TypeFeedbackVectors are strongly rooted by a closure. However, in modern
JavaScript closures are created and abandoned more freely. An important
closure may not be present in the root-set at time of garbage collection,
even though we've cached optimized code and use it regularly. For
example, consider leaf functions in an event dispatching system. They may
well be "hot," but tragically non-present when we collect the heap.
Until now, we've relied on a weak root to cache the feedback vector in
this case. Since there is no way to signal intent or relative importance,
this weak root is as susceptible to clearing as any other weak root at
garbage collection time.
Meanwhile, the feedback vector has become more important. All of our
ICs store their data there. Literal and regex boilerplates are stored there.
If we lose the vector, then we not only lose optimized code built from
it, we also lose the very feedback which allowed us to create that optimized
code. Therefore it's vital to express that dependency through the root
set.
This CL does this by creating a strong link to a feedback
vector at the instantiation site of the function closure.
This instantiation site is in the code and feedback vector
of the outer closure.
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2674593003
Cr-Commit-Position: refs/heads/master@{#42953}
The property details of a LookupIterator are not accessible,
if the iterator state is interceptor. Instead, use the
property attributes.
Fixes a crash in Node.js tests in Debug mode, see
c2c6ae52ea
BUG=
Review-Url: https://codereview.chromium.org/2675993002
Cr-Commit-Position: refs/heads/master@{#42941}
Apparently it happens quite easily that different NaNs are produced in
the interpreter than in the execution of the compiled code. This
non-determinism caused problems for the fuzzer which compares the
equality of the results of the interpreter and the compiled code.
I decided therefore to refactor the detection of non-determinism in the
interpreter. Instead of tracking whether potentially non-deterministic
NaNs were produced, I track now whether potentially non-deterministic
NaNs could have been observed. The only way the NaN non-determinism can
be observed is by observing the non-deterministic bit pattern of the
NaN. AFAICT the only way to observe the bit pattern is with a
I(32|64)_REINTERPRET_F(32|64) instruction or with a F(32|64)_STORE
followed by a load. Therefore I flag an execution as potentially
non-deterministic when either a NaN is reinterpreted to an int, or when
a NaN is stored to memory.
R=titzer@chromium.org, eholk@chromium.org
BUG=682180
Review-Url: https://codereview.chromium.org/2671803002
Cr-Commit-Position: refs/heads/master@{#42917}
Turns out is_hidden is not the right condition for "scope should be present in
the preparse data". For now, replaced it with "is hidden leaf scope" (i.e.,
doesn't contain any non-hidden scopes). That's probably not the right condition
either; will be fixed once there's more data to decide what the right condition
is.
BUG=v8:5516
R=vogelheim@chromium.org
Review-Url: https://codereview.chromium.org/2669163002
Cr-Commit-Position: refs/heads/master@{#42909}
We don't need Code::CALL_IC for anything now that the CallICStub is
migrated and no longer hooks into the traditional IC system.
R=yangguo@chromium.org
BUG=v8:5049
Review-Url: https://codereview.chromium.org/2669193002
Cr-Commit-Position: refs/heads/master@{#42890}
- Remove obsolete BreakLocatorType.
- Perform PrepareStepOnThrow after OnException event, in case stepping
was scheduled in the exception event.
- Use frame count instead of frame pointer for stepping. Frame pointer
is not reliable due to possible deopts.
- Consistently check for inlined functions in inlined frames.
- Use SharedFunctionInfo in FloodWithOneshot and EnsureDebugInfo.
R=jgruber@chromium.org
BUG=v8:5901
Review-Url: https://codereview.chromium.org/2664793002
Cr-Commit-Position: refs/heads/master@{#42878}
- Adds vqadd.s/u, vqsub.s/u for all integer lane sizes.
- Refactors disassembler and simulator, using switches instead
of long if-else chains.
LOG=N
BUG=v8:4124
Review-Url: https://codereview.chromium.org/2649323012
Cr-Commit-Position: refs/heads/master@{#42865}
This is a step towards encoding all the necessary information in
the feedback slot kind instead of storing it in the IC dispatcher's
code object flags.
BUG=v8:5849, v8:5917
Review-Url: https://codereview.chromium.org/2662113005
Cr-Commit-Position: refs/heads/master@{#42859}
I removed some constant folding optimizations for float instruction in
https://codereview.chromium.org/2647353007 because they were incorrect
if the input was a signalling NaN. Removing these optimizations, however
had an unexpectedly big impact on asm.js performance. With this CL I
restore the optimizations again when the source origin is not wasm. In
JavaScript signalling NaNs are not observable and therefore the
optimizations are correct.
R=titzer@chromium.org
BUG=chromium:686654
Review-Url: https://codereview.chromium.org/2666903002
Cr-Commit-Position: refs/heads/master@{#42850}
The String.prototype was altered after snapshot time (during
experimental natives setup), invalidating the stored map used for
fast-path checks.
BUG=
Review-Url: https://codereview.chromium.org/2663303003
Cr-Commit-Position: refs/heads/master@{#42842}
Rename to Construct and ConstructWithSpread, to match the names of
the JSOperators used.
Unfortunately, I can't find a way for auto-formatting to stay happy unless we
change the indentation for the whole BYTECODE_LIST macro.
Review-Url: https://codereview.chromium.org/2663963003
Cr-Commit-Position: refs/heads/master@{#42840}
Throw a syntax error on "new import(1)" expression. Adds a new error msg as well.
BUG=v8:5785
Review-Url: https://codereview.chromium.org/2661113002
Cr-Commit-Position: refs/heads/master@{#42827}
If hidden scopes are ignored, the scopes and variable produced by the PreParser
already matched the scopes and variables produced by the Parser.
BUG=v8:5516
R=vogelheim@chromium.org
Review-Url: https://codereview.chromium.org/2663043003
Cr-Commit-Position: refs/heads/master@{#42824}
Rewrites import expression into a runtime call. Uses peekahead to
determine if parsing an import declaration or import expression.
The runtime call doesn't actually do the import yet, will be added in
follow on patch.
Adds a new --harmony-dynamic-import flag.
Adds a ignore_error_msg parameter to the test runner to ignore the
discrepancy in the error messages while parsing import expression with
parser and pre parser. This discrepancy will actually never happen in
real code.
BUG=v8:5785
Review-Url: https://codereview.chromium.org/2661933003
Cr-Commit-Position: refs/heads/master@{#42820}
The existing unit test explicitly checked for this case, but was - under
the right circumstances - defeated by the optimization to not re-run the
whole position search if we were close enough.
BUG=chromium:685618
Review-Url: https://codereview.chromium.org/2663883002
Cr-Commit-Position: refs/heads/master@{#42794}
They have the same lifetime. It's a match!
Both structures are native context dependent and dealt with (creation,
clearing, gathering feedback) at the same time. By treating the spaces used
for literal boilerplates as feedback vector slots, we no longer have to keep
track of the materialized literal count elsewhere.
A follow-on CL removes even more parser infrastructure related to this count.
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2655853010
Cr-Commit-Position: refs/heads/master@{#42771}
This unifies the behavior of Scope::DeclareVariableName with
Scope::DeclareVariable.
BUG=v8:5516
Review-Url: https://codereview.chromium.org/2658063005
Cr-Commit-Position: refs/heads/master@{#42744}
The mentioned asserts did not work properly with interpreted and turbofanned functions.
To fix this issue %GetOptimizationStatus() now returns a set of flags instead of a single value.
This CL also adds more helper functions to mjsunit, like isNeverOptimize(), isAlwaysOptimize(),
isOptimized(fun), etc.
BUG=v8:5890
Review-Url: https://codereview.chromium.org/2654733004
Cr-Original-Commit-Position: refs/heads/master@{#42703}
Committed: d1ddec7857
Review-Url: https://codereview.chromium.org/2654733004
Cr-Commit-Position: refs/heads/master@{#42731}
Always execute the implicit return if we fell off the function bytes.
This is not considered an additional "step" as it is not executing a
wasm instruction.
Otherwise, we might pause at an invalid position (one after the
function bytes).
R=titzer@chromium.org
BUG=v8:5822
Review-Url: https://codereview.chromium.org/2650293003
Cr-Commit-Position: refs/heads/master@{#42730}
Previously, when restarting a frame, we would rewrite all frames
between the debugger activation and the frame to restart to squash
them, and replace the return address with that of a builtin to
leave that rewritten frame, and restart the function by calling it.
We now simply remember the frame to drop to, and upon returning
from the debugger, we check whether to drop the frame, load the
new FP, and restart the function.
R=jgruber@chromium.org, mstarzinger@chromium.org
BUG=v8:5587
Review-Url: https://codereview.chromium.org/2636913002
Cr-Commit-Position: refs/heads/master@{#42725}
The mentioned asserts did not work properly with interpreted and turbofanned functions.
To fix this issue %GetOptimizationStatus() now returns a set of flags instead of a single value.
This CL also adds more helper functions to mjsunit, like isNeverOptimize(), isAlwaysOptimize(),
isOptimized(fun), etc.
BUG=v8:5890
Review-Url: https://codereview.chromium.org/2654733004
Cr-Commit-Position: refs/heads/master@{#42703}
According to the WebAssembly spec no arithmetic operation should ever
return a signalling NaN. With the constant folding in V8, however, it
was possible that some arithmetic operations were elided, and if the
input of the arithmetic operation was a signalling NaN, then also the
result was the same signalling NaN. This CL removes some constant
folding optimizations and adjusts others so that even with constant
folding the result of an arithmetic operation is never a signalling NaN.
R=titzer@chromium.org, rossberg@chromium.org, bmeurer@chromium.org
Review-Url: https://codereview.chromium.org/2647353007
Cr-Commit-Position: refs/heads/master@{#42694}
- Declaring a variable called "this" for preparsed functions was unnecessary;
DeclarationScope ctor already adds the variable.
- "arguments" for preparsed scopes need to be declared after parsing the
function, like it's done in the parser.
- Now arguments_ can be the dummy variable, so adapted code to it.
- A previous refactoring CL ( https://codereview.chromium.org/2638333002 ) was
incomplete; it had added ParserBase::ParseFunctionBody but
PreParser::ParseFunction didn't call it. This CL completes that work. This is
needed for getting "arguments" declared properly for preparsed functions.
- AllocateVariablesRecursively is already called for preparsed scopes (without
this CL, that is), and it bails out early. However, before the bailout it used
to dcheck num_stack_slots_ == 0; that is no longer true since we've done scope
analysis for preparsed scopes.
- Test fix: we cannot have any lazy inner functions in the test, except the
topmost lazy inner function. Such functions would also be lazy in the parser
case, and the parser would just throw away their variables. Then the test
tries to verify the preparsed data against the scopes without variables and fails.
- Disabled a test w/ a sloppy block function, will get that working again in the
upcoming CLs.
BUG=v8:5516
Review-Url: https://codereview.chromium.org/2655623005
Cr-Commit-Position: refs/heads/master@{#42685}
The InterpreterTester class cobbles together a JSFunction from
a manually created bytecode and feedback vector. However it's
fragile against design changes in the way literal arrays and
feedback vectors are handled. It's better to let it hand in
a feedback vector metadata object, and allow the system to
create the vector from that.
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2652893010
Cr-Commit-Position: refs/heads/master@{#42684}
In practice, Emscripten seems to emit cond?+a:+b type return
expressions. This is not allowed by the spec or errata, but we need
to support it for compatibility.
Similar patterns with ints / signed, do not seem to be supported.
BUG=v8:5891
R=mtrofin@chromium.org,aseemgarg@chromium.org
Review-Url: https://codereview.chromium.org/2648353010
Cr-Commit-Position: refs/heads/master@{#42677}
