This prefixes the escape analysis flag with "experimental", thereby
making sure the flag in question is not being fuzzed. It will reduce
noise levels on ClusterFuzz again.
R=jarin@chromium.org
BUG=chromium:603653
LOG=n
Review URL: https://codereview.chromium.org/1894513002
Cr-Commit-Position: refs/heads/master@{#35521}
This moves the responsibility of preparing full-codegen code with
deoptimization support into the backends. This avoids generating such
code when optimization can be done directly from existing bytecode.
R=bmeurer@chromium.org
BUG=v8:4280
LOG=n
Review URL: https://codereview.chromium.org/1883403002
Cr-Commit-Position: refs/heads/master@{#35517}
We have trouble with Math.fround(MEM[...]). Since we now properly type
LoadBuffer (it can produce undefined), lowering of fround has stopped
triggering (as it requires Number type). This CL is a quick fix for this issue
- we simply trigger the lowering for NumberOrUndefined and let representation
selection/truncation analysis deal with this.
Ultimately, we would want to insert some 'simplified' ToNumber conversion
that would be optimized as much as possible during representation
selection.
BUG=chromium:603802
LOG=n
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1893483003
Cr-Commit-Position: refs/heads/master@{#35513}
The current context is stored as a stack slot on the interpreter frame
and therefore we don't need to also maintain a machine register for the
context. Removes this register from bytecode handlers.
In the process modifies this frees up a register on ia32 to keep the
dispatch table pointer in a register rather than on a stack slot on
ia32.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1887493004
Cr-Commit-Position: refs/heads/master@{#35511}
port 974721c661 (r35283)
original commit message:
Introduce a ResumeGeneratorTrampoline, which does the actual stack state
reconstruction (currently always restores a fullcodegen frame), and
introduce appropriate TurboFan builtins for %GeneratorPrototype%.next,
%GeneratorPrototype%.return and %GeneratorPrototype%.throw based on
this native builtin.
Also unify the flooding in case of step-in to always work based on
JSFunction and remove the special casing for JSGeneratorObject.
BUG=
Review URL: https://codereview.chromium.org/1889083002
Cr-Commit-Position: refs/heads/master@{#35510}
port 3dd3beb066 (r35199)
original commit message:
Currently, if the size of two cmp or test operands is a byte or a word, we sign-extend or zero-extend each of them into a 32-bit register before doing the comparison, even when the conditions
for the use of a memory operand are met.
This CL makes it possible to load only one of them into a register and address the other as a memory operand.
The tricky bit is that, unlike as in the x64 counterpart http://crrev.com/1780193003, not all registers can be accessed as bytes.
BUG=
Review URL: https://codereview.chromium.org/1883373002
Cr-Commit-Position: refs/heads/master@{#35508}
The CL #35176 (https://codereview.chromium.org/1843983002) exposed one hidden bug in x87 turbofan code generation for kX87Float64ToUint32.
The current kX87Float64ToUint32 code generation will destroy the input value in X87 FPU stack which will be used by the following code.
This CL fixed this bug.
BUG=
Review URL: https://codereview.chromium.org/1884403002
Cr-Commit-Position: refs/heads/master@{#35507}
The current code for testing the VEX.L flag, indicating whether
128-bit or 256-bit registers are being accessed, was erroneous
and always returned true (i.e. indicated 128-bit registers).
This patch fixes this behaviour and checks the flag correctly.
Ref: https://github.com/nodejs/node/issues/6151
BUG=
Review URL: https://codereview.chromium.org/1875323002
Cr-Commit-Position: refs/heads/master@{#35506}
Rolling v8/buildtools to 5378d73123b64907773cc5c1bb027b2f765ff00a
Rolling v8/tools/clang to 41bff4c5ba97022c0fe69a59d8892a6c45fb0867
TBR=machenbach@chromium.org,vogelheim@chromium.org,hablich@chromium.org
Review URL: https://codereview.chromium.org/1889043002
Cr-Commit-Position: refs/heads/master@{#35505}
The builtin context is not a thing anymore. This means we don't have to
worry about being able to deserialize it when optimizing top-level code.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1891793002
Cr-Commit-Position: refs/heads/master@{#35502}
Port 0c05e02f25
Original commit message:
Modifies Ignition to store code entry addresses in the dispatch table
rather than code objects. This allows the interpreter to avoid
calculating the code entry address from the code object on every
dispatch and provides a ~5-7% performance improvement on Octane with
Ignition.
This change adds ArchOpcode::kArchTailCallAddress to TurboFan to enable
tail call dispatch using these code addresses. It also adds a Dispatch
linkage creator (distinct from the stub linkage type used previously) to
allow targetting a code address target (which will diverge further from
the stub linkage type when we remove the context machine register in
Ignition).
R=rmcilroy@chromium.org, joransiu@ca.ibm.com, jyan@ca.ibm.com, michael_dawson@ca.ibm.com, bjaideep@ca.ibm.com
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1888053002
Cr-Commit-Position: refs/heads/master@{#35501}
We will move this back to start of incremental marking when we make faster progress.
Perf sheriffs: This CL may cause a regression on benchmarks that improved earlier by enabling black allocation.
BUG=chromium:599174
LOG=n
Review URL: https://codereview.chromium.org/1889853002
Cr-Commit-Position: refs/heads/master@{#35500}
This changes closure creation to lower to inline allocations when
possible instead of going through the FastNewClosureStub. It allows us
to leverage all advantages of inline allocations on closures. Note that
it is only safe to embed the raw entry point of the compile lazy stub
into the code, because that stub is immortal and immovable.
R=mvstanton@chromium.org
Review URL: https://codereview.chromium.org/1573153002
Cr-Commit-Position: refs/heads/master@{#35499}
This allows us to remove the turbofan bailout that we introduced
as a response to crbug.com/589792.
BUG=chromium:589792
LOG=n
Review URL: https://codereview.chromium.org/1884713003
Cr-Commit-Position: refs/heads/master@{#35493}
Otherwise the megamorphic stubcache will miss. It only probes for
non-keyed handlers.
BUG=
Review URL: https://codereview.chromium.org/1883533003
Cr-Commit-Position: refs/heads/master@{#35492}
Reason for revert:
performance impact
Original issue's description:
> Correctly annotate eval origin.
>
> There were a couple of issues with it:
> - interpreter is not supported
> - the source position was just accidentally correct for full-codegen
> - the eval origin could have been cached
>
> Also fixes a few other places to use AbstractCode.
>
> R=mstarzinger@chromium.org
>
> Committed: https://crrev.com/2f3a171adc9e620c2235bf0562145b9d4eaba66d
> Cr-Commit-Position: refs/heads/master@{#35257}
>
> Committed: https://crrev.com/ad4e8a27963b704bb70ec8bac0991c57296b1d16
> Cr-Commit-Position: refs/heads/master@{#35481}
TBR=mstarzinger@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1888013002
Cr-Commit-Position: refs/heads/master@{#35491}
The parser no longer disables optimization. This is done solely by the
renumbering stage. It is sufficient to mark a SharedFunctionInfo as
disabled for optimization right after the renumbering stage.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1891753002
Cr-Commit-Position: refs/heads/master@{#35490}
- Get rid of fixing up pointers on the main thread
- Get rid of sweeping on the main thread
Instead:
- Record (and process afterwards) slots in parallel
- Add the pages to the concurrent sweeper as pointers have already been fixed
This reverts commit 6df04b296b.
BUG=chromium:581412
LOG=N
TEST=cctest/test-compaction/*
Review URL: https://codereview.chromium.org/1891683002
Cr-Commit-Position: refs/heads/master@{#35485}
This removes the flag in question. It just duplicates the corresponding
compilation hint in the underlying SharedFunctionInfo object. Now that
the backends should have a SharedFunctionInfo available most of the time
it is safe to use the cannonical predicate.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1883583002
Cr-Commit-Position: refs/heads/master@{#35482}
Modifies Ignition to store code entry addresses in the dispatch table
rather than code objects. This allows the interpreter to avoid
calculating the code entry address from the code object on every
dispatch and provides a ~5-7% performance improvement on Octane with
Ignition.
This change adds ArchOpcode::kArchTailCallAddress to TurboFan to enable
tail call dispatch using these code addresses. It also adds a Dispatch
linkage creator (distinct from the stub linkage type used previously) to
allow targetting a code address target (which will diverge further from
the stub linkage type when we remove the context machine register in
Ignition).
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1882073002
Cr-Commit-Position: refs/heads/master@{#35480}
Replaces the HCallRuntime instruction in HGraphBuilder::BuildBinaryOperation with a call to the corresponding TurboFan stub.
Review URL: https://codereview.chromium.org/1884103002
Cr-Commit-Position: refs/heads/master@{#35479}
port e6b6e55453 (r35163)
original commit message:
In the earlier implementation of GenerateDoubleToObject the context
is loaded from the parent's frame. rsi is clobbered because it is used
to store kHoleNan constnat. It is not always safe to peek at
the parents frame. Bytecode handlers have TypedFrame and the type of
frame is stored at FP + 1. GenerateDoubleToObject expects context
to be store at that place. In the current implementation rsi is pushed
onto the stack and is popped when exiting this function.
BUG=
Review URL: https://codereview.chromium.org/1884873007
Cr-Commit-Position: refs/heads/master@{#35478}
This makes sure that when cached optimized code is found while doing
lazy compilation via Compiler::Compile installs any existing literals
array as well.
R=mvstanton@chromium.org
Review URL: https://codereview.chromium.org/1879973005
Cr-Commit-Position: refs/heads/master@{#35477}
Removes skips for two tests that no longer fail on windows system.
Updates description of another failing test with more details.
BUG=v8:4280,v8:4680
LOG=N
Review URL: https://codereview.chromium.org/1878893004
Cr-Commit-Position: refs/heads/master@{#35475}
To keep track of memory characteristics of a JS program use
GetHeapSpaceStatistics from a combination of
- a GC epilogue callback registered using AddGCEpilogueCallback
- an interrupt registered using RequestInterrupt
Rationale:
The feature is currently broken as those callbacks are (also) executed from
background threads, breaking our implicit contract for only performing callbacks
on the foreground thread. Furhtermore, it's not clear for the embedder whether
these allocations originate in JS, or are required for a GC.
Note:
- The already deprecated non-isolate versions are removed with this CL.
R=jochen@chromium.org
BUG=v8:4813
LOG=Y
Review URL: https://codereview.chromium.org/1886573006
Cr-Commit-Position: refs/heads/master@{#35474}
port 53d51c52f3 (r35139)
original commit message:
Removed Frame::needs_frame and the function-wide logic using it in
favor of FrameAccessState::has_frame, which can be set on a more
granular level, and driving it block by block.
BUG=
Review URL: https://codereview.chromium.org/1886963002
Cr-Commit-Position: refs/heads/master@{#35473}
This removes one of the duplicated pipeline implementation from the
compiler. By now we can reuse the existing CompileForDebugging for all
compilations being kicked off for debugging.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1885063002
Cr-Commit-Position: refs/heads/master@{#35472}
