GH-108362: Revert "GH-108362: Incremental GC implementation (GH-108038)" (#115132)

Revert "GH-108362: Incremental GC implementation (GH-108038)"

This reverts commit 36518e69d7.

Doc/whatsnew/3.13.rst

@@ -92,10 +92,6 @@ Interpreter improvements:
 New Features
 ============
 
-* The cyclic garbage collector is now incremental.
-  This means that maximum pause times are reduced,
-  by an order of magnitude or more for larger heaps.
-
 Improved Error Messages
 -----------------------
 
@@ -105,13 +101,6 @@ Improved Error Messages
   variables. See also :ref:`using-on-controlling-color`.
   (Contributed by Pablo Galindo Salgado in :gh:`112730`.)
 
-Incremental Garbage Collection
-------------------------------
-
-* The cycle garbage collector is now incremental.
-  This means that maximum pause times are reduced
-  by an order of magnitude or more for larger heaps.
-
 Other Language Changes
 ======================
 
@@ -257,29 +246,6 @@ fractions
   sign handling, minimum width and grouping. (Contributed by Mark Dickinson
   in :gh:`111320`.)
 
-gc
---
-* The cyclic garbage collector is now incremental, which changes the meanings
-  of the results of :meth:`gc.get_threshold` and :meth:`gc.get_threshold` as
-  well as :meth:`gc.get_count` and :meth:`gc.get_stats`.
-* :meth:`gc.get_threshold` returns a three-tuple for backwards compatibility,
-  the first value is the threshold for young collections, as before, the second
-  value determines the rate at which the old collection is scanned; the
-  default is 10 and higher values mean that the old collection is scanned more slowly.
-  The third value is meangless and is always zero.
-* :meth:`gc.set_threshold` ignores any items after the second.
-* :meth:`gc.get_count` and :meth:`gc.get_stats`.
-  These functions return the same format of results as before.
-  The only difference is that instead of the results refering to
-  the young, aging and old generations, the results refer to the
-  young generation and the aging and collecting spaces of the old generation.
-
-In summary, code that attempted to manipulate the behavior of the cycle GC may
-not work as well as intended, but it is very unlikely to harmful.
-All other code will work just fine.
-Uses should avoid calling :meth:`gc.collect` unless their workload is episodic,
-but that has always been the case to some extent.
-
 glob
 ----
 

Include/internal/pycore_gc.h

@@ -88,15 +88,11 @@ static inline void _PyObject_GC_SET_SHARED(PyObject *op) {
 
 /* Bit flags for _gc_prev */
 /* Bit 0 is set when tp_finalize is called */
-#define _PyGC_PREV_MASK_FINALIZED  1
+#define _PyGC_PREV_MASK_FINALIZED  (1)
 /* Bit 1 is set when the object is in generation which is GCed currently. */
-#define _PyGC_PREV_MASK_COLLECTING 2
+#define _PyGC_PREV_MASK_COLLECTING (2)
-
-/* Bit 0 is set if the object belongs to old space 1 */
-#define _PyGC_NEXT_MASK_OLD_SPACE_1    1
-
 /* The (N-2) most significant bits contain the real address. */
-#define _PyGC_PREV_SHIFT           2
+#define _PyGC_PREV_SHIFT           (2)
 #define _PyGC_PREV_MASK            (((uintptr_t) -1) << _PyGC_PREV_SHIFT)
 
 /* set for debugging information */
@@ -122,13 +118,11 @@ typedef enum {
 // Lowest bit of _gc_next is used for flags only in GC.
 // But it is always 0 for normal code.
 static inline PyGC_Head* _PyGCHead_NEXT(PyGC_Head *gc) {
-    uintptr_t next = gc->_gc_next & _PyGC_PREV_MASK;
+    uintptr_t next = gc->_gc_next;
     return (PyGC_Head*)next;
 }
 static inline void _PyGCHead_SET_NEXT(PyGC_Head *gc, PyGC_Head *next) {
-    uintptr_t unext = (uintptr_t)next;
+    gc->_gc_next = (uintptr_t)next;
-    assert((unext & ~_PyGC_PREV_MASK) == 0);
-    gc->_gc_next = (gc->_gc_next & ~_PyGC_PREV_MASK) | unext;
 }
 
 // Lowest two bits of _gc_prev is used for _PyGC_PREV_MASK_* flags.
@@ -136,7 +130,6 @@ static inline PyGC_Head* _PyGCHead_PREV(PyGC_Head *gc) {
     uintptr_t prev = (gc->_gc_prev & _PyGC_PREV_MASK);
     return (PyGC_Head*)prev;
 }
-
 static inline void _PyGCHead_SET_PREV(PyGC_Head *gc, PyGC_Head *prev) {
     uintptr_t uprev = (uintptr_t)prev;
     assert((uprev & ~_PyGC_PREV_MASK) == 0);
@@ -222,13 +215,6 @@ struct gc_generation {
                   generations */
 };
 
-struct gc_collection_stats {
-    /* number of collected objects */
-    Py_ssize_t collected;
-    /* total number of uncollectable objects (put into gc.garbage) */
-    Py_ssize_t uncollectable;
-};
-
 /* Running stats per generation */
 struct gc_generation_stats {
     /* total number of collections */
@@ -250,8 +236,8 @@ struct _gc_runtime_state {
     int enabled;
     int debug;
     /* linked lists of container objects */
-    struct gc_generation young;
+    struct gc_generation generations[NUM_GENERATIONS];
-    struct gc_generation old[2];
+    PyGC_Head *generation0;
     /* a permanent generation which won't be collected */
     struct gc_generation permanent_generation;
     struct gc_generation_stats generation_stats[NUM_GENERATIONS];
@@ -264,20 +250,22 @@ struct _gc_runtime_state {
     /* This is the number of objects that survived the last full
        collection. It approximates the number of long lived objects
        tracked by the GC.
+
        (by "full collection", we mean a collection of the oldest
        generation). */
     Py_ssize_t long_lived_total;
-
+    /* This is the number of objects that survived all "non-full"
-    Py_ssize_t work_to_do;
+       collections, and are awaiting to undergo a full collection for
-    /* Which of the old spaces is the visited space */
+       the first time. */
-    int visited_space;
+    Py_ssize_t long_lived_pending;
 };
 
 
 extern void _PyGC_InitState(struct _gc_runtime_state *);
 
-extern Py_ssize_t _PyGC_Collect(PyThreadState *tstate, int generation, _PyGC_Reason reason);
+extern Py_ssize_t _PyGC_Collect(PyThreadState *tstate, int generation,
-extern void _PyGC_CollectNoFail(PyThreadState *tstate);
+                                _PyGC_Reason reason);
+extern Py_ssize_t _PyGC_CollectNoFail(PyThreadState *tstate);
 
 /* Freeze objects tracked by the GC and ignore them in future collections. */
 extern void _PyGC_Freeze(PyInterpreterState *interp);

Include/internal/pycore_object.h

@@ -125,7 +125,19 @@ static inline void _Py_RefcntAdd(PyObject* op, Py_ssize_t n)
 }
 #define _Py_RefcntAdd(op, n) _Py_RefcntAdd(_PyObject_CAST(op), n)
 
-extern void _Py_SetImmortal(PyObject *op);
+static inline void _Py_SetImmortal(PyObject *op)
+{
+    if (op) {
+#ifdef Py_GIL_DISABLED
+        op->ob_tid = _Py_UNOWNED_TID;
+        op->ob_ref_local = _Py_IMMORTAL_REFCNT_LOCAL;
+        op->ob_ref_shared = 0;
+#else
+        op->ob_refcnt = _Py_IMMORTAL_REFCNT;
+#endif
+    }
+}
+#define _Py_SetImmortal(op) _Py_SetImmortal(_PyObject_CAST(op))
 
 // Makes an immortal object mortal again with the specified refcnt. Should only
 // be used during runtime finalization.
@@ -313,12 +325,11 @@ static inline void _PyObject_GC_TRACK(
                           filename, lineno, __func__);
 
     PyInterpreterState *interp = _PyInterpreterState_GET();
-    PyGC_Head *generation0 = &interp->gc.young.head;
+    PyGC_Head *generation0 = interp->gc.generation0;
     PyGC_Head *last = (PyGC_Head*)(generation0->_gc_prev);
     _PyGCHead_SET_NEXT(last, gc);
     _PyGCHead_SET_PREV(gc, last);
     _PyGCHead_SET_NEXT(gc, generation0);
-    assert((gc->_gc_next & _PyGC_NEXT_MASK_OLD_SPACE_1) == 0);
     generation0->_gc_prev = (uintptr_t)gc;
 #endif
 }

Include/internal/pycore_runtime_init.h

@@ -162,12 +162,12 @@ extern PyTypeObject _PyExc_MemoryError;
         }, \
         .gc = { \
             .enabled = 1, \
-            .young = { .threshold = 2000, }, \
+            .generations = { \
-            .old = { \
+                /* .head is set in _PyGC_InitState(). */ \
+                { .threshold = 700, }, \
+                { .threshold = 10, }, \
                 { .threshold = 10, }, \
-                { .threshold = 0, }, \
             }, \
-            .work_to_do = -5000, \
         }, \
         .object_state = _py_object_state_INIT(INTERP), \
         .dtoa = _dtoa_state_INIT(&(INTERP)), \

Lib/test/test_gc.py

@@ -383,11 +383,19 @@ class GCTests(unittest.TestCase):
         # each call to collect(N)
         x = []
         gc.collect(0)
-        # x is now in the old gen
+        # x is now in gen 1
         a, b, c = gc.get_count()
-        # We don't check a since its exact values depends on
+        gc.collect(1)
+        # x is now in gen 2
+        d, e, f = gc.get_count()
+        gc.collect(2)
+        # x is now in gen 3
+        g, h, i = gc.get_count()
+        # We don't check a, d, g since their exact values depends on
         # internal implementation details of the interpreter.
         self.assertEqual((b, c), (1, 0))
+        self.assertEqual((e, f), (0, 1))
+        self.assertEqual((h, i), (0, 0))
 
     def test_trashcan(self):
         class Ouch:
@@ -838,6 +846,16 @@ class GCTests(unittest.TestCase):
         self.assertFalse(
                 any(l is element for element in gc.get_objects(generation=2))
         )
+        gc.collect(generation=1)
+        self.assertFalse(
+                any(l is element for element in gc.get_objects(generation=0))
+        )
+        self.assertFalse(
+                any(l is element for element in  gc.get_objects(generation=1))
+        )
+        self.assertTrue(
+                any(l is element for element in gc.get_objects(generation=2))
+        )
         gc.collect(generation=2)
         self.assertFalse(
                 any(l is element for element in gc.get_objects(generation=0))

Misc/NEWS.d/next/Core and Builtins/2024-01-07-04-22-51.gh-issue-108362.oB9Gcf.rst

@@ -1,13 +0,0 @@
-Implements an incremental cyclic garbage collector. By collecting the old
-generation in increments, there is no need for a full heap scan. This can
-hugely reduce maximum pause time for programs with large heaps.
-
-Reduces the number of generations from three to two. The old generation is
-split into two spaces, "aging" and "collecting".
-
-Collection happens in two steps:: * First, the young generation is scanned
-and the survivors moved to the end of the aging space. * Then objects are
-taken from the collecting space,    at such a rate that all cycles are
-collected eventually.    Those objects are then scanned and the survivors
-moved to    the end of the aging space.    When the collecting space becomes
-empty, the two spaces are swapped.

Modules/gcmodule.c

@@ -158,12 +158,17 @@ gc_set_threshold_impl(PyObject *module, int threshold0, int group_right_1,
 {
     GCState *gcstate = get_gc_state();
 
-    gcstate->young.threshold = threshold0;
+    gcstate->generations[0].threshold = threshold0;
     if (group_right_1) {
-        gcstate->old[0].threshold = threshold1;
+        gcstate->generations[1].threshold = threshold1;
     }
     if (group_right_2) {
-        gcstate->old[1].threshold = threshold2;
+        gcstate->generations[2].threshold = threshold2;
+
+        /* generations higher than 2 get the same threshold */
+        for (int i = 3; i < NUM_GENERATIONS; i++) {
+            gcstate->generations[i].threshold = gcstate->generations[2].threshold;
+        }
     }
     Py_RETURN_NONE;
 }
@@ -180,9 +185,9 @@ gc_get_threshold_impl(PyObject *module)
 {
     GCState *gcstate = get_gc_state();
     return Py_BuildValue("(iii)",
-                         gcstate->young.threshold,
+                         gcstate->generations[0].threshold,
-                         gcstate->old[0].threshold,
+                         gcstate->generations[1].threshold,
-                         0);
+                         gcstate->generations[2].threshold);
 }
 
 /*[clinic input]
@@ -197,9 +202,9 @@ gc_get_count_impl(PyObject *module)
 {
     GCState *gcstate = get_gc_state();
     return Py_BuildValue("(iii)",
-                         gcstate->young.count,
+                         gcstate->generations[0].count,
-                         gcstate->old[gcstate->visited_space].count,
+                         gcstate->generations[1].count,
-                         gcstate->old[gcstate->visited_space^1].count);
+                         gcstate->generations[2].count);
 }
 
 /*[clinic input]

Objects/object.c

@@ -2387,21 +2387,6 @@ _Py_NewReferenceNoTotal(PyObject *op)
     new_reference(op);
 }
 
-void
-_Py_SetImmortal(PyObject *op)
-{
-    if (PyObject_IS_GC(op) && _PyObject_GC_IS_TRACKED(op)) {
-        _PyObject_GC_UNTRACK(op);
-    }
-#ifdef Py_GIL_DISABLED
-    op->ob_tid = _Py_UNOWNED_TID;
-    op->ob_ref_local = _Py_IMMORTAL_REFCNT_LOCAL;
-    op->ob_ref_shared = 0;
-#else
-    op->ob_refcnt = _Py_IMMORTAL_REFCNT;
-#endif
-}
-
 void
 _Py_ResurrectReference(PyObject *op)
 {

Objects/structseq.c

@@ -603,9 +603,6 @@ _PyStructSequence_InitBuiltinWithFlags(PyInterpreterState *interp,
                                        PyStructSequence_Desc *desc,
                                        unsigned long tp_flags)
 {
-    if (Py_TYPE(type) == NULL) {
-        Py_SET_TYPE(type, &PyType_Type);
-    }
     Py_ssize_t n_unnamed_members;
     Py_ssize_t n_members = count_members(desc, &n_unnamed_members);
     PyMemberDef *members = NULL;
@@ -621,7 +618,7 @@ _PyStructSequence_InitBuiltinWithFlags(PyInterpreterState *interp,
         }
         initialize_static_fields(type, desc, members, tp_flags);
 
-        _Py_SetImmortal((PyObject *)type);
+        _Py_SetImmortal(type);
     }
 #ifndef NDEBUG
     else {

Python/gc_free_threading.c

@@ -616,7 +616,7 @@ void
 _PyGC_InitState(GCState *gcstate)
 {
     // TODO: move to pycore_runtime_init.h once the incremental GC lands.
-    gcstate->young.threshold = 2000;
+    gcstate->generations[0].threshold = 2000;
 }
 
 
@@ -911,8 +911,8 @@ cleanup_worklist(struct worklist *worklist)
 static bool
 gc_should_collect(GCState *gcstate)
 {
-    int count = _Py_atomic_load_int_relaxed(&gcstate->young.count);
+    int count = _Py_atomic_load_int_relaxed(&gcstate->generations[0].count);
-    int threshold = gcstate->young.threshold;
+    int threshold = gcstate->generations[0].threshold;
     if (count <= threshold || threshold == 0 || !gcstate->enabled) {
         return false;
     }
@@ -920,7 +920,7 @@ gc_should_collect(GCState *gcstate)
     // objects. A few tests rely on immediate scheduling of the GC so we ignore
     // the scaled threshold if generations[1].threshold is set to zero.
     return (count > gcstate->long_lived_total / 4 ||
-            gcstate->old[0].threshold == 0);
+            gcstate->generations[1].threshold == 0);
 }
 
 static void
@@ -1031,15 +1031,10 @@ gc_collect_main(PyThreadState *tstate, int generation, _PyGC_Reason reason)
 
     /* update collection and allocation counters */
     if (generation+1 < NUM_GENERATIONS) {
-        gcstate->old[generation].count += 1;
+        gcstate->generations[generation+1].count += 1;
     }
     for (i = 0; i <= generation; i++) {
-        if (i == 0) {
+        gcstate->generations[i].count = 0;
-            gcstate->young.count = 0;
-        }
-        else {
-            gcstate->old[i-1].count = 0;
-        }
     }
 
     PyInterpreterState *interp = tstate->interp;
@@ -1362,7 +1357,7 @@ _PyGC_Collect(PyThreadState *tstate, int generation, _PyGC_Reason reason)
     return gc_collect_main(tstate, generation, reason);
 }
 
-void
+Py_ssize_t
 _PyGC_CollectNoFail(PyThreadState *tstate)
 {
     /* Ideally, this function is only called on interpreter shutdown,
@@ -1371,7 +1366,7 @@ _PyGC_CollectNoFail(PyThreadState *tstate)
        during interpreter shutdown (and then never finish it).
        See http://bugs.python.org/issue8713#msg195178 for an example.
        */
-    gc_collect_main(tstate, NUM_GENERATIONS - 1, _Py_GC_REASON_SHUTDOWN);
+    return gc_collect_main(tstate, NUM_GENERATIONS - 1, _Py_GC_REASON_SHUTDOWN);
 }
 
 void
@@ -1495,7 +1490,7 @@ _PyObject_GC_Link(PyObject *op)
 {
     PyThreadState *tstate = _PyThreadState_GET();
     GCState *gcstate = &tstate->interp->gc;
-    gcstate->young.count++;
+    gcstate->generations[0].count++;
 
     if (gc_should_collect(gcstate) &&
         !_Py_atomic_load_int_relaxed(&gcstate->collecting))
@@ -1610,8 +1605,8 @@ PyObject_GC_Del(void *op)
 #endif
     }
     GCState *gcstate = get_gc_state();
-    if (gcstate->young.count > 0) {
+    if (gcstate->generations[0].count > 0) {
-        gcstate->young.count--;
+        gcstate->generations[0].count--;
     }
     PyObject_Free(((char *)op)-presize);
 }

Python/import.c

@@ -1030,7 +1030,7 @@ _extensions_cache_set(PyObject *filename, PyObject *name, PyModuleDef *def)
     if (!already_set) {
         /* We assume that all module defs are statically allocated
            and will never be freed.  Otherwise, we would incref here. */
-        _Py_SetImmortal((PyObject *)def);
+        _Py_SetImmortal(def);
     }
     res = 0;
 

Tools/gdb/libpython.py

@@ -1753,11 +1753,8 @@ class Frame(object):
             return (name == 'take_gil')
 
     def is_gc_collect(self):
-        '''Is this frame a collector within the garbage-collector?'''
+        '''Is this frame gc_collect_main() within the garbage-collector?'''
-        return self._gdbframe.name() in (
+        return self._gdbframe.name() in ('collect', 'gc_collect_main')
-            'collect', 'gc_collect_full', 'gc_collect_main',
-            'gc_collect_young', 'gc_collect_increment'
-        )
 
     def get_pyop(self):
         try: