[PATCH] D149588: clang: Start emitting intrinsic for __builtin_ldexp*

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: clang/lib/CodeGen/CGBuiltin.cpp:3057-3058
+  case Builtin::BI__builtin_ldexpf:
+  case Builtin::BI__builtin_ldexpl:
+  case Builtin::BI__builtin_ldexpf16: {
+    return RValue::get(emitBinaryExpMaybeConstrainedFPBuiltin(
----------------
kpn wrote:
> tra wrote:
> > Should those builtins that map to an experimental intrinsic be `experimental`, too?
> > 
> > Alternatively, if the intrinsic functionality is something we intend to officially provide via clang builtin, should we drop the `experimental` part?
> > 
> They only map to experimental intrinsics if you use the still fairly new FENV_ACCESS support. So I'd say no to marking them experimental. Most users of them will not encounter any experimental support.
> 
> I'm still working on getting the IR Verifier changes into the tree. At the very least I don't think we should be dropping the experimental from the names of the constrained intrinsics when there's no checking at all for violations of the rules of using them.
It's not a constrained builtin, it's constrained handling for a builtin. Certainly part of the experimental part is that we only have patchwork support for strictfp for all of the builtins.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D149588/new/

https://reviews.llvm.org/D149588