Fix auto roll update.

Auto roll requires an additional "git fetch" now that it uses
a workdir. Before, the checkout used to be updated
externally.

BUG=408523
LOG=n
TEST=script_test.py
R=tandrii@chromium.org
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/695323002

Cr-Commit-Position: refs/heads/master@{#25079}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25079 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Introduce separate SelectLowering reducer.

Split lowering of Select nodes into a separate graph reducer and be more
clever when lowering to diamonds, i.e. reuse diamonds that have the same
condition and only add more phis to it.

TEST=unittests
R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/702463002

Cr-Commit-Position: refs/heads/master@{#25078}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25078 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix uninitialized fields in the BinaryOperation ast node.

R=jkummerow@chromium.org, svenpanne@chromium.org
BUG=chromium:429194
LOG=n

Review URL: https://codereview.chromium.org/701533002

Cr-Commit-Position: refs/heads/master@{#25077}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25077 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Don't double-check elements in the prototype chain in array builtins

BUG=
R=mvstanton@chromium.org

Review URL: https://codereview.chromium.org/701513002

Cr-Commit-Position: refs/heads/master@{#25076}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25076 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Fix Typer::Rangify to handle integral bitset types.

R=bmeurer@chromium.org
BUG=

Review URL: https://codereview.chromium.org/688413003

Cr-Commit-Position: refs/heads/master@{#25074}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25074 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

build fix after 25072

TBR=bmeurer@chromium.org

BUG=

Review URL: https://codereview.chromium.org/688323005

Cr-Commit-Position: refs/heads/master@{#25073}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25073 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

build fix after r25071

TBR=bmeurer@chromium.org

BUG=

Review URL: https://codereview.chromium.org/684523003

Cr-Commit-Position: refs/heads/master@{#25072}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25072 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] initial framework for unittesting of register allocator

BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/683133005

Cr-Commit-Position: refs/heads/master@{#25071}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25071 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix svn commit for deleted files in release scripts.

BUG=408523
LOG=n
TEST=script_test.py
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/685623004

Cr-Commit-Position: refs/heads/master@{#25070}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25070 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "[turbofan] Fix bug in Rangify."

This reverts commit r25067 for breaking Win64.

TBR=jarin@chromium.org

Review URL: https://codereview.chromium.org/697093002

Cr-Commit-Position: refs/heads/master@{#25069}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25069 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Fix bug in Rangify.

R=jarin@chromium.org

Review URL: https://codereview.chromium.org/699603002

Cr-Commit-Position: refs/heads/master@{#25067}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25067 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Enforce allocation of register for imull/mull.

R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/697073002

Cr-Commit-Position: refs/heads/master@{#25066}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25066 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Positive dividends are heavily favored for Int32Mod.

Mark negative dividend case as deferred, as modulus by power of two with
negative dividend almost never appears in practice.

R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/699573002

Cr-Commit-Position: refs/heads/master@{#25065}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25065 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Make generic algorithm less generic.

R=jarin@chromium.org

Review URL: https://codereview.chromium.org/701473002

Cr-Commit-Position: refs/heads/master@{#25064}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25064 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Select tbz/tbnz when possible on ARM64.

R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/697653002

Cr-Commit-Position: refs/heads/master@{#25063}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25063 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Now with more checkings! Skip the CallFunctionStub when the callee function can be statically determined.

R=mstarzinger@chromium.org
BUG=

Review URL: https://codereview.chromium.org/696223002

Cr-Commit-Position: refs/heads/master@{#25062}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25062 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Also optimize unsigned division by constant.

TEST=cctest,mjsunit,unittests
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/697663003

Cr-Commit-Position: refs/heads/master@{#25061}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25061 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix constant in Math.tan implementation.

R=jkummerow@chromium.org, rtoy@chromium.org
BUG=chromium:427468
LOG=Y

Review URL: https://codereview.chromium.org/695263002

Cr-Commit-Position: refs/heads/master@{#25060}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25060 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix workdir feature for release scripts.

BUG=408523
LOG=n
TEST=script_test.py
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/700493002

Cr-Commit-Position: refs/heads/master@{#25058}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25058 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix for bug 429168, PdfJs regression. We pay a very high cost for AllocationResult being a > kPointerSize struct. This can be avoided by using Smis to indicate failure with retry spaces.

BUG=429168
LOG=N
R=yangguo@chromium.org

Review URL: https://codereview.chromium.org/699473002

Cr-Commit-Position: refs/heads/master@{#25057}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25057 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add workdir feature to release scripts.

BUG=408523
LOG=n
TEST=script_test.py
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/701463002

Cr-Commit-Position: refs/heads/master@{#25056}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25056 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

More retries when tagging in release scripts.

BUG=408523
LOG=n
TEST=script_test.py
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/697433003

Cr-Commit-Position: refs/heads/master@{#25055}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25055 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[x86] Fix register constraints for multiply high and modulus.

R=jarin@chromium.org
TEST=mjsunit/compiler/regress-register-allocator2

Review URL: https://codereview.chromium.org/697053002

Cr-Commit-Position: refs/heads/master@{#25054}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25054 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Do not use the generic graph algorithm for widening in the typer.

This change uses an explicit queue for type-widening instead of the
generic algorithm. The trouble with the generic algorithm was that it
called the visitor on the same phi many times in a row (and thus caused
unnecessary retyping). I also think that the queue-based fixpoint is
more readable.

The CL cuts running time of the nbody-java benchmark from ~19s to ~15s,
the time spent in the typer goes from 4.5s to 1s. This is still a lot
- the root cause appears to be slow handling of union subtyping
(m*n for unions of sizes m and n). I see a re-typing of a
single phi node taking > 100ms. I will work on a fix with Andreas,
hopefully we can come up with some canonical representation
of unions at least for the common cases (union of Smi constants).

I have also changed the initial typer run to always compute a type, even
if we already had a type for the node. This fixes one assert failure
where context specialization updates a node without updating the type,
which confuses the typer when widening (as some types suddenly narrow).

BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/689403002

Cr-Commit-Position: refs/heads/master@{#25053}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25053 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

IA: Double arithmetic binops support memory operand

BUG=
R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/662813002

Patch from Weiliang Lin <weiliang.lin@intel.com>.

Cr-Commit-Position: refs/heads/master@{#25052}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25052 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix push script for autoroll account.

Push tags to the repo not to the remote - which is a cache
checkout on the bots.

Add new files when doing pure svn commits. Otherwise, new
files are silently ignored and not committed.

BUG=408523
LOG=n
TEST=script_test.py
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/699493004

Cr-Commit-Position: refs/heads/master@{#25049}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25049 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Teach push script to commit with roll account.

BUG=410721
LOG=n
TEST=script_test.py
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/693333002

Cr-Commit-Position: refs/heads/master@{#25046}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25046 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Avoid unnecessary (u)int32<->float64 changes in simplified lowering.

BUG=
R=bmeurer@chromium.org, titzer@chromium.org

Review URL: https://codereview.chromium.org/683873002

Cr-Commit-Position: refs/heads/master@{#25045}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25045 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "Skip the CallFunctionStub when the callee function can be statically determined."

This reverts commit 9845dfadd2ad4a83fa00bf43e0c2791ccf7eacad (r25042) for failing tests.

TBR=titzer@chromium.org
BUG=

Review URL: https://codereview.chromium.org/693103004

Cr-Commit-Position: refs/heads/master@{#25044}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25044 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS64: Fix simulator arguments handling on generated code entry.

Removed workaround for this issue.

TEST=mjsunit, cctest on sim64
BUG=
R=paul.lind@imgtec.com

Review URL: https://codereview.chromium.org/695933002

Cr-Commit-Position: refs/heads/master@{#25043}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25043 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip the CallFunctionStub when the callee function can be statically determined.

R=mstarzinger@chromium.org
BUG=

Review URL: https://codereview.chromium.org/694773002

Cr-Commit-Position: refs/heads/master@{#25042}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25042 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Minor cleanup about unscheduled use count for fixed nodes.

R=jarin@chromium.org

Review URL: https://codereview.chromium.org/696643002

Cr-Commit-Position: refs/heads/master@{#25041}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25041 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Changing the aging mechanism for script and eval caches.
Instead of using multiple generations for the code, first only store the hash that gets aged. Once a hash matched on a next probe, actually cache the code. Use regular code aging to remove entries from the cache.

BUG=
R=ulan@chromium.org

Review URL: https://codereview.chromium.org/675013004

Cr-Commit-Position: refs/heads/master@{#25040}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25040 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Break allocations in the code serializer into correct chunk sizes.

This change has been inspired by Slava Chigrin <vchigrin@yandex-team.ru> (https://codereview.chromium.org/689663002/)

R=mvstanton@chromium.org, vchigrin@yandex-team.ru

Review URL: https://codereview.chromium.org/686103004

Cr-Commit-Position: refs/heads/master@{#25039}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25039 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Properly handle stack overflows in the AST graph builder.

R=jarin@chromium.org
BUG=chromium:429159
TEST=mjsunit/regress/regress-crbug-429159
LOG=N

Review URL: https://codereview.chromium.org/697473006

Cr-Commit-Position: refs/heads/master@{#25037}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25037 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Upgrade test262-es6

TBR=machenbach@chromium.org
BUG=

Review URL: https://codereview.chromium.org/695763003

Cr-Commit-Position: refs/heads/master@{#25036}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25036 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip tests for mips.

TBR=paul.lind@imgtec.com

Review URL: https://codereview.chromium.org/698513003

Cr-Commit-Position: refs/heads/master@{#25035}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25035 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Switch auto push script to git.

BUG=410721
LOG=n
TEST=script_test.py
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/698513002

Cr-Commit-Position: refs/heads/master@{#25033}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25033 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Clear old backing store of WeakCollection on updates.

Not clearing can lead to a crash under following conditions:
1. Backing store of a weak map is allocated in large object space.
2. The backing store is marked incrementaly via the weak map.
3. The weak map is updated and gets a new backing store.
4. The store buffer overflows and marks the chunk of the old backing store as
"scan on scavenge."
5. Mark-compact collection kills some elements of the weak map. Note that the
old backing store survives because it was marked incrementally, but its dead
elements are not cleared.
6. Scavenger iterates over the old backing store, tries to move a dead object
and crashes.

BUG=v8:3631
LOG=N
TEST=cctest/test-heap/Regress3631
R=jkummerow@chromium.org

Review URL: https://codereview.chromium.org/686783003

Cr-Commit-Position: refs/heads/master@{#25032}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25032 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Scanner: remove PushBack calls when we're going to return ILLEGAL.

This simplifies escape handling and makes it easier to extend escapes for ES6.

PushBack just before detecting ILLEGAL is unnecessary, since we will abort the
scanning / parsing anyway at that point, and it doesn't matter where the cursor
exactly is. The error messages w/ PushBack are not any better or more correct
than without.

In addition: remove a comment about handling invalid escapes gracefully when we
no longer do. (*)

This CL includes a behavioral change: For input "var r = /foobar/g\urrrr;" we
used to report "unexpected_token: ILLEGAL" for "\u", but now we report
malformed_regexp_flags which is a more correct error message. (Note that the
code for reporting invalid_regexp_flags was dead, and invalid_regexp_flags is
not the right error message.)

Note that the V8 is more relaxed about unicode escapes in regexp flags than ES6
(see
http://people.mozilla.org/~jorendorff/es6-draft.html#sec-regular-expressions )
and this CL doesn't change it. (V8 accepts any \uxxxx, ES6 spec says only a
certain value range is acceptable.)

(*) Code archaeology:

Originally, doing PushBack in ScanHexEscape made sense (see e.g., here
https://codereview.chromium.org/5063003/diff/6001/src/prescanner.h ), since we
wouldn't return ILLEGAL but treat an invalid escape sequence "\uxxxx" as
"uxxxx".

(The repo at that point contains another instance of the same function, from the
initial commit. The logic is the same.)

This behavior was changed in a "renaming" commit
https://codereview.chromium.org/7739020.

BUG=
R=rossberg@chromium.org

Review URL: https://codereview.chromium.org/684873002

Cr-Commit-Position: refs/heads/master@{#25031}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25031 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

convert BitVector to use pointer size blocks

additionally rename data-flow.* to bit-vector.* as at some point these file became very inaccurately named

BUG=
R=svenpanne@chromium.org

Review URL: https://codereview.chromium.org/683243005

Cr-Commit-Position: refs/heads/master@{#25030}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25030 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Assert that unoptimized code does not embed context-specific objects.

R=verwaest@chromium.org

Review URL: https://codereview.chromium.org/690713003

Cr-Commit-Position: refs/heads/master@{#25029}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25029 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Ensure we don't try to inline raw access to indexed interceptor receivers.

BUG=chromium:419220
LOG=y
R=mvstanton@chromium.org

Review URL: https://codereview.chromium.org/697643002

Cr-Commit-Position: refs/heads/master@{#25028}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25028 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

X64: replace explicit zero-extend with runtime assert

Since the input float64 is in the range of [0, 2^32-1], the upper 32 bits of output register
should be zero.

BUG=
R=bmeurer@chromium.org, titzer@chromium.org

Review URL: https://codereview.chromium.org/644993002

Patch from Weiliang Lin <weiliang.lin@intel.com>.

Cr-Commit-Position: refs/heads/master@{#25027}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25027 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add turbocheck Makefile target.

The turbocheck target is similar to quickcheck, but runs only the
turbofan variant.

R=machenbach@chromium.org

Review URL: https://codereview.chromium.org/697573002

Cr-Commit-Position: refs/heads/master@{#25026}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25026 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Lower NumberModulus to Uint32Mod if both inputs are Unsigned32.

TEST=cctest/test-simplified-lowering
R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/685993003

Cr-Commit-Position: refs/heads/master@{#25025}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25025 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] First step towards correctified 64-bit addressing.

Also remove the LEA matching from x64, since it was never really
effective. We'll optimize that once we're correct.

TEST=cctest,unittests
R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/652363006

Cr-Commit-Position: refs/heads/master@{#25024}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25024 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix bug in optimization of Uint32LessThan.

R=jarin@chromium.org
BUG=

Review URL: https://codereview.chromium.org/689883003

Cr-Commit-Position: refs/heads/master@{#25023}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25023 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Make releases script more git friendly.

BUG=410721
LOG=n
TEST=script_test.py
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/682863006

Cr-Commit-Position: refs/heads/master@{#25022}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25022 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Introduce v8::Exception::GetMessage to find location of an error object.

API=v8::Exception::GetMessage
BUG=chromium:427954
R=yangguo@chromium.org
LOG=Y

Committed: https://code.google.com/p/v8/source/detail?r=25015

Review URL: https://codereview.chromium.org/687253002

Cr-Commit-Position: refs/heads/master@{#25021}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25021 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Do not compile with Turbofan if we cannot deopt for debugging.

R=jarin@chromium.org
BUG=v8:3660
LOG=N

Review URL: https://codereview.chromium.org/677843005

Cr-Commit-Position: refs/heads/master@{#25020}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25020 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Do not embed array objects in unoptimized code.

R=verwaest@chromium.org

Review URL: https://codereview.chromium.org/685393002

Cr-Commit-Position: refs/heads/master@{#25019}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25019 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add floor, ceil, round (truncate) instructions for ia32, x64 (if SSE4.1) and
add floor, ceil, round (truncate and away from zero) for arm64.

R=bmeurer@chromium.org, dcarney@chromium.org, mstarzinger@chromium.org, rodolph.perfetta@arm.com
TEST=test/mjsunit/asm/math-floor.js,test/mjsunit/asm/math-ceil.js,test/unittest/compiler/js-builtin-reducer-unittest.cc

Review URL: https://codereview.chromium.org/677433002

Cr-Commit-Position: refs/heads/master@{#25018}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25018 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Reverting r25015 and r25016 for broken build.

TBR=yangguo@chromium.org, machenbach@chromium.org
LOG=N

Review URL: https://codereview.chromium.org/688813002

Cr-Commit-Position: refs/heads/master@{#25017}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25017 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Build fix after r25015

TBR=yangguo@chromium.org, machenbach@chromium.org
LOG=N

Review URL: https://codereview.chromium.org/688803003

Cr-Commit-Position: refs/heads/master@{#25016}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25016 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Introduce v8::Exception::GetMessage to find location of an error object.

API=v8::Exception::GetMessage
BUG=chromium:427954
R=yangguo@chromium.org
LOG=Y

Review URL: https://codereview.chromium.org/687253002

Cr-Commit-Position: refs/heads/master@{#25015}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25015 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip tests with TF.

TBR=mstarzinger@chromium.org,bmeurer@chromium.org

Review URL: https://codereview.chromium.org/691623004

Cr-Commit-Position: refs/heads/master@{#25014}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25014 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add vrint{a,n,p,m,z} instructions to arm assembler. These instructions are only available on ARMv8.

R=rodolph.perfetta@gmail.com, ulan@chromium.org, bmeurer@chromium.org, rodolph.perfetta@arm.com

Review URL: https://codereview.chromium.org/682643002

Cr-Commit-Position: refs/heads/master@{#25013}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25013 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

remove a bunch of isolate::current in api.cc

R=svenpanne@chromium.org

BUG=

Review URL: https://codereview.chromium.org/686053003

Cr-Commit-Position: refs/heads/master@{#25011}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25011 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] move Node to vreg mapping to InstructionSelector

BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/683933004

Cr-Commit-Position: refs/heads/master@{#25010}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25010 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skipt tests with TF on windows.

TBR=bmeurer@chromium.org,mstarzinger@chromium.org

Review URL: https://codereview.chromium.org/692823002

Cr-Commit-Position: refs/heads/master@{#25009}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25009 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] add configuration parameters for register allocator

T=jarin@chromium.org

BUG=
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/688633002

Cr-Commit-Position: refs/heads/master@{#25008}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25008 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip test on win debug.

TBR=bmeurer@chromium.org
BUG=chromium:428315
LOG=n

Review URL: https://codereview.chromium.org/693473002

Cr-Commit-Position: refs/heads/master@{#25007}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25007 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Changed free(buffer) to delete [] buffer.

This bug (mismatch between new[] and free) was found by running VC++'s /analyze on all of Chrome.

BUG=chromium:427616
LOG=N
R=yangguo@chromium.org

Review URL: https://codereview.chromium.org/686193004

Patch from Bruce Dawson <brucedawson@chromium.org>.

Cr-Commit-Position: refs/heads/master@{#25003}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25003 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add test case for replacing turbofan code for debugging.

R=jarin@chromium.org
BUG=v8:3660
LOG=N

Review URL: https://codereview.chromium.org/665833003

Cr-Commit-Position: refs/heads/master@{#25002}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25002 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix assertion scope in Runtime_GetScript.

The HeapIterator implies DisallowHeapAllocation, but Script::GetWrapper
may allocate.

LOG=N
R=jkummerow@chromium.org
BUG=chromium:410033

Review URL: https://codereview.chromium.org/680283002

Cr-Commit-Position: refs/heads/master@{#25001}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25001 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

X87: EmitCreateIteratorResult loads map from function's context

port r24987

original commit message:

  EmitCreateIteratorResult loads map from function's context

  Caching or serialization can cause full-codegen output to be shared
  between contexts. CreateIteratorResult, however, was doing the wrong
  thing by creating results with the map that was current when the code
  was generated. Instead, we should chase pointers to load the right map
  from the function's context.

BUG=
R=weiliang.lin@intel.com

Review URL: https://codereview.chromium.org/674073004

Patch from Chunyang Dai <chunyang.dai@intel.com>.

Cr-Commit-Position: refs/heads/master@{#25000}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25000 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

X87: Classes: Add super support in methods and accessors

port r24976.

original commit message:

  Classes: Add super support in methods and accessors
  This is done by installing the [[HomeObject]] on the method and the
  accessor functions.

BUG=
R=weiliang.lin@intel.com

Review URL: https://codereview.chromium.org/687273005

Patch from Chunyang Dai <chunyang.dai@intel.com>.

Cr-Commit-Position: refs/heads/master@{#24999}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24999 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix lo_space initialization for external snapshot

R=rmcilroy@chromium.org

Review URL: https://codereview.chromium.org/684263002

Patch from André Baixo <baixo@google.com>.

Cr-Commit-Position: refs/heads/master@{#24998}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24998 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Fix input count in Uint32Mod/Div reduction.

BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/684193002

Cr-Commit-Position: refs/heads/master@{#24997}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24997 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Minor cleanup using BooleanConstant in graph builder.

R=titzer@chromium.org

Review URL: https://codereview.chromium.org/687103004

Cr-Commit-Position: refs/heads/master@{#24996}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24996 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Inline trivial OperatorProperties methods.

R=mstarzinger@chromium.org
BUG=

Review URL: https://codereview.chromium.org/686213002

Cr-Commit-Position: refs/heads/master@{#24995}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24995 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add a few missing overrides found by a new clang warning.

Namely, -Winconsistent-missing-override. No behavior change.

BUG=v8:3658
LOG=N
R=jkummerow@chromium.org

Review URL: https://codereview.chromium.org/688533002

Cr-Commit-Position: refs/heads/master@{#24994}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24994 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS: EmitCreateIteratorResult loads map from function's context.

Port r24987 (cfc4713)

Original commit message:
Caching or serialization can cause full-codegen output to be shared
between contexts.  CreateIteratorResult, however, was doing the wrong
thing by creating results with the map that was current when the code
was generated.  Instead, we should chase pointers to load the right map
from the function's context.

BUG=v8:3656
LOG=N
R=paul.lind@imgtec.com

Review URL: https://codereview.chromium.org/691563002

Cr-Commit-Position: refs/heads/master@{#24993}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24993 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Add deopt for %StringMatch

R=mstarzinger@chromium.org
BUG=

Review URL: https://codereview.chromium.org/639763004

Cr-Commit-Position: refs/heads/master@{#24992}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24992 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Cleanup unscheduled use count for controls of coupled nodes.

R=titzer@chromium.org, jarin@chromium.org

Review URL: https://codereview.chromium.org/687133003

Cr-Commit-Position: refs/heads/master@{#24991}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24991 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Various cleanup/simplification in object-observe.js

The biggest change is the removal of the map wrapper objects:
we now operate directly on the observation weak map, since there
are already Get/GetOrCreate/Set functions for each info map.
Various other small cleanups as well, including the deletion of
unnecessary forwarding functions and making use of standard macros.

This is a reland of r24972, retaining GetObservationStateJS() to
keep snapshotting working properly.

R=rossberg@chromium.org

Review URL: https://codereview.chromium.org/663253006

Cr-Commit-Position: refs/heads/master@{#24990}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24990 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Bug fix in arm64 ubfx selection.

R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/687193002

Cr-Commit-Position: refs/heads/master@{#24989}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24989 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS: Remove madd.d usage for r6.

TEST=/benchmarks/octane/crypto
BUG=
R=paul.lind@imgtec.com

Review URL: https://codereview.chromium.org/688693002

Cr-Commit-Position: refs/heads/master@{#24988}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24988 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

EmitCreateIteratorResult loads map from function's context

Caching or serialization can cause full-codegen output to be shared
between contexts.  CreateIteratorResult, however, was doing the wrong
thing by creating results with the map that was current when the code
was generated.  Instead, we should chase pointers to load the right map
from the function's context.

R=verwaest@chromium.org
BUG=v8:3656
LOG=N

Review URL: https://codereview.chromium.org/686063002

Cr-Commit-Position: refs/heads/master@{#24987}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24987 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Run ControlReducer early after graph building, then again later.

The justification for doing this is to reduce the size of the graph and
therefore speedup later phases of compilation. The control reducer also
obviates the need to run the PhiReducer, since it subsumes it.

R=mstarzinger@chromium.org
BUG=

Review URL: https://codereview.chromium.org/681263004

Cr-Commit-Position: refs/heads/master@{#24986}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24986 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip some tests with TF on gc stress.

TBR=mstarzinger@chromium.org

Review URL: https://codereview.chromium.org/652193007

Cr-Commit-Position: refs/heads/master@{#24985}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24985 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix presubmit.

TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/688013002

Cr-Commit-Position: refs/heads/master@{#24984}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24984 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Move input/output counts directly into Operators, simplying OperatorProperties.

This is a first step to refactoring OperatorProperties out of existence.
The next step is to inline OperatorProperties::GetXXXCount into the callers.

R=rossberg@chromium.org
BUG=

Review URL: https://codereview.chromium.org/680313003

Cr-Commit-Position: refs/heads/master@{#24983}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24983 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Switch push-to-trunk to git mode.

BUG=410721
LOG=n
TEST=script_test.py
R=tandrii@chromium.org
TBR=tandrii@chromium.org

Review URL: https://codereview.chromium.org/670263004

Cr-Commit-Position: refs/heads/master@{#24981}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24981 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Introduce new Select operator to improve bounds checking.

TEST=mjsunit,unittests
R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/691513002

Cr-Commit-Position: refs/heads/master@{#24980}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24980 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS: Classes: Add super support in methods and accessors.

Port r24976 (5cb743f)

Original commit message:
This is done by installing the [[HomeObject]] on the method and the
accessor functions.

BUG=v8:3330
LOG=Y
R=dusan.milosavljevic@imgtec.com

Review URL: https://codereview.chromium.org/691503002

Cr-Commit-Position: refs/heads/master@{#24979}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24979 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] cleanup register allocator interface a little

BUG=
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/671043004

Cr-Commit-Position: refs/heads/master@{#24978}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24978 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "Various cleanup/simplification in object-observe.js"

It broke the context snapshot by trying to get ahold of the
per-isolate observation_state object too early.

This reverts commit a137cb74fb8f89fabcb0396bebb0bdf248c4b0ae.

TBR=yangguo@chromium.org

Review URL: https://codereview.chromium.org/692473002

Cr-Commit-Position: refs/heads/master@{#24977}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24977 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Classes: Add super support in methods and accessors

This is done by installing the [[HomeObject]] on the method and the
accessor functions.

BUG=v8:3330
LOG=Y
R=dslomov@chromium.org

Review URL: https://codereview.chromium.org/683893002

Cr-Commit-Position: refs/heads/master@{#24976}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24976 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

harmony-scoping: Correct dynamic lookups on a top level.

R=rossberg@chromium.org

Review URL: https://codereview.chromium.org/650663003

Cr-Commit-Position: refs/heads/master@{#24975}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24975 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[arm64] Turn Word64And into tst instruction when possible.

Also add corresponding unit tests.

R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/684813004

Cr-Commit-Position: refs/heads/master@{#24974}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24974 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

allow disabling of ArrayBuffer neutering

BUG=
R=dslomov@chromium.org

Review URL: https://codereview.chromium.org/684103002

Cr-Commit-Position: refs/heads/master@{#24973}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24973 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Various cleanup/simplification in object-observe.js

The biggest change is the removal of the map wrapper objects: we now operate
directly on the observation weak map, since there are already Get/GetOrCreate/Set
functions for each info map. Various other small cleanups as well, including the
deletion of unnecessary forwarding functions and making use of standard macros.

R=arv@chromium.org, rossberg@chromium.org

Review URL: https://codereview.chromium.org/686773002

Cr-Commit-Position: refs/heads/master@{#24972}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24972 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Eager compile for debugging when debug is active.

Scenario:
On reload, Devtools restores previously set break points. At this point,
since Devtools is already opened, all code is compiled with debug break
slots. No break points exist yet, so we would lazily compile inner
functions, even ones that cannot be compiled lazily without a context.
So when we use Debug::FindSharedFunctionInfoInScript to find the break
positions, those lazily compiled functions are skipped.

By eagerly compiling when debug is active, we make sure that whenever
Devtools is open, functions that cannot be compiled lazily without a
context are always compiled with its outer function.

R=ulan@chromium.org
BUG=chromium:424142
LOG=N

Review URL: https://codereview.chromium.org/683413003

Cr-Commit-Position: refs/heads/master@{#24971}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24971 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Reland "In PrepareForBreakPoints, also purge shared function info not referenced by functions."

BUG=chromium:424142
LOG=N
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/692453002

Cr-Commit-Position: refs/heads/master@{#24970}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24970 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add performance tests for Map/Set with String and Object keys

The big change here is to split out each key type into its own
benchmark suite, since optimizations for different key types
are likely to be disjoint. ForEach tests have also been split
into separate "Iteration" suites, again with the thought that
iteration performance is likely separable from key type.

As part of adding the new tests, rejiggered the existing tests:
  - Map.set and Set.add (along with their Weak variants) no longer
    include construction costs in their scores
  - Moved key logic into the shared common.js

R=arv@chromium.org, machenbach@chromium.org

Review URL: https://codereview.chromium.org/685753004

Cr-Commit-Position: refs/heads/master@{#24969}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24969 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Switch scheduler to iterative floating control placement.

R=jarin@chromium.org

Review URL: https://codereview.chromium.org/685773002

Cr-Commit-Position: refs/heads/master@{#24968}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24968 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Reland "X87: Classes: Add basic support for properties."

TBR=machenbach@chromium.org

Review URL: https://codereview.chromium.org/687163003

Cr-Commit-Position: refs/heads/master@{#24967}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24967 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "In PrepareForBreakPoints, also purge shared function info not referenced by functions."

This reverts commit r24964.

TBR=machenbach@chromium.org

Review URL: https://codereview.chromium.org/687163002

Cr-Commit-Position: refs/heads/master@{#24966}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24966 ce2b1a6d-e550-0410-aec6-3dcde31c8c00