Allow build-time customization of the max semispace size. Building a
version of V8 with snapshots and with a non-default max semi-space
size is much easier when you can set the max semispace size in the
build environment.
Review URL: http://codereview.chromium.org/4937001
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5823 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Currently weak handles retain an object for another GC round (oftem times,
major GC round.) Instrumenting Chromium shows that navigation leaves
many global objects which are only collected in next go. Let's
attempt to collect more garbage when approacing OOM condition.
This is a better version of rolled out r5455: now it's correctly
rebuilds object groups between additional GCs.
Review URL: http://codereview.chromium.org/4295004
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5761 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
An attempt to retrieve security context for a function may fail if the
destination heap space is in an incomplete state. To fix this, we only
record unknown functions discovered at GC object moves, and then
register them after GC completes.
BUG=crbug/59627
Review URL: http://codereview.chromium.org/3763012
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5667 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Since 2.6.31 perf_events interface has been available in the
kernel. There's a nice tool called "perf" (linux-2.6/tools/perf) that
uses this interface and provides capabilities similar to oprofile. The
simplest form of its usage is just dumping the raw log (trace) of
events generated by the kernel. In this patch I'm adding a script
(tools/ll_prof.py) to build profiles based on perf trace and our code
log. All the heavy-lifting is done by perf. Compared to oprofile agent
this approach does not require recompilation and supports code moving
garbage collections.
Expected usage is documented in the ll_prof's help. Basically one
should run V8 under perf passing --ll-prof flag and then the produced
logs can be analyzed by tools/ll_prof.py.
The new --ll-prof flag enables logging of generated code object
locations and names (like --log-code), and also of their bodies, which
can be later disassembled and annotated by the script.
Review URL: http://codereview.chromium.org/3831002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5663 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The object's space in Page starts after Page header and is aligned to kMapAlignment which is 32 bytes on 32-bit and 8 bytes on 64-bit.
In case of 64-bit target, the current page header size is exactly 32 bytes so we get the code magically aligned at 32 bytes but it is better to have a separate CODE_POINTER_ALIGN macro to make sure the object space in Page is aligned properly for both maps and code.
There could be a small waste of bytes sometimes (since both Page header and Code header sizes are aligned separately) but it seems the optimal one would involve cross-dependencies between .h files and not clear if it's worth it.
This is a back-port from Isolates branch.
Review URL: http://codereview.chromium.org/3461021
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5526 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The cause for missing functions is that some of them are created
from compiled code (see FastNewClosureStub), and thus not get
registered in profiler's code map.
My solution is to hook on GC visitor to provide JS functions
addresses to profiler, only if it is enabled.
BUG=858
TEST=
Review URL: http://codereview.chromium.org/3417019
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5523 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The number of inobject properties used to be derived from the number
of this property assignments in the constructor (and increased by 2 to
allow for properties added later). This very often leads to wasted inobject
slots.
This patch reclaims some of the unused inobject space by the following method:
- for each constructor function the first several objects are allocated using the initial
("generous) instance size estimation (this is called 'tracking phase').
- during the tracking phase map transitions are tracked and actual property counts are collected.
- at the end of the tracking phase instance sizes in the maps are decreased if necessary
(starting with the function's initial map and traversing the transition tree).
- all further allocation use more realistic instance size estimation.
Shrinking generously allocated objects without costly heap traversal is made possible
by initializing their inobject properties with one_pointer_filler_map (instead of undefined).
The initial slack for the generous allocation is increased from 2 to 6 which really helps some tests.
Review URL: http://codereview.chromium.org/3329019
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5510 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Currently weak handles retain an object for another GC round (oftem times,
major GC round.) Instrumenting Chromium shows that navigation leaves
many global objects which are only collected in next go. Let's
attempt to collect more garbage when approacing OOM condition.
Review URL: http://codereview.chromium.org/3327021
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5455 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
That could improve chances for commit success as currently,
if we moved free pages out of order, we cannot shrink spaces.
However, when we experience problems commiting from space back, we should
use most of resources at our disposal.
Also get rid of currently unused parameter to DeallocateFunction.
Review URL: http://codereview.chromium.org/3260001
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5372 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
the code object's instructions.
This allows us to find a code object using just the pc. This approach
uses a cache (PcToCodeCache) to make sure we don't continuously have
to iterate heap pages.
This change eliminates the need for cooking and uncooking of stack frames.
Review URL: http://codereview.chromium.org/3226014
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5369 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
r5147 wrongly assumed that a code cache for a slow case map is always empty.
This patch solves this: whenever we attempt to add a stub to a map's code cache
we check that this map is cached. If it is we give the object its own copy
of the map and only then modify the map.
Review URL: http://codereview.chromium.org/3134027
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5342 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Object model changes
----------------------------------------
New fixed_cow_array_map is used for the elements array of a JSObject
to mark it as COW. The JSObject's map and other fields are not
affected. The JSObject's map still has the "fast elements" bit set. It
means we can do only the receiver map check in keyed loads and the
receiver and the elements map checks in keyed stores. So introducing
COW arrays doesn't hurt performance of these operations. But note that
the elements map check is necessary in all mutating operations because
the "has fast elements" bit now means "has fast elements for reading".
EnsureWritableFastElements can be used in runtime functions to perform
the necessary lazy copying.
Generated code changes
----------------------------------------
Generic keyed load is updated to only do the receiver map check (this
could have been done earlier). FastCloneShallowArrayStub now has two
modes: clone elements and use COW elements. AssertFastElements macro
is added to check the elements when necessary. The custom call IC
generators for Array.prototype.{push,pop} are updated to avoid going
to the slow case (and patching the IC) when calling the builtin should
work.
COW enablement
----------------------------------------
Currently we only put shallow and simple literal arrays in the COW
mode. This is done by the parser.
Review URL: http://codereview.chromium.org/3144002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
That could allow us to understand why commit of from space sometimes fails.
Another option would be start a separate structure with OS-related info, but
as it's a single field, let's put it into HeapStats, at least for now.
Review URL: http://codereview.chromium.org/3118013
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5261 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
in allowing more flexible compilation and to simplify builtins lookup.
This changes a number of places where code objects are assigned to
SharedFunctionInfo objects to also assign this code object to the
JSFunction. In addition, the code flushing is changed slightly to
accomodate this (we need to flush the code from functions pointing to
SharedFunctionInfo objects that has already been flushed).
Review URL: http://codereview.chromium.org/3120006
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5233 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
To trace objects between snapshots, an external map of object tags is
maintained. After the first heap snapshot has been taken, the map is
updated by reporting object moves from the GC. If no snapshots were
taken, there is no overhead (except for flag checking).
I considered graph comparison algorithms that doesn't require using
object tags, but they are all of a high computational complexity, and
will still fail to detect object moves properly, even for trivial
cases, so using tags looks like unavoidable.
Review URL: http://codereview.chromium.org/3020002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5078 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
The static ScopeInfo members moved into this class.
The new class is named ScopeInfoObject which I am not proud of,
better ideas are very welcome.
Also got rid of the sentinels in the serialized scope info which saves 3 words per function
and is not slower.
Review URL: http://codereview.chromium.org/2908009
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5067 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Use virtually dispatched specialized scavengers instead of single generic ScavengeObjectSlow implementation.
Rollback of r5041 with assertion checking callback alignment removed. Map space is iterated in a special fashion during scavenges so special callback alignment is not required.
Review URL: http://codereview.chromium.org/2950003
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5047 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
A potential issue with this change is creating lots of maps when
objects flip between fast/slow elements modes. We could add special
transitions to avoid this. Yet testing this on our benchmarks, gmail,
and wave seems to indicate that this is not a real problem.
Review URL: http://codereview.chromium.org/2870018
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4941 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Stable high survival rates of young objects both during partial and
full collection indicate that mutator is either building or modifying
a structure with a long lifetime.
In this case we aggressively raise old generation memory limits to
postpone subsequent mark-sweep collection and trade memory space
for the mutation speed.
Review URL: http://codereview.chromium.org/2809032
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4938 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
If a two-byte string only contains ascii characters, then we can save
memory when flattening a cons string containing it. Similarly we can
use this in Array.prototype.join implementation. To track this a new
bit is added to instance type. This bit is used as a hint in generated
code and in runtime functions.
To enable testing a new V8 extension is added controlled by
--expose-externalize-string flag.
Review URL: http://codereview.chromium.org/2762008
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4894 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Changed the disposal of external string resources to call a virtual Dispose method on the resource. The default inplementation of Dispose deletes the object and will capture the delete operator matching the new operator used to allocate the object.
Review URL: http://codereview.chromium.org/2658008
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4816 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This adds an additional step to full gc, removing code from functions
that are no longer in the compilation cache. The code is replaced with
a lazy compile version enabling us to recompile the function in case
we do actually need it again.
Review URL: http://codereview.chromium.org/2632003
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4814 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
- New сardmarking write barrier handles large objects and normal objects in a similar fashion (no more additional space for pointer tracking is required, no conditional branches in WB code).
- Changes to enable oldspaces iteration without maps decoding:
-- layout change for FixedArrays: length is stored as a smis (initial patch by
Kevin Millikin)
-- layout change for SharedFunctionInfo: integer fields are stored as smi on
arm, ia32 and rearranged on x64.
-- layout change for String: meaning of LSB bit is fliped (1 now means hash not
computed); on x64 padding is added.
-- layout of maps is _not_ changed. Map space is currently iterated in a special
way.
Review URL: http://codereview.chromium.org/2144006
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4715 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
-- layout change for FixedArrays: length is stored as a smis (initial patch by Kevin Millikin)
-- layout change for SharedFunctionInfo: integer fields are stored as smi on arm, ia32 and rearranged on x64.
-- layout change for String: meaning of LSB bit is fliped (1 now means hash not computed); on x64 padding is added.
-- layout of maps is _not_ changed. Map space is currently iterated in a special way.
- Cardmarking write barrier. New barrier handles large objects and normal objects in a similar fashion (no more additional space for pointer tracking is required, no conditional branches in WB code).
Review URL: http://codereview.chromium.org/2101002
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4685 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
* Faster hashing for sequential strings.
* When adding short external two-byte strings try to convert them
back to ascii. Chances are high the embedder uses two-byte
representation even for ascii strings. This optimization saves
memory and makes hashing faster.
Review URL: http://codereview.chromium.org/1444001
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4300 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
registration of external references in Proxy objects).
I moved the declaration of the two functions to stub-cache.h
because with all the types they use it's hard to declare them
anywhere else. But the actual definition is still in runtime.cc
near to the place where they are used.
Review URL: http://codereview.chromium.org/1079012
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4231 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
A separate object type for the code cache have been added. This object has two different code caches. The first one (default_cache) is a fixed array organized in the same way as the as the code cache was before. The second cache (global_access_cache) is for code stubs to access the global object. This cache is organized as a hash table taking the property name and code flags as the key.
The reason for separating the global access stubs into a hash table representation is that the number of these is not bounded in the same was as the other types.
This is a remake of r3952 (http://codereview.chromium.org/652119) which have the additional ability to look for the index of code stubs for access to the global object.
BUG=http://code.google.com/p/v8/issues/detail?id=613
Review URL: http://codereview.chromium.org/717001
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4066 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
A separate object type for the code cache have been added. This object has two different code caches. The first one (default_cache) is a fixed array organized in the same way as the as the code cache was before. The second cache (global_access_cache) is for code stubs to access the global object. This cache is organized as a hash table taking the property name and code flags as the key.
The reason for separating the global access stubs into a hash table representation is that the number of these is not bounded in the same was as the other types.
BUG=613
Review URL: http://codereview.chromium.org/652119
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3952 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Moved all the logic to a function on SharedFunctionInfo (including the flag check) to make things more readable.
Changed the check for setter to do a lookup for a named setter for each of the properties assigned in the constructor.
Added tests using accessors and interseptors set through the API.
Added fast case objects to the mjsunit test.
TEST=test/mjsunit/setter-on-constructor-prototype.js
TEST=test/cctest/test-api/SetterOnConstructorPrototype
TEST=test/cctest/test-api/InterceptorOnConstructorPrototype
Review URL: http://codereview.chromium.org/619006
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3893 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Simple objects which are constructed without calling the actual constructor function did not take setters defined on prototype chain of the new object into account.
Constructing objects this way is now not done if there are setters involved on the prototype chain of the new object.
This only fixes the case where the setter is found when the first object from a constructor is created. If the prototype chain is changed new objects will on take any change to setters into account.
TEST=test/mjsunit/setter-on-constructor-prototype.js
Review URL: http://codereview.chromium.org/606062
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3879 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
NumberToString in runtime JavaScript is inlined through a call to a stub. Currently the stub only checks the number string cache and only if the number is a smi. Code is shared with the inlining of number string cache lookup when adding a smi to a string.
Review URL: http://codereview.chromium.org/604062
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3865 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
When functions only have simple assignments of the form this.x = ... the object is created in generated code without actually calling the constructor. In this case the initial map for the function already contains the properties assigned in the constructor. The field descriptors in this initial map now has an enumeration index assigned to make property enumeration order the insertion order. The insertion order here is the order of the this.x assignments in the code.
BUG=http://crbug.com/3867
TEST=test/mjsunit/regress/regress-crbug-3867.js
Review URL: http://codereview.chromium.org/566016
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3768 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
1. Avoid using SKIP_WRITE_BARRIER when we don't have to (smis).
2. Check and document the remaining uses of SKIP_WRITE_BARRIER.
3. Only allow GetWriteBarrierMode when in an AssertNoAllocation scope.
The only functional change should be in DeepCopyBoilerplate where we
no longer use the write barrier mode (because of allocations). I'm
running benchmarks to see if this has a measurable impact on performance.
Review URL: http://codereview.chromium.org/558041
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3743 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
objects in the startup heap from a partial snapshot. This happens
through the partial snapshot cache. A startup snapshot and a
partial snapshot are created together so that the startup snapshot
contains the partial snapshot cache entries needed.
Review URL: http://codereview.chromium.org/548149
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3713 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
true. The rules are:
1. Heap::AllocateRaw can normally handle allocation requests in new
space even when always_allocate() is true. It properly retries
failed allocation in the second 'retry' space.
2. Heap::Allocate can normally handle allocation requests in new
space.
3. We only need to check always_allocate() when explicitly requesting
allocation in new space via Heap::new_space().AllocateRaw().
4. The exception to these rules is fixed arrays with size such that
MaxObjectSizeInPagedSpace < size <= MaxObjectSizeInNewSpace (ie,
those that will be allocated in new space and promoted to large
object space). They cannot be allocated in new space via
Heap::Allocate or Heap::AllocateRaw, because the retry logic does
not know to allocate extra remembered set bits when retrying in
large object space.
Review URL: http://codereview.chromium.org/518007
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3535 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
that attempt to run with a small heap. Additionally, it can
potentially keep a lot of string data alive and it is never flushed.
Can we make it grow dynamically if used so that we can still start the
VM with a small heap size?
Review URL: http://codereview.chromium.org/503081
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3517 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
On 32-bit the maps are now aligned on a 32-byte boundary in order to encode more maps during compacting GC. The actual size of a map on 32-bit is 28 bytes making this change waste 4 bytes per map.
On 64-bit the encoding for compacting GC is now using more than 32-bits and the maps here are still pointer size aligned. The actual size of a map on 64-bit is 48 bytes and this change does not intruduce any waste.
My choice of 16 bits for kMapPageIndexBits for 64-bit should give the same maximum number of pages (8K) for map space. As maps on 64-bit are larger than on 32-bit the total number of maps on 64-bit will be smaller than on 32-bit. We could consider raising this to 17 or 18.
I moved the kPageSizeBits to globals.h as the calculation of the encoding really depended on this.
There are still an #ifdef/#endif in objects.h and this constant could be moved to globaks.h as well, but I kept it together with the related constants.
All the tests run in debug mode with additional options --gc-global --always-compact as well (except for a few tests on which also fails before this change when run with --gc-global --always-compact).
BUG=http://code.google.com/p/v8/issues/detail?id=524
BUG=http://crbug.com/29428
TEST=test/mjsunit/regress/regress-524.js
Review URL: http://codereview.chromium.org/504026
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@3481 ce2b1a6d-e550-0410-aec6-3dcde31c8c00