[llvm] [X86] Don't fold very large offsets into addr displacements during ISel (PR #121678)

[Simon Pilgrim via llvm-commits]

================
@@ -1852,7 +1852,10 @@ bool X86DAGToDAGISel::foldOffsetIntoAddress(uint64_t Offset,
     if (Subtarget->isTarget64BitILP32() && !isUInt<31>(Val) &&
----------------
RKSimon wrote:

Do we know why its matching for unsigned here?

https://github.com/llvm/llvm-project/pull/121678