[llvm] Use range attribute to constant fold comparisons with constant values. (PR #84627)

[Andreas Jonson via llvm-commits]

================
@@ -3776,24 +3796,14 @@ static Value *simplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS,
 
   // If both operands have range metadata, use the metadata
   // to simplify the comparison.
-  if (isa<Instruction>(RHS) && isa<Instruction>(LHS)) {
-    auto RHS_Instr = cast<Instruction>(RHS);
-    auto LHS_Instr = cast<Instruction>(LHS);
-
-    if (Q.IIQ.getMetadata(RHS_Instr, LLVMContext::MD_range) &&
-        Q.IIQ.getMetadata(LHS_Instr, LLVMContext::MD_range)) {
-      auto RHS_CR = getConstantRangeFromMetadata(
-          *RHS_Instr->getMetadata(LLVMContext::MD_range));
-      auto LHS_CR = getConstantRangeFromMetadata(
-          *LHS_Instr->getMetadata(LLVMContext::MD_range));
-
-      if (LHS_CR.icmp(Pred, RHS_CR))
+  if (std::optional<ConstantRange> RhsCr = getRange(RHS, Q.IIQ))
+    if (std::optional<ConstantRange> LhsCr = getRange(LHS, Q.IIQ)) {
+      if (LhsCr->icmp(Pred, *RhsCr))
         return ConstantInt::getTrue(RHS->getContext());
----------------
andjo403 wrote:

```suggestion
        return ConstantInt::getTrue(ITy);
```
I noticed that this simplification crashes for vectors types unless this change is applied.

https://github.com/llvm/llvm-project/pull/84627