gh-148171: Convert CALL_BUILTIN_CLASS to leave arguments on the stack (gh-148381)

Author: corona10

Include/internal/pycore_ceval.h

@@ -461,7 +461,7 @@ _PyCallMethodDescriptorFastWithKeywords_StackRef(
     int total_args);
 
 PyAPI_FUNC(PyObject *)
-_Py_CallBuiltinClass_StackRefSteal(
+_Py_CallBuiltinClass_StackRef(
     _PyStackRef callable,
     _PyStackRef *arguments,
     int total_args);

Include/internal/pycore_opcode_metadata.h

@@ -4572,23 +4572,24 @@ dummy_func(
             EXIT_IF(tp->tp_vectorcall == NULL);
         }
 
-        op(_CALL_BUILTIN_CLASS, (callable, self_or_null, args[oparg] -- res)) {
+        op(_CALL_BUILTIN_CLASS, (callable, self_or_null, args[oparg] -- callable, self_or_null, args[oparg])) {
             int total_args = oparg;
             _PyStackRef *arguments = args;
             if (!PyStackRef_IsNull(self_or_null)) {
                 arguments--;
                 total_args++;
             }
             STAT_INC(CALL, hit);
-            PyObject *res_o = _Py_CallBuiltinClass_StackRefSteal(
+            PyObject *res_o = _Py_CallBuiltinClass_StackRef(
                 callable,
                 arguments,
                 total_args);
-            DEAD(args);
-            DEAD(self_or_null);
-            DEAD(callable);
-            ERROR_IF(res_o == NULL);
-            res = PyStackRef_FromPyObjectSteal(res_o);
+            if (res_o == NULL) {
+                ERROR_NO_POP();
+            }
+            _PyStackRef temp = callable;
+            callable = PyStackRef_FromPyObjectSteal(res_o);
+            PyStackRef_CLOSE(temp);
         }
 
         macro(CALL_BUILTIN_CLASS) =
@@ -4597,6 +4598,8 @@ dummy_func(
             unused/2 +
             _GUARD_CALLABLE_BUILTIN_CLASS +
             _CALL_BUILTIN_CLASS +
+            _POP_TOP_OPARG +
+            POP_TOP +
             _CHECK_PERIODIC_AT_END;
 
         op(_GUARD_CALLABLE_BUILTIN_O, (callable, self_or_null, args[oparg] -- callable, self_or_null, args[oparg])) {

Include/internal/pycore_uop_ids.h

@@ -894,30 +894,20 @@ _PyCallMethodDescriptorFastWithKeywords_StackRef(
 }
 
 PyObject *
-_Py_CallBuiltinClass_StackRefSteal(
+_Py_CallBuiltinClass_StackRef(
     _PyStackRef callable,
     _PyStackRef *arguments,
     int total_args)
 {
     PyObject *res;
     STACKREFS_TO_PYOBJECTS(arguments, total_args, args_o);
     if (CONVERSION_FAILED(args_o)) {
-        res = NULL;
-        goto cleanup;
+        return NULL;
     }
     PyTypeObject *tp = (PyTypeObject *)PyStackRef_AsPyObjectBorrow(callable);
     res = tp->tp_vectorcall((PyObject *)tp, args_o, total_args | PY_VECTORCALL_ARGUMENTS_OFFSET, NULL);
     STACKREFS_TO_PYOBJECTS_CLEANUP(args_o);
     assert((res != NULL) ^ (PyErr_Occurred() != NULL));
-cleanup:
-    // arguments is a pointer into the GC visible stack,
-    // so we must NULL out values as we clear them.
-    for (int i = total_args-1; i >= 0; i--) {
-        _PyStackRef tmp = arguments[i];
-        arguments[i] = PyStackRef_NULL;
-        PyStackRef_CLOSE(tmp);
-    }
-    PyStackRef_CLOSE(callable);
     return res;
 }
 

Include/internal/pycore_uop_metadata.h

@@ -1364,6 +1364,10 @@ dummy_func(void) {
         callable = sym_new_not_null(ctx);
     }
 
+    op(_CALL_BUILTIN_CLASS, (callable, self_or_null, args[oparg] -- callable, self_or_null, args[oparg])) {
+        callable = sym_new_not_null(ctx);
+    }
+
     op(_GUARD_CALLABLE_METHOD_DESCRIPTOR_O, (callable, self_or_null, args[oparg] -- callable, self_or_null, args[oparg])) {
         PyObject *callable_o = sym_get_const(ctx, callable);
         if (callable_o && sym_matches_type(callable, &PyMethodDescr_Type) &&