gh-112529: Make the GC scheduling thread-safe (#114880)

The GC keeps track of the number of allocations (less deallocations)
since the last GC. This buffers the count in thread-local state and uses
atomic operations to modify the per-interpreter count. The thread-local
buffering avoids contention on shared state.

A consequence is that the GC scheduling is not as precise, so
"test_sneaky_frame_object" is skipped because it requires that the GC be
run exactly after allocating a frame object.

2 files changed, 8 insertions, 0 deletions

diff --git a/Include/internal/pycore_gc.h b/Include/internal/pycore_gc.h
index 582a16b..a98864f 100644
--- a/Include/internal/pycore_gc.h
+++ b/Include/internal/pycore_gc.h
@@ -260,6 +260,13 @@ struct _gc_runtime_state {
     Py_ssize_t long_lived_pending;
 };
 
+#ifdef Py_GIL_DISABLED
+struct _gc_thread_state {
+    /* Thread-local allocation count. */
+    Py_ssize_t alloc_count;
+};
+#endif
+
 
 extern void _PyGC_InitState(struct _gc_runtime_state *);
 
diff --git a/Include/internal/pycore_tstate.h b/Include/internal/pycore_tstate.h
index 97aa85a..7fb9ab2 100644
--- a/Include/internal/pycore_tstate.h
+++ b/Include/internal/pycore_tstate.h
@@ -28,6 +28,7 @@ typedef struct _PyThreadStateImpl {
     PyThreadState base;
 
 #ifdef Py_GIL_DISABLED
+    struct _gc_thread_state gc;
     struct _mimalloc_thread_state mimalloc;
     struct _Py_object_freelists freelists;
     struct _brc_thread_state brc;