gh-94673: Add Per-Interpreter Storage for Static Builtin Types (#95255)

This is the last precursor to storing tp_subclasses (and tp_weaklist) on the interpreter state for static builtin types.

Here we add per-type storage on PyInterpreterState, but only for the static builtin types.  This involves the following:

* add PyInterpreterState.types
   * move PyInterpreterState.type_cache to it
   * add a "num_builtins_initialized" field
   * add a "builtins" field (a static array big enough for all the static builtin types)
* add _PyStaticType_GetState() to look up a static builtin type's state
* (temporarily) add PyTypeObject.tp_static_builtin_index (to hold the type's index into PyInterpreterState.types.builtins)

We will be eliminating tp_static_builtin_index in a later change.

Author: ericsnowcurrently

Diff for: Include/cpython/object.h

@@ -227,6 +227,7 @@ struct _typeobject {
 
     destructor tp_finalize;
     vectorcallfunc tp_vectorcall;
+    size_t tp_static_builtin_index;  /* 0 means "not initialized" */
 };
 
 /* This struct is used by the specializer

Diff for: Include/internal/pycore_interp.h

@@ -173,7 +173,7 @@ struct _is {
     struct _Py_exc_state exc_state;
 
     struct ast_state ast;
-    struct type_cache type_cache;
+    struct types_state types;
     struct callable_cache callable_cache;
 
     /* The following fields are here to avoid allocation during init.

Diff for: Include/internal/pycore_typeobject.h

@@ -39,9 +39,25 @@ struct type_cache {
 #endif
 };
 
+/* For now we hard-code this to a value for which we are confident
+   all the static builtin types will fit (for all builds). */
+#define _Py_MAX_STATIC_BUILTIN_TYPES 200
+
+typedef struct {
+    PyTypeObject *type;
+} static_builtin_state;
+
+struct types_state {
+    struct type_cache type_cache;
+    size_t num_builtins_initialized;
+    static_builtin_state builtins[_Py_MAX_STATIC_BUILTIN_TYPES];
+};
+
+
 extern PyStatus _PyTypes_InitSlotDefs(void);
 
 extern int _PyStaticType_InitBuiltin(PyTypeObject *type);
+extern static_builtin_state * _PyStaticType_GetState(PyTypeObject *);
 extern void _PyStaticType_Dealloc(PyTypeObject *type);
 
 

Diff for: Lib/test/test_sys.py

@@ -1507,7 +1507,7 @@ def delx(self): del self.__x
         check((1,2,3), vsize('') + 3*self.P)
         # type
         # static type: PyTypeObject
-        fmt = 'P2nPI13Pl4Pn9Pn12PIP'
+        fmt = 'P2nPI13Pl4Pn9Pn12PIPI'
         s = vsize('2P' + fmt)
         check(int, s)
         # class

Diff for: Objects/typeobject.c

@@ -67,6 +67,88 @@ slot_tp_setattro(PyObject *self, PyObject *name, PyObject *value);
 
 static inline PyTypeObject * subclass_from_ref(PyObject *ref);
 
+
+/* helpers for for static builtin types */
+
+static inline int
+static_builtin_index_is_set(PyTypeObject *self)
+{
+    return self->tp_static_builtin_index > 0;
+}
+
+static inline size_t
+static_builtin_index_get(PyTypeObject *self)
+{
+    assert(static_builtin_index_is_set(self));
+    /* We store a 1-based index so 0 can mean "not initialized". */
+    return self->tp_static_builtin_index - 1;
+}
+
+static inline void
+static_builtin_index_set(PyTypeObject *self, size_t index)
+{
+    assert(index < _Py_MAX_STATIC_BUILTIN_TYPES);
+    /* We store a 1-based index so 0 can mean "not initialized". */
+    self->tp_static_builtin_index = index + 1;
+}
+
+static inline void
+static_builtin_index_clear(PyTypeObject *self)
+{
+    self->tp_static_builtin_index = 0;
+}
+
+static inline static_builtin_state *
+static_builtin_state_get(PyInterpreterState *interp, PyTypeObject *self)
+{
+    return &(interp->types.builtins[static_builtin_index_get(self)]);
+}
+
+/* For static types we store some state in an array on each interpreter. */
+static_builtin_state *
+_PyStaticType_GetState(PyTypeObject *self)
+{
+    assert(self->tp_flags & _Py_TPFLAGS_STATIC_BUILTIN);
+    PyInterpreterState *interp = _PyInterpreterState_GET();
+    return static_builtin_state_get(interp, self);
+}
+
+static void
+static_builtin_state_init(PyTypeObject *self)
+{
+    /* Set the type's per-interpreter state. */
+    PyInterpreterState *interp = _PyInterpreterState_GET();
+
+    /* It should only be called once for each builtin type. */
+    assert(!static_builtin_index_is_set(self));
+
+    static_builtin_index_set(self, interp->types.num_builtins_initialized);
+    interp->types.num_builtins_initialized++;
+
+    static_builtin_state *state = static_builtin_state_get(interp, self);
+    state->type = self;
+}
+
+static void
+static_builtin_state_clear(PyTypeObject *self)
+{
+    /* Reset the type's per-interpreter state.
+       This basically undoes what static_builtin_state_init() did. */
+    PyInterpreterState *interp = _PyInterpreterState_GET();
+
+    static_builtin_state *state = static_builtin_state_get(interp, self);
+    state->type = NULL;
+    static_builtin_index_clear(self);
+
+    assert(interp->types.num_builtins_initialized > 0);
+    interp->types.num_builtins_initialized--;
+}
+
+// Also see _PyStaticType_InitBuiltin() and _PyStaticType_Dealloc().
+
+/* end static builtin helpers */
+
+
 /*
  * finds the beginning of the docstring's introspection signature.
  * if present, returns a pointer pointing to the first '('.
@@ -206,7 +288,7 @@ static struct type_cache*
 get_type_cache(void)
 {
     PyInterpreterState *interp = _PyInterpreterState_GET();
-    return &interp->type_cache;
+    return &interp->types.type_cache;
 }
 
 
@@ -225,7 +307,7 @@ type_cache_clear(struct type_cache *cache, PyObject *value)
 void
 _PyType_InitCache(PyInterpreterState *interp)
 {
-    struct type_cache *cache = &interp->type_cache;
+    struct type_cache *cache = &interp->types.type_cache;
     for (Py_ssize_t i = 0; i < (1 << MCACHE_SIZE_EXP); i++) {
         struct type_cache_entry *entry = &cache->hashtable[i];
         assert(entry->name == NULL);
@@ -242,7 +324,7 @@ _PyType_InitCache(PyInterpreterState *interp)
 static unsigned int
 _PyType_ClearCache(PyInterpreterState *interp)
 {
-    struct type_cache *cache = &interp->type_cache;
+    struct type_cache *cache = &interp->types.type_cache;
 #if MCACHE_STATS
     size_t total = cache->hits + cache->collisions + cache->misses;
     fprintf(stderr, "-- Method cache hits        = %zd (%d%%)\n",
@@ -274,11 +356,17 @@ PyType_ClearCache(void)
 void
 _PyTypes_Fini(PyInterpreterState *interp)
 {
-    struct type_cache *cache = &interp->type_cache;
+    struct type_cache *cache = &interp->types.type_cache;
     type_cache_clear(cache, NULL);
     if (_Py_IsMainInterpreter(interp)) {
         clear_slotdefs();
     }
+
+    assert(interp->types.num_builtins_initialized == 0);
+    // All the static builtin types should have been finalized already.
+    for (size_t i = 0; i < _Py_MAX_STATIC_BUILTIN_TYPES; i++) {
+        assert(interp->types.builtins[i].type == NULL);
+    }
 }
 
 
@@ -4247,6 +4335,8 @@ clear_static_tp_subclasses(PyTypeObject *type)
 void
 _PyStaticType_Dealloc(PyTypeObject *type)
 {
+    assert(!(type->tp_flags & Py_TPFLAGS_HEAPTYPE));
+
     type_dealloc_common(type);
 
     Py_CLEAR(type->tp_dict);
@@ -4261,6 +4351,11 @@ _PyStaticType_Dealloc(PyTypeObject *type)
     }
 
     type->tp_flags &= ~Py_TPFLAGS_READY;
+
+    if (type->tp_flags & _Py_TPFLAGS_STATIC_BUILTIN) {
+        static_builtin_state_clear(type);
+        /* We leave _Py_TPFLAGS_STATIC_BUILTIN set on tp_flags. */
+    }
 }
 
 
@@ -6679,7 +6774,13 @@ _PyStaticType_InitBuiltin(PyTypeObject *self)
 {
     self->tp_flags = self->tp_flags | _Py_TPFLAGS_STATIC_BUILTIN;
 
-    return PyType_Ready(self);
+    static_builtin_state_init(self);
+
+    int res = PyType_Ready(self);
+    if (res < 0) {
+        static_builtin_state_clear(self);
+    }
+    return res;
 }