[llvm] [X86] Compute `CTLZ` using fp math for `v4i32` on SSE2 targets (PR #190930)

[Simon Pilgrim via llvm-commits]

================
@@ -10870,6 +10870,49 @@ SDValue TargetLowering::expandCTLZ(SDNode *Node, SelectionDAG &DAG) const {
   return DAG.getNode(ISD::CTPOP, dl, VT, Op);
 }
 
+SDValue TargetLowering::expandCTLZWithFP(SDNode *Node,
+                                         SelectionDAG &DAG) const {
+  // pseudocode :
+  // v = v & ~(v >> 8)
+  // v = (f32)v
+  // v = bitcast<i32>(v)
+  // v = v >> 23
+  // v = 158 - v
+  // v = min(v, 32)
+
+  SDLoc dl(Node);
+  EVT VT = Node->getValueType(0);
+  SDValue Op = Node->getOperand(0);
+
+  EVT EltVT = VT.getVectorElementType();
----------------
RKSimon wrote:

No need to make this vector specific:
```suggestion
  EVT SVT = VT.getScalarType();
```

https://github.com/llvm/llvm-project/pull/190930