Skip to content

Commit 9fd1624

Browse files
byrootbyroot
committed
Lazily create objspace->id_to_obj_tbl
This inverse table is only useful if `ObjectSpace._id2ref` is used, which is extremely rare. The only notable exception is the `drb` gem and even then it has an option not to rely on `_id2ref`. So if we assume this table will never be looked up, we can just not maintain it, and if it turns out `_id2ref` is called, we can lock the VM and re-build it. ``` compare-ruby: ruby 3.5.0dev (2025-04-10T09:44:40Z master 684cfa4) +YJIT +PRISM [arm64-darwin24] built-ruby: ruby 3.5.0dev (2025-04-10T10:13:43Z lazy-id-to-obj d3aa962) +YJIT +PRISM [arm64-darwin24] warming up.. | |compare-ruby|built-ruby| |:----------|-----------:|---------:| |baseline | 26.364M| 25.974M| | | 1.01x| -| |object_id | 10.293M| 14.202M| | | -| 1.38x| ```
1 parent 3628e9e commit 9fd1624

5 files changed

Lines changed: 74 additions & 20 deletions

File tree

‎benchmark/object_id.yml‎

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,4 @@
1+
benchmark:
2+
baseline: "Object.new"
3+
object_id: "Object.new.object_id"
4+
# loop_count: 100000

‎gc.c‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1794,11 +1794,11 @@ id2ref(VALUE objid)
17941794
return ptr;
17951795
}
17961796
else {
1797-
rb_raise(rb_eRangeError, "%p is not symbol id value", (void *)ptr);
1797+
rb_raise(rb_eRangeError, "%p is not a symbol id value", (void *)ptr);
17981798
}
17991799
}
18001800

1801-
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is not id value", rb_int2str(objid, 10));
1801+
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is not an id value", rb_int2str(objid, 10));
18021802
}
18031803
}
18041804

@@ -1807,7 +1807,7 @@ id2ref(VALUE objid)
18071807
return obj;
18081808
}
18091809
else {
1810-
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is id of the unshareable object on multi-ractor", rb_int2str(objid, 10));
1810+
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is the id of an unshareable object on multi-ractor", rb_int2str(objid, 10));
18111811
}
18121812
}
18131813

‎gc/default/default.c‎

Lines changed: 36 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -1585,30 +1585,50 @@ rb_gc_impl_object_id(void *objspace_ptr, VALUE obj)
15851585
objspace->next_object_id += OBJ_ID_INCREMENT;
15861586

15871587
st_insert(objspace->obj_to_id_tbl, (st_data_t)obj, (st_data_t)id);
1588-
st_insert(objspace->id_to_obj_tbl, (st_data_t)id, (st_data_t)obj);
1588+
if (RB_UNLIKELY(objspace->id_to_obj_tbl)) {
1589+
st_insert(objspace->id_to_obj_tbl, (st_data_t)id, (st_data_t)obj);
1590+
}
15891591
FL_SET(obj, FL_SEEN_OBJ_ID);
15901592
}
15911593
rb_gc_vm_unlock(lev);
15921594

15931595
return id;
15941596
}
15951597

1598+
static int
1599+
build_id_to_obj_i(st_data_t key, st_data_t value, st_data_t data)
1600+
{
1601+
st_table *id_to_obj_tbl = (st_table *)data;
1602+
st_insert(id_to_obj_tbl, value, key);
1603+
return ST_CONTINUE;
1604+
}
1605+
15961606
VALUE
15971607
rb_gc_impl_object_id_to_ref(void *objspace_ptr, VALUE object_id)
15981608
{
15991609
rb_objspace_t *objspace = objspace_ptr;
16001610

1611+
unsigned int lev = rb_gc_vm_lock();
1612+
1613+
if (!objspace->id_to_obj_tbl) {
1614+
objspace->id_to_obj_tbl = st_init_table_with_size(&object_id_hash_type, st_table_size(objspace->obj_to_id_tbl));
1615+
st_foreach(objspace->obj_to_id_tbl, build_id_to_obj_i, (st_data_t)objspace->id_to_obj_tbl);
1616+
}
1617+
16011618
VALUE obj;
1602-
if (st_lookup(objspace->id_to_obj_tbl, object_id, &obj) &&
1603-
!rb_gc_impl_garbage_object_p(objspace, obj)) {
1619+
bool found = st_lookup(objspace->id_to_obj_tbl, object_id, &obj) && !rb_gc_impl_garbage_object_p(objspace, obj);
1620+
1621+
rb_gc_vm_unlock(lev);
1622+
1623+
if (found) {
16041624
return obj;
16051625
}
16061626

16071627
if (rb_funcall(object_id, rb_intern(">="), 1, ULL2NUM(objspace->next_object_id))) {
1608-
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is not id value", rb_funcall(object_id, rb_intern("to_s"), 1, INT2FIX(10)));
1628+
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is not an id value", rb_funcall(object_id, rb_intern("to_s"), 1, INT2FIX(10)));
16091629
}
16101630
else {
1611-
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is recycled object", rb_funcall(object_id, rb_intern("to_s"), 1, INT2FIX(10)));
1631+
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is a recycled object", rb_funcall(object_id, rb_intern("to_s"), 1, INT2FIX(10)));
16121632
}
16131633
}
16141634

@@ -2644,7 +2664,9 @@ obj_free_object_id(rb_objspace_t *objspace, VALUE obj)
26442664

26452665
if (st_delete(objspace->obj_to_id_tbl, &o, &id)) {
26462666
GC_ASSERT(id);
2647-
st_delete(objspace->id_to_obj_tbl, &id, NULL);
2667+
if (RB_UNLIKELY(objspace->id_to_obj_tbl)) {
2668+
st_delete(objspace->id_to_obj_tbl, &id, NULL);
2669+
}
26482670
}
26492671
else {
26502672
rb_bug("Object ID seen, but not in mapping table: %s", rb_obj_info(obj));
@@ -7168,7 +7190,9 @@ gc_update_references(rb_objspace_t *objspace)
71687190
}
71697191
}
71707192
gc_ref_update_table_values_only(objspace->obj_to_id_tbl);
7171-
gc_update_table_refs(objspace->id_to_obj_tbl);
7193+
if (RB_UNLIKELY(objspace->id_to_obj_tbl)) {
7194+
gc_update_table_refs(objspace->id_to_obj_tbl);
7195+
}
71727196
gc_update_table_refs(finalizer_table);
71737197

71747198
rb_gc_update_vm_references((void *)objspace);
@@ -9282,7 +9306,10 @@ rb_gc_impl_objspace_free(void *objspace_ptr)
92829306
heap->total_slots = 0;
92839307
}
92849308

9285-
st_free_table(objspace->id_to_obj_tbl);
9309+
9310+
if (objspace->id_to_obj_tbl) {
9311+
st_free_table(objspace->id_to_obj_tbl);
9312+
}
92869313
st_free_table(objspace->obj_to_id_tbl);
92879314

92889315
free_stack_chunks(&objspace->mark_stack);
@@ -9423,7 +9450,7 @@ rb_gc_impl_objspace_init(void *objspace_ptr)
94239450
heap_page_alloc_use_mmap = INIT_HEAP_PAGE_ALLOC_USE_MMAP;
94249451
#endif
94259452
objspace->next_object_id = OBJ_ID_INITIAL;
9426-
objspace->id_to_obj_tbl = st_init_table(&object_id_hash_type);
9453+
objspace->id_to_obj_tbl = NULL;
94279454
objspace->obj_to_id_tbl = st_init_numtable();
94289455
#if RGENGC_ESTIMATE_OLDMALLOC
94299456
objspace->rgengc.oldmalloc_increase_limit = gc_params.oldmalloc_limit_min;

‎gc/mmtk/mmtk.c‎

Lines changed: 30 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -366,7 +366,9 @@ rb_mmtk_update_obj_id_tables(void)
366366
struct objspace *objspace = rb_gc_get_objspace();
367367

368368
st_foreach(objspace->obj_to_id_tbl, rb_mmtk_update_obj_id_tables_obj_to_id_i, 0);
369-
st_foreach(objspace->id_to_obj_tbl, rb_mmtk_update_obj_id_tables_id_to_obj_i, 0);
369+
if (objspace->id_to_obj_tbl) {
370+
st_foreach(objspace->id_to_obj_tbl, rb_mmtk_update_obj_id_tables_id_to_obj_i, 0);
371+
}
370372
}
371373

372374
static int
@@ -1091,7 +1093,7 @@ static const struct st_hash_type object_id_hash_type = {
10911093
static void
10921094
objspace_obj_id_init(struct objspace *objspace)
10931095
{
1094-
objspace->id_to_obj_tbl = st_init_table(&object_id_hash_type);
1096+
objspace->id_to_obj_tbl = NULL;
10951097
objspace->obj_to_id_tbl = st_init_numtable();
10961098
objspace->next_object_id = OBJ_ID_INITIAL;
10971099
}
@@ -1119,30 +1121,51 @@ rb_gc_impl_object_id(void *objspace_ptr, VALUE obj)
11191121
objspace->next_object_id += OBJ_ID_INCREMENT;
11201122

11211123
st_insert(objspace->obj_to_id_tbl, (st_data_t)obj, (st_data_t)id);
1122-
st_insert(objspace->id_to_obj_tbl, (st_data_t)id, (st_data_t)obj);
1124+
if (RB_UNLIKELY(objspace->id_to_obj_tbl)) {
1125+
st_insert(objspace->id_to_obj_tbl, (st_data_t)id, (st_data_t)obj);
1126+
}
11231127
FL_SET(obj, FL_SEEN_OBJ_ID);
11241128
}
11251129
rb_gc_vm_unlock(lev);
11261130

11271131
return id;
11281132
}
11291133

1134+
static int
1135+
build_id_to_obj_i(st_data_t key, st_data_t value, st_data_t data)
1136+
{
1137+
st_table *id_to_obj_tbl = (st_table *)data;
1138+
st_insert(id_to_obj_tbl, value, key);
1139+
return ST_CONTINUE;
1140+
}
1141+
11301142
VALUE
11311143
rb_gc_impl_object_id_to_ref(void *objspace_ptr, VALUE object_id)
11321144
{
11331145
struct objspace *objspace = objspace_ptr;
11341146

1147+
1148+
unsigned int lev = rb_gc_vm_lock();
1149+
1150+
if (!objspace->id_to_obj_tbl) {
1151+
objspace->id_to_obj_tbl = st_init_table_with_size(&object_id_hash_type, st_table_size(objspace->obj_to_id_tbl));
1152+
st_foreach(objspace->obj_to_id_tbl, build_id_to_obj_i, (st_data_t)objspace->id_to_obj_tbl);
1153+
}
1154+
11351155
VALUE obj;
1136-
if (st_lookup(objspace->id_to_obj_tbl, object_id, &obj) &&
1137-
!rb_gc_impl_garbage_object_p(objspace, obj)) {
1156+
bool found = st_lookup(objspace->id_to_obj_tbl, object_id, &obj) && !rb_gc_impl_garbage_object_p(objspace, obj);
1157+
1158+
rb_gc_vm_unlock(lev);
1159+
1160+
if (found) {
11381161
return obj;
11391162
}
11401163

11411164
if (rb_funcall(object_id, rb_intern(">="), 1, ULL2NUM(objspace->next_object_id))) {
1142-
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is not id value", rb_funcall(object_id, rb_intern("to_s"), 1, INT2FIX(10)));
1165+
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is not an id value", rb_funcall(object_id, rb_intern("to_s"), 1, INT2FIX(10)));
11431166
}
11441167
else {
1145-
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is recycled object", rb_funcall(object_id, rb_intern("to_s"), 1, INT2FIX(10)));
1168+
rb_raise(rb_eRangeError, "%+"PRIsVALUE" is a recycled object", rb_funcall(object_id, rb_intern("to_s"), 1, INT2FIX(10)));
11461169
}
11471170
}
11481171

‎test/ruby/test_objectspace.rb‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -69,7 +69,7 @@ def test_id2ref_invalid_symbol_id
6969
# RB_STATIC_SYM_P checks for static symbols by checking that the bottom
7070
# 8 bits of the object is equal to RUBY_SYMBOL_FLAG, so we need to make
7171
# sure that the bottom 8 bits remain unchanged.
72-
msg = /is not symbol id value/
72+
msg = /is not a symbol id value/
7373
assert_raise_with_message(RangeError, msg) { ObjectSpace._id2ref(:a.object_id + 256) }
7474
end
7575

0 commit comments

Comments
 (0)