The existing implementation assumes that return nodes have exactly one
real value input. This assumption does not hold for WebAssembly. To
avoid incorrect behavior, this CL turns of the reduction of returns
with a value input count != 1.
R=titzer@chromium.org, mstarzinger@chromium.org
Review-Url: https://codereview.chromium.org/2638053002
Cr-Commit-Position: refs/heads/master@{#42425}
There were multiple bugs in the apply builtins on mips and mips64 which
caused them to always take the slow path by calling into the runtime to
create the array. The other bugs went undiscovered because those paths
were never taken.
Review-Url: https://codereview.chromium.org/2634393002
Cr-Commit-Position: refs/heads/master@{#42424}
Port b83c3d29a9
Original Commit Message:
When a function is declared in global scope, the closure is created
by the DeclareGlobals runtime service. It needs a pointer to the
literals array, already allocated in the feedback vector. This
fixes a bug where it's behavior wasn't in sync with CreateClosure,
which accepts the literals from the vector.
This enables a follow-on performance improvement in the CompileLazy
builtin.
R=mvstanton@chromium.org, joransiu@ca.ibm.com, jyan@ca.ibm.com, michael_dawson@ca.ibm.com
BUG=680637
LOG=N
Review-Url: https://codereview.chromium.org/2634423002
Cr-Commit-Position: refs/heads/master@{#42421}
Uses the structure of an IEEE float to speed up
DoubleToUint32IfEqualToSelf, similar to FastD2UI. Micro-benchmarks show
a ~1.2x-2x speed-up, depending on the input.
Review-Url: https://codereview.chromium.org/2636453003
Cr-Commit-Position: refs/heads/master@{#42420}
Port 95f210d562
Original commit message:
Improve CheckedLoad and Store bounds checking for arrays with power of two
length.
BUG=
Review-Url: https://codereview.chromium.org/2632213002
Cr-Commit-Position: refs/heads/master@{#42414}
Enable enqueueing of eager inner function compilation onto the compiler
dispatcher. This enables these tasks to be performed in parallel to
compilation of the outer functio (only for Ignition functions).
We currently synchronize to ensure all inner function compilations
are complete before executing the outer function - future work will
allow outer function execution to happenin parallel to inner function
compilation.
BUG=v8:5203,v8:5215
Review-Url: https://codereview.chromium.org/2611313002
Cr-Commit-Position: refs/heads/master@{#42413}
Reduce duplication of the inner function declaration + now we get the
test for inner arrow functions for free (if we later implement lazy
inner arrow functions).
BUG=v8:5501, v8:5516
Review-Url: https://codereview.chromium.org/2637003005
Cr-Commit-Position: refs/heads/master@{#42412}
Background: the first page of each space is implicitly immovable.
Recently, our builtin code objects have reached a size at which we
fill up the first page of code space during initialization. Once
that occurs, newly requested allocations of immovable code are
allocated in a large object space page of 512K.
This CL mitigates these effects by simply marking pages as immovable
during snapshot creation instead of going into LO space.
On snapshot builds, this should just work: deserialized pages are
trimmed and marked immovable when deserialization finishes.
However, non-snapshot builds and allocations of immovable CEntryStub
code at runtime are still affected.
BUG=v8:5831
Review-Url: https://codereview.chromium.org/2635973002
Cr-Commit-Position: refs/heads/master@{#42411}
This is just a bug. CompileLazy knows how to look in the
SharedFunctionInfo for code (or bytecode, essentially) if it
finds nothing in the OptimizedCodeMap. Ensure that behavior.
BUG=681543
R=leszeks@chromium.org
Review-Url: https://codereview.chromium.org/2636143002
Cr-Commit-Position: refs/heads/master@{#42409}
When a function is declared in global scope, the closure is created
by the DeclareGlobals runtime service. It needs a pointer to the
literals array, already allocated in the feedback vector. This
fixes a bug where it's behavior wasn't in sync with CreateClosure,
which accepts the literals from the vector.
This enables a follow-on performance improvement in the CompileLazy
builtin.
BUG=680637
Review-Url: https://codereview.chromium.org/2634283003
Cr-Commit-Position: refs/heads/master@{#42408}
The resume trampolin used to call the generator function with the context of the
last suspension rather than the closure's context. While that was fine for
Ignition, Turbofan got utterly confused. With this CL, the resume trampolin
always passes in the closure's context (like in the very first call of the
generator function). The generator function itself then restores its previously
current context by reading it from the generator object and doing a
PushContext.
BUG=chromium:681171
Review-Url: https://codereview.chromium.org/2639533002
Cr-Commit-Position: refs/heads/master@{#42407}
Instead, it is supposed to just return an empty context if it failed.
Also don't invoke interceptors (we don't for the parts that deserialize
from the snapshot anyways).
BUG=v8:5830
R=yangguo@chromium.org
Review-Url: https://codereview.chromium.org/2636903002
Cr-Commit-Position: refs/heads/master@{#42404}
This internalization was not necessary, since the rewriting does not use
the .result name string.
The subsequent internalization is still needed, so to simplify later
refactoring, this CL also adds "releasing" of the disallow scopes and
uses them here immediately before the second internalize. Notably, this
means that the rewriting is now also in the disallow scopes.
Driveby: Remove isolate from the rewriter's processor constructor.
BUG=v8:5832
Review-Url: https://codereview.chromium.org/2635913002
Cr-Commit-Position: refs/heads/master@{#42403}
This makes sure all call sites allocating and installing the feedback
vector metadata are within the compilation pipeline and avoids spreading
them accross components.
R=leszeks@chromium.org
Review-Url: https://codereview.chromium.org/2631253002
Cr-Commit-Position: refs/heads/master@{#42399}
This makes it clearer which places are creating variables which are
something else than NORMAL_VARIABLE + kCreatedInitialized.
BUG=
Review-Url: https://codereview.chromium.org/2631173002
Cr-Commit-Position: refs/heads/master@{#42395}
Creates an AstStringConstants container which pre-initializes the
string constants used by AstValueFactory. This ensures that all
AstValueFactories will produce the same AstValue objects for constants,
and so they can be used by the BytecodeGenerator without having to pass
the AstValueFactory to it, enabling construction off-thread.
BUG=v8:5203
Review-Url: https://codereview.chromium.org/2630343002
Cr-Original-Commit-Position: refs/heads/master@{#42381}
Committed: d611496b8e
Review-Url: https://codereview.chromium.org/2630343002
Cr-Commit-Position: refs/heads/master@{#42394}
A GC could happen when creating the Module string, which would cause the
JSModuleNamespace to be seen in an incomplete state. Instead make this
string a root object so that we don't need to allocate it during
NewJSModuleNamespace construction.
BUG=v8:1569
Review-Url: https://codereview.chromium.org/2631223002
Cr-Commit-Position: refs/heads/master@{#42393}
KillFields of an object should remove its cache from all the fields.
Currently, the cache in the front field is kept which is not expected.
This patch fixes it.
Review-Url: https://codereview.chromium.org/2618273002
Cr-Commit-Position: refs/heads/master@{#42388}
- Adds vmin, vmax for FP and integer vectors, both signed and unsigned.
- Regularizes switching logic in disasm and simulator for special codes
4 and 6.
- Factors vrecpe, vrsqrte, vrecps, vrsqrts into helper fns.
LOG=N
BUG=v8:4124
Review-Url: https://codereview.chromium.org/2623993006
Cr-Commit-Position: refs/heads/master@{#42385}
Reason for revert:
Seems to break modules-namespace2 on gcstress.
Original issue's description:
> [Parser] Introduce AstStringConstants to share constants across AstValueFactory
>
> Creates an AstStringConstants container which pre-initializes the
> string constants used by AstValueFactory. This ensures that all
> AstValueFactories will produce the same AstValue objects for constants,
> and so they can be used by the BytecodeGenerator without having to pass
> the AstValueFactory to it, enabling construction off-thread.
>
> BUG=v8:5203
>
> Review-Url: https://codereview.chromium.org/2630343002
> Cr-Commit-Position: refs/heads/master@{#42381}
> Committed: d611496b8eTBR=ahaas@chromium.org,marja@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:5203
Review-Url: https://codereview.chromium.org/2638783002
Cr-Commit-Position: refs/heads/master@{#42382}
Creates an AstStringConstants container which pre-initializes the
string constants used by AstValueFactory. This ensures that all
AstValueFactories will produce the same AstValue objects for constants,
and so they can be used by the BytecodeGenerator without having to pass
the AstValueFactory to it, enabling construction off-thread.
BUG=v8:5203
Review-Url: https://codereview.chromium.org/2630343002
Cr-Commit-Position: refs/heads/master@{#42381}
Our .clang-format is derived on the Google style, which sets
PointerAlignment to left (e.g. "Type* name"), but sets
DerivePointerAlignment to true. Once we started with the wrong style,
this made all new code in wasm-debug.cc use PointerAlignment=right,
resulting in lots of code using the wrong style.
For this CL, I ran
clang-format -style="{DerivePointerAlignment: false, BasedOnStyle: \
Google}" -i src/wasm/wasm-debug.cc
R=titzer@chromium.org
Review-Url: https://codereview.chromium.org/2635003002
Cr-Commit-Position: refs/heads/master@{#42380}
