[llvm] expandIS_FPCLASS: Support fcNegative and fcPositive (PR #184788)

[Matt Arsenault via llvm-commits]

================
@@ -9306,6 +9306,67 @@ SDValue TargetLowering::expandIS_FPCLASS(EVT ResultVT, SDValue Op,
   // Tests that involve more than one class should be processed first.
   SDValue PartialRes;
 
+  // Handle sign bit tests first (fcPositive/fcNegative).
+  // These test only the sign bit, if not NaN.
+  // On 32-bit platforms with 64-bit floats, we need to be careful about
+  // integer comparisons. We use FP_ROUND to convert to a smaller float type
+  // that matches ResultVT's size, then compare with 0.
+  if (Test == fcPositive || Test == fcNegative) {
+    SDValue SignBitResult;
+    unsigned MaxLegalIntBits = 32;
+    if (isTypeLegal(MVT::i64))
+      MaxLegalIntBits = 64;
----------------
arsenm wrote:

Don't hardcode this, query from the legal type functions 

https://github.com/llvm/llvm-project/pull/184788