gh-123923: Defer refcounting for `f_executable` in `_PyInterpreterFrame` (#123924)

Use a `_PyStackRef` and defer the reference to `f_executable` when
possible. This avoids some reference count contention in the common case
of executing the same code object from multiple threads concurrently in
the free-threaded build.

2 files changed, 13 insertions, 10 deletions

diff --git a/Objects/frameobject.c b/Objects/frameobject.c
index 85c2455..b567327 100644
--- a/Objects/frameobject.c
+++ b/Objects/frameobject.c
@@ -1625,8 +1625,6 @@ frame_dealloc(PyFrameObject *f)
     }
 
     Py_TRASHCAN_BEGIN(f, frame_dealloc);
-    PyObject *co = NULL;
-
     /* GH-106092: If f->f_frame was on the stack and we reached the maximum
      * nesting depth for deallocations, the trashcan may have delayed this
      * deallocation until after f->f_frame is freed. Avoid dereferencing
@@ -1635,9 +1633,7 @@ frame_dealloc(PyFrameObject *f)
 
     /* Kill all local variables including specials, if we own them */
     if (f->f_frame == frame && frame->owner == FRAME_OWNED_BY_FRAME_OBJECT) {
-        /* Don't clear code object until the end */
-        co = frame->f_executable;
-        frame->f_executable = NULL;
+        PyStackRef_CLEAR(frame->f_executable);
         Py_CLEAR(frame->f_funcobj);
         Py_CLEAR(frame->f_locals);
         _PyStackRef *locals = _PyFrame_GetLocalsArray(frame);
@@ -1652,7 +1648,6 @@ frame_dealloc(PyFrameObject *f)
     Py_CLEAR(f->f_extra_locals);
     Py_CLEAR(f->f_locals_cache);
     PyObject_GC_Del(f);
-    Py_XDECREF(co);
     Py_TRASHCAN_END;
 }
 
diff --git a/Objects/genobject.c b/Objects/genobject.c
index b281af8..5dc8f92 100644
--- a/Objects/genobject.c
+++ b/Objects/genobject.c
@@ -55,6 +55,14 @@ gen_traverse(PyGenObject *gen, visitproc visit, void *arg)
             return err;
         }
     }
+    else {
+        // We still need to visit the code object when the frame is cleared to
+        // ensure that it's kept alive if the reference is deferred.
+        int err = _PyGC_VisitStackRef(&gen->gi_iframe.f_executable, visit, arg);
+        if (err) {
+            return err;
+        }
+    }
     /* No need to visit cr_origin, because it's just tuples/str/int, so can't
        participate in a reference cycle. */
     Py_VISIT(gen->gi_exc_state.exc_value);
@@ -139,6 +147,9 @@ gen_dealloc(PyGenObject *gen)
            and GC_Del. */
         Py_CLEAR(((PyAsyncGenObject*)gen)->ag_origin_or_finalizer);
     }
+    if (PyCoro_CheckExact(gen)) {
+        Py_CLEAR(((PyCoroObject *)gen)->cr_origin_or_finalizer);
+    }
     if (gen->gi_frame_state != FRAME_CLEARED) {
         _PyInterpreterFrame *frame = &gen->gi_iframe;
         gen->gi_frame_state = FRAME_CLEARED;
@@ -147,10 +158,7 @@ gen_dealloc(PyGenObject *gen)
         _PyErr_ClearExcState(&gen->gi_exc_state);
     }
     assert(gen->gi_exc_state.exc_value == NULL);
-    if (_PyGen_GetCode(gen)->co_flags & CO_COROUTINE) {
-        Py_CLEAR(((PyCoroObject *)gen)->cr_origin_or_finalizer);
-    }
-    Py_DECREF(_PyGen_GetCode(gen));
+    PyStackRef_CLEAR(gen->gi_iframe.f_executable);
     Py_CLEAR(gen->gi_name);
     Py_CLEAR(gen->gi_qualname);