platform/upstream/v8.git
14 years agoMake lint happy.
vitalyr@chromium.org [Mon, 4 Oct 2010 15:21:38 +0000 (15:21 +0000)]
Make lint happy.

TBR=kmillikin@chromium.org

Review URL: http://codereview.chromium.org/3609008

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5585 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAPI: expose RegExp.
vitalyr@chromium.org [Mon, 4 Oct 2010 15:04:16 +0000 (15:04 +0000)]
API: expose RegExp.

Review URL: http://codereview.chromium.org/3585010

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5584 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMore refactoring of class Compiler's interface.
kmillikin@chromium.org [Mon, 4 Oct 2010 14:30:43 +0000 (14:30 +0000)]
More refactoring of class Compiler's interface.

Change more functions used by the Compiler class to have a uniform
interface: they get passed as argument an input/output pointer to a
CompilationInfo that they mutate if they succeed, and they return a
flag telling whether they succeeded.

Also, remove some unnecessary timers.

Review URL: http://codereview.chromium.org/3561012

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5583 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPreviously, there was a bug there, fixed in r5581. But having an assert still makes...
mikhail.naganov@gmail.com [Mon, 4 Oct 2010 12:54:31 +0000 (12:54 +0000)]
Previously, there was a bug there, fixed in r5581. But having an assert still makes sense.

Also, add functions registration to CPU profiler in case
when closure code wasn't compiled yet.

Review URL: http://codereview.chromium.org/3557007

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5582 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoBegin a more aggressive refactoring of the Compiler interface.
kmillikin@chromium.org [Mon, 4 Oct 2010 11:35:46 +0000 (11:35 +0000)]
Begin a more aggressive refactoring of the Compiler interface.

The plan is to use the CompilationInfo class to communicate inputs and
outputs to compilation pipeline phases, which each return a boolean
success/failure flag.

The intent is to make it easier to compose small pieces of the
pipeline without having to grow a custom function each time, each
taking a half dozen arguments.

This change modifies the very front end (the parser).

Review URL: http://codereview.chromium.org/3586006

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5581 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPrepare push to trunk. Now working on version 2.4.9.
ricow@chromium.org [Mon, 4 Oct 2010 07:57:07 +0000 (07:57 +0000)]
Prepare push to trunk. Now working on version 2.4.9.

Review URL: http://codereview.chromium.org/3538009

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5579 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix a pair of compilation failure bugs in test files due to r5576.
kmillikin@chromium.org [Fri, 1 Oct 2010 15:32:32 +0000 (15:32 +0000)]
Fix a pair of compilation failure bugs in test files due to r5576.

1. parser.h depends on ast.h, but only gets it included implicitly
   everywhere parser.h is included (except for tests).  Include ast.h in
   parser.h.

2. Regular expression tests test the free functions that have just been
   moved into class Parser.

TBR=fschneider@chromium.org

Review URL: http://codereview.chromium.org/3602007

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5577 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMake some free functions into static members of class Parser.
kmillikin@chromium.org [Fri, 1 Oct 2010 14:10:47 +0000 (14:10 +0000)]
Make some free functions into static members of class Parser.

Review URL: http://codereview.chromium.org/3518005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5576 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix property array length calculation in TransformPropertiesToFastFor.
kaznacheev@chromium.org [Fri, 1 Oct 2010 12:40:30 +0000 (12:40 +0000)]
Fix property array length calculation in TransformPropertiesToFastFor.

It was silently assumed that inobject_properties value is not too large.
Recent introduction of inobject slack tracking made the assumption false
and debug tests with no snapshot failed.

Review URL: http://codereview.chromium.org/3584004

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5575 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRename include guard to match file name
sgjesse@chromium.org [Fri, 1 Oct 2010 09:30:42 +0000 (09:30 +0000)]
Rename include guard to match file name
Review URL: http://codereview.chromium.org/3532006

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5574 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix bad cast in string-search.h.
lrn@chromium.org [Fri, 1 Oct 2010 09:08:52 +0000 (09:08 +0000)]
Fix bad cast in string-search.h.

Review URL: http://codereview.chromium.org/3530004

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5572 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFine-tune inobject slack tracking.
kaznacheev@chromium.org [Fri, 1 Oct 2010 08:28:28 +0000 (08:28 +0000)]
Fine-tune inobject slack tracking.

The current parameters (initial slack = 6, kGenerousAllocationsCount = 16)
caused 15-20% performance regression on 3d-raytrace on x64
(visible both in standalone shell and in the web harness). This is very
likely some kind of an alignment  problem (it goes away completely with --prof).
A similar problem can be observed on ia32 with initial slack=2.

Increased initial slack to 8 and decreased constructions count to 8.
This made the regression go away (and did not harm any of the other tests).

Review URL: http://codereview.chromium.org/3595002

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5571 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix HeapSnapshotsDiff test, diff implementation, and a bug introduced
mikhail.naganov@gmail.com [Fri, 1 Oct 2010 07:19:23 +0000 (07:19 +0000)]
Fix HeapSnapshotsDiff test, diff implementation, and a bug introduced
during snapshot size optimization.

Sorry, now I figured out that the diff implementation itself was also
incorrect.  Reachable nodes must be filtered from the beginning,
otherwise, an object that is already disconnected, but not discarded
yet, will not appear as a deleted (thankfully, this bug for some
reason had appeared on the x64 port.)

BUG=868
TEST=HeapSnapshotRootPreservedAfterSorting

Review URL: http://codereview.chromium.org/3531005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5570 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoUse existing global cell status as a hint when generating loads.
vitalyr@chromium.org [Thu, 30 Sep 2010 17:39:31 +0000 (17:39 +0000)]
Use existing global cell status as a hint when generating loads.

Review URL: http://codereview.chromium.org/3537003

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5569 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRevert last commit due to check fail on Linux.
mikhail.naganov@gmail.com [Thu, 30 Sep 2010 12:35:03 +0000 (12:35 +0000)]
Revert last commit due to check fail on Linux.

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5568 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix HeapSnapshotsDiff test and a bug introduced during snapshot size optimization.
mikhail.naganov@gmail.com [Thu, 30 Sep 2010 12:24:56 +0000 (12:24 +0000)]
Fix HeapSnapshotsDiff test and a bug introduced during snapshot size optimization.

The bug wasn't revealed because the test was disabled.

BUG=848
TEST=HeapSnapshotsDiff,HeapSnapshotRootPreservedAfterSorting

Review URL: http://codereview.chromium.org/3572003

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5567 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRename some x64 macros to be more precise about their semantics.
antonm@chromium.org [Thu, 30 Sep 2010 11:48:03 +0000 (11:48 +0000)]
Rename some x64 macros to be more precise about their semantics.

Review URL: http://codereview.chromium.org/3574002

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5566 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd fflush after writing to log (Issue 868).
mikhail.naganov@gmail.com [Thu, 30 Sep 2010 10:50:39 +0000 (10:50 +0000)]
Add fflush after writing to log (Issue 868).

BUG=868

Review URL: http://codereview.chromium.org/3605001

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5565 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPrepare push to trunk. Now working on version 2.4.8.
kmillikin@chromium.org [Thu, 30 Sep 2010 09:35:50 +0000 (09:35 +0000)]
Prepare push to trunk.  Now working on version 2.4.8.

Review URL: http://codereview.chromium.org/3585003

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5562 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoCleanup of the parser.
kmillikin@chromium.org [Thu, 30 Sep 2010 09:28:58 +0000 (09:28 +0000)]
Cleanup of the parser.

The lazy parsing functions took a host of arguments that can all be
derived from the SharedFunctionInfo, and the SharedFunctionInfo is
always available when parsing lazily.  Change the interface to take a
single CompilationInfo or SharedFunctionInfo argument.

Also remove a flag in the parser that was always false when it was read.

Review URL: http://codereview.chromium.org/3538005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5561 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoIntroduce subclasses of class CompilationInfo.
kmillikin@chromium.org [Thu, 30 Sep 2010 08:48:37 +0000 (08:48 +0000)]
Introduce subclasses of class CompilationInfo.

It was a wart that we had three handle fields, exactly one of which
was non-null; and that we had three overloaded constructors.  Instead,
introduce subclasses and virtual methods.

Remove some unused fields from class CompilationInfo.

Review URL: http://codereview.chromium.org/3566003

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5560 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoNew attempt to make the allocation routines 64 bit clean.
erik.corry@gmail.com [Thu, 30 Sep 2010 07:22:53 +0000 (07:22 +0000)]
New attempt to make the allocation routines 64 bit clean.
This one has been approved by the 64 bit compiler in MSVC
2005 so I hope it also passes the 2008 version.

The --max-new-space-size option is now in kBytes.
The --max-old-space-size option is now in MBytes.

Some issues remain with 64 bit heaps and the counters.  See
http://code.google.com/p/v8/issues/detail?id=887
Review URL: http://codereview.chromium.org/3573005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5559 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd another check to be sure non-negative smi fits into Uint32 range.
antonm@chromium.org [Wed, 29 Sep 2010 18:15:36 +0000 (18:15 +0000)]
Add another check to be sure non-negative smi fits into Uint32 range.

Review URL: http://codereview.chromium.org/3546003

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5558 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRemove some unused fields from class CompilationInfo.
kmillikin@chromium.org [Wed, 29 Sep 2010 17:38:37 +0000 (17:38 +0000)]
Remove some unused fields from class CompilationInfo.

Review URL: http://codereview.chromium.org/3533007

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5557 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoDo not invoke indexed interceptor getters for negative indices.
antonm@chromium.org [Wed, 29 Sep 2010 15:08:38 +0000 (15:08 +0000)]
Do not invoke indexed interceptor getters for negative indices.

BUG=https://bugs.webkit.org/show_bug.cgi?id=46689

Review URL: http://codereview.chromium.org/3520006

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5553 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd cast to make Win-64 compile without warnings. This is a commit of http://coderev...
erik.corry@gmail.com [Wed, 29 Sep 2010 14:06:30 +0000 (14:06 +0000)]
Add cast to make Win-64 compile without warnings.  This is a commit of codereview.chromium.org/3521002 for lrn.

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5552 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdded missing file. Mea culpa.
lrn@chromium.org [Wed, 29 Sep 2010 12:12:45 +0000 (12:12 +0000)]
Added missing file. Mea culpa.

Review URL: http://codereview.chromium.org/3586001

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5551 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRefactored string search code.
lrn@chromium.org [Wed, 29 Sep 2010 10:57:23 +0000 (10:57 +0000)]
Refactored string search code.
Made string search state explicit for repreated calls (a StringSearch class).

Review URL: http://codereview.chromium.org/3467010

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5550 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd an AST node type enum to AST nodes.
kmillikin@chromium.org [Wed, 29 Sep 2010 07:51:47 +0000 (07:51 +0000)]
Add an AST node type enum to AST nodes.

When inspecting an AST node in the debugger, it can be difficult to
find its concrete type.  Add a "node_type" virtual function to all the
concrete AST nodes.

Use a macro to generate the definition of node_type as well as Accept
and the AsXXX type testing/casting functions.

Review URL: http://codereview.chromium.org/3431026

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5549 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMark ptest instruction as requiring SSE4.1.
vitalyr@chromium.org [Tue, 28 Sep 2010 14:56:36 +0000 (14:56 +0000)]
Mark ptest instruction as requiring SSE4.1.

I also added a flag like the ones we have for other SSE versions.

Review URL: http://codereview.chromium.org/3505003

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5548 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix external array iteration used in array concat.
vitalyr@chromium.org [Tue, 28 Sep 2010 14:49:29 +0000 (14:49 +0000)]
Fix external array iteration used in array concat.

We should use Factory and not Heap there.

Review URL: http://codereview.chromium.org/3380031

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5547 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRevert attempt to make heap size 32/64 clean. This change needs to
erik.corry@gmail.com [Tue, 28 Sep 2010 12:49:25 +0000 (12:49 +0000)]
Revert attempt to make heap size 32/64 clean.  This change needs to
be done from Windows where the compiler is stricter about truncating
changes.
Review URL: http://codereview.chromium.org/3454035

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5545 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMore changes to 32/64 bit correctness.
erik.corry@gmail.com [Tue, 28 Sep 2010 12:18:19 +0000 (12:18 +0000)]
More changes to 32/64 bit correctness.
TBR=vegorov
Review URL: http://codereview.chromium.org/3412038

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5544 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMore correctness around PrintF and 32/64 bit values.
erik.corry@gmail.com [Tue, 28 Sep 2010 11:44:30 +0000 (11:44 +0000)]
More correctness around PrintF and 32/64 bit values.
Fix test after 64 bit heap size change.
Review URL: http://codereview.chromium.org/3432032

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5543 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoUse intptr_t instead of int for heap sizes. This is a step towards
erik.corry@gmail.com [Tue, 28 Sep 2010 10:35:47 +0000 (10:35 +0000)]
Use intptr_t instead of int for heap sizes.  This is a step towards
removing the 1Gbyte limit.
Review URL: http://codereview.chromium.org/3418035

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5542 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAvoid logging preparse-data inside lazily compiled functions.
lrn@chromium.org [Tue, 28 Sep 2010 07:50:32 +0000 (07:50 +0000)]
Avoid logging preparse-data inside lazily compiled functions.
Reduces size of preparser data significantly when there are nested functions.
Also allows us to drop the "skip" fields of function entries,
that tells us how much preparse-data to skip when skipping the function source.

Review URL: http://codereview.chromium.org/3412034

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5541 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoCustom call IC for Math.abs.
vitalyr@chromium.org [Mon, 27 Sep 2010 23:41:41 +0000 (23:41 +0000)]
Custom call IC for Math.abs.

Review URL: http://codereview.chromium.org/3446024

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5538 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd Debug::CancelDebugBreak method to api
podivilov@chromium.org [Mon, 27 Sep 2010 15:09:55 +0000 (15:09 +0000)]
Add Debug::CancelDebugBreak method to api

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5535 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPrepare push of version 2.4.6 to trunk. Now working on version 2.4.7.
whesse@chromium.org [Mon, 27 Sep 2010 10:58:18 +0000 (10:58 +0000)]
Prepare push of version 2.4.6 to trunk.  Now working on version 2.4.7.
Review URL: http://codereview.chromium.org/3470016

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5531 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix more GC unsafe places
vegorov@chromium.org [Mon, 27 Sep 2010 10:29:25 +0000 (10:29 +0000)]
Fix more GC unsafe places

Review URL: http://codereview.chromium.org/3499001

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5530 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix error in x64 inlined optimized shift operators.
whesse@chromium.org [Mon, 27 Sep 2010 07:24:01 +0000 (07:24 +0000)]
Fix error in x64 inlined optimized shift operators.
Review URL: http://codereview.chromium.org/3496007

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5529 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agofixed
mikhail.naganov@gmail.com [Mon, 27 Sep 2010 05:25:31 +0000 (05:25 +0000)]
fixed

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5528 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd CODE_POINTER_ALIGN, use it in Page to align generated code.
dimich@chromium.org [Fri, 24 Sep 2010 21:48:44 +0000 (21:48 +0000)]
Add CODE_POINTER_ALIGN, use it in Page to align generated code.

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

14 years agoImplemented missing instructions in ia32 and x64 disassembler.
kaznacheev@chromium.org [Fri, 24 Sep 2010 12:55:17 +0000 (12:55 +0000)]
Implemented missing instructions in ia32 and x64 disassembler.

ia32: fld(i), fldpi
x64: fld(i), fldpi, cvtsd2si, cvttsd2si

Review URL: http://codereview.chromium.org/3471011

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5525 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix presubmit error.
mikhail.naganov@gmail.com [Fri, 24 Sep 2010 11:50:50 +0000 (11:50 +0000)]
Fix presubmit error.

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5524 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoProvide more functions to CPU profiler (fix issue 858).
mikhail.naganov@gmail.com [Fri, 24 Sep 2010 11:45:12 +0000 (11:45 +0000)]
Provide more functions to CPU profiler (fix issue 858).

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

14 years agoRevert 5520 due to compilation failure.
erik.corry@gmail.com [Fri, 24 Sep 2010 09:35:19 +0000 (09:35 +0000)]
Revert 5520 due to compilation failure.

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5521 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoSmall tweaks to make the x64 full code generator generate
erik.corry@gmail.com [Fri, 24 Sep 2010 08:34:52 +0000 (08:34 +0000)]
Small tweaks to make the x64 full code generator generate
more compact code.
Review URL: http://codereview.chromium.org/3443022

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5520 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix some inconsistent formatting.
kmillikin@chromium.org [Fri, 24 Sep 2010 08:25:31 +0000 (08:25 +0000)]
Fix some inconsistent formatting.

I noticed we sometimes had extra spaces before and after the "const"
keyword.  Probably the result of a search and replace gone wrong.

This is a whitespace only change.

Review URL: http://codereview.chromium.org/3427021

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5519 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPrevent modification of cached normalized maps.
kaznacheev@chromium.org [Fri, 24 Sep 2010 08:18:33 +0000 (08:18 +0000)]
Prevent modification of cached normalized maps.

Finally sovles the problem that r5342 attempted to solve.
When adding a stub to a map's code cache we need to make
sure that this map is not used by object that do not need
this stub.

Existing solution had 2 flaws:
1. It checked that the map is cached by asking the current context.
If the object escaped into another context then NormalizedMapCache::Contains
returns false negative.

2. If a map gets evicted from the cache we should not try to modify it
even though Contains returns false.

This patch implements much less fragile solution of the same problem:
A map now has a flag (is_shared) that is set once the map is added
to a cache, stays set even after the cache eviction, and is cleared
if the object goes back to fast mode.

Added a regression test.

Review URL: http://codereview.chromium.org/3472006

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5518 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoClean up some messiness in Scopes.
kmillikin@chromium.org [Fri, 24 Sep 2010 07:53:59 +0000 (07:53 +0000)]
Clean up some messiness in Scopes.

For some reason, the scope's arguments and arguments shadow were
variable proxies, which resulted in all references to the arguments
shadow being shared in the AST.  This makes it hard to put per-node
state on the AST nodes.

I took the opportunity to remove Variable::AsVariable which has
confused people in the past, and to rename Variable::slot to the more
accurate Variable::AsSlot.

Review URL: http://codereview.chromium.org/3432022

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5517 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoSupport dec_b instruction in ia32 and x64 disassembler.
kaznacheev@chromium.org [Thu, 23 Sep 2010 19:01:56 +0000 (19:01 +0000)]
Support dec_b instruction in ia32 and x64 disassembler.

Review URL: http://codereview.chromium.org/3412028

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5515 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd an assert to ensure we never allocate second empty fixed array.
vitalyr@chromium.org [Thu, 23 Sep 2010 12:49:59 +0000 (12:49 +0000)]
Add an assert to ensure we never allocate second empty fixed array.

Review URL: http://codereview.chromium.org/3442012

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5514 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix copy-on-write assert by setting the new array map early.
vitalyr@chromium.org [Thu, 23 Sep 2010 12:23:35 +0000 (12:23 +0000)]
Fix copy-on-write assert by setting the new array map early.

BUG=876

Review URL: http://codereview.chromium.org/3466013

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5513 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix getOwnPropertyDescriptor() support for index properties.
vegorov@chromium.org [Thu, 23 Sep 2010 11:25:01 +0000 (11:25 +0000)]
Fix getOwnPropertyDescriptor() support for index properties.

Add support for index properties with getters, setters or indexed interceptors.

For indexed interceptor case only fix crashes, do not guarantee any semantic soundness. Separate issue opened for this http://code.google.com/p/v8/issues/detail?id=877

BUG=http://code.google.com/p/v8/issues/detail?id=874

Review URL: http://codereview.chromium.org/3462005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5512 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoCleanup of contexts in the full code generator.
erik.corry@gmail.com [Thu, 23 Sep 2010 09:22:45 +0000 (09:22 +0000)]
Cleanup of contexts in the full code generator.
Review URL: http://codereview.chromium.org/3449004

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5511 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoDynamically determine optimal instance size.
kaznacheev@chromium.org [Thu, 23 Sep 2010 09:15:26 +0000 (09:15 +0000)]
Dynamically determine optimal instance size.

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

14 years agoFix possible evaluation order problems.
vegorov@chromium.org [Thu, 23 Sep 2010 08:27:51 +0000 (08:27 +0000)]
Fix possible evaluation order problems.

We should not allow handle dereference and GC inside the same expression because order of subexpression evalution are not defined.

Review URL: http://codereview.chromium.org/3398014

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5509 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoUnuse labels when bailing out of StubCompiler methods.
kaznacheev@chromium.org [Thu, 23 Sep 2010 08:06:19 +0000 (08:06 +0000)]
Unuse labels when bailing out of StubCompiler methods.

There are 3 methods where early return happen before the
miss label is bound. This is harmless in Release mode,
in Debug an assertion fails.

Review URL: http://codereview.chromium.org/3405022

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5508 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix a build for Win64 with VS2008.
antonm@chromium.org [Wed, 22 Sep 2010 13:57:00 +0000 (13:57 +0000)]
Fix a build for Win64 with VS2008.

It complaints of type conversions.

Review URL: http://codereview.chromium.org/3396015

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5507 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix Win64 build with VS2008.
antonm@chromium.org [Wed, 22 Sep 2010 13:24:14 +0000 (13:24 +0000)]
Fix Win64 build with VS2008.

strlen returns size_t and VS2008 is now pretty picky about that.

Review URL: http://codereview.chromium.org/3384019

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5506 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRegExp: Fix caching to correctly set lastIndex.
lrn@chromium.org [Wed, 22 Sep 2010 11:22:57 +0000 (11:22 +0000)]
RegExp: Fix caching to correctly set lastIndex.

BUG: 52801

Review URL: http://codereview.chromium.org/3389022

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5504 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPrepare push to trunk. Now working on version 2.4.6.
fschneider@chromium.org [Wed, 22 Sep 2010 09:35:56 +0000 (09:35 +0000)]
Prepare push to trunk. Now working on version 2.4.6.

Review URL: http://codereview.chromium.org/3466009

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5501 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoTry fix win64 build.
vitalyr@chromium.org [Tue, 21 Sep 2010 13:04:42 +0000 (13:04 +0000)]
Try fix win64 build.

TBR=erik.corry

Review URL: http://codereview.chromium.org/3446015

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5500 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoCustom call IC for Math.floor.
vitalyr@chromium.org [Tue, 21 Sep 2010 12:54:12 +0000 (12:54 +0000)]
Custom call IC for Math.floor.

Review URL: http://codereview.chromium.org/3327022

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5499 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoUse //@ sourceURL when formatting stack trace
yurys@chromium.org [Mon, 20 Sep 2010 15:55:28 +0000 (15:55 +0000)]
Use //@ sourceURL when formatting stack trace

BUG=672
Review URL: http://codereview.chromium.org/3444011

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5498 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoGenerate inline code for contextual loads.
vitalyr@chromium.org [Mon, 20 Sep 2010 13:50:27 +0000 (13:50 +0000)]
Generate inline code for contextual loads.

Contextual load requires only a map check followed by a cell hole
check so we can generate pretty compact code for that. The fact that
we have inlined code is marked by mov ecx, offset instruction after
the IC call. Inlining is only enabled inside loops and in non-builtin
functions.

The generated code size increase is about 3%. This descreased the
pc-to-code cache hit rate in some of the benchmarks that trigger
GC. To compensate we now have 4 times as much entries in the cache.

Review URL: http://codereview.chromium.org/3402014

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5497 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix ProfileNodeFindOrAddChildForSameFunction test on Win32 debug.
mikhail.naganov@gmail.com [Mon, 20 Sep 2010 12:25:45 +0000 (12:25 +0000)]
Fix ProfileNodeFindOrAddChildForSameFunction test on Win32 debug.

Test was failing is due to a fact that MSVC in debug mode uses
separate instances for identical string literals.

TBR=sgjesse@chromium.org

Review URL: http://codereview.chromium.org/3381013

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5496 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix build-breakage on Mac. Initial variable eventhough it is initialized in all reach...
floitschV8@gmail.com [Mon, 20 Sep 2010 10:58:41 +0000 (10:58 +0000)]
Fix build-breakage on Mac. Initial variable eventhough it is initialized in all reachable branches.
Review URL: http://codereview.chromium.org/3471001

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5494 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoChanged the RegExp benchmark to exercise the regexp engine on different
sandholm@chromium.org [Mon, 20 Sep 2010 10:36:21 +0000 (10:36 +0000)]
Changed the RegExp benchmark to exercise the regexp engine on different
inputs by scrambling the input strings.
Review URL: http://codereview.chromium.org/3435012

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5493 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoShow RegExp calls in the profile.
mikhail.naganov@gmail.com [Mon, 20 Sep 2010 09:29:12 +0000 (09:29 +0000)]
Show RegExp calls in the profile.

It turns out they were filtered out. But when I unfiltered them, I
discovered another issue: when DevTools run, regexp literals get
recompiled each time they called (looks like this is concerned with
switching to full compiler), so I ended up having multiple entries for
the same regexp. To fix this, I changed the way of how code entries
equivalence is considered.

BUG=crbug/55999
TEST=cctest/test-profile-generator/ProfileNodeFindOrAddChildForSameFunction
(the test isn't for the whole issue, but rather for equivalence testing)

Review URL: http://codereview.chromium.org/3426008

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5492 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdded precision mode to fast-dtoa.
floitschV8@gmail.com [Mon, 20 Sep 2010 09:18:00 +0000 (09:18 +0000)]
Added precision mode to fast-dtoa.
Review URL: http://codereview.chromium.org/2000004

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5491 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRemove unused variable from Scanner class.
lrn@chromium.org [Mon, 20 Sep 2010 08:43:35 +0000 (08:43 +0000)]
Remove unused variable from Scanner class.

BUG=None
TEST=compiles

Patch from Thiago Farina <tfarina@chromium.org>

Review URL: http://codereview.chromium.org/3413015

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5490 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMake preparsing data reusable.
lrn@chromium.org [Fri, 17 Sep 2010 12:55:27 +0000 (12:55 +0000)]
Make preparsing data reusable.
Fix bug in writing symbol ids over 128.

Review URL: http://codereview.chromium.org/3416010

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5485 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd missing x87 instructions fisttp and fst to the disassembler.
fschneider@chromium.org [Fri, 17 Sep 2010 12:08:50 +0000 (12:08 +0000)]
Add missing x87 instructions fisttp and fst to the disassembler.

Review URL: http://codereview.chromium.org/3408011

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5484 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix direct loading of global function prototypes:
vitalyr@chromium.org [Fri, 17 Sep 2010 09:56:47 +0000 (09:56 +0000)]
Fix direct loading of global function prototypes:
We must also check the current context has not changed.

The bug reported by Florian.

Review URL: http://codereview.chromium.org/3410014

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5483 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMake some runtime arguments checks be RUNTIME_ASSERT, not ASSERT.
lrn@chromium.org [Fri, 17 Sep 2010 08:34:53 +0000 (08:34 +0000)]
Make some runtime arguments checks be RUNTIME_ASSERT, not ASSERT.

Review URL: http://codereview.chromium.org/3411013

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5482 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd breakOnCaughtException and breakOnUncaughtException flags
peter.rybin@gmail.com [Thu, 16 Sep 2010 21:40:42 +0000 (21:40 +0000)]
Add breakOnCaughtException and breakOnUncaughtException flags

Review URL: http://codereview.chromium.org/3275011

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5480 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPrevent inline constructor generation when duplicate properties are present in the...
kaznacheev@chromium.org [Thu, 16 Sep 2010 10:55:37 +0000 (10:55 +0000)]
Prevent inline constructor generation when duplicate properties are present in the constructor.

Currenly the constructor like this:
function f() {
  this.a = 0;
  this.a = 1;
  this.a = 2;
}
creates a map with duplicate desciptors which is bad in many ways.

Review URL: http://codereview.chromium.org/3434004

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5476 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix a regression in character-at stub when doing a keyed load on a string.
fschneider@chromium.org [Thu, 16 Sep 2010 09:18:08 +0000 (09:18 +0000)]
Fix a regression in character-at stub when doing a keyed load on a string.

Loading from out-of-range has to go to the runtime system to check if there
exists a property with that index in the prototype.

Review URL: http://codereview.chromium.org/3410011

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5471 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoClean up the insertion of nops (signalling non-inlined loads and
kasperl@chromium.org [Thu, 16 Sep 2010 09:10:34 +0000 (09:10 +0000)]
Clean up the insertion of nops (signalling non-inlined loads and
stores) in the full codegens.
Review URL: http://codereview.chromium.org/3431010

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5470 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoReally fix Mac build, and lint.
mikhail.naganov@gmail.com [Thu, 16 Sep 2010 08:51:13 +0000 (08:51 +0000)]
Really fix Mac build, and lint.

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5469 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix compilation issue on Mac
mikhail.naganov@gmail.com [Thu, 16 Sep 2010 08:39:27 +0000 (08:39 +0000)]
Fix compilation issue on Mac

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5468 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoEnhance SafeStackFrameIterator to avoid triggering assertions in debug mode.
mikhail.naganov@gmail.com [Thu, 16 Sep 2010 08:23:34 +0000 (08:23 +0000)]
Enhance SafeStackFrameIterator to avoid triggering assertions in debug mode.

When running profiling in debug mode, several assertions in frame
iterators that are undoubtedly useful when iterator is started from a
VM thread in a known "good" state, may fail when running over a stack
of a suspended VM thread. This patch makes SafeStackFrameIterator
to proactively check addresses and bail out from iteration early,
before an assertion will be triggered.

BUG=crbug/55565

Review URL: http://codereview.chromium.org/3436006

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5467 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoFix x64 compilation error. Patch from tfarina@chromium.org
mikhail.naganov@gmail.com [Thu, 16 Sep 2010 08:17:46 +0000 (08:17 +0000)]
Fix x64 compilation error. Patch from tfarina@chromium.org

http://codereview.chromium.org/3442004/show

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5466 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoX64: Templating Smi-macros to use both Label and NearLabel.
lrn@chromium.org [Thu, 16 Sep 2010 07:24:15 +0000 (07:24 +0000)]
X64: Templating Smi-macros to use both Label and NearLabel.
Added some more uses of NearLabel.

Review URL: http://codereview.chromium.org/3381005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5465 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMade the use of past tense more consistent in change log.
erik.corry@gmail.com [Wed, 15 Sep 2010 12:17:41 +0000 (12:17 +0000)]
Made the use of past tense more consistent in change log.
Review URL: http://codereview.chromium.org/3425005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5461 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd support for near labels.
ricow@chromium.org [Wed, 15 Sep 2010 11:43:12 +0000 (11:43 +0000)]
Add support for near labels.

This change introduces near labels in the assembler, allowing us to
uptimize forward jumps (conditional and unconditional) if we can
guarantee that the jump is witin range -128 to +127.

I changed a large fractions of the existing Labels to NearLabels, and
left out cases where it was not immediately clear if it could be used
or not (not immediately clear means labels covering a large code
block, or used in function calls which we could potentially change to
accept near labels).

Review URL: http://codereview.chromium.org/3388004

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5460 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPrepare push to trunk. We are now working on version 2.4.5.
erik.corry@gmail.com [Wed, 15 Sep 2010 10:58:25 +0000 (10:58 +0000)]
Prepare push to trunk.  We are now working on version 2.4.5.
Review URL: http://codereview.chromium.org/3429006

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5459 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMade predata smaller by storing symbol data in variable length base-128.
lrn@chromium.org [Wed, 15 Sep 2010 10:54:35 +0000 (10:54 +0000)]
Made predata smaller by storing symbol data in variable length base-128.
Remove position from symbol data - they must come in the correct order anyway.

Review URL: http://codereview.chromium.org/3384003

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5458 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoReplace 2 ARM ldr instructions with one ldrd in the code generated
erik.corry@gmail.com [Wed, 15 Sep 2010 10:22:55 +0000 (10:22 +0000)]
Replace 2 ARM ldr instructions with one ldrd in the code generated
for a SubStringStub and StringCompareStub in the ARM backend.  This
is a commit of http://codereview.chromium.org/3341012 for Andreas
Anyuru.
Review URL: http://codereview.chromium.org/3387003

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5457 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMake the CompareStub and the UnaryOpStub accept smi inputs.
fschneider@chromium.org [Wed, 15 Sep 2010 10:14:25 +0000 (10:14 +0000)]
Make the CompareStub and the UnaryOpStub accept smi inputs.

The stubs get an additional flag for including the smi code
inside the stub. This allows us to generate more compact code
if we don't want to inline the smi case outside the stub.

Review URL: http://codereview.chromium.org/3388005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5456 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAttempt to collect more garbage before panicking with out of memory.
antonm@chromium.org [Tue, 14 Sep 2010 17:48:56 +0000 (17:48 +0000)]
Attempt to collect more garbage before panicking with out of memory.

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

14 years agoAdd a few missing nops to signal no inlined code was generated.
vitalyr@chromium.org [Tue, 14 Sep 2010 16:38:02 +0000 (16:38 +0000)]
Add a few missing nops to signal no inlined code was generated.

Review URL: http://codereview.chromium.org/3399005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5454 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAllow List::sort, with an integer comparison function, to sort 64-bit pointers in...
whesse@chromium.org [Tue, 14 Sep 2010 15:16:32 +0000 (15:16 +0000)]
Allow List::sort, with an integer comparison function, to sort 64-bit pointers in profile-generator.  Change a static const int member to be declared and defined only inside the class declaration in class Runtime.
Review URL: http://codereview.chromium.org/3424002

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5453 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoAdd support for abortion in v8::OutputStream.
mikhail.naganov@gmail.com [Tue, 14 Sep 2010 15:11:24 +0000 (15:11 +0000)]
Add support for abortion in v8::OutputStream.

It's a good idea to allow receiver to interrupt data transmission.

Review URL: http://codereview.chromium.org/3409002

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5452 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoMove inlined function declarations and support from codegen.* to runtime.*.
whesse@chromium.org [Tue, 14 Sep 2010 14:52:53 +0000 (14:52 +0000)]
Move inlined function declarations and support from codegen.* to runtime.*.
Review URL: http://codereview.chromium.org/3293002

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5451 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoImplement heap snapshots serialization into JSON. API is designed
mikhail.naganov@gmail.com [Tue, 14 Sep 2010 11:49:06 +0000 (11:49 +0000)]
Implement heap snapshots serialization into JSON. API is designed
to avoid storing serialized snapshot on VM, instead it is emitted
using output stream interface.

The size of JSON emitted is roughly equal to used heap size
(when stored as an ASCII string).

Now a whole heap snapshot can be serialized and transmitted outside
VM. This makes possible:
  - implementing non-async UI for heap snapshots inspection;
  - storing heap snapshots for further inspection;
  - remote profiling (we can even implement a snapshotting mode
    where a snapshot isn't even stored in VM, only transmitted --
    good for mobile devices);
  - creating tools for outside heap snapshots processing, e.g.
    converting to HPROF.

Review URL: http://codereview.chromium.org/3311028

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5450 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoRecalibrate benchmarks.
sandholm@chromium.org [Mon, 13 Sep 2010 13:55:26 +0000 (13:55 +0000)]
Recalibrate benchmarks.
Review URL: http://codereview.chromium.org/3300031

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5449 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

14 years agoPrepare push to trunk. We are now working on version 2.4.4.
sgjesse@chromium.org [Mon, 13 Sep 2010 09:10:47 +0000 (09:10 +0000)]
Prepare push to trunk. We are now working on version 2.4.4.
Review URL: http://codereview.chromium.org/3308028

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5446 ce2b1a6d-e550-0410-aec6-3dcde31c8c00