[PATCH] D50975: [FPEnv] Support constrained FREM intrinsic

[Cameron McInally via Phabricator via llvm-commits]

cameron.mcinally added inline comments.


================
Comment at: lib/CodeGen/SelectionDAG/LegalizeDAG.cpp:4203
     break;
   case ISD::FADD:
     Results.push_back(ExpandFPLibCall(Node, RTLIB::ADD_F32, RTLIB::ADD_F64,
----------------
cameron.mcinally wrote:
> craig.topper wrote:
> > Should STRICT_FADD be here? Same question for MUL and SUB
> Good call. I'll add those in and see if I can find a test case that expands on x86. Maybe FP128 or something...
Ah, selecting the runtime FP128 calls is going to take good deal of work. Too much work to quickly tack on to this FREM change. 

Craig, are you okay with me addressing those changes (ADD/SUB/MUL/DIV) in a separate Diff? 


Repository:
  rL LLVM

https://reviews.llvm.org/D50975