More tests for toplevel lexical scope

R=dslomov@chromium.org
BUG=v8:2198
LOG=N

Review URL: https://codereview.chromium.org/716523002

Cr-Commit-Position: refs/heads/master@{#25240}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25240 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "[turbofan] Fix select lowering."

This reverts commit cf08e8d70ff0fa35dd315ddc430b743bb04c89cf for
breaking Embenchen bullet.

TBR=jarin@chromium.org

Review URL: https://codereview.chromium.org/684413003

Cr-Commit-Position: refs/heads/master@{#25238}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25238 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Upgrade version and change log after branch.

R=jkummerow@chromium.org
TBR=jkummerow@chromium.org

Review URL: https://codereview.chromium.org/681113011

Cr-Commit-Position: refs/heads/master@{#25237}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25237 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Fix select lowering.

Select lowering must not merge Select nodes that depend on each other,
because the resulting graph is not schedulable.

TEST=unittests
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/717473002

Cr-Commit-Position: refs/heads/master@{#25236}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25236 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add expected outcomes to json test output.

BUG=
R=tandrii@chromium.org

Review URL: https://codereview.chromium.org/713463002

Cr-Commit-Position: refs/heads/master@{#25235}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25235 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "Fix stepping in for-loops."

TBR=mvstanton@chromium.org

Review URL: https://codereview.chromium.org/713813002

Cr-Commit-Position: refs/heads/master@{#25233}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25233 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix stepping in for-loops.

R=ulan@chromium.org
BUG=v8:3634
LOG=N

Review URL: https://codereview.chromium.org/682413004

Cr-Commit-Position: refs/heads/master@{#25231}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25231 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

X87: Try avoiding MISS for callic monomorphic case.

port 6412c8ba70f2331aae0f21b968dd051b2320b947 (r25207)

original commit message:

  Try avoiding MISS for callic monomorphic case.

BUG=
R=weiliang.lin@intel.com

Review URL: https://codereview.chromium.org/713743003

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

Cr-Commit-Position: refs/heads/master@{#25230}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25230 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Resolve race condition with GC in CompilationCacheTable::Put.

The trouble is that the AsHandle method can allocate and thus
invoke GC, which could empty the table while we are updating it.

BUG=chromium:430880
LOG=N
R=verwaest@chromium.org

Review URL: https://codereview.chromium.org/700913003

Cr-Commit-Position: refs/heads/master@{#25229}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25229 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[arm] Recognize SXTB, SXTH, UXTB and UXTH.

TEST=cctest,msjunit/asm,unittests
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/709123005

Cr-Commit-Position: refs/heads/master@{#25228}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25228 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

X87: Turn ToNumberStub into a PlatformCodeStub again.

port b6737c56eaf51d63ab83b6529f657c15f834623c (r25187)

original commit message:

  Turn ToNumberStub into a PlatformCodeStub again.

  The HydrogenCodeStub is too expensive and there's no easy way to reduce
  this cost, so turning it into a PlatformCodeStub solves that problem until
  we can use TurboFan for code stubs.

BUG=
R=weiliang.lin@intel.com

Review URL: https://codereview.chromium.org/714573003

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

Cr-Commit-Position: refs/heads/master@{#25227}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25227 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

X87: fix one GC issue.

 The incorrect parameter number is set when recording safepoint.

BUG=
R=weiliang.lin@intel.com

Review URL: https://codereview.chromium.org/716463003

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

Cr-Commit-Position: refs/heads/master@{#25226}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25226 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Arm64: Remove forced csp alignment to 16 byte values for Nvidia chips.

Remove the forced alignment of csp to 16 byte values on Nvidia chips.
Benchmarks on current devices show that this is no longer required.

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

Review URL: https://codereview.chromium.org/710613002

Cr-Commit-Position: refs/heads/master@{#25225}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25225 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "Avoid some unnecessary fast-properties map creations."

This reverts commit e1f23eab4255d63344011dfb885b8e8962cb60e2.

Broke cctest/test-heap/ReleaseOverReservedPages on a bunch of builders

http://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20debug/builds/928/steps/Check/logs/ReleaseOverReservedPa..

BUG=None
LOG=N
TBR=jkummerow@chromium.org

Review URL: https://codereview.chromium.org/709123002

Cr-Commit-Position: refs/heads/master@{#25224}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25224 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Optimize add operations to use 'leal' instruction on x64

Add MemoryOperandMatcher that recognizes node clusters in the form
[%r1 + %r2*SCALE + OFFSET] and explicit support in the x64 Int32Add
selector to use it to translate complex adds to 'leal' instructions.

R=titzer@chromium.org

Review URL: https://codereview.chromium.org/704713003

Cr-Commit-Position: refs/heads/master@{#25223}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25223 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Classes: Partial fix for constructor not calling super

Introduce two new function kind, one for default constructor and one
for default constructor call super. Then when we are about to pares
these we just generate the correct AST in source.

BUG=v8:3661, v8:3672
LOG=Y
R=dslomov@chromium.org

Review URL: https://codereview.chromium.org/700523003

Cr-Commit-Position: refs/heads/master@{#25222}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25222 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Avoid some unnecessary fast-properties map creations.

(1) When we have just normalized and re-fastified a map, we don't need to copy it again to set the is_prototype bit.
(2) When defining accessors causes a non-prototype object to go slow, don't force re-fastification.

BUG=v8:3267
LOG=n
R=verwaest@chromium.org

Review URL: https://codereview.chromium.org/706243002

Cr-Commit-Position: refs/heads/master@{#25221}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25221 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

harmony_scoping: Implement lexical bindings at top level

This implements correct semantics for "extensible" top level lexical scope.
The entire lexical scope is represented at runtime by GlobalContextTable, reachable from native context and accumulating global contexts from every script loaded into the context.

When the new script starts executing, it does the following validation:
- checks the GlobalContextTable and global object (non-configurable own) properties against the set of declarations it introduces and reports potential conflicts.
- invalidates the conflicting PropertyCells on global object, so that any code depending on them will miss/deopt causing any contextual lookups to be reexecuted under the new bindings
- adds the lexical bindings it introduces to the GlobalContextTable

Loads and stores for contextual lookups are modified so that they check the GlobalContextTable before looking up properties on global object, thus implementing the shadowing of global object properties by lexical declarations.

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

Review URL: https://codereview.chromium.org/705663004

Cr-Commit-Position: refs/heads/master@{#25220}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25220 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add FLAG_trace_maps

R=verwaest@chromium.org

Review URL: https://codereview.chromium.org/694533003

Cr-Commit-Position: refs/heads/master@{#25219}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25219 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS: Try avoiding MISS for callic monomorphic case.

Port c01061044028f7a31ba51496e44900481005eb38
Port r25207

BUG=
R=mvstanton@chromium.org

Review URL: https://codereview.chromium.org/684653007

Cr-Commit-Position: refs/heads/master@{#25218}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25218 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Make scheduler handle floating non-naked loops.

R=jarin@chromium.org
TEST=cctest/test-scheduler/NestedFloatingDiamondWithLoop

Review URL: https://codereview.chromium.org/706123003

Cr-Commit-Position: refs/heads/master@{#25217}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25217 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Teach TurboFan to call C functions with result_size > 1.

Happily, this fixes WIN64 failures too.

BUG=
R=titzer@chromium.org

Review URL: https://codereview.chromium.org/709033002

Cr-Commit-Position: refs/heads/master@{#25216}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25216 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Scanner: disallow unicode escapes in regexp flags.

The spec explicitly forbids them. V8 never handled them properly either, just
the Scanner accepted them (it had code to add them literally to the
LiteralBuffer) and later on, Regexp constructor disallowed them.

According to the spec, unicode escapes in regexp flags should be an early error
("It is a Syntax Error if IdentifierPart contains a Unicode escape sequence.").

Note that Scanner is still more relaxed about regexp flags than the
spec. Especially, it accepts any identifier parts (not just a small set of
letters) and doesn't check for duplicates.

R=rossberg@chromium.org

Review URL: https://codereview.chromium.org/700373003

Cr-Commit-Position: refs/heads/master@{#25215}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25215 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Fix deoptimization of uint8, uint16 inputs.

BUG=
R=mstarzinger@chromium.org

Review URL: https://codereview.chromium.org/707443003

Cr-Commit-Position: refs/heads/master@{#25214}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25214 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Remove heap size filter for context disposal garbage collection.

BUG=
R=ulan@chromium.org

Review URL: https://codereview.chromium.org/712563002

Cr-Commit-Position: refs/heads/master@{#25213}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25213 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Use current time to calculate context disposal rate.

BUG=
R=ulan@chromium.org

Review URL: https://codereview.chromium.org/679363003

Cr-Commit-Position: refs/heads/master@{#25212}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25212 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert r24922 "Windows: use SystemTimeToTzSpecificLocalTime instead of localtime_s."

SystemTimeToTzSpecificLocalTime doesn't work correctly inside Chrome sandbox.

BUG=chromium:417640
LOG=Y
R=yangguo@chromium.org

Review URL: https://codereview.chromium.org/710643002

Cr-Commit-Position: refs/heads/master@{#25211}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25211 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan][arm64] Optimize shifts combined with truncations or extensions.

R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/705983002

Cr-Commit-Position: refs/heads/master@{#25210}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25210 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[arm64] Bug fix in sapefoint recording.

cp is handled by the register allocator and doesn't necessarily hold the
context pointer on arm64.

R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/695323005

Cr-Commit-Position: refs/heads/master@{#25209}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25209 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Perform context disposal garbage collections only if contexts disposals happen at a high rate.

BUG=
R=ulan@chromium.org

Review URL: https://codereview.chromium.org/710603003

Cr-Commit-Position: refs/heads/master@{#25208}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25208 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Try avoiding MISS for callic monomorphic case.

BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/704143003

Cr-Commit-Position: refs/heads/master@{#25207}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25207 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Enable inlining tests that pass with new scheduler.

R=titzer@chromium.org
TEST=test-run-inlining/InlineTwiceDependentDiamond

Review URL: https://codereview.chromium.org/704293002

Cr-Commit-Position: refs/heads/master@{#25205}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25205 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Enable ES6 numeric literals by default

BUG=v8:2783
LOG=Y
R=dslomov@chromium.org

Review URL: https://codereview.chromium.org/703943002

Cr-Commit-Position: refs/heads/master@{#25204}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25204 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Disable ranging for modulus in hopes it will fix arm.

TBR=mstarzinger@chromium.org
BUG=

Review URL: https://codereview.chromium.org/705113002

Cr-Commit-Position: refs/heads/master@{#25203}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25203 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS: Turn ToNumberStub into a PlatformCodeStub again.

Port 724a15e8bf3fb8697a950c8e0f96158c8e56ec71
Port r25187

The HydrogenCodeStub is too expensive and there's no easy way to reduce
this cost, so turning it into a PlatformCodeStub solves that problem
until we can use TurboFan for code stubs.

TEST=mjsunit
BUG=
R=dusan.milosavljevic@imgtec.com

Review URL: https://codereview.chromium.org/707863002

Cr-Commit-Position: refs/heads/master@{#25202}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25202 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] move label generation to code generator

R=bmeurer@chromium.org

BUG=

Review URL: https://codereview.chromium.org/707803002

Cr-Commit-Position: refs/heads/master@{#25201}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25201 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Allocate a small amount of slack space to graph builder environments.

R=mstarzinger@chromium.org
BUG=

Review URL: https://codereview.chromium.org/709463003

Cr-Commit-Position: refs/heads/master@{#25200}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25200 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Lower asm.js truncations to JSToNumber.

- JSBitwiseOr(x, 0) => Word32Or(NumberToInt32(JSToNumber(x)), 0)
- JSMultiply(x, 1) => NumberMultiply(JSToNumber(x), 1)

R=jarin@chromium.org

Review URL: https://codereview.chromium.org/706863003

Cr-Commit-Position: refs/heads/master@{#25198}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25198 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix O(n^2) successor traversal in ControlReducer.

R=mstarzinger@chromium.org
BUG=

Review URL: https://codereview.chromium.org/703163002

Cr-Commit-Position: refs/heads/master@{#25197}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25197 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Combine JSToNumber and ChangeTaggedToFloat64 for fast SMI case.

This does the following optimization during ChangeLowering:

 ChangeTaggedToFloat64(JSToNumber(x)) =>
   if IsSmi(x) then ChangeSmiToFloat64(x)
   else let y = JSToNumber(x) in
     if IsSmi(y) then ChangeSmiToFloat64(y)
     else LoadHeapNumberValue(y)

Also cleanup/rearrange some code duplication in ChangeLowering.

R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/703103003

Cr-Commit-Position: refs/heads/master@{#25195}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25195 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "Remove workaround for successors on end block from scheduler."

TBR=titzer@chromium.org

Review URL: https://codereview.chromium.org/705043002

Cr-Commit-Position: refs/heads/master@{#25194}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25194 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] (reland) Compute tighter ranges for modulus in Typer.

R=jarin@chromium.org
BUG=

Review URL: https://codereview.chromium.org/694703004

Cr-Commit-Position: refs/heads/master@{#25193}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25193 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix marking time in cumulative gc stats.

BUG=
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/708483004

Cr-Commit-Position: refs/heads/master@{#25192}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25192 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] phis cannot take registers as inputs

BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/704153002

Cr-Commit-Position: refs/heads/master@{#25191}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25191 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Allow JSArray fast moving elements even if the array's proto isn't Array.prototype in original state
Otherwise array builtins don't work on internal arrays.

BUG=v8:3681
LOG=n
R=machenbach@chromium.org, mvstanton@chromium.org

Review URL: https://codereview.chromium.org/706703005

Cr-Commit-Position: refs/heads/master@{#25190}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25190 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Push TruncateFloat64ToInt32 into phis.

TEST=unittests
R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/707683003

Cr-Commit-Position: refs/heads/master@{#25189}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25189 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Remove workaround for successors on end block from scheduler.

R=titzer@chromium.org
TEST=cctest/test-scheduler/RPOLoop

Review URL: https://codereview.chromium.org/708763002

Cr-Commit-Position: refs/heads/master@{#25188}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25188 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Turn ToNumberStub into a PlatformCodeStub again.

The HydrogenCodeStub is too expensive and there's no easy way to reduce
this cost, so turning it into a PlatformCodeStub solves that problem
until we can use TurboFan for code stubs.

TEST=mjsunit
R=mvstanton@chromium.org

Review URL: https://codereview.chromium.org/683913008

Cr-Commit-Position: refs/heads/master@{#25187}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25187 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip json2 on 64 bits debug.

BUG=3681
LOG=n
TBR=mvstanton@chromium.org

Review URL: https://codereview.chromium.org/701353003

Cr-Commit-Position: refs/heads/master@{#25186}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25186 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix for an assertion failure in Map::FindTransitionToField(...). Appeared after r25136.

BUG=chromium:430846
LOG=N
R=verwaest@chromium.org

Review URL: https://codereview.chromium.org/704183002

Cr-Commit-Position: refs/heads/master@{#25185}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25185 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Reuse RPO traversal stack in the scheduler.

R=jarin@chromium.org

Review URL: https://codereview.chromium.org/695303003

Cr-Commit-Position: refs/heads/master@{#25184}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25184 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip tests that timeout on arm64

TBR=machenbach@chromium.org

Review URL: https://codereview.chromium.org/709443002

Cr-Commit-Position: refs/heads/master@{#25183}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25183 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

The idea behind of this solution is to use the existing "relocation info" instead of consumption the CodeLinePosition events emitted by the V8 compilers.
During generation code and relocation info are generated simultaneously.
When code generation is done you each code object has associated "relocation info".
Relocation information lets V8 to mark interesting places in the generated code: the pointers that might need to be relocated (after garbage collection),
correspondences between the machine program counter and source locations for stack walking.

This patch:
1. Add more source positions info in reloc info to make it suitable for source level mapping.
The amount of data should not be increased dramatically because (1) V8 already marks interesting places in the generated code and
(2) V8 does not write redundant information (it writes a pair (pc_offset, pos) only if pos is changed and skips other).
I measured it on Octane benchmark - for unoptimized code the number of source positions may achieve 2x ('lin_solve' from NavierStokes benchmark).

2. When a sample happens, CPU profiler finds a code object by pc, then use its reloc info to match the sample to a source line.
If a source line is found that hit counter is increased by one for this line.

3. Add a new public V8 API to get the hit source lines by CDT CPU profiler.
Note that it's expected a minor patch in Blink to pack the source level info in JSON to be shown.

4.Add a test that checks how the samples are distributed through source lines.
It tests two cases: (1) relocation info created during code generation and (2) relocation info associated with precompiled function's version.

Patch from Denis Pravdin <denis.pravdin@intel.com>;

R=svenpanne@chromium.org, yurys@chromium.org

Review URL: https://codereview.chromium.org/682143003

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

Cr-Commit-Position: refs/heads/master@{#25182}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25182 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Avoid redundant work in scheduler loop header/depth calculation.

R=jarin@chromium.org

TEST=cctest/test-scheduler/LoopedFloatingDiamond2

Review URL: https://codereview.chromium.org/702683002

Cr-Commit-Position: refs/heads/master@{#25181}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25181 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Turn various diamonds into selects.

TEST=cctest/test-changes-lowering,mjsunit/asm,unittests
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/704463004

Cr-Commit-Position: refs/heads/master@{#25180}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25180 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Retry: Parser & internalization fix: ensure no heap allocs during GetString(Handle<String>).

The bug has always been there: when the parser is operating in the "immediately
internalize" mode and calls GetString, we get FlatContent of a string and then
do heap allocation.

The bug was uncovered by https://codereview.chromium.org/693803004/ (which put
the parser to the "immediately internalize" mode more often), but looking at the
code, it's possible that it can happen in other cases too.

This CL makes AstValueFactory handle this situation gracefully: it won't try to
internalize inside GetString(Handle<String>); it's unnecessary anyway since we
have the Handle<String> already.

BUG=
R=rossberg@chromium.org

Review URL: https://codereview.chromium.org/706533005

Cr-Commit-Position: refs/heads/master@{#25179}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25179 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] extend register allocator testing with control flow

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

BUG=

Review URL: https://codereview.chromium.org/699083003

Cr-Commit-Position: refs/heads/master@{#25178}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25178 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS: Skip embenchen tests on big-endian.

Emscripten requires little-endian arch. It has assertion for endianness:

assert(HEAPU8[0] === 255 && HEAPU8[3] === 0,
    'Typed arrays 2 must be run on a little-endian system');

TEST=mjsunit/asm/embenchen/*
BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/708663003

Patch from Paul Lind <paul.lind@imgtec.com>.

Cr-Commit-Position: refs/heads/master@{#25177}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25177 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Transform x * -1.0 to -0.0 - x.

TEST=msjunit/asm,unittests
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/683753004

Cr-Commit-Position: refs/heads/master@{#25176}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25176 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS: Fix uint32div bug after r25103.

The div and mod instructions on MIPS are not safe for rhs 0.

TEST=test/mjsunit/asm/uint32div.js
BUG=
R=akos.palfi@imgtec.com

Review URL: https://codereview.chromium.org/704013002

Cr-Commit-Position: refs/heads/master@{#25174}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25174 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Whitespace change to trigger bots (5)

Cr-Commit-Position: refs/heads/master@{#25173}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25173 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Whitespace change to trigger bots (4)

Cr-Commit-Position: refs/heads/master@{#25172}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25172 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Whitespace change to trigger bots (3)

Cr-Commit-Position: refs/heads/master@{#25171}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25171 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Whitespace change to trigger bots (2)

Cr-Commit-Position: refs/heads/master@{#25170}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25170 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Whitespace change to trigger bots (1)

Cr-Commit-Position: refs/heads/master@{#25169}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25169 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

MIPS: Fix instruction selection test expectations after r25120.

TEST=unittests/InstructionSelectorCmpTest.Parameter,
     InstructionSelectorTest.Word32EqualWithZero
BUG=
R=paul.lind@imgtec.com

Review URL: https://codereview.chromium.org/706633003

Cr-Commit-Position: refs/heads/master@{#25167}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25167 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip tests that timout on arm64

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

Review URL: https://codereview.chromium.org/702163002

Cr-Commit-Position: refs/heads/master@{#25166}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25166 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix printing and verification of RPO computation.

R=jarin@chromium.org

Review URL: https://codereview.chromium.org/700153002

Cr-Commit-Position: refs/heads/master@{#25165}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25165 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "[turbofan] Compute tighter ranges for modulus in Typer."

This reverts commit 482586747ca6bc0c94cc23f916fd190931d3a836.

TBR=jarin@chromium.org
BUG=

Review URL: https://codereview.chromium.org/702893002

Cr-Commit-Position: refs/heads/master@{#25164}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25164 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

fix assert in arm64 gap resolver

BUG=
R=ulan@chromium.org

Review URL: https://codereview.chromium.org/708473002

Cr-Commit-Position: refs/heads/master@{#25163}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25163 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Compute tighter ranges for modulus in Typer.

R=jarin@chromium.org
BUG=

Review URL: https://codereview.chromium.org/689133003

Cr-Commit-Position: refs/heads/master@{#25162}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25162 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip test in release mode for arm64

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

Review URL: https://codereview.chromium.org/701013006

Cr-Commit-Position: refs/heads/master@{#25161}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25161 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Make slow tests on arm64

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

Review URL: https://codereview.chromium.org/701023003

Cr-Commit-Position: refs/heads/master@{#25160}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25160 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "Parser & internalization fix: ensure no heap allocs during GetString(Handle<String>)."

This reverts r25155.

Reason: breaks webkit unit tests.

BUG=
TBR=marja@chromium.org

Review URL: https://codereview.chromium.org/707483002

Cr-Commit-Position: refs/heads/master@{#25159}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25159 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Parser & internalization fix: ensure no heap allocs during GetString(Handle<String>).

The bug has always been there: when the parser is operating in the "immediately
internalize" mode and calls GetString, we get FlatContent of a string and then
do heap allocation.

The bug was uncovered by https://codereview.chromium.org/693803004/ (which put
the parser to the "immediately internalize" mode more often), but looking at the
code, it's possible that it can happen in other cases too.

This CL makes AstValueFactory handle this situation gracefully: it won't try to
internalize inside GetString(Handle<String>); it's unnecessary anyway since we
have the Handle<String> already.

R=rossberg@chromium.org
BUG=

Review URL: https://codereview.chromium.org/699343004

Cr-Commit-Position: refs/heads/master@{#25155}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25155 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Make merge_to_branch more git-friendly.

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

Review URL: https://codereview.chromium.org/702843002

Cr-Commit-Position: refs/heads/master@{#25154}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25154 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

HPhis with HParameter inputs must have Tagged representation

BUG=v8:3670
LOG=n
R=verwaest@chromium.org

Review URL: https://codereview.chromium.org/702053002

Cr-Commit-Position: refs/heads/master@{#25153}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25153 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Skip box2d for gc stress.

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

Review URL: https://codereview.chromium.org/696213003

Cr-Commit-Position: refs/heads/master@{#25152}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25152 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Fix deopt for assignments in non-effect context.

BUG=
R=mstarzinger@chromium.org

Review URL: https://codereview.chromium.org/701853002

Cr-Commit-Position: refs/heads/master@{#25151}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25151 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix fast-variants feature in test driver.

BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/701933003

Cr-Commit-Position: refs/heads/master@{#25150}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25150 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] optimize hot loop in ResolveControlFlow

R=jarin@chromium.org

BUG=

Review URL: https://codereview.chromium.org/694473002

Cr-Commit-Position: refs/heads/master@{#25149}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25149 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Replace C++ bitfields with our own BitFields

Shave this yak from orbit, it's the only way to be sure.

BUG=chromium:427616
LOG=n
R=svenpanne@chromium.org

Review URL: https://codereview.chromium.org/700963002

Cr-Commit-Position: refs/heads/master@{#25148}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25148 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Use compiler barrier instead of memory barrier for release/acquire atomic operations on mac.

BUG=
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/694703003

Cr-Commit-Position: refs/heads/master@{#25147}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25147 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

ARM: optimize inlined doubles

Use 'vmov Dn[x], ip' instead of 'vmov Sn, ip' to load double immediates.
This patch is only useful for Turbofan as Crankshaft loads double immediates from the constant pool.
This give a little improvement on asm.js benchmarks

R=bmeurer@chromium.org, ulan@chromium.org

BUG=

Review URL: https://codereview.chromium.org/704543002

Cr-Commit-Position: refs/heads/master@{#25146}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25146 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Revert "Reland "Optimize function across closures.""

This reverts commit r25142.

TBR=ishell@chromium.org

Review URL: https://codereview.chromium.org/702853002

Cr-Commit-Position: refs/heads/master@{#25145}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25145 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Use load/store indexes as 64-bit registers on arm64.

Since r25024 load/store indexes are 64-bit value but arm64 backend was still
treating them as 32-bit value.

R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/698383003

Cr-Commit-Position: refs/heads/master@{#25144}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25144 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Handle store buffer slot overwrite during object promotion.

The bad scenario this fix handles:

We have a slot in a free list, then promote the object pointed-to by
the slot during scavenge. When allocating the space for the promoted
object, we overwrite the slot with the free list entry map if the
object is allocated just before the slot. After the allocation,
ScavengingVisitor::PromoteObject overwrites the slot with the
address of the allocated object, thus corrupting the free list.

Unfortunately, we do not have a way to construct a reliable repro
case because we would need to somehow craft a free list and store
buffer slot to be in the right configuration.

R=hpayer@chromium.org
BUG=

Review URL: https://codereview.chromium.org/695213004

Cr-Commit-Position: refs/heads/master@{#25143}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25143 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Reland "Optimize function across closures."

R=mvstanton@chromium.org

Review URL: https://codereview.chromium.org/703603003

Cr-Commit-Position: refs/heads/master@{#25142}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25142 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Propagate "deferredness" to dominated basic blocks.

TEST=cctest/test-scheduler
R=mstarzinger@chromium.org

Review URL: https://codereview.chromium.org/686273005

Cr-Commit-Position: refs/heads/master@{#25141}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25141 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Don't use C++11's std::trunc and std::round, use the traditional C functions.

R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/703893002

Cr-Commit-Position: refs/heads/master@{#25140}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25140 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add fast-variants feature to test driver.

BUG=
R=bmeurer@chromium.org

Review URL: https://codereview.chromium.org/704723002

Cr-Commit-Position: refs/heads/master@{#25139}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25139 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Make special RPO computation iterative during scheduling.

This contains the following changes squashed together:
- Switch BasicBlock::loop_end to be a basic block instead of an RPO.
- Switch ScheduleLate to use dominator depth instead of RPO.
- Switch ScheduleEarly to use dominator depth instead of RPO.
- Push out absolute RPO ordering everywhere else in the scheduler.
- Keep linked list of blocks in RPO order while scheduling.
- Switch from RPO number to depth for dominator calculation.

R=jarin@chromium.org

Review URL: https://codereview.chromium.org/696363002

Cr-Commit-Position: refs/heads/master@{#25138}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25138 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

TransitionArray now uses <is_data_property, name, attributes> tuple as a key, which allows to have several entries for the same property name.

R=verwaest@chromium.org

Review URL: https://codereview.chromium.org/661133002

Cr-Commit-Position: refs/heads/master@{#25136}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25136 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Add test cases based on the programs from the Embenchen benchmark suite.

TEST=mjsunit/asm/embenchen
R=dcarney@chromium.org

Review URL: https://codereview.chromium.org/704653004

Cr-Commit-Position: refs/heads/master@{#25135}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25135 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

x32: disable turbofan for x32 port

BUG=
R=titzer@chromium.org

Review URL: https://codereview.chromium.org/697353003

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

Cr-Commit-Position: refs/heads/master@{#25133}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25133 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Fix enum in debug.h

TBR=ishell@chromium.org

Review URL: https://codereview.chromium.org/700103002

Cr-Commit-Position: refs/heads/master@{#25131}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25131 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Introduce new stepping mode to step into another frame.

R=aandrey@chromium.org
BUG=chromium:267592
LOG=Y

Review URL: https://codereview.chromium.org/690263004

Cr-Commit-Position: refs/heads/master@{#25130}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25130 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Don't crash if the tm_zone field returned by localtime is NULL

BUG=chromium:158355
R=svenpanne@chromium.org
LOG=n

Review URL: https://codereview.chromium.org/662093003

Cr-Commit-Position: refs/heads/master@{#25129}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25129 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

[turbofan] Fix missing machine type for float32/float64 array accesses.

TEST=mjsunit/asm
R=svenpanne@chromium.org

Review URL: https://codereview.chromium.org/703883002

Cr-Commit-Position: refs/heads/master@{#25128}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25128 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

Truncate strings synchronized.

BUG=
R=jarin@chromium.org

Review URL: https://codereview.chromium.org/702743002

Cr-Commit-Position: refs/heads/master@{#25127}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25127 ce2b1a6d-e550-0410-aec6-3dcde31c8c00