gh-93911: Specialize LOAD_ATTR for custom __getattr__ and __getattribute__

Include/cpython/object.h

@@ -234,6 +234,8 @@ struct _typeobject {
  * by code other than the specializer and interpreter. */
 struct _specialization_cache {
     PyObject *getitem;
+    PyObject *getattribute;
+    PyObject *getattr;
 };
 
 /* The *real* layout of a type object when allocated on the heap */

Include/internal/pycore_typeobject.h

@@ -44,6 +44,9 @@ extern PyStatus _PyTypes_InitSlotDefs(void);
 extern void _PyStaticType_Dealloc(PyTypeObject *type);
 
 
+extern PyObject *_Py_slot_tp_getattro(PyObject *self, PyObject *name);
+extern PyObject *_Py_slot_tp_getattr_hook(PyObject *self, PyObject *name);
+extern PyObject *_Py_call_attribute(PyObject *self, PyObject *attr, PyObject *name);
 #ifdef __cplusplus
 }
 #endif

Lib/opcode.py

@@ -285,6 +285,8 @@ def jabs_op(name, op):
         "LOAD_ATTR_ADAPTIVE",
         # These potentially push [NULL, bound method] onto the stack.
         "LOAD_ATTR_CLASS",
+        "LOAD_ATTR_GETATTRIBUTE_GETATTR_OVERRIDDEN",
+        "LOAD_ATTR_GETATTRIBUTE_GETATTR_PY_OVERRIDDEN",
         "LOAD_ATTR_INSTANCE_VALUE",
         "LOAD_ATTR_MODULE",
         "LOAD_ATTR_PROPERTY",

Lib/test/test_sys.py

@@ -1518,7 +1518,7 @@ def delx(self): del self.__x
                   '10P'                 # PySequenceMethods
                   '2P'                  # PyBufferProcs
                   '6P'
-                  '1P'                  # Specializer cache
+                  '3P'                  # Specializer cache
                   )
         class newstyleclass(object): pass
         # Separate block for PyDictKeysObject with 8 keys and 5 entries

Misc/NEWS.d/next/Core and Builtins/2022-06-18-17-00-33.gh-issue-93911.y286of.rst

@@ -0,0 +1 @@
+Specialize ``LOAD_ATTR`` for objects with custom ``__getattr__`` and ``__getattribute__``. Patch by Ken Jin.

Objects/typeobject.c

@@ -7766,24 +7766,24 @@ slot_tp_call(PyObject *self, PyObject *args, PyObject *kwds)
 
 /* There are two slot dispatch functions for tp_getattro.
 
-   - slot_tp_getattro() is used when __getattribute__ is overridden
+   - _Py_slot_tp_getattro() is used when __getattribute__ is overridden
      but no __getattr__ hook is present;
 
-   - slot_tp_getattr_hook() is used when a __getattr__ hook is present.
+   - _Py_slot_tp_getattr_hook() is used when a __getattr__ hook is present.
 
-   The code in update_one_slot() always installs slot_tp_getattr_hook(); this
-   detects the absence of __getattr__ and then installs the simpler slot if
-   necessary. */
+   The code in update_one_slot() always installs _Py_slot_tp_getattr_hook();
+   this detects the absence of __getattr__ and then installs the simpler
+   slot if necessary. */
 
-static PyObject *
-slot_tp_getattro(PyObject *self, PyObject *name)
+PyObject *
+_Py_slot_tp_getattro(PyObject *self, PyObject *name)
 {
     PyObject *stack[2] = {self, name};
     return vectorcall_method(&_Py_ID(__getattribute__), stack, 2);
 }
 
-static inline PyObject *
-call_attribute(PyObject *self, PyObject *attr, PyObject *name)
+PyObject *
+_Py_call_attribute(PyObject *self, PyObject *attr, PyObject *name)
 {
     PyObject *res, *descr = NULL;
 
@@ -7807,8 +7807,8 @@ call_attribute(PyObject *self, PyObject *attr, PyObject *name)
     return res;
 }
 
-static PyObject *
-slot_tp_getattr_hook(PyObject *self, PyObject *name)
+PyObject *
+_Py_slot_tp_getattr_hook(PyObject *self, PyObject *name)
 {
     PyTypeObject *tp = Py_TYPE(self);
     PyObject *getattr, *getattribute, *res;
@@ -7817,19 +7817,19 @@ slot_tp_getattr_hook(PyObject *self, PyObject *name)
        method fully for each attribute lookup for classes with
        __getattr__, even when the attribute is present. So we use
        _PyType_Lookup and create the method only when needed, with
-       call_attribute. */
+       _Py_call_attribute. */
     getattr = _PyType_Lookup(tp, &_Py_ID(__getattr__));
     if (getattr == NULL) {
         /* No __getattr__ hook: use a simpler dispatcher */
-        tp->tp_getattro = slot_tp_getattro;
-        return slot_tp_getattro(self, name);
+        tp->tp_getattro = _Py_slot_tp_getattro;
+        return _Py_slot_tp_getattro(self, name);
     }
     Py_INCREF(getattr);
     /* speed hack: we could use lookup_maybe, but that would resolve the
        method fully for each attribute lookup for classes with
        __getattr__, even when self has the default __getattribute__
        method. So we use _PyType_Lookup and create the method only when
-       needed, with call_attribute. */
+       needed, with _Py_call_attribute. */
     getattribute = _PyType_Lookup(tp, &_Py_ID(__getattribute__));
     if (getattribute == NULL ||
         (Py_IS_TYPE(getattribute, &PyWrapperDescr_Type) &&
@@ -7838,12 +7838,12 @@ slot_tp_getattr_hook(PyObject *self, PyObject *name)
         res = PyObject_GenericGetAttr(self, name);
     else {
         Py_INCREF(getattribute);
-        res = call_attribute(self, getattribute, name);
+        res = _Py_call_attribute(self, getattribute, name);
         Py_DECREF(getattribute);
     }
     if (res == NULL && PyErr_ExceptionMatches(PyExc_AttributeError)) {
         PyErr_Clear();
-        res = call_attribute(self, getattr, name);
+        res = _Py_call_attribute(self, getattr, name);
     }
     Py_DECREF(getattr);
     return res;
@@ -8182,10 +8182,10 @@ static slotdef slotdefs[] = {
            PyWrapperFlag_KEYWORDS),
     TPSLOT("__str__", tp_str, slot_tp_str, wrap_unaryfunc,
            "__str__($self, /)\n--\n\nReturn str(self)."),
-    TPSLOT("__getattribute__", tp_getattro, slot_tp_getattr_hook,
+    TPSLOT("__getattribute__", tp_getattro, _Py_slot_tp_getattr_hook,
            wrap_binaryfunc,
            "__getattribute__($self, name, /)\n--\n\nReturn getattr(self, name)."),
-    TPSLOT("__getattr__", tp_getattro, slot_tp_getattr_hook, NULL, ""),
+    TPSLOT("__getattr__", tp_getattro, _Py_slot_tp_getattr_hook, NULL, ""),
     TPSLOT("__setattr__", tp_setattro, slot_tp_setattro, wrap_setattr,
            "__setattr__($self, name, value, /)\n--\n\nImplement setattr(self, name, value)."),
     TPSLOT("__delattr__", tp_setattro, slot_tp_setattro, wrap_delattr,