If the array's map is the initial FastHoley array map, and the array prototype
chain is undisturbed and empty of elements, then keyed loads can convert the
load of a hole to undefined.
BUG=
Review URL: https://codereview.chromium.org/1100083002
Cr-Commit-Position: refs/heads/master@{#28056}
Implements the strong mode proposal's restrictions on
implicit conversions for binary arithmetic operations, not
including the + special case. Adds some infrastructure
for future implementation of the restrictions for other
operators.
BUG=v8:3956
LOG=N
Review URL: https://codereview.chromium.org/1092353002
Cr-Commit-Position: refs/heads/master@{#28045}
Calling new Array(JSObject::kInitialMaxFastElementArray) in optimized code
makes a stub call that bails out due to the length. Currently, the bailout
code a) doesn't have the allocation site, and b) wouldn't use it if it did
because the length is perceived to be too high.
This CL passes the allocation site to the stub call (rather than undefined),
and alters the bailout code to utilize the feedback.
BUG=
Review URL: https://codereview.chromium.org/1086873003
Cr-Commit-Position: refs/heads/master@{#27857}
This adds a missing bailout id to a ForInStatement for when retrieving
and filtering a property name deoptimizes. This can happen with proxies
that have a getPropertyDescriptor trap.
R=jarin@chromium.org
TEST=mjsunit/for-in-opt
Review URL: https://codereview.chromium.org/1086083002
Cr-Commit-Position: refs/heads/master@{#27846}
They are content with a dummy vector, as MISSES won't result in
changing the real vector/slot at all.
BUG=
Review URL: https://codereview.chromium.org/1067573003
Cr-Commit-Position: refs/heads/master@{#27788}
This tweak was already present in CrankShaft for the non-AVX case. As it
turns out, it's also relevant even with AVX. Now the same optimization
is applied in case of TurboFan as well.
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/1081033003
Cr-Commit-Position: refs/heads/master@{#27774}
Reason for revert:
Causes test failures on ARM bots related to cells and write barriers.
Original issue's description:
> Merge cellspace into old pointer space
>
> BUG=
>
> Committed: https://crrev.com/4e7163ce05f135918205c7855ae60a48e5d46cc5
> Cr-Commit-Position: refs/heads/master@{#27707}
TBR=hpayer@chromium.org,balazs.kilvady@imgtec.com,yangguo@chromium.org,verwaest@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/1053243003
Cr-Commit-Position: refs/heads/master@{#27711}
This enables eager optimization of top-level code with TurboFan and
extends test coverage by triggering it with the --always-opt flag.
Script contexts are now also properly allocated in TurboFan.
R=titzer@chromium.org
Review URL: https://codereview.chromium.org/1053063003
Cr-Commit-Position: refs/heads/master@{#27633}
Optimistically pushing a lot of arguments can run into the stack limit of the process, at least on operating systems where this limit is close to the limit that V8 sets for itself.
BUG=chromium:469768
LOG=y
Review URL: https://codereview.chromium.org/1056913003
Cr-Commit-Position: refs/heads/master@{#27614}
Previously, the only optimized code path for Maps and Sets was for String keys.
This was achieved through an implementation of various complex operations
in Hydrogen. This approach was neither scalable nor forward-compatible.
This patch adds the necessary intrinsics to implement Maps and Sets almost entirely
in JS. The added intrinsics are:
%_FixedArrayGet
%_FixedArraySet
%_TheHole
%_JSCollectionGetTable
%_StringGetRawHashField
With these additions, as well as a few changes to what's exposed as runtime functions,
most of the C++ code backing Maps and Sets is gone (including both runtime code in
objects.cc and Crankshaft in hydrogen.cc).
Review URL: https://codereview.chromium.org/947683002
Cr-Commit-Position: refs/heads/master@{#27605}
A bug allows JSObject literals with elements to have the elements in the
boilerplate modified.
BUG=466993
LOG=N
Review URL: https://codereview.chromium.org/1037273002
Cr-Commit-Position: refs/heads/master@{#27511}
This adds the basics necessary to support float32 operations in TurboFan.
The actual functionality required to detect safe float32 operations will
be added based on this later. Therefore this does not affect production
code except for some cleanup/refactoring.
In detail, this patchset contains the following features:
- Add support for float32 operations to arm, arm64, ia32 and x64
backends.
- Add float32 machine operators.
- Add support for float32 constants to simplified lowering.
- Handle float32 representation for phis in simplified lowering.
In addition, contains the following (related) cleanups:
- Fix/unify naming of backend instructions.
- Use AVX comparisons when available.
- Extend ArchOpcodeField to 9 bits (required for arm64).
- Refactor some code duplication in instruction selectors.
BUG=v8:3589
LOG=n
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/1044793002
Cr-Commit-Position: refs/heads/master@{#27509}
This switches full-codegen to no longer push and pop StackHandler
markers onto the operand stack, but relies on a range-based handler
table instead. We only use StackHandlers in JSEntryStubs to mark the
transition from C to JS code.
Note that this makes deoptimization and OSR from within any try-block
work out of the box, makes the non-exception paths faster and should
overall be neutral on the memory footprint (pros).
On the other hand it makes the exception paths slower and actually
throwing and exception more expensive (cons).
R=yangguo@chromium.org
TEST=cctest/test-run-jsexceptions/DeoptTry
Review URL: https://codereview.chromium.org/1010883002
Cr-Commit-Position: refs/heads/master@{#27440}
Handlers should be in charge of this work. The change uncovered a bug in
vector-ics related to keyed loads into strings. It's important for
StringCharCodeAtGenerator, a helper used in full code and in
LoadIndexedStringStub (a handler) to protect the vector and slot registers
when it makes a runtime call to convert a HeapNumber to a Smi.
It's still possible for the handler to MISS after this call, perhaps due
to out of bounds access. In that case, the vector and slot registers need
to be delivered safely to the MISS handler.
BUG=
Review URL: https://codereview.chromium.org/1028093002
Cr-Commit-Position: refs/heads/master@{#27377}
I found some strange split in deopt entry points generator.
The code for table entry generator had two classes.
It is safe to join these classes together and drop virtual.
BUG=
LOG=n
Review URL: https://codereview.chromium.org/1010413003
Cr-Commit-Position: refs/heads/master@{#27264}
This relands commit 96f79568a9.
This makes the Isolate::Throw logic not depend on a prediction of
whether an exception is caught or uncaught. Such a prediction is
inherently undecidable because a finally block can decide between
consuming or re-throwing an exception depending on arbitray control
flow.
There still is a conservative prediction mechanism in place that
components like the debugger or tracing can use for reporting.
With this change we can get rid of the StackHandler::kind field, a
pre-requisite to do table-based lookups of exception handlers.
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/997213003
Cr-Commit-Position: refs/heads/master@{#27263}
This is the fifth part of https://codereview.chromium.org/1012633002
In this part we collect the offsets of deopt calls and save it into
an inlined function info.
On the Next:
Later when deopt happens we will get the offset of deopt call and
search it among inlined infos.
BUG=chromium:452067
LOG=n
Review URL: https://codereview.chromium.org/1011113004
Cr-Commit-Position: refs/heads/master@{#27258}
This ensures that there is only one stub that deals with unwinding the
stack. Having more than one place containing that logic is brittle and
error prone, especially when it is a corner case only for RangeErrors.
R=titzer@chromium.org
TEST=mjsunit/regress/regress-crbug-467047
BUG=chromium:467047
LOG=N
Review URL: https://codereview.chromium.org/1012103002
Cr-Commit-Position: refs/heads/master@{#27243}
A hydrogen code stub is not the best approach because it builds a frame
and doesn't have the technology to discard roots at tail call exits.
Platform-specific stubs provide much better performance at this point.
R=verwaest@chromium.org
BUG=
Review URL: https://codereview.chromium.org/988653003
Cr-Commit-Position: refs/heads/master@{#27235}
the third part of the patch https://codereview.chromium.org/1012633002
this patch
1) moves DeoptInfo builder code to platform independent file lithium-codegen.cc
2) adds inlining_id property to HEnterInlined so we can use it on lithium level.
BUG=chromium:452067
LOG=n
Review URL: https://codereview.chromium.org/1011733005
Cr-Commit-Position: refs/heads/master@{#27231}
