diff --git a/include/gc/gc.h b/include/gc/gc.h
index ad2f054c4..847700302 100644
--- a/include/gc/gc.h
+++ b/include/gc/gc.h
@@ -205,6 +205,10 @@ GC_API GC_ATTR_DEPRECATED int GC_finalize_on_demand;
 GC_API void GC_CALL GC_set_finalize_on_demand(int);
 GC_API int GC_CALL GC_get_finalize_on_demand(void);
 
+/* Returns the total number of finalizers that have been run so far     */
+/* by the collector.                                                    */
+GC_API size_t GC_CALL GC_finalized_total(void);
+
 /* Mark objects reachable from finalizable objects in a separate        */
 /* post-pass.  This makes it a bit safer to use                         */
 /* non-topologically-ordered finalization.  Default value is determined */
diff --git a/include/private/gc_priv.h b/include/private/gc_priv.h
index c82cf223a..e2c0f2f1e 100644
--- a/include/private/gc_priv.h
+++ b/include/private/gc_priv.h
@@ -380,11 +380,12 @@ typedef struct hblkhdr hdr;
 EXTERN_C_BEGIN
 
 #ifndef GC_NO_FINALIZATION
+  GC_INNER void GC_maybe_wake_finalizer_thread(void);
   /* If GC_finalize_on_demand is not set, invoke eligible           */
   /* finalizers.  Otherwise: call (*GC_finalizer_notifier)() if     */
   /* there are finalizers to be run, and we have not called this    */
   /* procedure yet this collection cycle.                           */
-# define GC_INVOKE_FINALIZERS() GC_notify_or_invoke_finalizers()
+# define GC_INVOKE_FINALIZERS() GC_maybe_wake_finalizer_thread()
   GC_INNER void GC_notify_or_invoke_finalizers(void);
 
   /* Perform all indicated finalization actions on unmarked         */
@@ -1368,7 +1369,8 @@ struct _GC_arrays {
   /* running.  Used to approximate size of memory explicitly            */
   /* deallocated by finalizers.                                         */
   word _finalizer_bytes_freed;
-
+  /* Number of finalizers that have been run so far.      */
+  size_t _finalizers_run;
   /* Pointer to the first (lowest address) bottom_index; assumes the    */
   /* allocator lock is held.                                            */
   bottom_index *_all_bottom_indices;
@@ -1676,6 +1678,7 @@ GC_API_PRIV GC_FAR struct _GC_arrays GC_arrays;
 #define GC_composite_in_use GC_arrays._composite_in_use
 #define GC_excl_table GC_arrays._excl_table
 #define GC_finalizer_bytes_freed GC_arrays._finalizer_bytes_freed
+#define GC_finalizers_run GC_arrays._finalizers_run
 #define GC_heapsize GC_arrays._heapsize
 #define GC_large_allocd_bytes GC_arrays._large_allocd_bytes
 #define GC_large_free_bytes GC_arrays._large_free_bytes
diff --git a/reclaim.c b/reclaim.c
index 821c7c435..b5e55d09e 100644
--- a/reclaim.c
+++ b/reclaim.c
@@ -25,6 +25,14 @@
 /* originally on free lists which we had to drop.                       */
 GC_INNER signed_word GC_bytes_found = 0;
 
+#ifdef GC_PTHREADS
+static pthread_mutex_t flzr_mtx = PTHREAD_MUTEX_INITIALIZER;
+static pthread_cond_t flzr_t_has_work = PTHREAD_COND_INITIALIZER;
+static int flzr_can_work = 0;
+#elif
+#error "This fork of BDWGC only supports POSIX threads"
+#endif
+
 #if defined(PARALLEL_MARK)
   /* Number of threads currently building free lists without holding    */
   /* the allocator lock.  It is not safe to collect if this is nonzero. */
@@ -53,6 +61,8 @@ STATIC unsigned GC_n_leaked = 0;
   STATIC void GC_reclaim_unconditionally_marked(void);
 #endif
 
+STATIC unsigned GC_finalizer_thread_exists = 0;
+
 GC_INLINE void GC_add_leaked(ptr_t leaked)
 {
     GC_ASSERT(I_HOLD_LOCK());
@@ -880,3 +890,40 @@ GC_API void GC_CALL GC_enumerate_reachable_objects_inner(
   ed.client_data = client_data;
   GC_apply_to_all_blocks(GC_do_enumerate_reachable_objects, &ed);
 }
+
+static void* init_finalize_thread(void *arg)
+{
+  while(1) {
+    pthread_mutex_lock(&flzr_mtx);
+    while (flzr_can_work == 0) {
+      pthread_cond_wait(&flzr_t_has_work, &flzr_mtx);
+    }
+    flzr_can_work = 0;
+    pthread_mutex_unlock(&flzr_mtx);
+    GC_finalizers_run += GC_invoke_finalizers();
+  }
+  return arg;
+}
+
+GC_INNER void GC_maybe_wake_finalizer_thread()
+{
+  if (!GC_finalizer_thread_exists) {
+    pthread_t t;
+    pthread_create(&t, NULL, init_finalize_thread, NULL /* arg */);
+    GC_finalizer_thread_exists = 1;
+    return;
+  }
+
+  if (GC_should_invoke_finalizers() == 0)
+    return;
+
+  pthread_mutex_lock(&flzr_mtx);
+  flzr_can_work = 1;
+  pthread_cond_signal(&flzr_t_has_work);
+  pthread_mutex_unlock(&flzr_mtx);
+}
+
+GC_API size_t GC_finalized_total()
+{
+  return GC_finalizers_run;
+}