false-positive results for calls that a backend > decides to emit as normal calls. Well, I mean in the frontend. I certainly wouldn't expect the backend to recognize the pattern I described and somehow turn it into a tail call! > In any case, this patch can be useful not only for tail calls: `trunc + zext` > sequence generated to round a return value can be problematic for other cases > as well. Sure, I can imagine that it's hard to eliminate the extra zext in the backend. Maybe we should have an undef_extend? You should get backend sign-off before making Swift generate non-target-legal return types. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D100591/new/ https://reviews.llvm.org/D100591 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
