before marking. This fixes the following test failure
introduced in r51126:
make test-all TESTOPTS='--gc-stress ruby/test_refinement.rb'
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51591 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(such as ST_CONTINUE) for enum rb_id_table_iterator_result.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51544 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
[Feature #11420]
This table only manage ID->VALUE table to reduce overhead of st.
Some functions prefixed rb_id_table_* are provided.
* id_table.c: implement rb_id_table_*.
There are several algorithms to implement it.
Now, there are roughly 4 types:
* st
* array
* hash (implemented by Yura Sokolov)
* mix of array and hash
The macro ID_TABLE_IMPL can choose implementation.
You can see detailes about them at the head of id_table.c.
At the default, I choose 34 (mix of list and hash).
This is not final decision.
Please report your suitable parameters or
your data structure.
* symbol.c: introduce rb_id_serial_t and rb_id_to_serial()
to represent ID by serial number.
* internal.h: use id_table for method tables.
* class.c, gc.c, marshal.c, vm.c, vm_method.c: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51541 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* include/ruby/ruby.h (ALLOCV_N): check integer overflow, as well
as ruby_xmalloc2. pointed out by Paul <pawlkt AT gmail.com>.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51533 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
All contents of previous rb_iseq_t is in rb_iseq_t::body.
Remove rb_iseq_t::self because rb_iseq_t is an object.
RubyVM::InstructionSequence is wrapper object points T_IMEMO/iseq.
So RubyVM::ISeq.of(something) method returns different wrapper
objects but they point the same T_IMEMO/iseq object.
This patch is big, but most of difference is replacement of
iseq->xxx to iseq->body->xxx.
(previous) rb_iseq_t::compile_data is also located to
rb_iseq_t::compile_data.
It was moved from rb_iseq_body::compile_data.
Now rb_iseq_t has empty two pointers.
I will split rb_iseq_body data into static data and dynamic data.
* compile.c: rename some functions/macros.
Now, we don't need to separate iseq and iseqval (only VALUE).
* eval.c (ruby_exec_internal): `n' is rb_iseq_t (T_IMEMO/iseq).
* ext/objspace/objspace.c (count_imemo_objects): count T_IMEMO/iseq.
* gc.c: check T_IMEMO/iseq.
* internal.h: add imemo_type::imemo_iseq.
* iseq.c: define RubyVM::InstructionSequnce as T_OBJECT.
Methods are implemented by functions named iseqw_....
* load.c (rb_load_internal0): rb_iseq_new_top() returns
rb_iseq_t (T_IMEMO/iesq).
* method.h (rb_add_method_iseq): accept rb_iseq_t (T_IMEMO/iseq).
* vm_core.h (GetISeqPtr): removed because it is not T_DATA now.
* vm_core.h (struct rb_iseq_body): remove padding for
[Bug #10037][ruby-core:63721].
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51327 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
ObjectSpace.each_object. [Bug #11360]
* test/ruby/test_objectspace.rb: add a test about it.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51320 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (run_finalizer): set and restore safe level here to reduce
nested EXEC_TAGs.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51291 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This may cause compatibility problems for some users, but maybe
nobody tracked it closely in the past because it is disabled by
default.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51151 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
rb_control_frame_t::klass.
[Bug #11278], [Bug #11279]
rb_method_entry_t data belong to modules/classes.
rb_method_entry_t::owner points defined module or class.
module M
def foo; end
end
In this case, owner is M.
rb_callable_method_entry_t data belong to only classes.
For modules, MRI creates corresponding T_ICLASS internally.
rb_callable_method_entry_t can also belong to T_ICLASS.
rb_callable_method_entry_t::defined_class points T_CLASS or
T_ICLASS.
rb_method_entry_t data for classes (not for modules) are also
rb_callable_method_entry_t data because it is completely same data.
In this case, rb_method_entry_t::owner == rb_method_entry_t::defined_class.
For example, there are classes C and D, and incldues M,
class C; include M; end
class D; include M; end
then, two T_ICLASS objects for C's super class and D's super class
will be created.
When C.new.foo is called, then M#foo is searcheed and
rb_callable_method_t data is used by VM to invoke M#foo.
rb_method_entry_t data is only one for M#foo.
However, rb_callable_method_entry_t data are two (and can be more).
It is proportional to the number of including (and prepending)
classes (the number of T_ICLASS which point to the module).
Now, created rb_callable_method_entry_t are collected when
the original module M was modified. We can think it is a cache.
We need to select what kind of method entry data is needed.
To operate definition, then you need to use rb_method_entry_t.
You can access them by the following functions.
* rb_method_entry(VALUE klass, ID id);
* rb_method_entry_with_refinements(VALUE klass, ID id);
* rb_method_entry_without_refinements(VALUE klass, ID id);
* rb_resolve_refined_method(VALUE refinements, const rb_method_entry_t *me);
To invoke methods, then you need to use rb_callable_method_entry_t
which you can get by the following APIs corresponding to the
above listed functions.
* rb_callable_method_entry(VALUE klass, ID id);
* rb_callable_method_entry_with_refinements(VALUE klass, ID id);
* rb_callable_method_entry_without_refinements(VALUE klass, ID id);
* rb_resolve_refined_method_callable(VALUE refinements, const rb_callable_method_entry_t *me);
VM pushes rb_callable_method_entry_t, so that rb_vm_frame_method_entry()
returns rb_callable_method_entry_t.
You can check a super class of current method by
rb_callable_method_entry_t::defined_class.
* method.h: renamed from rb_method_entry_t::klass to
rb_method_entry_t::owner.
* internal.h: add rb_classext_struct::callable_m_tbl to cache
rb_callable_method_entry_t data.
We need to consider abotu this field again because it is only
active for T_ICLASS.
* class.c (method_entry_i): ditto.
* class.c (rb_define_attr): rb_method_entry() does not takes
defiend_class_ptr.
* gc.c (mark_method_entry): mark RCLASS_CALLABLE_M_TBL() for T_ICLASS.
* cont.c (fiber_init): rb_control_frame_t::klass is removed.
* proc.c: fix `struct METHOD' data structure because
rb_callable_method_t has all information.
* vm_core.h: remove several fields.
* rb_control_frame_t::klass.
* rb_block_t::klass.
And catch up changes.
* eval.c: catch up changes.
* gc.c: ditto.
* insns.def: ditto.
* vm.c: ditto.
* vm_args.c: ditto.
* vm_backtrace.c: ditto.
* vm_dump.c: ditto.
* vm_eval.c: ditto.
* vm_insnhelper.c: ditto.
* vm_method.c: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51126 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Fill internal object information into passed buffer.
* gc.h: declare rb_raw_obj_info().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51107 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
a rb_method_entry_t data (me).
Normally, `me' points `def'. Some Ruby objects pointed from `def'
and objects are marked by `me' (mark_method_entry() in gc.c).
However, `def' is built before making a `me', then nobody can mark
objects pointed from `def' before making (and pointing from) `me'.
I hope this patch solve #11244.
* vm_method.c: remove `rb_' prefix from some static functions.
* method.h (rb_method_entry_create): constify
* gc.c (mark_method_entry): add checking `def' and
`def->body.iseq.iseqptr' availability because they can be NULL.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51026 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
rich obj_info() output.
At the default, the value of RGENGC_OBJ_INFO is
(RGENGC_DEBUG | RGENGC_CHECK_MODE).
* gc.c (RGENGC_OBJ_INFO): force enable it to debug #11244.
* gc.c (gc_mark_ptr): print more details with obj_info().
* gc.c (gc_mark_children): remove useless debug prints.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51018 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
It should be removed later. But we can remain this check
because it is only a branch.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51010 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* internal.h (rb_fstring_lit): new macro to make a fstring from a
string literal.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51008 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Special constants are all frozen since [Feature #8923] and cannot
support ivars. Remove some unused code we had for supporting them.
* variable.c (special_generic_ivar): remove flag
(givar_i, rb_mark_generic_ivar_tbl): remove functions
(rb_free_generic_ivar, rb_ivar_lookup, rb_ivar_delete,
generic_ivar_set, rb_ivar_set, rb_ivar_defined,
rb_copy_generic_ivar, rb_ivar_foreach, rb_ivar_count,
rb_obj_remove_instance_variable):
adjust for lack of ivar support in special constants
* test/ruby/test_variable.rb: test ivars for special consts
* internal.h: remove rb_mark_generic_ivar_tbl decl
* gc.c (gc_mark_roots): remove rb_mark_generic_ivar_tbl call
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50758 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
While making a r50728, iseqval is needed (to mark correctly),
but now just iseqptr is enough.
* class.c: catch up this fix.
* gc.c: ditto.
* proc.c: ditto.
* vm_method.c: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50731 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Motivation and new data structure are described in [Bug #11203].
This patch also solve the following issues.
* [Bug #11200] Memory leak of method entries
* [Bug #11046] __callee__ returns incorrect method name in orphan
proc
* test/ruby/test_method.rb: add a test for [Bug #11046].
* vm_core.h: remvoe rb_control_frame_t::me. me is located at value
stack.
* vm_core.h, gc.c, vm_method.c: remove unlinked_method... codes
because method entries are simple VALUEs.
* method.h: Now, all method entries has own independent method
definititons. Strictly speaking, this change is not essential,
but for future changes.
* rb_method_entry_t::flag is move to rb_method_definition_t::flag.
* rb_method_definition_t::alias_count is now
rb_method_definition_t::alias_count_ptr, a pointer to the counter.
* vm_core.h, vm_insnhelper.c (rb_vm_frame_method_entry) added to
search the current method entry from value stack.
* vm_insnhelper.c (VM_CHECK_MODE): introduced to enable/disable
assertions.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50728 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_mark_roots): stress_to_class is also a GC root.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50648 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (newobj_of): debug feature to fail allocation of particular
classes.
* gc.c (rb_gcdebug_add_stress_to_class): add classes to the list.
* gc.c (rb_gcdebug_remove_stress_to_class): remove classes from
the list.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50647 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* include/ruby/ruby.h (rb_data_typed_object_alloc),
(rb_data_object_alloc): warn use of old names.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50558 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (rb_data_object_wrap, rb_data_typed_object_wrap): rename
alloc as wrap. these functions do not allocate data pointers
but just wrap the given pointers.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50506 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* include/ruby/ruby.h (Data_Make_Struct, TypedData_Make_Struct):
allocate wrapper data object before allocating DATA_PTR to get
rid of possible memory leak when the former failed.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50464 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_mark_children): call dmark function for non-NULL
pointers only, so that DATA_PTR can be NULL safely now.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50462 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
with regards to `:TOTAL` key, with patch by @schneems [Fixes GH-871]
https://github.com/ruby/ruby/pull/871 [Bug #11067] [ci skip]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50303 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* skip on incremental marking because not sure what happen :p
* rb_gc_writebarrier_remember() is enough to mark children.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50024 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
If RGENGC_OLD_NEWOBJ_CHECK > 0, then create old objects
(not new objects) periodically.
Create one old objects per RGENGC_OLD_NEWOBJ_CHECK objects are
created.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50013 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* not inline on RGENGC_CHECK_MODE > 0.
* check SPECIAL_CONST_P(obj) first.
* add a check that remembered_bit is only TRUE when old (age == 3).
* gc.c (RVALUE_DEMOTE): should clear RVALUE_REMEMBERED bit.
remembered_bit should be TRUE only for old (age == 3) objects.
Actually there are no effect because demoted objects will be
uncollectible WB unprotected objects (marked at the begginig of
every minor GC).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50011 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* long_lived -> uncollectible:
because this bitmap does not mean "long lived objects in past",
but means "prohibit collection thse objects until next major GC".
Uncollectible objects consist of two types objects, one is old
objects (WB protected objects which age == 3) and another is
uncollectible WB unprotected objects which are referred from old
objects
* remembered_wb_unprotected_objects ->
uncollectible_wb_unprotected_objects:
because uncollectible objects does not mean remembered objects.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50009 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
barrier (WB) implementation.
When finding reference from [Old obj] to [New obj] by WB, current
implementation marks [Old obj] as remembered old objects and marks
chilldren of [Old obj] at the beggining of marking.
Added (but disabled) code changes current behaviour. This fix promote
[New obj] to old and marks as a remembered old object. We can assume
"new objects referred from old objects are maybe long-lived old
objects".
Disadvantage of added algorithm is we may promote unwilling
short-lived objects. For example, consider many new objects push and
pop to an old stack object. All of new objects (short-lived objects)
promote to old objects unexpectedly.
To compare these behaviour, I add this new code (but disabled it).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50006 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
For incremental GC, we need to get a pointer to the objspace.
We can share this pointer for the following WB process.
And considering icache hit ratio, prcess in the GC.
* gc.c (rb_gc_writebarrier): added.
* gc.c (gc_writebarrier_generational, gc_writebarrier_incremental):
make them NOINLINE because inlining them into rb_gc_writebarrier()
makes a prologue code of rb_gc_writebarrier() longer (storing callee
save registers).
This patch improve the performance of WB on micro-benchmarks.
name ruby 2.1 trunk modified
vm1_gc_wb_ary* 0.511 0.632 0.532
vm1_gc_wb_ary_promoted* 0.578 0.701 0.674
vm1_gc_wb_obj* 0.419 0.575 0.492
vm1_gc_wb_obj_promoted* 0.537 0.664 0.618
(sec)
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49987 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
memo->v1 and memo->v2 is WB protected values.
So use MEMO_V1/V2_SET() macros to set these values.
memo->u3 is ambiguous (sometimes a VALUE, sometimes an integer
value), so use gc_mark_maybe() in gc.c to mark it.
Rename NEW_MEMO() to MEMO_NEW().
Move MEMO_FOR and NEW_MEMO_FOF macros from node.h.
Export a rb_imemo_new() function for ext/ripper.
* node.h: remove NODE_MEMO.
* enum.c: catch up these change.
* enumerator.c: ditto.
* load.c: ditto.
* ext/objspace/objspace.c (count_nodes): ditto.
* gc.c (gc_mark_children): mark imemo_memo type.
* parse.y (new_args_gen): use T_IMEMO.
(I'm not sure it is working correctly...)
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49942 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Add THROW_DATA_NEW().
* internal.h: move defnition of `struct THROW_DATA'
from vm_insnhelper.h to internal.h.
Rename `THROW_DATA' to `vm_throw_data'.
* eval_intern.h (THROW_DATA_P): move to internal.h.
THROW_DATA is no longer T_NODE, so check T_IMEMO.
* gc.c (gc_mark_children): mark THROW_DATA.
* vm.c: catch up these changes.
* vm_eval.c: ditto.
* vm_insnhelper.c: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49936 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* internal.h, vm_insnhelper.h: move definition `struct SVAR'
from vm_insnhelper.h to internal.h. And rename it to strcut vm_svar.
new imemo_type imemo_svar is added.
* gc.c (gc_mark_children): mark imemo_svar.
* node.c (rb_gc_mark_node): remove useless marking.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49935 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
T_IMEMO is Internal Memo type, internal use only.
T_IMEMO has same purpose of NODE_MEMO.
To insert T_IMEMO, type numbers are modified a little.
* internal.h: define struct RIMemo. Each RIMemo objects
has imemo_type. We can observe it by the imemo_type() function.
* gc.c (rb_imemo_new): added.
* node.h: remove NODE_CREF and NEW_CREF().
* node.c (rb_gc_mark_node): ditto.
* vm.c (vm_cref_new): use rb_imem_new().
* vm_eval.c: ditto.
* vm_eval.c (eval_string_with_cref):
* vm_eval.c (rb_type_str):
* vm_insnhelper.c: use RIMemo objects for CREF.
* ext/objspace/objspace.c: support T_IMEMO.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49932 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
It is easy to reproduce with such script:
module M; def bar; end; end
loop{
Class.new do
def foo; end
prepend M
end
}
* gc.c (obj_free): free T_ICLASS::m_tbl if it is created by prepend.
To recognize it, check RICLASS_IS_ORIGIN flag.
* gc.c (gc_mark_children): T_ICLASS objects only need to mark
T_ICLASS::m_tbl if RICLASS_IS_ORIGIN is set.
* gc.c (obj_memsize_of): count T_ICLASS if RICLASS_IS_ORIGIN is set.
* internal.h (RCLASS_SET_ORIGIN): add to set RCLASS_SET_ORIGIN.
TODO: The word `origin' seems not good name. We need to invent
another good name.
* class.c: use RCLASS_SET_ORIGIN().
* class.c (class_alloc): zero clear rb_classext_t.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49931 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_core.h, method.h: remove rb_iseq_t::cref_stack. CREF is stored
to rb_method_definition_t::body.iseq_body.cref.
* vm_insnhelper.c: modify SVAR usage.
When calling ISEQ type method, push CREF information onto method
frame, SVAR located place. Before this fix, SVAR is simply nil.
After this patch, CREF (or NULL == Qfalse for not iseq methods)
is stored at the method invocation.
When SVAR is requierd, then put NODE_IF onto SVAR location,
and NDOE_IF::nd_reserved points CREF itself.
* vm.c (vm_cref_new, vm_cref_dump, vm_cref_new_toplevel): added.
* vm_insnhelper.c (vm_push_frame): accept CREF.
* method.h, vm_method.c (rb_add_method_iseq): added. This function
accepts iseq and CREF.
* class.c (clone_method): use rb_add_method_iseq().
* gc.c (mark_method_entry): mark method_entry::body.iseq_body.cref.
* iseq.c: remove CREF related codes.
* insns.def (getinlinecache/setinlinecache): CREF should be cache key
because a different CREF has a different namespace.
* node.c (rb_gc_mark_node): mark NODE_IF::nd_reserved for SVAR.
* proc.c: catch up changes.
* struct.c: ditto.
* insns.def: ditto.
* vm_args.c (raise_argument_error): ditto.
* vm_eval.c: ditto.
* test/ruby/test_class.rb: add a test.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49874 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
struct method_table_wrapper was introduced to avoid duplicate marking
for method tables.
For example, `module M1; def foo; end; end` make one method table
(mtbl) contains a method `foo`. M1 (T_MODULE) points mtbl.
Classes C1 and C2 includes M1, then two T_ICLASS objects are created
and they points mtbl too. In this case, three objects (one T_MODULE
and two T_ICLASS objects) points same mtbl. On marking phase, these
three objects mark same mtbl. To avoid such duplication, struct
method_table_wrapper was introduced.
However, created two T_ICLASS objects have same or shorter lifetime
than M1 (T_MODULE) object. So that we only need to mark mtbl from M1,
not from T_ICLASS objects. This patch tries marking only from M1.
Note that one `Module#prepend` call creates two T_ICLASS objects.
One for refering to a prepending Module object, same as
`Module#include`. We don't nedd to care this T_ICLASS.
One for moving original mtbl from a prepending class. We need to
mark such mtbl from this T_ICLASS object. To mark the mtbl,
we need to use `RCLASS_ORIGIN(klass)` on marking from a prepended
class `klass`.
* class.c: ditto.
* eval.c (rb_using_refinement): ditto.
* gc.c: ditto.
* include/ruby/ruby.h: define m_tbl directly. The definition of
struct RClass should be moved to (srcdir)/internal.h.
* method.h: remove decl of rb_free_m_tbl_wrapper().
* object.c: use RCLASS_M_TBL() directly.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49862 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
force firnalizations at the end of interpreter process.
[Bug #10768]
1) Prohibit incremental GC while running Ruby-level finalizers
to avoid any danger.
2) Prohibit GC while invoking T_DATA/T_FILE data structure
because these operations break object relations consistency.
This patch can introduce another memory consuming issue because
Ruby-level finalizers can run after (2), GC is disabled.
However, basically object consistency was broken at (2) as I
described above. So that running Ruby-level finalizers contains
danger originally. Because of this point, I need to suggest to
remove these 3 lines (invoking remaining finalizers). And add a
rule to add that finalizers should not add new finalizers, or
say there is no guarantee to invoke finalizers that added by
another finalizer.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49684 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (rb_objspace_free): cause rb_bug if lazy sweep is in progress
during rb_objspace_free. Adds extra protection for r46340.
Patch by Vicent Marti. [Bug #10768] [ruby-core:67734]
* gc.c (rb_objspace_call_finalizer): Ensure GC is completed after
finalizers have run. We already call gc_rest() before invoking
finalizers, but finalizer can allocate new objects and start new GC
cycle, so we call gc_rest() again after finalizers are complete.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49474 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Fix SEGV during finilize of WeakRef on Solaris (though the SEGV
could occur on all OS/platforms). [ruby-dev:48779] [Bug #10646]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48999 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (GC_HEAP_FREE_SLOTS): move definition to match use order
(RUBY_GC_HEAP_GROWTH_SLOTS): s/factor/number of slots/
* man/ruby.1: add section for GC environment variables
[Feature #10197]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48855 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This prevents excessive memory growth when a WeakRef
is repeatedly created
* gc.c (define_final0): avoid duplicate blocks
[Bug #10537]
* test/test_weakref.rb (test_repeated_object_leak): new test
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48820 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Class and Module objects can be living long life.
* iseq.c: Same for ISeq objects.
* gc.c (RVALUE_AGE_RESET): added.
* gc.c (newobj_of): allow to generate (age != 0) objects.
* gc.c (rb_copy_wb_protected_attribute): reset age for wb unprotected
objects.
* include/ruby/ruby.h: add RUBY_TYPED_PROMOTED1 as an unrecommended
flag.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48771 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
use 0 for rb_data_type_t::reserved instead of NULL, since its type
may be changed in the future and possibly not a pointer type.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48662 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(malloc_increase) to make GC incrementally.
This change can increase memory consumption. Report us if you find
any problem.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48603 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (default_proc_for_compat_func): check arguments number and
type, and get rid of reentering this default proc.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48425 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
From Ruby 2.2, keys of GC.stat are changed [Feature #9924].
To provide compatible layer, GC.stat add a default_proc
(if default_proc of given Hash object is not set).
At first use of this compatible layer of interpreter process,
show a warning message like that:
program: GC.stat[:total_allocated_object]
warning message: "warning: GC.stat keys were changed from Ruby
2.1. In this case, you refer to obsolete `total_allocated_object'
(new key is `total_allocated_objects').
Please check <https://bugs.ruby-lang.org/issues/9924>
for more information."
Pleaes correct my English message :)
* hash.c (rb_hash_set_default_proc): export (in internal).
* internal.h: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48423 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
to hide unused codes.
* gc.c: similar to GC_ENABLE_LAZY_SWEEP.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48419 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Before this patch, gc_marks_step() marks (white -> grey) fixed
number objects. However, this strategy does not fit practical
cases, for example too slow to make progres.
This patch changes this strategy how many objects the
gc_marks_step() should handle.
We can estimate how many times gc_marks_step() is called during
this major marking (== C) with the free slot number in pooled
pages. We also can estimate the living object number (== L)
using last marked_slots value. We can solve this problem (how
many objects should be process in gc_marks_step()) by L/C.
* gc.c (rb_objspace_t): add rb_objspace_t::rincgc::pooled_sltos and
step_slots.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48413 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
rb_objspace_t:🎏:during_incremental_marking is not defined
when GC_ENABLE_INCREMENTAL_MARK is 0.
* gc.c (will_be_incremental_marking, is_full_marking): similar fix.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48335 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
use `&&' with GC_ENABLE_INCREMENTAL_MARK intead of using
#if/#else/#endif.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48157 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_sweep_rest): sweep rest pages regardless of whether
lazy sweep is enabled or not. based on the patch by Masahiro
Ide at [ruby-dev:48706]. [Bug #10431]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48150 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
marking.
(rb_gc_mark_machine_stack) [__mc68000__]: Also handle it here.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@48065 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
make and register special exceptions.
* vm.c (rb_vm_mark): do not need to mark special exceptions
because they are registerd by rb_gc_register_mark_object().
* eval.c (Init_eval): use rb_vm_register_special_exception().
* gc.c (Init_GC): ditto.
* proc.c (Init_Proc): ditto.
* thread.c (Init_Thread): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47534 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This function is similar to rb_gc_mark_locations(), but not
conservertive.
* internal.h: ditto.
* vm.c (env_mark): use rb_gc_mark_values() because env values should
be Ruby VALUEs.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47533 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_mark): add gc_mark_ptr() to skip is_markable_object()
check. gc_mark_maybe() can use gc_mark_ptr() directly because
passed pointer is checked by is_pointer_to_heap().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47532 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (should_be_callable): preserve encoding of class name in
error messages.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47514 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
ref: [Feature #9924]
* test/ruby/test_gc.rb: add constraints test for gc stat information.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47493 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
newobj_of() also touch:
(4) increment total_allocated_object_num
(5) check hook_events
And gather fields related to marking phase.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47469 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
to make accesssing both parameters easy.
* gc.c (heap_get_freeobj): add LIKELY() hint.
* gc.c (heap_get_freeobj_from_next_freepage): ditto.
* gc.c (newobj_of): check both parameters at once for exceptional
case.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47467 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
ruby_gc_stressfull macro.
Rename objspace->gc_stress to objspace->gc_stress_mode.
If objspace->gc_stress_mode is true (!nil and !false) then
ruby_gc_stressfull becomes TRUE.
ruby_gc_stressfull will speedup newobj_of() slightly.
* gc.c: initialize ruby_gc_stress(full|_mode) by gc_params.gc_stress
even if ENABLE_VM_OBJSPACE is false.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47466 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Please refer this ticket for details.
This change also introduces the following changes.
* Remove RGENGC_AGE2_PROMOTION and introduce object age (0 to 3).
Age can be count with FL_PROMOTE0 and FL_PROMOTE1 flags in
RBasic::flags (2 bit). Age == 3 objects become old objects.
* WB_PROTECTED flag in RBasic to WB_UNPROTECTED bitmap.
* LONG_LIVED bitmap to represent living objects while minor GCs
It specifies (1) Old objects and (2) remembered shady objects.
* Introduce rb_objspace_t::marked_objects which counts marked
objects in current marking phase. marking count is needed to
introduce incremental marking.
* rename mark related function and sweep related function to
gc_(marks|sweep)_(start|finish|step|rest|continue).
* rename rgengc_report() to gc_report().
* Add obj_info() function to get cstr of object details.
* Add MEASURE_LINE() macro to measure execution time of specific line.
* and many small fixes.
* include/ruby/ruby.h: add flag USE_RINCGC.
Now USE_RINCGC can be set only with USE_RGENGC.
* include/ruby/ruby.h: introduce FL_PROMOTED0 and add FL_PROMOTED1
to count object age.
* include/ruby/ruby.h: rewrite write barriers for incremental marking.
* debug.c: catch up flag name changes.
* internal.h: add rb_gc_writebarrier_remember() instead of
rb_gc_writebarrier_remember_promoted().
* array.c (ary_memcpy0): use rb_gc_writebarrier_remember().
* array.c (rb_ary_modify): ditto.
* hash.c (rb_hash_keys): ditto.
* hash.c (rb_hash_values): ditto.
* object.c (init_copy): use rb_copy_wb_protected_attribute() because
FL_WB_PROTECTED is moved from RBasic::flags.
* test/objspace/test_objspace.rb: catch up ObjectSpace.dump() changes.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47444 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
when called from check_gen_consistency. It fixes segmentation
fault on RGENGC_CHECK_MODE >= 1 introduced by r47188.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47208 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
to objspace::rgengc::parent_object (VALUE).
Use Qfalse or RVALUE pointer instead of FALSE and TRUE.
* gc.c (gc_marks_body): should clear parent_object just before
gc_mark_roots() because there are no parents objects
for root objects.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46939 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* add GPR_FLAG_MAJOR_BY_FORCE, which indicates
major GC by METHOD, CAPI and so on (see GC_BY).
* remove GPR_FLAG_MAJOR_BY_RESCAN because not used.
* remove GPR_FLAG_MAJOR_BY_STRESS, use FORCE instead.
* test/ruby/test_gc.rb: catch up.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46927 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm.c: remove mark function for RubyVM object because
RubyVM object marked manually.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46832 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
fstrings refered by static symbols and pinned dynamic symbols
are registerd by rb_gc_register_mark_object().
frstring refered by dynamic symbols (not pinned symbols)
are refered from global_symbols.dsymbol_fstr_hash (Hash object).
Note that fstrings refered from dynamic symbols must live loger
than symbol objects themselves because rb_gc_free_dsymbol() uses
fstring to remove from symbol tables.
This is why we can not mark fstrings from dynamic symbols.
This technique reduces root objects for GC marking.
* gc.c (gc_mark_roots): ditto.
* internal.h: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46772 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.h: rb_objspace_garbage_object_p() as an exported function.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46726 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This function is not opposite against is_live_object()
because is_dying_object() does *not* check object type.
* gc.c (is_dying_object): change condition.
* gc.c (is_live_object): use T_NONE instead of 0.
* gc.c (rb_objspace_dying_object_p): added.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46716 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
From single array, to array of arrays. Each array only has 1024
entries.
* vm.c (Init_vm_objects): change default capa from 1 to 128.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46714 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Extend heap only at
(1) after major GC
or
(2) after several (two times, at current) minor GC
Details in https://bugs.ruby-lang.org/issues/9607#note-9
[Bug #9607]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46387 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_page_sweep): should not set, but add final_slots into
sweep_page->final_slots.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46356 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
clear counter at the beggining of every GC and
count promoted (infant->young) objects.
Some promotions (infant->young) are transition of promoting to old
objects. We should not count such promotions.
With this technique, we don't need to check young objects
at obj_free().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46336 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
gc_verify_internal_consistency() counts all
- live objects
- young objects (if age2 promotion)
- old objects
in all pages and compares with objspace managing counters.
* gc.c (gc_after_sweep): do gc_verify_internal_consistency()
when RGENGC_CHECK_MODE >= 2.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46335 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c: use int for simple predicate functions instead of VALUE.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46334 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
of RVALUE_OLD_P(). clang fails to compile it because is_old is
`int' but RVALUE_OLD_P() returns VALUE.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46332 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (RVALUE_DEMOTE_FROM_OLD): decrement old object count.
* gc.c (RVALUE_DEMOTE_FROM_YOUNG): decrement young object count.
* gc.c (rb_gc_resurrect): increment old object count.
* gc.c (gc_marks_body): should not add old object count.
This code is completely my misunderstanding.
* gc.c (rb_gc_force_recycle): decrement young or old object count
correctly.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46327 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (rb_free_m_tbl): mark function as static
* method.h (rb_free_m_tbl): remove prototype
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46055 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
at the major GC because AGE2Promotion changes all old objects into
young objects at major GC.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45960 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
heap_pages_increment.
For example, GC by exceeding malloc_limit can remain
heap_pages_increment.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45959 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
on RGENGC_AGE2_PROMOTION.
* gc.c (RVALUE_PROMOTE_YOUNG): decrement young object count on
YOUNG->OLD.
* gc.c (obj_free): decrement young object count when young object
freed.
* gc.c (gc_marks): should not clear young object count.
* gc.c (gc_stat_internal): GC.stat :young_object information.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45925 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Tomb heap pages are freed pages here, so expanding heap is
not required.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45760 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
when memop type is MEMOP_TYPE_REALLOC.
GC at realloc is not well maintained.
We need a time to make it safe.
[ruby-dev:48117]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45656 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (objspace_malloc_increase): run full mark if 0x04 bit is
set in ruby_gc_stress. [ruby-core:62103] [Feature #9761]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45655 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (objspace_malloc_increase): run GC after realloc not only
malloc and calloc by GC.stress. [ruby-core:62103] [Feature #9761]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45653 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (garbage_collect_body): name magic numbers.
* gc.c (gc_stress_set): GC.stress accepts not only boolean but
also Fixnum.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45644 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (rb_gc_writebarrier): drop special case for big hash/array
[Bug #9518]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45638 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c: no newline in messages for rb_bug, it outputs a newline
after the message.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45560 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_verify_internal_consistency): always do nothing unless
USE_RGENGC is set, no local variable needed.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45547 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This reverts a hunk of r40703 by ko1.
This fixes [ruby-dev:48098] [Bug #9717].
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45542 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (rgengc_rememberset_mark): don't promote, but remain in
remember set for infant objects.
* gc.c (RVALUE_PROMOTE_INFANT, RVALUE_PROMOTE_YOUNG): count numbers
in these functions.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45531 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Add a new (hidden) C-API to iterate objspace snapshot.
This API is not safe to call any C-APIs in a given callback
function. Be careful to use this C-API.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45505 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_info_decode): get rid of inadvertent dynamic symbol
pin-down, and preserve encoding in error messages. also should
not use RSTRING_PTR macro on function calls.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45480 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
See this ticket about Symbol GC.
* include/ruby/ruby.h:
Declare few functions.
* rb_sym2id: almost same as old SYM2ID but support dynamic symbols.
* rb_id2sym: almost same as old ID2SYM but support dynamic symbols.
* rb_sym2str: almost same as `rb_id2str(SYM2ID(sym))` but not
pin down a dynamic symbol.
Declare a new struct.
* struct RSymbol: represents a dynamic symbol as object in
Ruby's heaps.
Add few macros.
* STATIC_SYM_P: check a static symbol.
* DYNAMIC_SYM_P: check a dynamic symbol.
* RSYMBOL: cast to RSymbol
* gc.c: declare RSymbol. support T_SYMBOL.
* internal.h: Declare few functions.
* rb_gc_free_dsymbol: free up a dynamic symbol. GC call this
function at a sweep phase.
* rb_str_dynamic_intern: convert a string to a dynamic symbol.
* rb_check_id_without_pindown: not pinning function.
* rb_sym2id_without_pindown: ditto.
* rb_check_id_cstr_without_pindown: ditto.
* string.c (Init_String): String#intern and String#to_sym use
rb_str_dynamic_intern.
* template/id.h.tmpl: use LSB of ID as a flag for determining a
static symbol, so we shift left other ruby_id_types.
* string.c: use rb_sym2str instead `rb_id2str(SYM2ID(sym))` to
avoid pinning.
* load.c: use xx_without_pindown function at creating temporary ID
to avoid pinning.
* object.c: ditto.
* sprintf.c: ditto.
* struct.c: ditto.
* thread.c: ditto.
* variable.c: ditto.
* vm_method.c: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45426 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
garbage_collect_with_gvl() here on non-ruby threads.
Should just ignore the malloc_increase.
This issue is pointed by Eric Wong [ruby-core:61519].
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45370 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This hopefully works on all platforms with malloc_usable_size.
This may also trigger bugs in places which did not expect GC, too;
so maybe some existing code will need RB_GC_GUARD.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45347 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
add links to `Object#hash` to each #`hash` methods rdocs.
[Fixes GH-567]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45329 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
at first.
And return immediately if we don't touch sorted list any more.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45182 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* include/ruby/ruby.h (RB_GC_GUARD):
use rb_gc_guarded_ptr_val on non-GCC/MSC
* gc.c (rb_gc_guarded_ptr_val): rename and adjust argument.
RB_GC_GUARD should be robust enough for any compiler.
[ruby-core:60816] [Bug #7805]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45064 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
"RUBY_GC_HEAP_OLDOBJECT_LIMIT_FACTOR" to control major/minor GC
frequency.
Do full GC when the number of old objects is more than R * N
where R is this factor and
N is the number of old objects just after last full GC.
* test/ruby/test_gc.rb: add a test.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45021 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This function *ONLY* works just after marking phase,
before any sweeping.
This function is highly depending current GC implementation
and can be removed future version.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44886 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (ruby_gc_params_t, get_envparam_size): use size_t for
integer environment parameters for sizes.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44871 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(changed by last commit) because "" or "foo" (not a number) strings
are parsed as 0. They should be rejected.
* gc.c (get_envparam_double): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44861 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
then set objspace->rgengc.oldmalloc_increase_limit.
Without this fix, the env variable RUBY_GC_OLDMALLOC_LIMIT
does not work.
* gc.c (get_envparam_int): accept a value equals to lowerbounds.
* gc.c (get_envparam_double): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44853 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_core.h (rb_thread_struct): aggregate cpu stuff into a struct,
so that a debugger can show its content at once.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44722 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
RUBY_FREE_MIN/RUBY_HEAP_MIN_SLOTS if
RUBY_GC_HEAP_FREE_SLOTS/RUBY_GC_HEAP_INIT_SLOTS are given.
[Bug #9276]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44327 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (heap_get_freeobj_from_next_freepage): replace with
heap_get_freepage(). It returns freeobj instead of freepage.
This is not on hot path.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44287 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (heap_get_freeobj): remove redundant assignment. heap->freelist
is set after the while() loop already.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44282 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
can be OR of multiple reasons.
* gc.c (gc_profile_dump_on): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44278 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c: check objspace->profile.current_record before inserting
profiling record by new macro gc_prof_enabled().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44276 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_finalize_deferred_register): define in prototype style,
instead of old K&R style.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44256 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
internal data structure.
Now this method only checks geneartion (old/young) consistency.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44228 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_info_decode): Use :major_by=>:nofree as fallback reason
when other trigger conditions are present.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44171 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (wmap_mark): disable deletion of dead objects by default,
so that WeakMap can be non-shady.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44149 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (wmap_keys): return keys for live objects only, like as
wmap_values.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44142 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (wmap_aset): use st_update instead of st_lookup and
st_insert.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44118 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (wmap_aset): check if both arguments are able to finalize
before setting finalizers.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44108 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* ext/objspace/objspace_dump.c (dump_object): include fstring flag on
strings. include gc flags (old, remembered, wb_protected) on all objects.
* ext/objspace/objspace_dump.c (Init_objspace_dump): initialize lazy
IDs before first use.
* gc.c (rb_obj_gc_flags): new function to retrieve object flags
* internal.h (RB_OBJ_GC_FLAGS_MAX): maximum flags allowed for one obj
* test/objspace/test_objspace.rb (test_dump_flags): test for above
* test/objspace/test_objspace.rb (test_trace_object_allocations):
resolve name before dump (for rb_class_path_cached)
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44105 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Increase oldmalloc_increase with malloc_increase
instead of using obj_memsize_of().
This change will avoid the danger of memory full without major GC.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44040 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
oldmalloc_increase_limit at Init_heap.
rb_objspace_alloc() is not called on some platforms.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44038 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
care about underflow.
* gc.c (objspace_malloc_increase): use it.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44036 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_insnhelper.c (rb_get_kwargs): get keyword argument values from an
option hash, not only checking keys.
* dir.c (dir_initialize): use rb_get_kwargs.
* gc.c (gc_start_internal): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44035 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
heap_pages_free_unused_pages() here.
It was done in after_sweep().
* gc.c (rb_gc): The reason is now GPR_FLAG_CAPI.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44029 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_start_internal): GC.start() now accepts two optional
keyword arguments. These can be used to disable full_mark (minor
mark only) or disable immediate_sweep (use lazy sweep). These new
options are useful for benchmarking GC behavior, or performing minor
GC out-of-band.
* test/ruby/test_gc.rb (class TestGc): tests for new options.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44027 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (struct rb_objspace): rename internal last_collection_flags to
latest_gc_info
* gc.c (gc_latest_collection_info): add GC.latest_gc_info() with similar
behavior to GC.stat()
* gc.c (rb_gc_latest_gc_info): new c-api for above
* gc.c (gc_stat_internal): remove :last_collection_flags from GC.stat
* gc.c (gc_profile_decode_flags): remove GC::Profiler.decode_flags
* include/ruby/intern.h (rb_gc_latest_gc_info): export new c-api
* test/ruby/test_gc.rb (class TestGc): test for new behavior
* NEWS: note about new api
* gc.c (gc_stat_internal): raise TypeError on wrong type
* gc.c (gc_stat): fix error message
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44008 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* include/ruby/intern.h: add rb_gc_stat() for access to GC.stat
variables from c-api
* gc.c (rb_gc_stat): new c-api method. accepts either VALUE hash like
GC.stat, or VALUE symbol key and returns size_t directly. the second
form is useful to avoid allocations, i.e. for usage inside
INTERNAL_EVENT_GC tracepoints.
* gc.c (gc_stat): add GC.stat(:key) to return single value instead of hash
* gc.c (gc_stat_internal): helper method to retrieve single or all stat values
* test/ruby/test_gc.rb (class TestGc): test for new behavior
* NEWS: note about this new api
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44005 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
return malloc_usable_size() if possible.
* gc.c (MALLOC_ALLOCATED_SIZE): add new setting macro to enable
GC.allocated_size.
If platform supports `malloc_usable_size()' (or similar one),
GC.allocated_size can be implemented with this function.
Default is 0.
* gc.c (vm_xmalloc, vm_xrealloc, vm_xfree): use vm_malloc_size()
to detect collect allocated size.
* gc.c (vm_malloc_increase): refactoring.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43998 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* include/ruby/ruby.h: remove INTERNAL_EVENT_GC_END and replace with
two new events: GC_END_MARK and GC_END_SWEEP
* gc.c (gc_after_sweep): emit GC_END_SWEEP after lazy sweep is done
* gc.c (gc_marks_body): emit GC_END_MARK at end of minor/major mark
* ext/-test-/tracepoint/tracepoint.c (struct tracepoint_track): tests
for new events.
* test/-ext-/tracepoint/test_tracepoint.rb (class TestTracepointObj):
ditto.
* NEWS: remove ObjectSpace.after_gc_*_hook. These are only a sample,
and will be removed before ruby 2.1.
* ext/objspace/gc_hook.c: remove ObjectSpace.after_gc_end_hook=
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43997 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (finalize_deferred): flush all deferred finalizers while other
finalizers can get ready to run newly by lazy sweep.
[ruby-core:58833] [Bug #9205]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43994 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (rb_gc_set_params): define as separate function.
RUBY_ALIAS_FUNCTION is for simple alias only, the third parameter
will be ignored on gcc.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43993 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (ruby_gc_set_params): Accept safe_level argument so GC tuning
settings can be applied before rb_safe_level() is available.
* internal.h (rb_gc_set_params): ditto.
* ruby.c (process_options): Apply GC tuning early during boot process
so boot-time allocations can benefit. This also benefits any code
loaded in via `ruby -r`.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43991 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* configure.in: check malloc_size() available on BSD.
* gc.c: use malloc_size() with malloc/malloc.h on BSD.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43983 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
RClass->m_tbl with serial. This prevents double marking method
tables, since many classes/modules can share the same method table.
This improves minor mark time in a large application by 30%.
* internal.h (struct method_table_wrapper): Define new
wrapper struct with additional serial.
* internal.h (RCLASS_M_TBL_INIT): New macro for initializing method
table wrapper and st_table.
* method.h (void rb_sweep_method_entry): Rename rb_free_m_table to
rb_free_m_tbl for consistentcy
* .gdbinit (define rb_method_entry): Update rb_method_entry gdb helper
for new method table structure.
* class.c: Use RCLASS_M_TBL_WRAPPER and
RCLASS_M_TBL_INIT macros.
* class.c (rb_include_class_new): Share WRAPPER between module and
iclass, so serial can prevent double marking.
* eval.c (rb_prepend_module): ditto.
* eval.c (rb_using_refinement): ditto.
* gc.c: Mark and free new wrapper struct.
* gc.c (obj_memsize_of): Count size of additional wrapper struct.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43973 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_stat): Add :last_collection_flags for reason/trigger/type of
last GC run.
* gc.c (gc_prof_sweep_timer_stop): Record HAVE_FINALIZE GPR even
without GC_PROFILE_MORE_DETAIL.
* gc.c (gc_profile_flags): Add GC::Profiler.decode_flags to make sense
of GC.stat[:last_collection_flags]
* test/ruby/test_gc.rb (class TestGc): Test for above.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43893 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
heap_tomb_page_length.
* test/ruby/test_gc.rb: fix to use GC.stat[:heap_eden_page_length]
instead of GC.stat[:heap_length].
This test expects `heap_eden_page_length' (used pages size).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43874 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
if HAVE_MALLOC_USABLE_SIZE (or _WIN32) is defined.
We don't need these function if malloc_usable_size() is available.
* gc.c: catch up this change.
* gc.c: define HAVE_MALLOC_USABLE_SIZE on _WIN32.
* array.c (ary_resize_capa): do not use ruby_sized_xfree() with
local variable to avoid "unused local variable" warning.
This change only has few impact.
* string.c (rb_str_resize): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43839 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
RGENGC_ESTIMATE_OLDSPACE -> RGENGC_ESTIMATE_OLDMALLOC.
* gc.c: add a new major GC reason GPR_FLAG_MAJOR_BY_OLDMALLOC.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43837 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* heap_swept_slot
* malloc_increase
* malloc_limit
* remembered_shady_object
* remembered_shady_object_limit
* old_object
* old_object_limit
* oldmalloc_increase
* oldmalloc_limit
* gc.c (gc_stat): rename names.
* heap_live_num -> heap_live_slot
* heap_free_num -> heap_free_slot
* heap_final_slot -> heap_final_slot
Quote from RDoc of GC.stat():
"The contents of the hash are implementation specific and may
be changed in the future."
* test/ruby/test_gc.rb: catch up this change.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43835 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(oldspace -> oldmalloc for variable names)
OLDSPACE is confusing because it is not includes slots.
To more clearly, rename such as (oldspace_limit -> oldmalloc_limit).
It is clear that it measures (estimates) malloc()'ed size.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43833 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Now we have following environments (and related variable names).
* RUBY_GC_HEAP_INIT_SLOTS
* RUBY_GC_HEAP_FREE_SLOTS
* RUBY_GC_HEAP_GROWTH_FACTOR (new from 2.1)
* RUBY_GC_HEAP_GROWTH_MAX_SLOTS (new from 2.1)
* obsolete
* RUBY_FREE_MIN -> RUBY_GC_HEAP_FREE_SLOTS (from 2.1)
* RUBY_HEAP_MIN_SLOTS -> RUBY_GC_HEAP_INIT_SLOTS (from 2.1)
* RUBY_GC_MALLOC_LIMIT
* RUBY_GC_MALLOC_LIMIT_MAX (new from 2.1)
* RUBY_GC_MALLOC_LIMIT_GROWTH_FACTOR (new from 2.1)
* RUBY_GC_OLDSPACE_LIMIT (new from 2.1)
* RUBY_GC_OLDSPACE_LIMIT_MAX (new from 2.1)
* RUBY_GC_OLDSPACE_LIMIT_GROWTH_FACTOR (new from 2.1)
* test/ruby/test_gc.rb: catch up this change.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43811 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (gc_after_sweep): allocate pages to allocate at least
RUBY_HEAP_MIN_SLOTS.
[Bug #9137]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43795 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
rb_gc_set_params because it's only used in ruby internal.
* internal.h (ruby_gc_set_params): Declare rb_gc_set_params's
alias function.
* gc.c: ditto.
* ruby.c: use ruby_gc_set_params.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43794 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
objspace.
* gc.c (ruby_mimfree): added. It is similar to ruby_mimmalloc().
* internal.h: ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43771 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
malloc_usable_size() is defined by malloc_np.h on FreeBSD.
* configure.in: check malloc.h and malloc_np.h.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43769 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (vm_xrealloc, vm_xfree): use malloc_usable_size() to obtain old
size if available.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43760 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
They are not only used initial values.
Chikanaga-san: Congratulations on RubyPrize!
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43757 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Without this option, some application consumes huge memory.
(and there are only a few performance down)
Introduced new environment variables:
* RUBY_GC_HEAP_OLDSPACE (default 16MB)
* RUBY_GC_HEAP_OLDSPACE_MAX (default 128 MB)
* RUBY_GC_HEAP_OLDSPACE_GROWTH_FACTOR (default 1.2)
* gc.c (initial_malloc_limit): rename to initial_malloc_limit_min.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43755 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Basically, make an object graph of all of living objects before and
after marking and check status.
[Before marking: check WB sanity]
If there is a non-old object `obj' pointed from old object
(`parent') then `parent' or `obj' should be remembered.
[After marking: check marking miss]
Traversible objects with the object graph should be marked.
(However, this alert about objects pointed by machine context
can be false positive. We only display alert.)
[Implementation memo]
objspace_allrefs() creates an object graph.
The object graph is represented by st_table, key is object (VALUE)
and value is referring objects. Referring objects are stored by
"struct reflist".
* gc.c (init_mark_stack): do not use push_mark_stack_chunk() at init.
This pre-allocation causes failure on is_mark_stask_empty()
without any pushing.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43744 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
rb_fstring() used rb_gc_mark() to avoid freeing used string.
However, rb_gc_mark() set mark bit *and* pushes mark_stack.
rb_gc_resurrect() does only set mark bit if it is before sweeping.
* string.c (rb_fstring): use rb_gc_resurrect.
* internal.h: add decl.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43718 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
To reduce memory usage, sweep as soon as possible.
This behavior is same as Ruby 2.0.0 and before.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43623 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
If GC_PROFILE_MORE_DETAIL && GC_PROFILE_DETAIL_MEMORY,
maxrss, minflt and majflt are added to each profile record.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43591 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
vm_malloc_increase() can be called without GVL.
However, gc_rest_sweep() assumes acquiring GVL.
To avoid this problem, check GVL before gc_rest_sweep().
[Bug #9090]
This workaround introduces possibility to set malloc_limit as
wrong value (*1). However, this may be rare case. So I commit it.
*1: Without rest_sweep() here, gc_rest_sweep() can decrease
malloc_increase due to ruby_sized_xfree().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43574 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
conservative for closing to memory consumption of ruby 2.0.
* gc.c (GC_MALLOC_LIMIT, GC_MALLOC_LIMIT_GROWTH_FACTOR):
Adjust parameters for new algorithm.
Example: make gcbench-rdoc on a pc
time maxrss
2.0.0p343 285.27 281853952
trunk before patch 207.19 690405376
trunk after patch 211.59 312500224
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43558 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This is another approach to solve an issue discussed at r43530.
This feature is diabled as default.
This feature measures an increment of memory consuption by oldgen
objects. It measures memory consumption for each objects when
the object is promoted. However, measurement of memory consumption
is not accurate now. So that this measurement is `estimation'.
To implement this feature, move memsize_of() function from
ext/objspace/objspace.c and expose rb_obj_memsize_of().
Some memsize() functions for T_DATA (T_TYPEDDATA) have problem to
measure memory size, so that we ignores T_DATA objects now.
For example, some functions skip NULL check for pointer.
The macro RGENGC_ESTIMATE_OLDSPACE enables/disables this feature,
and turned off as default.
We need to compare 3gen GC and this feature carefully.
(it is possible to enable both feature)
We need a help to compare them.
* internal.h: expose rb_obj_memsize_of().
* ext/objspace/objspace.c: use rb_obj_memsize_of() function.
* cont.c (fiber_memsize): fix to check NULL.
* variable.c (autoload_memsize): ditto.
* vm.c (vm_memsize): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43532 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
RGenGC is designed as 2 generational GC, young and old generation.
Young objects will be promoted to old objects after one GC.
Old objects are not collect until major (full) GC.
The issue of this approach is some objects can promoted as old
objects accidentally and not freed until major GC.
Major GC is not frequently so short-lived but accidentally becoming
old objects are not freed.
For example, the program "loop{Array.new(1_000_000)}" consumes huge
memories because short lived objects (an array which has 1M
elements) are promoted while GC and they are not freed before major
GC.
To solve this problem, generational GC with more generations
technique is known. This patch implements three generations gen GC.
At first, newly created objects are "Infant" objects.
After surviving one GC, "Infant" objects are promoted to "Young"
objects.
"Young" objects are promoted to "Old" objects after surviving
next GC.
"Infant" and "Young" objects are collected if it is not marked
while minor GC. So that this technique solves this problem.
Representation of generations:
* Infant: !FL_PROMOTED and !oldgen_bitmap [00]
* Young : FL_PROMOTED and !oldgen_bitmap [10]
* Old : FL_PROMOTED and oldgen_bitmap [11]
The macro "RGENGC_THREEGEN" enables/disables this feature, and
turned off as default because there are several problems.
(1) Failed sometimes (Heisenbugs).
(2) Performance down.
Especially on write barriers. We need to detect Young or Old
object by oldgen_bitmap. It is slower than checking flags.
To evaluate this feature on more applications, I commit this patch.
Reports are very welcome.
This patch includes some refactoring (renaming names, etc).
* include/ruby/ruby.h: catch up 3gen GC.
* .gdbinit: fix to show a prompt "[PROMOTED]" for promoted objects.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43530 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This flag represents that "this object is promoted at least once."
* gc.c, debug.c, object.c: catch up this change.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43527 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (obj_free): suppress a false shorten-64-to-32 warning,
RUBY_TYPED_FREE_IMMEDIATELY never exceed the limit of int.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43516 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
performance. Add before_sweep condition to heap_page structure.
* gc.c (rb_gc_force_recycle): Use before_sweep member.
* gc.c (heap_is_before_sweep, is_before_sweep): Remove. They has not
already been used.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43508 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (is_live_object): finalizer may not run because of lazy-sweep.
[ruby-dev:47786] [Bug #9069]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43502 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* RUBY_TYPED_FREE_IMMEDIATELY: free the data given by DATA_PTR()
with dfree function immediately. Otherwise (default), the data
freed at finalizaton point.
* RUBY_TYPED_WB_PROTECTED: make this object with FL_WB_PROTECT
(not shady).
* gc.c (obj_free): support RUBY_TYPED_FREE_IMMEDIATELY.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43463 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
We only need one sweep time measurement without lazy sweep.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43427 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (GC_MALLOC_LIMIT): change default value to 16MB.
* gc.c (GC_MALLOC_LIMIT_GROWTH_FACTOR): change default value to 2.0.
* gc.c (gc_before_sweep): change decrease ratio of `malloc_limit'
from 1/4 to 1/10.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43425 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
gc_rest_sweep() can reduce malloc_increase, so try it before GC.
Otherwise, malloc_increase can be less than malloc_limit at
gc_before_sweep(). This means that re-calculation of malloc_limit
may be wrong value.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43424 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
After rb_gc_force_recycle() for a object blonging to heap->freelist,
`heap->using_page->freelist' is not null.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43421 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Share `increment' information with heaps.
* gc.c: change ratio of heap_pages_free_min_page
to 0.80.
This change means slow down page freeing speed.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43397 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Tomb heap is where zombie objects and ghost (freed slot) lived in.
Separate from other heaps (now there is only eden heap) at sweeping
helps freeing pages more efficiently.
Before this patch, even if there is an empty page at former phase
of sweeping, we can't free it.
Algorithm:
(1) Sweeping all pages in a heap and move empty pages from the
heap to tomb_heap.
(2) Check all exsisting pages and free a page
if all slots of this page are empty and
there is enough empty slots (checking by swept_num)
To introduce this pach, there are several tuning of GC parameters.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43395 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
ko1
ktsj
nobu
hsbt
kazu
normal
akr
zzak
tmm1
ngoto
naruse
nagachika
svn
nari
marcandre
tarui
charliesome