Introduce a new TypedOptimization reducer that contains the type
based optimization reduction steps, which are not (directly)
related to lowering JavaScript operators based on types (which is
what JSTypedLowering is supposed to do).
This also addresses a chicken-and-egg problem that we see in the
Octane/Mandreel benchmark where type based constant folding isn't
applied to the numeric comparison operators introduced by the
JSTypedLowering itself, and thus gives up to 10% speedup for the
benchmark.
BUG=v8:5267
Review-Url: https://codereview.chromium.org/2280673003
Cr-Commit-Position: refs/heads/master@{#38928}
This patch:
1. Removes the unecessary inheritance of ParserBaseTraits<Impl>
in ParserBase<Impl>.
2. Flattens ParserBaseTraits<Impl> and renames it to
ParserTypes<Impl>. The Traits parameter/member is renamed to
Types.
3. Removes unecessary v8::internal:: qualifications from parser
types.
R=adamk@chromium.org, marja@chromium.org
BUG=
LOG=N
Review-Url: https://codereview.chromium.org/2279773002
Cr-Commit-Position: refs/heads/master@{#38927}
This interleaves setting names and values in the scope info. It's a little messy since globals and locals are interleaved, but afaiu globals is going away.
BUG=v8:5209
Review-Url: https://codereview.chromium.org/2272293004
Cr-Commit-Position: refs/heads/master@{#38925}
They are both accumulated at the same time, and either one already triggered
an error in ValidateExpression. Basically, there were no disjoint uses of
these two error types.
Review-Url: https://codereview.chromium.org/2268173005
Cr-Commit-Position: refs/heads/master@{#38922}
If the parser creates a catch scope, the variable is also added to
ordered_variables_ so just always do that.
Otherwise, it's not possible to create a ScopeInfo for that catch scope
R=verwaest@chromium.org
BUG=
Review-Url: https://codereview.chromium.org/2278963002
Cr-Commit-Position: refs/heads/master@{#38921}
Previously the calls to ExpressionClassifier::Accumulate() each chose
slightly different sets of productions to accumulate, and it turned
out that these were in some cases broader than needed and in some
cases less broad.
The existence of some grab-bag production bitmasks like
ExpressionClassifier::ExpressionProductions made this situation more
error-prone (for example, that production was missing AsyncArrowFormalParametersProduction).
This patch removes all "grab-bags" besides AllProductions. In some of
the places where code was using those grab-bags for convenience, it
switches them to use negation of AllProductions. In other, specifically
those having to do with expressions that are disallowed anywhere in
a sub-expression of a parameter list, I've added a new method on
ExpressionClassifier to centralize the logic.
The aforementioned centralization/addition of
AsyncArrowFormalParametersProduction fixes several cases where we were
failing to report an error for 'await' in some contexts; I've added
those test cases.
The patch also narrows all cases to exactly the set or productions
necessary, with a comment on each explaining the choice.
BUG=v8:4483
Review-Url: https://codereview.chromium.org/2271063002
Cr-Commit-Position: refs/heads/master@{#38918}
Decrease regular heap object size to 400k. In a follow up, we can now get rid of
the new space border page while keeping the 1M minimum new space size.
This reverts commit 1617043c10.
BUG=chromium:636331
Review-Url: https://codereview.chromium.org/2278653003
Cr-Commit-Position: refs/heads/master@{#38916}
Suitably scary warnings attached, as this will yet evolve (notably to handle
host objects, which are not currently handled).
Unit tests adjusted to use the public version of ValueSerializer, eliminating
any need they have to access v8::internal.
With this, Blink can begin using this code experimentally behind a flag as it
develops.
BUG=chromium:148757
Review-Url: https://codereview.chromium.org/2274693002
Cr-Commit-Position: refs/heads/master@{#38915}
Transferral is not included in this CL, nor is SharedArrayBuffer.
BUG=chromium:148757
Review-Url: https://codereview.chromium.org/2264403004
Cr-Commit-Position: refs/heads/master@{#38913}
Record details, such as cumulative duration, number of steps, and longest steps
in IncrementalMarkingDetails which get populated at a single callsite
(AddScopeSample). Remove member fields that thus become obsolete (unfortunately
not all of them).
Additional remove some dead code and refactor printing. Printing in a single
statement allows for using logcat on Android.
This should also address the regression in chromium:640524.
BUG=chromium:639818,chromium:640524
R=jochen@chromium.org
Review-Url: https://codereview.chromium.org/2269093002
Cr-Commit-Position: refs/heads/master@{#38912}
These tests were spliced out of changelist 2216353002 and extended.
BUG=
Review-Url: https://codereview.chromium.org/2245263003
Cr-Commit-Position: refs/heads/master@{#38906}
I could only test this with FreeBSD and OSX
(on the Node.js CI).
I don't know if the fix is correct for other BSD platforms.
Review-Url: https://codereview.chromium.org/2251603004
Cr-Commit-Position: refs/heads/master@{#38905}
DuplicateFinder isn't actually used by the Scanner, except for one
convenience function which we should probably remove, also.
BUG=
Review-Url: https://codereview.chromium.org/2281443002
Cr-Commit-Position: refs/heads/master@{#38904}
Unlike Crankshaft, Turbofan does not provide a context when trying to grow
elements. Depending on the code path we might end up updating transitioning
elements kinds in allocation sites for which we need access to the current
context. Unlike GrowCapacityAndConvert, the newly introduced GrowCapacity simply
returns false in cases where map transitions are involved.
BUG=chromium:637279
Patch by Camillo Bruni <cbruni@chromium.org>,
originally reviewed at https://codereview.chromium.org/2244983004/
Review-Url: https://codereview.chromium.org/2252393002
Cr-Commit-Position: refs/heads/master@{#38901}
During finalization, we create SharedFunctionInfos which in turn
will create ScopeInfos for the Scopes in the AST. The Scopes then
cache a handle to the ScopeInfos. However, once the scope is closed,
all those handles get zapped, and it's no longer possible to access
the scopes (even though we actually still need the AST).
R=rmcilroy@chromium.org
BUG=
Review-Url: https://codereview.chromium.org/2278933002
Cr-Commit-Position: refs/heads/master@{#38898}
Adds compile operations to the CompilerDispatcherJob interface. As such,
introduces Compiler::PrepareUnoptimizedCompilationJob and updates the
unoptimized compilation path to use CompilationJobs. Also unifies
FinalizeCompilationJob to deal with both optimized and unoptimized
compilation jobs.
A dummy FullCodegenCompilationJob is also introduced, where all the work
is done in the ExecuteJob phase, which cannot be run on a
background thread.
BUG=v8:5203
Review-Url: https://codereview.chromium.org/2251713002
Cr-Commit-Position: refs/heads/master@{#38897}
Try to narrow types of Phis further during JSTypedLowering, because
lowering based on types might create further opportunities for improving
the types.
R=jarin@chromium.org
BUG=v8:5267
Review-Url: https://codereview.chromium.org/2278903002
Cr-Commit-Position: refs/heads/master@{#38895}
Existing uses are correct but the return type was misleading.
Also clarify some related comments to make the difference between Bits
and BitField more obvious.
BUG=
Review-Url: https://codereview.chromium.org/2275973002
Cr-Commit-Position: refs/heads/master@{#38894}
Reason for revert:
Octane/Mandreel aborts with an exception now:
TypeError: __FUNCTION_TABLE__[(r2 >> 2)] is not a function
Original issue's description:
> [turbofan] Insert dummy values when changing from None type.
>
> Currently we choose the MachineRepresentation::kNone representation for
> values of Type::None, and when converting values from the kNone representation
> we use "impossible" conversions that will crash at runtime. This
> assumes that the impossible conversions should never be hit (the only
> way to produce the impossible values is to perform an always-failing
> runtime check on a value, such as Smi-checking a string). Note that
> this assumes that the runtime check is executed before the impossible
> convesrion.
>
> Introducing BitwiseOr type feedback broke this in two ways:
>
> - we always pick Word32 representation for bitwise-or, so the
> impossible conversion does not trigger (it only triggers with
> None representation), and we could end up with unsupported
> conversions from Word32.
>
> - even if we inserted impossible conversions, they are pure conversions.
> Since untagging, bitwise-or operations are also pure, we could hoist
> all these before the smi check of the inputs and we could hit the
> impossible conversions before we get to the smi check.
>
> This CL addresses this by just providing dummy values for conversions
> from the Type::None type. It also removes the impossible-to-* conversions.
>
> BUG=chromium:638132
>
> Committed: https://crrev.com/c83b21ab755f1420b6da85b3ff43d7e96ead9bbe
> Cr-Commit-Position: refs/heads/master@{#38883}
TBR=mstarzinger@chromium.org,jarin@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:638132
Review-Url: https://codereview.chromium.org/2280613002
Cr-Commit-Position: refs/heads/master@{#38893}
This patch moves the following methods from the traits objects to
the (pre)parser implementation objects:
- AddFormalParameter
- AddParameterInitializationBlock
- DeclareFormalParameter
- ExpressionListToExpression
- GetNonPatternList
- GetReportedErrorList
- IsTaggedTemplate
- MaterializeUnspreadArgumentsLiterals
- NoTemplateTag
- ParseArrowFunctionFormalParameterList
- ReindexLiterals
- SetFunctionNameFromIdentifierRef
- SetFunctionNameFromPropertyName
It moves the Void method from the preparser traits object to the
preparser implementation object. It also removes the traits zone
method and replaces it with that of ParserBase, which it turns to
public.
After all this, the traits objects contain just typedefs and the
delegate methods are no more necessary.
R=adamk@chromium.org, marja@chromium.org
BUG=
LOG=N
Review-Url: https://codereview.chromium.org/2277843002
Cr-Commit-Position: refs/heads/master@{#38892}
Used a BitField to for Variable fields instead of relying on the compiler, saving some memory probably.
This reduces sizeof(Variable) from 64 to 40 on x64
BUG=v8:5209
Review-Url: https://codereview.chromium.org/2257493002
Cr-Commit-Position: refs/heads/master@{#38891}
This patch moves the following methods from the traits objects to
the (pre)parser implementation objects:
- ExpressionFromIdentifier
- ExpressionFromLiteral
- ExpressionFromString
- FunctionSentExpression
- GetNextSymbol
- GetNumberAsSymbol
- GetSymbol
- NewExpressionList
- NewPropertyList
- NewStatementList
- NewSuperCallReference
- NewSuperPropertyReference
- NewTargetExpression
- ThisExpression
Also, the method GetIterator is specific only to the parser and is
removed from the preparser's implementation.
R=adamk@chromium.org, marja@chromium.org
BUG=
LOG=N
Review-Url: https://codereview.chromium.org/2274113002
Cr-Commit-Position: refs/heads/master@{#38890}
There's no point in running the LoadElimination on asm.js functions and
it would take serious amount of effort to actually make it correct for
the deprecated parts of the pipeline.
R=jarin@chromium.org
BUG=v8:5308
Review-Url: https://codereview.chromium.org/2276273002
Cr-Commit-Position: refs/heads/master@{#38884}
Currently we choose the MachineRepresentation::kNone representation for
values of Type::None, and when converting values from the kNone representation
we use "impossible" conversions that will crash at runtime. This
assumes that the impossible conversions should never be hit (the only
way to produce the impossible values is to perform an always-failing
runtime check on a value, such as Smi-checking a string). Note that
this assumes that the runtime check is executed before the impossible
convesrion.
Introducing BitwiseOr type feedback broke this in two ways:
- we always pick Word32 representation for bitwise-or, so the
impossible conversion does not trigger (it only triggers with
None representation), and we could end up with unsupported
conversions from Word32.
- even if we inserted impossible conversions, they are pure conversions.
Since untagging, bitwise-or operations are also pure, we could hoist
all these before the smi check of the inputs and we could hit the
impossible conversions before we get to the smi check.
This CL addresses this by just providing dummy values for conversions
from the Type::None type. It also removes the impossible-to-* conversions.
BUG=chromium:638132
Review-Url: https://codereview.chromium.org/2266823002
Cr-Commit-Position: refs/heads/master@{#38883}
For concurrent recompilation we created the CompilationHandleScope after
the CanonicalHandleScope, which basically disabled the canonicalization
because the deferred handle creation doesn't pay attention to the
canonicalization mode then. This meant that we did not canonicalize
handles properly as soon as concurrent recompilation was enabled.
R=jarin@chromium.org
BUG=v8:5267
Review-Url: https://codereview.chromium.org/2276953004
Cr-Commit-Position: refs/heads/master@{#38882}
This introduces appropriate unit tests to ensure that merging of
elements/fields information is correct for diamonds.
BUG=chromium:639210,v8:5266
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2278043002
Cr-Commit-Position: refs/heads/master@{#38881}
According to our style guide on Copyable and Movable Types,
copy/move operators should be disabled using = delete in the public: section.
Use consistent style for disabling new and delete.
BUG=
Review-Url: https://codereview.chromium.org/2276063002
Cr-Commit-Position: refs/heads/master@{#38880}
According to our style guide on Copyable and Movable Types,
copy/move operators should be disabled using = delete in the public: section,
not in the private: section.
BUG=
Review-Url: https://codereview.chromium.org/2272063002
Cr-Commit-Position: refs/heads/master@{#38879}
