[PATCH] D137241: [X86] Add ExpandLargeFpConvert Pass and enable for X86

[LuoYuanke via Phabricator via llvm-commits]

LuoYuanke added inline comments.


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:145
+  Value *FloatVal0 = FloatVal;
+  if (FloatVal->getType()->getFPMantissaWidth() == 64)
+    FloatVal0 =
----------------
FloatVal->getType()->isX86_FP80Ty()?


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:150
+      Builder.CreateBitCast(FloatVal0, Builder.getIntNTy(FloatWidth));
+  Value *ARep = Builder.CreateZExt(ARep0, FPToI->getType());
+  Value *Tobool_not = Builder.CreateICmpSGT(
----------------
Do we assume the _BitInt size is larger than 128-bit?


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:151
+  Value *ARep = Builder.CreateZExt(ARep0, FPToI->getType());
+  Value *Tobool_not = Builder.CreateICmpSGT(
+      ARep0, ConstantInt::getSigned(Builder.getIntNTy(FloatWidth), -1));
----------------
ToBoolNot?


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:158
+      Builder.CreateLShr(ARep, Builder.getIntN(BitWidth, FPMantissaWidth));
+  Value *Exponent_with_bias = Builder.CreateAnd(
+      And, Builder.getIntN(BitWidth, (1u << ExponentWidth) - 1));
----------------
ExponentWithBias.


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:164
+      Exponent_with_bias,
+      Builder.getIntN(BitWidth, (1u << (ExponentWidth - 1)) - 1));
+  Builder.CreateCondBr(Cmp, End, IfEnd);
----------------
`1u << (ExponentWidth - 1)) - 1` expression is used several times. Maybe to set `Bias = 1u << (ExponentWidth - 1)) - 1` to be more readable.


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:179
+  Builder.SetInsertPoint(IfThen5);
+  Value *Neg_one = Builder.CreateSExt(
+      ConstantInt::getSigned(Builder.getInt32Ty(), -1), IntTy);
----------------
NegOne


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:181
+      ConstantInt::getSigned(Builder.getInt32Ty(), -1), IntTy);
+  Value *Neg_inf =
+      Builder.CreateShl(ConstantInt::getSigned(IntTy, 1),
----------------
NegInf


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:184
+                        ConstantInt::getSigned(IntTy, BitWidth - 1));
+  Value *Pos_inf = Builder.CreateXor(Neg_one, Neg_inf, "inf_temp3");
+  Value *Cond8 = Builder.CreateSelect(Tobool_not, Pos_inf, Neg_inf, "cond8");
----------------
PosInf


================
Comment at: llvm/lib/CodeGen/ExpandLargeFpConvert.cpp:218
+  Builder.SetInsertPoint(End, End->begin());
+  PHINode *Retval_0 = Builder.CreatePHI(FPToI->getType(), 4);
+
----------------
Retval0 


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D137241/new/

https://reviews.llvm.org/D137241