gh-115999: Specialize CALL_KW in free-threaded builds (#127713)

* Enable specialization of CALL_KW

* Fix bug pushing frame in _PY_FRAME_KW

`_PY_FRAME_KW` pushes a pointer to the new frame onto the stack for
consumption by the next uop. When pushing the frame fails, we do not
want to push the result, `NULL`, to the stack because it is not
a valid stackref. This works in the default build because `PyStackRef_NULL`
 and `NULL` are the same value, so the `PyStackRef_XCLOSE()` in the error
handler ignores it. In the free-threaded build the values are not the same;
`PyStackRef_XCLOSE()` will attempt to decref a null pointer.

Author: mpage

Python/bytecodes.c

@@ -4311,17 +4311,18 @@ dummy_func(
             assert(Py_TYPE(callable_o) == &PyFunction_Type);
             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(
+            _PyInterpreterFrame *temp = _PyEvalFramePushAndInit(
                 tstate, callable[0], locals,
                 args, positional_args, kwnames_o, frame
             );
             PyStackRef_CLOSE(kwnames);
             // The frame has stolen all the arguments from the stack,
             // so there is no need to clean them up.
             SYNC_SP();
-            if (new_frame == NULL) {
+            if (temp == NULL) {
                 ERROR_NO_POP();
             }
+            new_frame = temp;
         }
 
         op(_CHECK_FUNCTION_VERSION_KW, (func_version/2, callable[1], self_or_null[1], unused[oparg], kwnames -- callable[1], self_or_null[1], unused[oparg], kwnames)) {
@@ -4372,15 +4373,15 @@ dummy_func(
             _PUSH_FRAME;
 
         specializing op(_SPECIALIZE_CALL_KW, (counter/1, callable[1], self_or_null[1], args[oparg], kwnames -- callable[1], self_or_null[1], args[oparg], kwnames)) {
-            #if ENABLE_SPECIALIZATION
+            #if ENABLE_SPECIALIZATION_FT
             if (ADAPTIVE_COUNTER_TRIGGERS(counter)) {
                 next_instr = this_instr;
                 _Py_Specialize_CallKw(callable[0], next_instr, oparg + !PyStackRef_IsNull(self_or_null[0]));
                 DISPATCH_SAME_OPARG();
             }
             OPCODE_DEFERRED_INC(CALL_KW);
             ADVANCE_ADAPTIVE_COUNTER(this_instr[1].counter);
-            #endif  /* ENABLE_SPECIALIZATION */
+            #endif  /* ENABLE_SPECIALIZATION_FT */
         }
 
         macro(CALL_KW) =

Python/executor_cases.c.h

@@ -2107,7 +2107,7 @@ specialize_py_call_kw(PyFunctionObject *func, _Py_CODEUNIT *instr, int nargs,
         return -1;
     }
     write_u32(cache->func_version, version);
-    instr->op.code = bound_method ? CALL_KW_BOUND_METHOD : CALL_KW_PY;
+    specialize(instr, bound_method ? CALL_KW_BOUND_METHOD : CALL_KW_PY);
     return 0;
 }
 
@@ -2202,10 +2202,9 @@ _Py_Specialize_CallKw(_PyStackRef callable_st, _Py_CODEUNIT *instr, int nargs)
 {
     PyObject *callable = PyStackRef_AsPyObjectBorrow(callable_st);
 
-    assert(ENABLE_SPECIALIZATION);
+    assert(ENABLE_SPECIALIZATION_FT);
     assert(_PyOpcode_Caches[CALL_KW] == INLINE_CACHE_ENTRIES_CALL_KW);
     assert(_Py_OPCODE(*instr) != INSTRUMENTED_CALL_KW);
-    _PyCallCache *cache = (_PyCallCache *)(instr + 1);
     int fail;
     if (PyFunction_Check(callable)) {
         fail = specialize_py_call_kw((PyFunctionObject *)callable, instr, nargs, false);
@@ -2221,19 +2220,11 @@ _Py_Specialize_CallKw(_PyStackRef callable_st, _Py_CODEUNIT *instr, int nargs)
         }
     }
     else {
-        instr->op.code = CALL_KW_NON_PY;
+        specialize(instr, CALL_KW_NON_PY);
         fail = 0;
     }
     if (fail) {
-        STAT_INC(CALL, failure);
-        assert(!PyErr_Occurred());
-        instr->op.code = CALL_KW;
-        cache->counter = adaptive_counter_backoff(cache->counter);
-    }
-    else {
-        STAT_INC(CALL, success);
-        assert(!PyErr_Occurred());
-        cache->counter = adaptive_counter_cooldown();
+        unspecialize(instr);
     }
 }