gh-124218: Avoid refcount contention on builtins module (GH-125847)

This replaces `_PyEval_BuiltinsFromGlobals` with
`_PyDict_LoadBuiltinsFromGlobals`, which returns a new reference
instead of a borrowed reference. Internally, the new function uses
per-thread reference counting when possible to avoid contention on the
refcount fields on the builtins module.

2 files changed, 29 insertions, 3 deletions

diff --git a/Include/internal/pycore_ceval.h b/Include/internal/pycore_ceval.h
index cff2b1f..411bbff 100644
--- a/Include/internal/pycore_ceval.h
+++ b/Include/internal/pycore_ceval.h
@@ -83,9 +83,6 @@ extern void _PyEval_Fini(void);
 
 
 extern PyObject* _PyEval_GetBuiltins(PyThreadState *tstate);
-extern PyObject* _PyEval_BuiltinsFromGlobals(
-    PyThreadState *tstate,
-    PyObject *globals);
 
 // Trampoline API
 
diff --git a/Include/internal/pycore_dict.h b/Include/internal/pycore_dict.h
index 1d18555..c5399ad 100644
--- a/Include/internal/pycore_dict.h
+++ b/Include/internal/pycore_dict.h
@@ -108,6 +108,9 @@ extern Py_ssize_t _PyDictKeys_StringLookup(PyDictKeysObject* dictkeys, PyObject
 PyAPI_FUNC(PyObject *)_PyDict_LoadGlobal(PyDictObject *, PyDictObject *, PyObject *);
 PyAPI_FUNC(void) _PyDict_LoadGlobalStackRef(PyDictObject *, PyDictObject *, PyObject *, _PyStackRef *);
 
+// Loads the __builtins__ object from the globals dict. Returns a new reference.
+extern PyObject *_PyDict_LoadBuiltinsFromGlobals(PyObject *globals);
+
 /* Consumes references to key and value */
 PyAPI_FUNC(int) _PyDict_SetItem_Take2(PyDictObject *op, PyObject *key, PyObject *value);
 extern int _PyDict_SetItem_LockHeld(PyDictObject *dict, PyObject *name, PyObject *value);
@@ -318,6 +321,8 @@ PyDictObject *_PyObject_MaterializeManagedDict_LockHeld(PyObject *);
 #ifndef Py_GIL_DISABLED
 #  define _Py_INCREF_DICT Py_INCREF
 #  define _Py_DECREF_DICT Py_DECREF
+#  define _Py_INCREF_BUILTINS Py_INCREF
+#  define _Py_DECREF_BUILTINS Py_DECREF
 #else
 static inline Py_ssize_t
 _PyDict_UniqueId(PyDictObject *mp)
@@ -341,6 +346,30 @@ _Py_DECREF_DICT(PyObject *op)
     Py_ssize_t id = _PyDict_UniqueId((PyDictObject *)op);
     _Py_THREAD_DECREF_OBJECT(op, id);
 }
+
+// Like `_Py_INCREF_DICT`, but also handles non-dict objects because builtins
+// may not be a dict.
+static inline void
+_Py_INCREF_BUILTINS(PyObject *op)
+{
+    if (PyDict_CheckExact(op)) {
+        _Py_INCREF_DICT(op);
+    }
+    else {
+        Py_INCREF(op);
+    }
+}
+
+static inline void
+_Py_DECREF_BUILTINS(PyObject *op)
+{
+    if (PyDict_CheckExact(op)) {
+        _Py_DECREF_DICT(op);
+    }
+    else {
+        Py_DECREF(op);
+    }
+}
 #endif
 
 #ifdef __cplusplus