[SubZero] lower float and double constants for MIPS

src/IceTargetLoweringMIPS32.h

Index: src/IceTargetLoweringMIPS32.h
diff --git a/src/IceTargetLoweringMIPS32.h b/src/IceTargetLoweringMIPS32.h
index 5cc412f2ad290bb27adcff09cc36b0c1945bf127..a027afdfbca89129dc0e8d5f77a6d233e63f3329 100644
--- a/src/IceTargetLoweringMIPS32.h
+++ b/src/IceTargetLoweringMIPS32.h
@@ -35,9 +35,8 @@ public:
 
   static void staticInit(GlobalContext *Ctx);
   static bool shouldBePooled(const Constant *C) {
-    if (auto *ConstDouble = llvm::dyn_cast<ConstantDouble>(C)) {
-      return !Utils::isPositiveZero(ConstDouble->getValue());

[Jim Stichnoth, 2016/09/19 20:50:14]

Question about this change.

Both ARM and x86 have idioms for loading a +0.0 pattern into a floating-point
register without requiring a memory load, usually something involving an xor
instruction.

Does MIPS32 have something similar?

If not, then this change is fine.

[jaydeep.patil, 2016/09/20 07:28:27]

Yes, for float/double constant 0 we can use mtc1 $zero.

-    }
+    if (llvm::isa<ConstantDouble>(C))
+      return true;
     if (llvm::isa<ConstantFloat>(C))
       return true;
     return false;
@@ -824,7 +823,6 @@ protected:
 
 private:
   ~TargetDataMIPS32() override = default;
-  template <typename T> static void emitConstantPool(GlobalContext *Ctx);
 };
 
 class TargetHeaderMIPS32 final : public TargetHeaderLowering {