[PATCH] Fix makeLibCall arguments for SoftenFloatRes_FRINT function

[Petar Jovanovic]

REPOSITORY
  rL LLVM

================
Comment at: include/llvm/Target/TargetLowering.h:1074
@@ -1073,1 +1073,3 @@
 
+ virtual bool shouldSignExtendFloat32InAbiCall() const {
+  return false;
----------------
Can you please add a comment before the function?

================
Comment at: lib/Target/Mips/MipsISelLowering.cpp:3034
@@ +3033,3 @@
+bool MipsTargetLowering::shouldSignExtendFloat32InAbiCall() const {
+  
+  return Subtarget.hasMips64();
----------------
Nit: remove empty line.

================
Comment at: lib/Target/Mips/MipsISelLowering.cpp:3035
@@ +3034,3 @@
+  
+  return Subtarget.hasMips64();
+    
----------------
I believe you should be checking for Subtarget.abiUsesSoftFloat() too.
Upper 32bits are undefined for hard float ABI.

================
Comment at: lib/Target/Mips/MipsISelLowering.cpp:3036
@@ +3035,3 @@
+  return Subtarget.hasMips64();
+    
+}
----------------
Remove empty line.

================
Comment at: lib/Target/Mips/MipsISelLowering.h:477
@@ -476,2 +476,3 @@
                         SDLoc dl, SelectionDAG &DAG) const override;
+    bool shouldSignExtendFloat32InAbiCall() const override;
 
----------------
Add an empty line above this.

http://reviews.llvm.org/D7791

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/