[NFC][X86] Fix Werror=extra error due to enum

[jsji]

This is one of the many PRs to fix errors with LLVM_ENABLE_WERROR=on. Built by GCC 11.

Fix warnings:
llvm-project/llvm/lib/Target/X86/X86FastISel.cpp: In member function ‘virtual bool {anonymous}::X86FastISel::fastLowerCall(llvm::FastISel::CallLoweringInfo&)’:
llvm-project/llvm/lib/Target/X86/X86FastISel.cpp:3547: error: enumerated and non-enumerated type in conditional expression [-Werror=extra]
3547 | MIB.addReg(Is64Bit ? X86::RIP : 0).addImm(1).addReg(0);

[llvmbot]

@llvm/pr-subscribers-backend-x86

Author: Jinsong Ji (jsji)

Changes

This is one of the many PRs to fix errors with LLVM_ENABLE_WERROR=on. Built by GCC 11.

Fix warnings:
llvm-project/llvm/lib/Target/X86/X86FastISel.cpp: In member function ‘virtual bool {anonymous}::X86FastISel::fastLowerCall(llvm::FastISel::CallLoweringInfo&)’:
llvm-project/llvm/lib/Target/X86/X86FastISel.cpp:3547: error: enumerated and non-enumerated type in conditional expression [-Werror=extra]
3547 | MIB.addReg(Is64Bit ? X86::RIP : 0).addImm(1).addReg(0);

---

Full diff: https://github.com/llvm/llvm-project/pull/112812.diff

2 Files Affected:

• (modified) llvm/lib/Target/X86/X86FastISel.cpp (+1-1)
• (modified) llvm/lib/Target/X86/X86ISelLowering.cpp (+1-1)

diff --git a/llvm/lib/Target/X86/X86FastISel.cpp b/llvm/lib/Target/X86/X86FastISel.cpp
index 4bf660b5e234ae..9e4e5547c642cc 100644
--- a/llvm/lib/Target/X86/X86FastISel.cpp
+++ b/llvm/lib/Target/X86/X86FastISel.cpp
@@ -3544,7 +3544,7 @@ bool X86FastISel::fastLowerCall(CallLoweringInfo &CLI) {
 
     MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, MIMD, TII.get(CallOpc));
     if (NeedLoad)
-      MIB.addReg(Is64Bit ? X86::RIP : 0).addImm(1).addReg(0);
+      MIB.addReg(Is64Bit ? X86::RIP : X86::NoRegister).addImm(1).addReg(0);
     if (Symbol)
       MIB.addSym(Symbol, OpFlags);
     else
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index db633d10edc49a..a469a5a554354f 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -29986,7 +29986,7 @@ static SDValue LowerShift(SDValue Op, const X86Subtarget &Subtarget,
           MVT::getVectorVT(NarrowScalarVT, WideNumElts), dl, AmtWideElts);
       AmtWide = DAG.getZExtOrTrunc(AmtWide, dl, WideVT);
       // Perform the actual shift.
-      unsigned LogicalOpc = Opc == ISD::SRA ? ISD::SRL : Opc;
+      unsigned LogicalOpc = Opc == ISD::SRA ? (unsigned)ISD::SRL : Opc;
       SDValue ShiftedR = DAG.getNode(LogicalOpc, dl, WideVT, RWide, AmtWide);
       // Now we need to construct a mask which will "drop" bits that get
       // shifted past the LSB/MSB. For a logical shift left, it will look

[phoebewang]

Why need (unsigned) here? If the ISD::XXX is not a unsigned value, I think we would have more warn/error when comparing with Opc.

[phoebewang]

Oh, maybe because we just use == and != here.

[jsji]

It is about Opc is unsigned and ISD::SRL is enum, and they are used in the ?: operator, so type mismatch.

[phoebewang]

LGTM.