[llvm] r338381 - AMDGPU: Don't handle FP16_TO_FP in isCanonicalized

[Matt Arsenault via llvm-commits]

Author: arsenm
Date: Tue Jul 31 07:15:16 2018
New Revision: 338381

URL: http://llvm.org/viewvc/llvm-project?rev=338381&view=rev
Log:
AMDGPU: Don't handle FP16_TO_FP in isCanonicalized

This needs more special handling to do correctly.
Fixes test in subsequent commit.

Modified:
    llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp

Modified: llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp?rev=338381&r1=338380&r2=338381&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp Tue Jul 31 07:15:16 2018
@@ -6800,10 +6800,6 @@ static bool isCanonicalized(SelectionDAG
     return Op.getOperand(0).getValueType().getScalarType() != MVT::f16 ||
            ST->hasFP16Denormals();
 
-  case ISD::FP16_TO_FP:
-  case ISD::FP_TO_FP16:
-    return ST->hasFP16Denormals();
-
   // It can/will be lowered or combined as a bit operation.
   // Need to check their input recursively to handle.
   case ISD::FNEG: