push stackrefs into tstate

Author: Fidget-Spinner

Include/internal/pycore_tstate.h

@@ -21,6 +21,9 @@ struct _gc_thread_state {
 };
 #endif
 
+/* How much scratch space to give stackref to PyObject* conversion. */
+#define MAX_STACKREF_SCRATCH 10
+
 // Every PyThreadState is actually allocated as a _PyThreadStateImpl. The
 // PyThreadState fields are exposed as part of the C API, although most fields
 // are intended to be private. The _PyThreadStateImpl fields not exposed.
@@ -47,6 +50,7 @@ typedef struct _PyThreadStateImpl {
     struct _qsbr_thread_state *qsbr;  // only used by free-threaded build
     struct llist_node mem_free_queue; // delayed free queue
 
+    PyObject *stackref_scratch[MAX_STACKREF_SCRATCH + 1];
 #ifdef Py_GIL_DISABLED
     // Stack references for the current thread that exist on the C stack
     struct _PyCStackRef *c_stack_refs;

Python/ceval_macros.h

@@ -393,13 +393,19 @@ do {                                                   \
 
 /* Stackref macros */
 
-/* How much scratch space to give stackref to PyObject* conversion. */
-#define MAX_STACKREF_SCRATCH 10
 
+#if Py_TAIL_CALL_INTERP && defined(_MSC_VER)
+#define STACKREFS_TO_PYOBJECTS(ARGS, ARG_COUNT, NAME) \
+    /* +1 because vectorcall might use -1 to write self */ \
+    _PyThreadStateImpl *_tstate = (_PyThreadStateImpl *)tstate; \
+    PyObject **NAME##_temp = _tstate->stackref_scratch; \
+    PyObject **NAME = _PyObjectArray_FromStackRefArray(ARGS, ARG_COUNT, NAME##_temp + 1);
+#else
 #define STACKREFS_TO_PYOBJECTS(ARGS, ARG_COUNT, NAME) \
     /* +1 because vectorcall might use -1 to write self */ \
     PyObject *NAME##_temp[MAX_STACKREF_SCRATCH+1]; \
     PyObject **NAME = _PyObjectArray_FromStackRefArray(ARGS, ARG_COUNT, NAME##_temp + 1);
+#endif
 
 #define STACKREFS_TO_PYOBJECTS_CLEANUP(NAME) \
     /* +1 because we +1 previously */ \