This moves the optimization for variables loads targeting lookup slots
in DYNAMIC_GLOBAL and DYNAMIC_LOCAL mode into the AstGraphBuilder. This
way we implicitly get all optimizations that target global loads and
context loads for free.
R=bmeurer@chromium.org
BUG=v8:4513
LOG=n
Review URL: https://codereview.chromium.org/1424943008
Cr-Commit-Position: refs/heads/master@{#31713}
This ports some code from chromium for using the bundled
toolchain.
BUG=chromium:548586
LOG=n
Review URL: https://codereview.chromium.org/1237803003
Cr-Commit-Position: refs/heads/master@{#31706}
port 4490ce8520 (r31701).
original commit message:
Original issue's description:
> [es6] Better support for built-ins subclassing.
>
> Create proper initial map for original constructor (new.target) instead of doing prototype
> transition on the base constructor's initial map. This approach fixes in-object slack tracking
> for subclass instances.
> This CL also fixes subclassing from String.
>
> BUG=v8:3101, v8:3330
> LOG=Y
>
> Committed: https://crrev.com/cd5f48302a502154a0106d12e3066bd563c6340c
> Cr-Commit-Position: refs/heads/master@{#31680}
It also fixes typed array map smashing done during typed array initialization.
BUG=
Review URL: https://codereview.chromium.org/1432483003
Cr-Commit-Position: refs/heads/master@{#31704}
The compiler can generate a named access for o[x] if x is a compile time
constant that can be turned into a name using ToName (limited to
primitive x values, because other ToName invocations might be observable),
or the KeyedLoadIC/KeyedStoreIC have gather constant name feedback for x
(i.e. the access always goes to the same symbol).
R=jarin@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1414013004
Cr-Commit-Position: refs/heads/master@{#31703}
Original issue's description:
> [es6] Better support for built-ins subclassing.
>
> Create proper initial map for original constructor (new.target) instead of doing prototype
> transition on the base constructor's initial map. This approach fixes in-object slack tracking
> for subclass instances.
> This CL also fixes subclassing from String.
>
> BUG=v8:3101, v8:3330
> LOG=Y
>
> Committed: https://crrev.com/cd5f48302a502154a0106d12e3066bd563c6340c
> Cr-Commit-Position: refs/heads/master@{#31680}
It also fixes typed array map smashing done during typed array initialization.
BUG=v8:3101, v8:3330, v8:4419
LOG=Y
Review URL: https://codereview.chromium.org/1413033006
Cr-Commit-Position: refs/heads/master@{#31701}
Return undefined for missing properties (or throw an exception in strong
mode). Also do a bit of code cleanup.
R=jarin@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1427913003
Cr-Commit-Position: refs/heads/master@{#31700}
This utility makes it possible to test TF graphs that accept parameters of any machine type (even int64 and float64), which are previously problematic due to the complexity of C calling conventions.
R=titzer@chromium.org
Review URL: https://codereview.chromium.org/1423133005
Cr-Commit-Position: refs/heads/master@{#31698}
Adds an optimization to emit JumpIfToBooleanTrue/False instead
of ToBoolean followed by JumpIfTrue/False if the value in the
accumulator is not boolean.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1426913002
Cr-Commit-Position: refs/heads/master@{#31697}
This introduces an AllocateMutableHeapNumberStub for the boxed double
field case, where we need to allocate a box in case of a transitioning
store first. We cannot use our inline allocations for this currently,
because mutable HeapNumber objects have certain alignment constraints,
and I don't want to mess up Allocate/AllocateInNewSpace eagerly.
Also refactor the PropertyAccessInfoFactory slightly to split the long
methods into simpler parts.
R=jarin@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1419173007
Cr-Commit-Position: refs/heads/master@{#31695}
This re-introduces the long lost concept of conditional builder methods
into the AstGraphBuilder that are allowed to return {nullptr} when a
certain optimization does not apply. This can be used to separate our
optimizations in the graph builder clearly from code required for
correctness.
R=bmeurer@chromium.org
BUG=v8:4513
LOG=n
Review URL: https://codereview.chromium.org/1414723004
Cr-Commit-Position: refs/heads/master@{#31694}
This makes FullCodeGenerator::VisitCall be independent of the target
architecture. Only the EmitPossiblyEvalCall case was dependent and is
hoisted into an emitter method.
R=rossberg@chromium.org
Review URL: https://codereview.chromium.org/1428953002
Cr-Commit-Position: refs/heads/master@{#31693}
Also changed the way that transitioning stores are represented in
a PropertyAccessInfo: There's no dedicated kind, but DataFields
have an optional transition map.
R=jarin@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1416973014
Cr-Commit-Position: refs/heads/master@{#31692}
If the CallIC collected a known target function for a callsite, add
a runtime check to ensure that the feedback remains the same and
specialize the JSCallFunction node to the known target function so that
inlining and typed lowering can pick up the feedback.
R=mstarzinger@chromium.org
BUG=v8:4470, v8:4493
LOG=n
Review URL: https://codereview.chromium.org/1428923002
Cr-Commit-Position: refs/heads/master@{#31689}
Adds support for switch statments to the interpreter.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1415093006
Cr-Commit-Position: refs/heads/master@{#31687}
This is currently hardcoded in the infra-side test runner
setup, but will be removed there, so that swarming triggers
don't need a custom environment.
BUG=chromium:535160
LOG=n
Review URL: https://codereview.chromium.org/1411733008
Cr-Commit-Position: refs/heads/master@{#31682}
This moves the cctest file for the interpreter to live in the same
namespace as the components it is testing. Hence we can avoid the
forbidden using directives pulling in entire namespaces.
From the Google C++ style guide: "You may not use a using-directive to
make all names from a namespace available". This would be covered by
presubmit linter checks if build/namespaces were not blacklisted.
R=rmcilroy@chromium.org
Review URL: https://codereview.chromium.org/1410993009
Cr-Commit-Position: refs/heads/master@{#31681}
Create proper initial map for original constructor (new.target) instead of doing prototype transition on the base constructor's initial map. This approach fixes in-object slack tracking for subclass instances.
This CL also fixes subclassing from String.
BUG=v8:3101, v8:3330
LOG=Y
Review URL: https://codereview.chromium.org/1427483002
Cr-Commit-Position: refs/heads/master@{#31680}
Reason for revert:
[Sheriff] Causes layout test failures.
Original issue's description:
> Remove RegExp.multiline accessors.
>
> This is non-standard and not even documented on MDN.
>
> On Firefox, setting RegExp.multiline to true adds the multiline flag to all
> newly created RegExp objects (both from constructor and from literal).
>
> In V8 this has no effect.
>
> Source archaelogy shows that this is from the initial commit.
>
> R=bmeurer@chromium.org
>
> Committed: https://crrev.com/e8f752ce0c2a488e88cd87fe75f3907b4303d0a0
> Cr-Commit-Position: refs/heads/master@{#31673}
TBR=bmeurer@chromium.org,yangguo@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1431433004
Cr-Commit-Position: refs/heads/master@{#31679}
This is in preparation of implementing Reflect.set.
Besides making SetSuperProperty and others return Maybe<bool>, this CL
also fixes some parts of my previous refactoring of SetProperty and
others: It doesn't make sense to take both a language_mode and a
should_throw argument. A strict language_mode should imply
THROW_ON_ERROR.
R=rossberg, verwaest@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1431443003
Cr-Commit-Position: refs/heads/master@{#31678}
This adds optimized lowering for JSConvertReceiver (in the general case)
and JSToObject in typed lowering. It also uses JSConvertReceiver for
direct calls in typed lowering.
R=mstarzinger@chromium.org
BUG=v8:4493
LOG=n
Review URL: https://codereview.chromium.org/1431543002
Cr-Commit-Position: refs/heads/master@{#31676}
Adds new Guard[Type] common operator, which takes value and control
inputs and records a guaranty that a certain value has a certain type
in that control path. This is some kind of ad-hoc SSI similar to what
we have to do in Crankshaft in some places.
Also introduces an ObjectIsNumber simplified operator, which checks
whether a certain value is a number (either a Smi or a HeapNumber).
This doesn't yet support transitioning stores to double fields, which
require support for allocating mutable heap numbers.
R=jarin@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1420283009
Cr-Commit-Position: refs/heads/master@{#31675}
This is non-standard and not even documented on MDN.
On Firefox, setting RegExp.multiline to true adds the multiline flag to all
newly created RegExp objects (both from constructor and from literal).
In V8 this has no effect.
Source archaelogy shows that this is from the initial commit.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1410993008
Cr-Commit-Position: refs/heads/master@{#31673}
This moves all cctest files for the compiler to live in the same
namespace as the components they are testing. Hence we can avoid the
forbidden using directives pulling in entire namespaces.
From the Google C++ style guide: "You may not use a using-directive to
make all names from a namespace available". This would be covered by
presubmit linter checks if build/namespaces were not blacklisted.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1424943004
Cr-Commit-Position: refs/heads/master@{#31671}
If you are not logged in with your Google account you are
faced with a blank screen. In order to make it more
clear for the Chromium sheriffs on how to close the
auto-roller I want to add this information.
R=machenbach@chromium.org
LOG=N
NOTRY=true
Review URL: https://codereview.chromium.org/1392823004
Cr-Commit-Position: refs/heads/master@{#31670}
In order to properly (lazy) bailout when converting the receiver for
sloppy mode functions (using the newly added JSConvertReceiver
operator), we need to have a bailout location right before every call
(also right before every %_Call and %_CallFunction), otherwise if the
JSConvertReceiver just reuses the lazy bailout frame state from the
JSCallFunction node, it will skip the whole function in case of lazy
bailout.
Note it should be impossible to trigger this currently because we do not
yet support AllocationSite code dependencies in TurboFan, which can
trigger this kind of lazy bailout; therefore it's not possible to write
a regression test (yet).
R=yangguo@chromium.org
BUG=v8:4493
LOG=n
Review URL: https://codereview.chromium.org/1425883004
Cr-Commit-Position: refs/heads/master@{#31668}
We have plans to create more ICs, and we are out of bits to represent the Kind
in the flags field of the code object. The InlineCacheState can lose a bit
because it no longer needs the DEFAULT state. That state existed as a way to
detect errors where code incorrectly looked at a vector IC stub's
InlineCacheState instead of correctly determining said state from a glance at
the vector. This really isn't a danger anymore.
So, with the horse trading, we could now represent up to 32 code kinds.
BUG=
Review URL: https://codereview.chromium.org/1427803003
Cr-Commit-Position: refs/heads/master@{#31666}
