Add scaling benchmark and fix refcount contention on type

Author: colesbury

Objects/object.c

@@ -1999,7 +1999,7 @@ _PyObject_GenericSetAttrWithDict(PyObject *obj, PyObject *name,
     }
 
     Py_INCREF(name);
-    Py_INCREF(tp);
+    _Py_INCREF_TYPE(tp);
 
     PyThreadState *tstate = _PyThreadState_GET();
     _PyCStackRef cref;
@@ -2074,7 +2074,7 @@ _PyObject_GenericSetAttrWithDict(PyObject *obj, PyObject *name,
     }
   done:
     _PyThreadState_PopCStackRef(tstate, &cref);
-    Py_DECREF(tp);
+    _Py_DECREF_TYPE(tp);
     Py_DECREF(name);
     return res;
 }

Tools/ftscalingbench/ftscalingbench.py

@@ -264,6 +264,15 @@ def deepcopy():
     for i in range(40 * WORK_SCALE):
         copy.deepcopy(x)
 
+@register_benchmark
+def setattr_non_interned():
+    prefix = sys.implementation.name
+    obj = MyObject()
+    for _ in range(1000 * WORK_SCALE):
+        setattr(obj, f"{prefix}_a", None)
+        setattr(obj, f"{prefix}_b", None)
+        setattr(obj, f"{prefix}_c", None)
+
 
 def bench_one_thread(func):
     t0 = time.perf_counter_ns()