[X86] Return UNDEF from LowerScalarImmediateShift when the shift amount is out of range.

[craig.topper]

Summary

I think we only turn out of range shiftss to undef when
all elements are out of range or the shift amount is a splat out
of range. I'm not sure which, I didn't check.

During lowering we can split a shift where some elements
are out of range into multiple shifts. This can create a
new shift with a splat shift amount that is out of range.

This patch returns undef for this case.

Fixes PR42615.

[RKSimon]

This is only ever coming from ISD::SHL/SRL/SRA so returning UNDEF should be fine.

Alternatively if we wanted to match the sse behaviour, then we'd keep the clamp for SRA and return 0 for SHL/SRL cases.

Overall I think I prefer the UNDEF route tbh.

[craig.topper]

Switch to returning undef

[RKSimon]

LGTM

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

auto ArithmeticShiftRight64 = [&](uint64_t ShiftAmt) {
     }
     return DAG.getBitcast(VT, Ex);
   };
 
   // Optimize shl/srl/sra with constant shift amount.
   APInt APIntShiftAmt;
   if (!isConstantSplat(Amt, APIntShiftAmt))
     return SDValue();
-  assert(APIntShiftAmt.ult(VT.getScalarSizeInBits()) &&
-         "Out of range shift amount");
+
+  // If the shift amount is out of range, return undef.
+  if (APIntShiftAmt.uge(VT.getScalarSizeInBits()))
+    return DAG.getUNDEF(VT);
+
   uint64_t ShiftAmt = APIntShiftAmt.getZExtValue();
 
   if (SupportedVectorShiftWithImm(VT, Subtarget, Op.getOpcode()))
     return getTargetVShiftByConstNode(X86Opc, dl, VT, R, ShiftAmt, DAG);
 
   // i64 SRA needs to be performed as partial shifts.
   if (((!Subtarget.hasXOP() && VT == MVT::v2i64) ||
        (Subtarget.hasInt256() && VT == MVT::v4i64)) &&