gh-123923: Defer refcounting for `f_funcobj` in `_PyInterpreterFrame` (#124026)

Use a `_PyStackRef` and defer the reference to `f_funcobj` 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.

1 files changed, 21 insertions, 24 deletions

diff --git a/Python/bytecodes.c b/Python/bytecodes.c
index 846404e..5f194ae 100644
--- a/Python/bytecodes.c
+++ b/Python/bytecodes.c
@@ -808,14 +808,13 @@ dummy_func(
             assert(code->co_argcount == 2);
             DEOPT_IF(!_PyThreadState_HasStackSpace(tstate, code->co_framesize));
             STAT_INC(BINARY_SUBSCR, hit);
-            Py_INCREF(getitem);
         }
 
         op(_BINARY_SUBSCR_INIT_CALL, (container, sub -- new_frame: _PyInterpreterFrame* )) {
             PyTypeObject *tp = Py_TYPE(PyStackRef_AsPyObjectBorrow(container));
             PyHeapTypeObject *ht = (PyHeapTypeObject *)tp;
             PyObject *getitem = ht->_spec_cache.getitem;
-            new_frame = _PyFrame_PushUnchecked(tstate, (PyFunctionObject *)getitem, 2, frame);
+            new_frame = _PyFrame_PushUnchecked(tstate, PyStackRef_FromPyObjectNew(getitem), 2, frame);
             SYNC_SP();
             new_frame->localsplus[0] = container;
             new_frame->localsplus[1] = sub;
@@ -1666,8 +1665,9 @@ dummy_func(
         inst(COPY_FREE_VARS, (--)) {
             /* Copy closure variables to free variables */
             PyCodeObject *co = _PyFrame_GetCode(frame);
-            assert(PyFunction_Check(frame->f_funcobj));
-            PyObject *closure = ((PyFunctionObject *)frame->f_funcobj)->func_closure;
+            assert(PyStackRef_FunctionCheck(frame->f_funcobj));
+            PyFunctionObject *func = (PyFunctionObject *)PyStackRef_AsPyObjectBorrow(frame->f_funcobj);
+            PyObject *closure = func->func_closure;
             assert(oparg == co->co_nfreevars);
             int offset = co->co_nlocalsplus - oparg;
             for (int i = 0; i < oparg; ++i) {
@@ -2170,8 +2170,7 @@ dummy_func(
             DEOPT_IF(code->co_argcount != 1);
             DEOPT_IF(!_PyThreadState_HasStackSpace(tstate, code->co_framesize));
             STAT_INC(LOAD_ATTR, hit);
-            Py_INCREF(fget);
-            new_frame = _PyFrame_PushUnchecked(tstate, f, 1, frame);
+            new_frame = _PyFrame_PushUnchecked(tstate, PyStackRef_FromPyObjectNew(fget), 1, frame);
             new_frame->localsplus[0] = owner;
         }
 
@@ -2202,8 +2201,8 @@ dummy_func(
             STAT_INC(LOAD_ATTR, hit);
 
             PyObject *name = GETITEM(FRAME_CO_NAMES, oparg >> 1);
-            Py_INCREF(f);
-            _PyInterpreterFrame *new_frame = _PyFrame_PushUnchecked(tstate, f, 2, frame);
+            _PyInterpreterFrame *new_frame = _PyFrame_PushUnchecked(
+                tstate, PyStackRef_FromPyObjectNew(f), 2, frame);
             // Manipulate stack directly because we exit with DISPATCH_INLINED().
             STACK_SHRINK(1);
             new_frame->localsplus[0] = owner;
@@ -3251,7 +3250,7 @@ dummy_func(
                 int code_flags = ((PyCodeObject*)PyFunction_GET_CODE(callable_o))->co_flags;
                 PyObject *locals = code_flags & CO_OPTIMIZED ? NULL : Py_NewRef(PyFunction_GET_GLOBALS(callable_o));
                 _PyInterpreterFrame *new_frame = _PyEvalFramePushAndInit(
-                    tstate, (PyFunctionObject *)PyStackRef_AsPyObjectSteal(callable), locals,
+                    tstate, callable, locals,
                     args, total_args, NULL, frame
                 );
                 // Manipulate stack directly since we leave using DISPATCH_INLINED().
@@ -3340,7 +3339,7 @@ dummy_func(
             int code_flags = ((PyCodeObject*)PyFunction_GET_CODE(callable_o))->co_flags;
             PyObject *locals = code_flags & CO_OPTIMIZED ? NULL : Py_NewRef(PyFunction_GET_GLOBALS(callable_o));
             new_frame = _PyEvalFramePushAndInit(
-                tstate, (PyFunctionObject *)PyStackRef_AsPyObjectSteal(callable), locals,
+                tstate, callable, locals,
                 args, total_args, NULL, frame
             );
             // The frame has stolen all the arguments from the stack,
@@ -3475,11 +3474,9 @@ dummy_func(
         }
 
         replicate(5) pure op(_INIT_CALL_PY_EXACT_ARGS, (callable, self_or_null[1], args[oparg] -- new_frame: _PyInterpreterFrame*)) {
-            PyObject *callable_o = PyStackRef_AsPyObjectBorrow(callable);
             int has_self = !PyStackRef_IsNull(self_or_null[0]);
             STAT_INC(CALL, hit);
-            PyFunctionObject *func = (PyFunctionObject *)callable_o;
-            new_frame = _PyFrame_PushUnchecked(tstate, func, oparg + has_self, frame);
+            new_frame = _PyFrame_PushUnchecked(tstate, callable, oparg + has_self, frame);
             _PyStackRef *first_non_self_local = new_frame->localsplus + has_self;
             new_frame->localsplus[0] = self_or_null[0];
             for (int i = 0; i < oparg; i++) {
@@ -3601,10 +3598,9 @@ dummy_func(
             assert(_PyCode_CODE(_PyFrame_GetCode(shim))[0].op.code == EXIT_INIT_CHECK);
             /* Push self onto stack of shim */
             shim->localsplus[0] = PyStackRef_DUP(self);
-            PyFunctionObject *init_func = (PyFunctionObject *)PyStackRef_AsPyObjectSteal(init);
             args[-1] = self;
             init_frame = _PyEvalFramePushAndInit(
-                tstate, init_func, NULL, args-1, oparg+1, NULL, shim);
+                tstate, init, NULL, args-1, oparg+1, NULL, shim);
             SYNC_SP();
             if (init_frame == NULL) {
                 _PyEval_FrameClearAndPop(tstate, shim);
@@ -4080,7 +4076,7 @@ dummy_func(
                 int code_flags = ((PyCodeObject*)PyFunction_GET_CODE(callable_o))->co_flags;
                 PyObject *locals = code_flags & CO_OPTIMIZED ? NULL : Py_NewRef(PyFunction_GET_GLOBALS(callable_o));
                 _PyInterpreterFrame *new_frame = _PyEvalFramePushAndInit(
-                    tstate, (PyFunctionObject *)PyStackRef_AsPyObjectSteal(callable), locals,
+                    tstate, callable, locals,
                     args, positional_args, kwnames_o, frame
                 );
                 PyStackRef_CLOSE(kwnames);
@@ -4148,7 +4144,7 @@ dummy_func(
             int code_flags = ((PyCodeObject*)PyFunction_GET_CODE(callable_o))->co_flags;
             PyObject *locals = code_flags & CO_OPTIMIZED ? NULL : Py_NewRef(PyFunction_GET_GLOBALS(callable_o));
             new_frame = _PyEvalFramePushAndInit(
-                tstate, (PyFunctionObject *)PyStackRef_AsPyObjectSteal(callable), locals,
+                tstate, callable, locals,
                 args, positional_args, kwnames_o, frame
             );
             PyStackRef_CLOSE(kwnames);
@@ -4332,9 +4328,9 @@ dummy_func(
                     int code_flags = ((PyCodeObject *)PyFunction_GET_CODE(func))->co_flags;
                     PyObject *locals = code_flags & CO_OPTIMIZED ? NULL : Py_NewRef(PyFunction_GET_GLOBALS(func));
 
-                    _PyInterpreterFrame *new_frame = _PyEvalFramePushAndInit_Ex(tstate,
-                                                                                (PyFunctionObject *)PyStackRef_AsPyObjectSteal(func_st), locals,
-                                                                                nargs, callargs, kwargs, frame);
+                    _PyInterpreterFrame *new_frame = _PyEvalFramePushAndInit_Ex(
+                        tstate, func_st, locals,
+                        nargs, callargs, kwargs, frame);
                     // Need to manually shrink the stack since we exit with DISPATCH_INLINED.
                     STACK_SHRINK(oparg + 3);
                     if (new_frame == NULL) {
@@ -4408,8 +4404,8 @@ dummy_func(
         }
 
         inst(RETURN_GENERATOR, (-- res)) {
-            assert(PyFunction_Check(frame->f_funcobj));
-            PyFunctionObject *func = (PyFunctionObject *)frame->f_funcobj;
+            assert(PyStackRef_FunctionCheck(frame->f_funcobj));
+            PyFunctionObject *func = (PyFunctionObject *)PyStackRef_AsPyObjectBorrow(frame->f_funcobj);
             PyGenObject *gen = (PyGenObject *)_Py_MakeCoro(func);
             if (gen == NULL) {
                 ERROR_NO_POP();
@@ -4771,8 +4767,9 @@ dummy_func(
         }
 
         tier2 op(_CHECK_FUNCTION, (func_version/2 -- )) {
-            assert(PyFunction_Check(frame->f_funcobj));
-            DEOPT_IF(((PyFunctionObject *)frame->f_funcobj)->func_version != func_version);
+            assert(PyStackRef_FunctionCheck(frame->f_funcobj));
+            PyFunctionObject *func = (PyFunctionObject *)PyStackRef_AsPyObjectBorrow(frame->f_funcobj);
+            DEOPT_IF(func->func_version != func_version);
         }
 
         /* Internal -- for testing executors */