Add MLX op handler for aten.bitwise_xor (#18927)

Add BitwiseXorNode to the MLX delegate, enabling element-wise bitwise
XOR for boolean and integer tensors via mlx::core::bitwise_xor.

Closes #18927

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: nanookclaw

backends/mlx/_generated_inspector.py

@@ -50,6 +50,7 @@
     AsStridedNode,
     AsTypeNode,
     Atan2Node,
+    BitwiseXorNode,
     BroadcastToNode,
     CeilNode,
     ClipNode,
@@ -490,6 +491,12 @@ def handler(P: MLXProgramBuilder, n: Node) -> Slot:
         "aten.ne",
         True,
     ),
+    (
+        [torch.ops.aten.bitwise_xor.Tensor, torch.ops.aten.bitwise_xor.Scalar],
+        BitwiseXorNode,
+        "aten.bitwise_xor",
+        True,
+    ),
 ]
 
 

backends/mlx/ops.py

@@ -1395,6 +1395,13 @@ exec_logical_or(const LogicalOrNode& n, ExecutionState& st, StreamOrDevice s) {
       n.out, logical_or(st.const_tensor_ref(n.a), st.const_tensor_ref(n.b), s));
 }
 
+inline void exec_bitwise_xor(
+    const BitwiseXorNode& n, ExecutionState& st, StreamOrDevice s) {
+  st.set_tensor(
+      n.out,
+      bitwise_xor(st.const_tensor_ref(n.a), st.const_tensor_ref(n.b), s));
+}
+
 inline void exec_tri(const TriNode& n, ExecutionState& st, StreamOrDevice s) {
   int rows = resolve_int(n.n, st);
   int cols = resolve_int(n.m, st);
@@ -2227,6 +2234,10 @@ class Interpreter {
       case OpCode::METAL_KERNEL:
         ops::exec_metal_kernel(std::get<MetalKernelNode>(instr.node), st, s);
         break;
+      case OpCode::BITWISE_XOR:
+        ops::exec_bitwise_xor(
+            std::get<BitwiseXorNode>(instr.node), st, s);
+        break;
       default:
         throw std::runtime_error(
             "Unknown opcode: " + std::to_string(static_cast<int>(instr.op)));