tra added inline comments.
================ Comment at: clang/lib/Headers/__clang_cuda_cmath.h:85 +// (note that we do not create implicit base functions here). To avoid +// this clash we add a new trait to some of them that is always true +// (this is LLVM after all ;)). It will only influence the mangled name ---------------- jdoerfert wrote: > tra wrote: > > If you just want to disable some existing declarations that get in the way, > > one way to do it would be to redeclare them with an > > `__arrtibute__((enable_if(false)))` > > > > Having overloads with different return types will be observable. > We need both overloads as we don't know what return type the system uses. I > modeled the test below this way, that is we don't know if `isnan` has a > `bool` or `int` return type. > > > Having overloads with different return types will be observable. > > Unsure what observable effect you expect, the variants are there, yes, but > they have different names (wrt the base function and the other variant > function). The variant without a base function is simply an unused internal > function. Could you elaborate what problem you expect? What will be the result of `sizeof(isinf(1.0f))` ? I would expect it to be the same on host and on the device. I'm not quite sure what the pragma would do, so it's possible I'm barking at the wrong tree here. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D85879/new/ https://reviews.llvm.org/D85879 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
