Google Git
Sign in
webkit / WebKit / fbe5c45bf292398985d58488f98252e9c7d0702a / . / Source / JavaScriptCore / ChangeLog
blob: d5a3a42f334c2aff7b75e21473ff27f5882b0312 [file] [log] [blame]
2017-04-16 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/built-ins/isNaN/toprimitive-not-callable-throws.js
https://bugs.webkit.org/show_bug.cgi?id=170888
Reviewed by Saam Barati.
* runtime/ExceptionHelpers.h:
* runtime/ExceptionHelpers.cpp:
(JSC::createInvalidInstanceofParameterErrorHasInstanceValueNotFunction):
Fix up this function name.
* runtime/JSObject.cpp:
(JSC::callToPrimitiveFunction):
When called with @@isPrimitive, bail on undefined or null and
throw a type error if the value is not callable.
(JSC::JSObject::toPrimitive):
Use throw scope to check for exception.
2017-04-16 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/language/expressions/tagged-template/template-object.js
https://bugs.webkit.org/show_bug.cgi?id=170878
Reviewed by Saam Barati.
* runtime/JSArray.cpp:
(JSC::JSArray::put):
The fast path for setting an Array's length should check if length is
writable before checking for and possibly throwing a RangeError.
2017-04-16 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/built-ins/Object/getOwnPropertyNames/15.2.3.4-4-44.js
https://bugs.webkit.org/show_bug.cgi?id=170879
Reviewed by Saam Barati.
* runtime/StringObject.h:
* runtime/StringObject.cpp:
(JSC::StringObject::getOwnPropertyNames):
(JSC::StringObject::getOwnNonIndexPropertyNames):
Ensure 'length' comes after all indexed properties by moving
it out to the getOwnNonIndexPropertyNames method which is called
inside of getOwnPropertyNames after JSObject handles indices.
2017-04-16 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/built-ins/Date/prototype/Symbol.toPrimitive/name.js
https://bugs.webkit.org/show_bug.cgi?id=170884
Reviewed by Yusuke Suzuki.
* runtime/DatePrototype.cpp:
(JSC::DatePrototype::finishCreation):
* runtime/FunctionPrototype.cpp:
(JSC::FunctionPrototype::addFunctionProperties):
* runtime/RegExpPrototype.cpp:
(JSC::RegExpPrototype::finishCreation):
* runtime/SymbolPrototype.cpp:
(JSC::SymbolPrototype::finishCreation):
Give symbol property functions proper function names.
This addresses function.name but not function.toString().
2017-04-15 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/language/global-code/new.target-arrow.js
https://bugs.webkit.org/show_bug.cgi?id=170872
Reviewed by Saam Barati.
* parser/Parser.cpp:
(JSC::Parser<LexerType>::Parser):
Mark the global code scope.
(JSC::Parser<LexerType>::parseMemberExpression):
If new.target is detected in an arrow function defined in global scope
throw a SyntaxError.
* parser/Parser.h:
(JSC::Scope::Scope):
(JSC::Scope::setIsGlobalCodeScope):
(JSC::Scope::isGlobalCodeScope):
Marker for a global code scope.
* parser/ParserModes.h:
(JSC::isModuleParseMode):
(JSC::isProgramParseMode):
(JSC::isProgramOrModuleParseMode):
Helper for detecting global code based on parse mode.
2017-04-14 Nikita Vasilyev <nvasilyev@apple.com>
Web Inspector: WebSockets: messages with non-latin letters are displayed incorrectly
https://bugs.webkit.org/show_bug.cgi?id=170760
Reviewed by Joseph Pecoraro.
Add payloadLength property, which is used to display size. When payloadLength is unavailable,
it is calculated from payloadData by Web Inspector frontend.
This fixes <webkit.org/b/170609> Web Inspector: WebSockets: Transferred size is incorrect.
* inspector/protocol/Network.json:
2017-04-14 Saam Barati <sbarati@apple.com>
ParseInt intrinsic in DFG backend doesn't properly flush its operands
https://bugs.webkit.org/show_bug.cgi?id=170865
Reviewed by Mark Lam and Geoffrey Garen.
The DFG backend code needed to first call .gpr()/.jsValueRegs()
before calling flushRegisters(), or the input JSValueOperand would
not be flushed.
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileParseInt):
2017-04-14 Mark Lam <mark.lam@apple.com>
Update architectures in xcconfig files.
https://bugs.webkit.org/show_bug.cgi?id=170867
<rdar://problem/31628104>
Reviewed by Joseph Pecoraro.
* Configurations/Base.xcconfig:
* Configurations/FeatureDefines.xcconfig:
* Configurations/JavaScriptCore.xcconfig:
* Configurations/ToolExecutable.xcconfig:
2017-04-14 Keith Miller <keith_miller@apple.com>
WebAssembly: B3IRGenerator should use phis for result types
https://bugs.webkit.org/show_bug.cgi?id=170863
Reviewed by Filip Pizlo.
Currently, we use variables for the result types of control flow in
Wasm. We did this originally since we weren't sure that the phis we
generated would be optimal. Since then, we have verified that the edges
in wasm control flow ensure that each upsilon will dominate its phi
so we don't need to use variables.
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::ControlData::ControlData):
(JSC::Wasm::B3IRGenerator::addTopLevel):
(JSC::Wasm::B3IRGenerator::addBlock):
(JSC::Wasm::B3IRGenerator::addLoop):
(JSC::Wasm::B3IRGenerator::unify):
2017-04-14 Alex Christensen <achristensen@webkit.org>
Fix Windows build after r215368.
https://bugs.webkit.org/show_bug.cgi?id=170641
* CMakeLists.txt:
Add new directory containing files needed in WebCore.
2017-04-14 Caitlin Potter <caitp@igalia.com>
[JSC] use ExpressionErrorClassifier for AwaitExpression operand
https://bugs.webkit.org/show_bug.cgi?id=170844
Reviewed by Saam Barati.
In parseAssignmentExpression(), several cover grammars are handled, and
use ExpressionErrorClassifier to record hints about which grammars to
try.
In parseAwaitExpression(), the hints recorded during parsing of the
operand need to be discarded, because if they propagate to the outer
parseAssignmentExpression(), the hints will lead the parser down invalid
branches that should be skipped.
This change adds an additional ExpressionErrorClassifier to
parseAwaitExpression(), in order to discard hints recorded trying to
parse the operand.
* parser/Parser.cpp:
(JSC::Parser<LexerType>::parseAwaitExpression):
2017-04-14 Saam Barati <sbarati@apple.com>
WebAssembly: There is a short window of time where a CodeBlock could be destroyed before all of its async compilation callbacks are called
https://bugs.webkit.org/show_bug.cgi?id=170641
Reviewed by Keith Miller.
There is an unlikely race when a CodeBlock compilation fails,
the module compiles a new CodeBlock for that memory mode, all while
the CodeBlock is notifying its callbacks that it has finished.
There is a chance that the Module could deref its failed CodeBlock
at that point, destroying it, before the callbacks were able to
grab a Ref to the CodeBlock. This patch fixes the race by having the
callbacks ref the CodeBlock.
This patch also has the Plan clear out all of its callbacks
once it gets completed. This adds an extra defense to anybody
that grabs refs to the Plan in the callback.
* wasm/WasmCodeBlock.cpp:
(JSC::Wasm::CodeBlock::CodeBlock):
(JSC::Wasm::CodeBlock::compileAsync):
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::complete):
2017-04-13 Filip Pizlo <fpizlo@apple.com>
Air::RegLiveness should be constraint-based
https://bugs.webkit.org/show_bug.cgi?id=170817
Reviewed by Saam Barati.
Previously, I changed the Air liveness analyses based on Air::Liveness<> to be
constraint-based and this was a significant speed-up. Now I'm adding the same
functionality to RegLiveness.
This is a 1% speed-up on wasm B3 -O1 compile times.
* b3/air/AirAllocateRegistersAndStackByLinearScan.cpp:
* b3/air/AirLivenessAdapter.h:
(JSC::B3::Air::LivenessAdapter::LivenessAdapter):
(JSC::B3::Air::LivenessAdapter::prepareToCompute):
(JSC::B3::Air::LivenessAdapter::actionsAt):
* b3/air/AirRegLiveness.cpp:
(JSC::B3::Air::RegLiveness::RegLiveness):
(JSC::B3::Air::RegLiveness::LocalCalcForUnifiedTmpLiveness::LocalCalcForUnifiedTmpLiveness):
(JSC::B3::Air::RegLiveness::LocalCalcForUnifiedTmpLiveness::execute):
(JSC::B3::Air::RegLiveness::LocalCalc::execute): Deleted.
* b3/air/AirRegLiveness.h:
(JSC::B3::Air::RegLiveness::Actions::Actions):
(JSC::B3::Air::RegLiveness::LocalCalcBase::LocalCalcBase):
(JSC::B3::Air::RegLiveness::LocalCalcBase::live):
(JSC::B3::Air::RegLiveness::LocalCalcBase::isLive):
(JSC::B3::Air::RegLiveness::LocalCalc::LocalCalc):
(JSC::B3::Air::RegLiveness::LocalCalc::execute):
(JSC::B3::Air::RegLiveness::LocalCalc::live): Deleted.
(JSC::B3::Air::RegLiveness::LocalCalc::isLive): Deleted.
2017-04-13 JF Bastien <jfbastien@apple.com>
WebAssembly: fix windows build
https://bugs.webkit.org/show_bug.cgi?id=170832
Reviewed by Mark Lam.
My previous patch re-declared isIOS which AssemblerCommon.h
already provided, and which was already included by Options.cpp.
* runtime/Options.cpp:
2017-04-13 Saam Barati <sbarati@apple.com>
WebAssembly: We should be able to postMessage a JSWebAssemblyModule
https://bugs.webkit.org/show_bug.cgi?id=170573
Reviewed by Filip Pizlo.
This patch adds a callback to JSRunLoopTimer to notify
clients that a timer has been set. This is used inside
WorkerRunLoop in WebCore so that its RunLoop can perform
an iteration when it sees that a timer got set.
* JavaScriptCore.xcodeproj/project.pbxproj:
* runtime/JSRunLoopTimer.cpp:
(JSC::JSRunLoopTimer::scheduleTimer):
(JSC::JSRunLoopTimer::addTimerSetNotification):
(JSC::JSRunLoopTimer::removeTimerSetNotification):
* runtime/JSRunLoopTimer.h:
* wasm/WasmCodeBlock.cpp:
(JSC::Wasm::CodeBlock::~CodeBlock):
* wasm/WasmCodeBlock.h:
* wasm/WasmModule.cpp:
(JSC::Wasm::Module::~Module):
(JSC::Wasm::Module::signatureIndexFromFunctionIndexSpace):
(JSC::Wasm::makeValidationCallback):
(JSC::Wasm::Module::validateSync):
(JSC::Wasm::Module::validateAsync):
(JSC::Wasm::Module::validateSyncImpl): Deleted.
(JSC::Wasm::Module::makeValidationCallback): Deleted.
* wasm/WasmModule.h:
(JSC::Wasm::Module::validateSync): Deleted.
(JSC::Wasm::Module::validateAsync): Deleted.
(JSC::Wasm::Module::signatureIndexFromFunctionIndexSpace): Deleted.
(JSC::Wasm::Module::nonNullCodeBlock): Deleted.
* wasm/js/JSWebAssemblyCodeBlock.cpp:
(JSC::JSWebAssemblyCodeBlock::create):
* wasm/js/JSWebAssemblyCodeBlock.h:
(JSC::JSWebAssemblyCodeBlock::create): Deleted.
* wasm/js/JSWebAssemblyModule.cpp:
(JSC::JSWebAssemblyModule::source):
* wasm/js/JSWebAssemblyModule.h:
(JSC::JSWebAssemblyModule::source): Deleted.
* wasm/js/WebAssemblyFunction.cpp:
(JSC::callWebAssemblyFunction):
* wasm/js/WebAssemblyModulePrototype.cpp:
2017-04-13 Mark Lam <mark.lam@apple.com>
Should use flushDirect() when flushing the scopeRegister due to needsScopeRegister().
https://bugs.webkit.org/show_bug.cgi?id=170661
<rdar://problem/31579046>
Reviewed by Filip Pizlo.
Previously, we were using flush() to flush the outermost frame's scopeRegister.
This is incorrect because flush() expects the VirtualRegister value passed to
it to be that of the top most inlined frame. In the event that we reach a
terminal condition while inside an inlined frame, flush() will end up flushing
the wrong register. The fix is simply to use flushDirect() instead.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::flush):
2017-04-13 Andy VanWagoner <thetalecrafter@gmail.com>
Change Intl prototypes to plain objects
https://bugs.webkit.org/show_bug.cgi?id=168178
Reviewed by JF Bastien.
* builtins/StringPrototype.js:
(localeCompare): Create default Collator once instead of using prototype.
* runtime/IntlCollatorPrototype.cpp:
(JSC::IntlCollatorPrototype::IntlCollatorPrototype):
* runtime/IntlCollatorPrototype.h:
* runtime/IntlDateTimeFormatPrototype.cpp:
(JSC::IntlDateTimeFormatPrototype::IntlDateTimeFormatPrototype):
* runtime/IntlDateTimeFormatPrototype.h:
* runtime/IntlNumberFormatPrototype.cpp:
(JSC::IntlNumberFormatPrototype::IntlNumberFormatPrototype):
* runtime/IntlNumberFormatPrototype.h:
* runtime/IntlObject.cpp:
(JSC::IntlObject::finishCreation): Don't set constructor on each prototype.
2017-04-13 Oliver Hunt <oliver@apple.com>
allocationSize should use safe arithmetic by default
https://bugs.webkit.org/show_bug.cgi?id=170804
Reviewed by JF Bastien.
Make all allocationSize() functions work in terms
of Checked<size_t>
* runtime/DirectArguments.h:
(JSC::DirectArguments::offsetOfSlot):
(JSC::DirectArguments::allocationSize):
* runtime/HashMapImpl.h:
(JSC::HashMapBuffer::allocationSize):
* runtime/JSArray.h:
(JSC::JSArray::allocationSize):
* runtime/JSArrayBufferView.h:
(JSC::JSArrayBufferView::allocationSize):
* runtime/JSAsyncFunction.h:
(JSC::JSAsyncFunction::allocationSize):
* runtime/JSFixedArray.h:
(JSC::JSFixedArray::allocationSize):
* runtime/JSFunction.h:
(JSC::JSFunction::allocationSize):
* runtime/JSGeneratorFunction.h:
(JSC::JSGeneratorFunction::allocationSize):
* runtime/JSModuleNamespaceObject.h:
* runtime/JSObject.h:
(JSC::JSFinalObject::allocationSize):
* runtime/JSWrapperObject.h:
(JSC::JSWrapperObject::allocationSize):
* runtime/ScopedArguments.h:
(JSC::ScopedArguments::allocationSize):
* runtime/VM.h:
(JSC::ScratchBuffer::allocationSize):
* wasm/js/JSWebAssemblyCodeBlock.h:
(JSC::JSWebAssemblyCodeBlock::offsetOfImportStubs):
(JSC::JSWebAssemblyCodeBlock::allocationSize):
* wasm/js/JSWebAssemblyInstance.h:
(JSC::JSWebAssemblyInstance::allocationSize):
2017-04-13 JF Bastien <jfbastien@apple.com>
WebAssembly: manage memory better
https://bugs.webkit.org/show_bug.cgi?id=170628
Reviewed by Keith Miller, Michael Saboff.
WebAssembly fast memories weren't managed very well. This patch
refactors it and puts us in a good position to further improve our
fast memory handling in the future.
We now cache fast memories at a process granularity, but make sure
that they don't consume dirty pages. We add a cap to the total
number of allocated fast memories to avoid ASLR degradation.
We teach the GC about memories as a kind of resource it should
care about because it didn't have visibility into the amount of
memory each represented. This allows benchmarks which allocate
memories back-to-back to reliably get fast memories 100% of the
time, even on a system under load, which wasn't the case
before. This reliability yields roughly 8% perf bump on x86-64
WasmBench.
The GC heuristic is as follows: each time we allocate a fast
memory we notify the GC, which then keeps track of the total
number of fast memories allocated since it last GC'd. We
separately keep track of the total number of fast memories which
have ever existed at any point in time (cached + allocated). This
is a monotonically-increasing high watermark. The GC will force a
full collection if, since it last ran, half or more of the high
watermark of fast memories was allocated.
At the same time, if we fail obtaining a fast memory from the
cache we do a GC to try to find one. If that fails we'll allocate
a new one (this can also fail, then we go to slow memory). This
can also be improved, but it's a good start.
This currently disables fast memories on iOS because getting fast
memories isn't a guaranteed thing. Rather, we get quite a few of
them and achieve significant speedups, but benchmarks which
allocate memories back-to-back end up falling behind because the
GC can conservatively hold onto memories, which then yields a perf
cliff. That cliff isn't reliable, WasmBench gets roughly 10 of 18
fast memories when in theory it should get all of them fast (as
MacOS does). The patch significantly improves the state of iOS
though, and in a follow-up we could re-enable fast memories.
Part of this good positioning is a facility to pre-allocate fast
memories very early at startup, before any fragmentation
occurs. This is currently disabled but worked extremely reliably
on iOS. Once we fix the above issues we'll want to re-visit and
turn on pre-allocation.
We also avoid locking for fast memory identification when
performing signal handling. I'm very nervous about acquiring locks
in a signal handler because in general signals can happen when
we've messed up. This isn't the case with fast memories: we're
raising a signal on purpose and handling it. However this doesn't
mean we won't mess up elsewhere! This will get more complicated
once we add support for multiple threads sharing memories and
being able to grow their memories. One example: the code calls
CRASH(), which executes the following code in release:
*(int *)(uintptr_t)0xbbadbeef = 0;
This is a segfault, which our fast memory signal handler tries to
handle. It does so by first figuring out whether 0xbbadbeef is in
a fast memory region, reqiring a lock. If we CRASH() while holding
the lock then our thread self-deadlocks, giving us no crash report
and a bad user experience.
Avoiding a lock therefore it's not about speed or reduced
contention. In fact, I'd use something else than a FIFO if these
were a concern. We're also doing syscalls, which dwarf any locking
cost.
We now only allocate 4GiB + redzone of 64k * 128 for fast memories
instead of 8GiB. This patch reuses the logic from
B3::WasmBoundsCheck to perform bounds checks when accesses could
exceed the redzone. We'll therefore benefit from CSE goodness when
it reaches WasmBoundsCheck. See bug #163469.
* b3/B3LowerToAir.cpp: fix a baaaaddd bug where unsigned->signed
conversion allowed out-of-bounds reads by -2GiB. I'll follow-up in
bug #170692 to prevent this type of bug once and for all.
(JSC::B3::Air::LowerToAir::lower):
* b3/B3Validate.cpp: update WasmBoundsCheck validation.
* b3/B3Value.cpp:
(JSC::B3::Value::effects): update WasmBoundsCheck effects.
* b3/B3WasmBoundsCheckValue.cpp:
(JSC::B3::WasmBoundsCheckValue::WasmBoundsCheckValue):
(JSC::B3::WasmBoundsCheckValue::redzoneLimit):
(JSC::B3::WasmBoundsCheckValue::dumpMeta):
* b3/B3WasmBoundsCheckValue.h:
(JSC::B3::WasmBoundsCheckValue::maximum):
* b3/air/AirCustom.cpp:
(JSC::B3::Air::WasmBoundsCheckCustom::isValidForm):
* b3/testb3.cpp:
(JSC::B3::testWasmBoundsCheck):
* heap/Heap.cpp:
(JSC::Heap::Heap):
(JSC::Heap::reportWebAssemblyFastMemoriesAllocated):
(JSC::Heap::webAssemblyFastMemoriesThisCycleAtThreshold):
(JSC::Heap::updateAllocationLimits):
(JSC::Heap::didAllocateWebAssemblyFastMemories):
(JSC::Heap::shouldDoFullCollection):
(JSC::Heap::collectIfNecessaryOrDefer):
* heap/Heap.h:
* runtime/InitializeThreading.cpp:
(JSC::initializeThreading):
* runtime/Options.cpp:
* runtime/Options.h:
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::fixupPointerPlusOffset):
(JSC::Wasm::B3IRGenerator::B3IRGenerator):
(JSC::Wasm::B3IRGenerator::emitCheckAndPreparePointer):
(JSC::Wasm::B3IRGenerator::emitLoadOp):
(JSC::Wasm::B3IRGenerator::emitStoreOp):
(JSC::Wasm::createJSToWasmWrapper):
* wasm/WasmFaultSignalHandler.cpp:
(JSC::Wasm::trapHandler):
* wasm/WasmMemory.cpp: Rewrite.
(JSC::Wasm::makeString):
(JSC::Wasm::Memory::initializePreallocations):
(JSC::Wasm::Memory::createImpl):
(JSC::Wasm::Memory::create):
(JSC::Wasm::Memory::~Memory):
(JSC::Wasm::Memory::fastMappedRedzoneBytes):
(JSC::Wasm::Memory::fastMappedBytes):
(JSC::Wasm::Memory::maxFastMemoryCount):
(JSC::Wasm::Memory::addressIsInActiveFastMemory):
(JSC::Wasm::Memory::grow):
* wasm/WasmMemory.h:
(Memory::maxFastMemoryCount):
(Memory::addressIsInActiveFastMemory):
* wasm/js/JSWebAssemblyInstance.cpp:
(JSC::JSWebAssemblyInstance::finishCreation):
(JSC::JSWebAssemblyInstance::visitChildren):
(JSC::JSWebAssemblyInstance::globalMemoryByteSize):
* wasm/js/JSWebAssemblyInstance.h:
* wasm/js/JSWebAssemblyMemory.cpp:
(JSC::JSWebAssemblyMemory::grow):
(JSC::JSWebAssemblyMemory::finishCreation):
(JSC::JSWebAssemblyMemory::visitChildren):
2017-04-13 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC] Use proper ifdef guard for code using MachineContext
https://bugs.webkit.org/show_bug.cgi?id=170800
Reviewed by Carlos Alberto Lopez Perez.
This patch drops MachineContext use if it is not available.
This situation can be considered like, building WebKit with musl.
In that case, we simply disable features that rely on MachineContext.
Examples are wasm fast memory, sampling profiler, and code profiling.
* runtime/Options.cpp:
(JSC::overrideDefaults):
* tools/CodeProfiling.cpp:
(JSC::CodeProfiling::begin):
(JSC::CodeProfiling::end):
Previously, PLATFORM(GTK) is excluded. But it is not obvious why it is excluded.
This patch just includes such platforms.
* wasm/WasmFaultSignalHandler.cpp:
(JSC::Wasm::enableFastMemory):
2017-04-12 Dan Bernstein <mitz@apple.com>
[Mac] Future-proof .xcconfig files
https://bugs.webkit.org/show_bug.cgi?id=170802
Reviewed by Tim Horton.
* Configurations/Base.xcconfig:
* Configurations/DebugRelease.xcconfig:
* Configurations/FeatureDefines.xcconfig:
* Configurations/Version.xcconfig:
2017-04-12 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/built-ins/NativeErrors/EvalError/proto.js
https://bugs.webkit.org/show_bug.cgi?id=170668
Reviewed by Keith Miller.
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
The [[Prototype]] of NativeError Constructor's should be the %Error%.
https://tc39.github.io/ecma262/#sec-properties-of-the-nativeerror-constructors
2017-04-12 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/language/literals/regexp/u-dec-esc.js
https://bugs.webkit.org/show_bug.cgi?id=170687
Reviewed by Michael Saboff.
* yarr/YarrParser.h:
(JSC::Yarr::Parser::parseEscape):
* yarr/YarrPattern.cpp:
(JSC::Yarr::YarrPattern::errorMessage):
(JSC::Yarr::YarrPattern::compile):
* yarr/YarrPattern.h:
In unicoe patterns, invalid backreferences are an error.
2017-04-12 Filip Pizlo <fpizlo@apple.com>
Move common stack allocation utilities out of AirAllocateStackByGraphColoring.cpp
https://bugs.webkit.org/show_bug.cgi?id=170799
Reviewed by Michael Saboff and Keith Miller.
When I added stack allocation to allocateRegistersByLinearScan, I reused a handful of
utility functions from AirAllocateStackByGraphColoring.cpp. I accomplished this by
putting their declarations in AirAllocateStackByGraphColoring.h.
That was pretty weird.
This patch moves a family of stack allocation helper functions out of
AirAllocateStackByGraphColoring.cpp and into the new AirStackAllocation.h|cpp. The
linear scan stack allocator no longer has to include the other stack allocator's
header, which addresses my OCD.
I moved the functions transitively reachable from the two functions that the linear
scan allocator needed. This forced me to give them better names (i.e. no "fooBarImpl")
and short descriptive comments. I think that such comments are useful in code that is
doing a convoluted version of some theoretical concept.
No behavior change.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/air/AirAllocateRegistersAndStackByLinearScan.cpp:
* b3/air/AirAllocateStackByGraphColoring.cpp:
(JSC::B3::Air::allocateStackByGraphColoring):
(JSC::B3::Air::allocateEscapedStackSlots): Deleted.
(JSC::B3::Air::updateFrameSizeBasedOnStackSlots): Deleted.
* b3/air/AirAllocateStackByGraphColoring.h:
* b3/air/AirStackAllocation.cpp: Added.
(JSC::B3::Air::attemptAssignment):
(JSC::B3::Air::assign):
(JSC::B3::Air::allocateAndGetEscapedStackSlotsWithoutChangingFrameSize):
(JSC::B3::Air::allocateEscapedStackSlots):
(JSC::B3::Air::updateFrameSizeBasedOnStackSlots):
* b3/air/AirStackAllocation.h: Added.
2017-04-12 Filip Pizlo <fpizlo@apple.com>
B3 -O1 should not allocateStackByGraphColoring
https://bugs.webkit.org/show_bug.cgi?id=170742
Reviewed by Keith Miller.
One of B3 -O1's longest running phases is allocateStackByGraphColoring. One approach to
this would be to make that phase cheaper. But it's weird that this phase reruns
liveness after register allocation already ran liveness. If only it could reuse the
liveness computed by register allocation then it would run a lot faster. At -O2, we do
not want this, since we run phases between register allocation and stack allocation,
and those phases are free to change the liveness of spill slots (in fact,
fixObviousSpills will both shorten and lengthen live ranges because of load and store
elimination, respectively). But at -O1, we don't really need to run any phases between
register and stack allocation.
This changes Air's backend in the following ways:
- Linear scan does stack allocation. This means that we don't need to run
allocateStackByGraphColoring at all. In reality, we reuse some of its innards, but
we don't run the expensive part of it (liveness->interference->coalescing->coloring).
This is a speed-up because we only run liveness once and reuse it for both register
and stack allocation.
- Phases that previously ran between register and stack allocation are taken care of,
each in its own special way:
-> handleCalleSaves: this is now a utility function called by both
allocateStackByGraphColoring and allocateRegistersAndStackByLinearScan.
-> fixObviousSpills: we didn't run this at -O1, so nothing needs to be done.
-> lowerAfterRegAlloc: this needed to be able to run before stack allocation because
it could change register usage (vis a vis callee saves) and it could introduce
spill slots. I changed this phase to have a secondary mode for when it runs after
stack allocation.
- The part of allocateStackByGraphColoring that lowered stack addresses and took care
of the call arg area is now a separate phase called lowerStackArgs. We run this phase
regardless of optimization level. It's a cheap and general lowering.
This also removes spillEverything, because we never use that phase, we never test it,
and it got in the way in this refactoring.
This is a 21% speed-up on wasm -O1 compile times. This does not significantly change
-O1 throughput. We had already disabled allocateStack's most important optimization
(spill coalescing). This probably regresses average stack frame size, but I didn't
measure by how much. Stack frame size is really not that important. The algorithm in
allocateStackByGraphColoring is about much more than optimal frame size; it also
tries to avoid having to zero-extend 32-bit spills, it kills dead code, and of course
it coalesces.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/B3Procedure.cpp:
(JSC::B3::Procedure::calleeSaveRegisterAtOffsetList):
(JSC::B3::Procedure::calleeSaveRegisters): Deleted.
* b3/B3Procedure.h:
* b3/B3StackmapGenerationParams.cpp:
(JSC::B3::StackmapGenerationParams::unavailableRegisters):
* b3/air/AirAllocateRegistersAndStackByLinearScan.cpp: Copied from Source/JavaScriptCore/b3/air/AirAllocateRegistersByLinearScan.cpp.
(JSC::B3::Air::allocateRegistersAndStackByLinearScan):
(JSC::B3::Air::allocateRegistersByLinearScan): Deleted.
* b3/air/AirAllocateRegistersAndStackByLinearScan.h: Copied from Source/JavaScriptCore/b3/air/AirAllocateRegistersByLinearScan.h.
* b3/air/AirAllocateRegistersByLinearScan.cpp: Removed.
* b3/air/AirAllocateRegistersByLinearScan.h: Removed.
* b3/air/AirAllocateStackByGraphColoring.cpp:
(JSC::B3::Air::allocateEscapedStackSlots):
(JSC::B3::Air::updateFrameSizeBasedOnStackSlots):
(JSC::B3::Air::allocateStackByGraphColoring):
* b3/air/AirAllocateStackByGraphColoring.h:
* b3/air/AirArg.cpp:
(JSC::B3::Air::Arg::stackAddr):
* b3/air/AirArg.h:
(JSC::B3::Air::Arg::stackAddr): Deleted.
* b3/air/AirCode.cpp:
(JSC::B3::Air::Code::addStackSlot):
(JSC::B3::Air::Code::setCalleeSaveRegisterAtOffsetList):
(JSC::B3::Air::Code::calleeSaveRegisterAtOffsetList):
(JSC::B3::Air::Code::dump):
* b3/air/AirCode.h:
(JSC::B3::Air::Code::setStackIsAllocated):
(JSC::B3::Air::Code::stackIsAllocated):
(JSC::B3::Air::Code::calleeSaveRegisters):
* b3/air/AirGenerate.cpp:
(JSC::B3::Air::prepareForGeneration):
(JSC::B3::Air::generate):
* b3/air/AirHandleCalleeSaves.cpp:
(JSC::B3::Air::handleCalleeSaves):
* b3/air/AirHandleCalleeSaves.h:
* b3/air/AirLowerAfterRegAlloc.cpp:
(JSC::B3::Air::lowerAfterRegAlloc):
* b3/air/AirLowerStackArgs.cpp: Added.
(JSC::B3::Air::lowerStackArgs):
* b3/air/AirLowerStackArgs.h: Added.
* b3/testb3.cpp:
(JSC::B3::testPinRegisters):
* ftl/FTLCompile.cpp:
(JSC::FTL::compile):
* jit/RegisterAtOffsetList.h:
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::parseAndCompile):
2017-04-12 Michael Saboff <msaboff@apple.com>
Implement Object.isFrozen() and Object.isSealed() per ECMA spec
https://bugs.webkit.org/show_bug.cgi?id=170753
Reviewed by Mark Lam.
* runtime/ObjectConstructor.cpp:
(JSC::testIntegrityLevel): Added local helper as described in the ECMA standard.
(JSC::objectConstructorSeal):
(JSC::objectConstructorFreeze):
Eliminated incomplete special handling of JSFinalObjects.
(JSC::objectConstructorIsSealed):
(JSC::objectConstructorIsFrozen):
Refactored to use the new testIntegrityLevel() helper.
2017-04-12 Yusuke Suzuki <utatane.tea@gmail.com>
Use HAVE(MACHINE_CONTEXT) instead of USE(MACHINE_CONTEXT)
https://bugs.webkit.org/show_bug.cgi?id=170770
Rubber stamped by Mark Lam.
* heap/MachineStackMarker.cpp:
(JSC::MachineThreads::MachineThread::Registers::framePointer):
(JSC::MachineThreads::MachineThread::Registers::instructionPointer):
(JSC::MachineThreads::MachineThread::Registers::llintPC):
* runtime/MachineContext.h:
(JSC::MachineContext::stackPointer):
(JSC::MachineContext::framePointer):
(JSC::MachineContext::instructionPointer):
(JSC::MachineContext::argumentPointer<1>):
(JSC::MachineContext::llintInstructionPointer):
2017-04-12 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC] Clean up heap/MachineStackMarker by introducing USE(MACHINE_CONTEXT)
https://bugs.webkit.org/show_bug.cgi?id=170770
Reviewed by Mark Lam.
We use USE(MACHINE_CONTEXT) to clean up runtime/MachineContext.h. And
we clean up heap/MachineStackMarker.cpp by using MachineContext functions.
* heap/MachineStackMarker.cpp:
(JSC::MachineThreads::MachineThread::Registers::stackPointer):
(JSC::MachineThreads::MachineThread::Registers::framePointer):
(JSC::MachineThreads::MachineThread::Registers::instructionPointer):
(JSC::MachineThreads::MachineThread::Registers::llintPC):
* heap/MachineStackMarker.h:
* runtime/MachineContext.h:
(JSC::MachineContext::stackPointer):
(JSC::MachineContext::framePointer):
(JSC::MachineContext::instructionPointer):
(JSC::MachineContext::argumentPointer<1>):
(JSC::MachineContext::llintInstructionPointer):
2017-04-12 Yusuke Suzuki <utatane.tea@gmail.com>
[WTF] Introduce Thread class and use RefPtr<Thread> and align Windows Threading implementation semantics to Pthread one
https://bugs.webkit.org/show_bug.cgi?id=170502
Reviewed by Mark Lam.
* API/tests/CompareAndSwapTest.cpp:
(testCompareAndSwap):
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/air/testair.cpp:
* b3/testb3.cpp:
(JSC::B3::run):
* bytecode/SuperSampler.cpp:
(JSC::initializeSuperSampler):
* dfg/DFGWorklist.cpp:
* disassembler/Disassembler.cpp:
* heap/Heap.cpp:
(JSC::Heap::lastChanceToFinalize):
(JSC::Heap::notifyIsSafeToCollect):
* heap/Heap.h:
* heap/MachineStackMarker.cpp:
(JSC::MachineThreads::~MachineThreads):
(JSC::MachineThreads::addCurrentThread):
(JSC::MachineThreads::removeThread):
(JSC::MachineThreads::removeThreadIfFound):
(JSC::MachineThreads::MachineThread::MachineThread):
(JSC::MachineThreads::MachineThread::getRegisters):
(JSC::MachineThreads::MachineThread::Registers::stackPointer):
(JSC::MachineThreads::MachineThread::Registers::framePointer):
(JSC::MachineThreads::MachineThread::Registers::instructionPointer):
(JSC::MachineThreads::MachineThread::Registers::llintPC):
(JSC::MachineThreads::MachineThread::captureStack):
(JSC::MachineThreads::tryCopyOtherThreadStack):
(JSC::MachineThreads::tryCopyOtherThreadStacks):
(pthreadSignalHandlerSuspendResume): Deleted.
(JSC::threadData): Deleted.
(JSC::MachineThreads::Thread::Thread): Deleted.
(JSC::MachineThreads::Thread::createForCurrentThread): Deleted.
(JSC::MachineThreads::Thread::operator==): Deleted.
(JSC::MachineThreads::machineThreadForCurrentThread): Deleted.
(JSC::MachineThreads::ThreadData::ThreadData): Deleted.
(JSC::MachineThreads::ThreadData::~ThreadData): Deleted.
(JSC::MachineThreads::ThreadData::suspend): Deleted.
(JSC::MachineThreads::ThreadData::resume): Deleted.
(JSC::MachineThreads::ThreadData::getRegisters): Deleted.
(JSC::MachineThreads::ThreadData::Registers::stackPointer): Deleted.
(JSC::MachineThreads::ThreadData::Registers::framePointer): Deleted.
(JSC::MachineThreads::ThreadData::Registers::instructionPointer): Deleted.
(JSC::MachineThreads::ThreadData::Registers::llintPC): Deleted.
(JSC::MachineThreads::ThreadData::freeRegisters): Deleted.
(JSC::MachineThreads::ThreadData::captureStack): Deleted.
* heap/MachineStackMarker.h:
(JSC::MachineThreads::MachineThread::suspend):
(JSC::MachineThreads::MachineThread::resume):
(JSC::MachineThreads::MachineThread::threadID):
(JSC::MachineThreads::MachineThread::stackBase):
(JSC::MachineThreads::MachineThread::stackEnd):
(JSC::MachineThreads::threadsListHead):
(JSC::MachineThreads::Thread::operator!=): Deleted.
(JSC::MachineThreads::Thread::suspend): Deleted.
(JSC::MachineThreads::Thread::resume): Deleted.
(JSC::MachineThreads::Thread::getRegisters): Deleted.
(JSC::MachineThreads::Thread::freeRegisters): Deleted.
(JSC::MachineThreads::Thread::captureStack): Deleted.
(JSC::MachineThreads::Thread::platformThread): Deleted.
(JSC::MachineThreads::Thread::stackBase): Deleted.
(JSC::MachineThreads::Thread::stackEnd): Deleted.
* jit/ICStats.cpp:
(JSC::ICStats::ICStats):
(JSC::ICStats::~ICStats):
* jit/ICStats.h:
* jsc.cpp:
(functionDollarAgentStart):
(startTimeoutThreadIfNeeded):
* runtime/JSLock.cpp:
(JSC::JSLock::lock):
* runtime/JSLock.h:
(JSC::JSLock::ownerThread):
(JSC::JSLock::currentThreadIsHoldingLock):
* runtime/SamplingProfiler.cpp:
(JSC::FrameWalker::isValidFramePointer):
(JSC::SamplingProfiler::SamplingProfiler):
(JSC::SamplingProfiler::createThreadIfNecessary):
(JSC::SamplingProfiler::takeSample):
* runtime/SamplingProfiler.h:
* runtime/VM.h:
(JSC::VM::ownerThread):
* runtime/VMTraps.cpp:
(JSC::findActiveVMAndStackBounds):
(JSC::VMTraps::SignalSender::send):
(JSC::VMTraps::fireTrap):
2017-04-11 Dean Jackson <dino@apple.com>
Disable outdated WritableStream API
https://bugs.webkit.org/show_bug.cgi?id=170749
<rdar://problem/31446233>
Reviewed by Tim Horton.
The API we implement is no longer accurate. Disable it until we
are compatible with the new specification
* Configurations/FeatureDefines.xcconfig:
2017-04-11 Yusuke Suzuki <utatane.tea@gmail.com>
Unreviewed, build fix for CF ports after r215241
https://bugs.webkit.org/show_bug.cgi?id=170725
* heap/GCActivityCallback.cpp:
(JSC::GCActivityCallback::nextFireTime):
2017-04-11 Yusuke Suzuki <utatane.tea@gmail.com>
[WebCore][JSC] ResourceUsageData.{timeOfNextEdenCollection,timeOfNextFullCollection} should be MonotonicTime
https://bugs.webkit.org/show_bug.cgi?id=170725
Reviewed by Sam Weinig.
This patch makes GCActivityCallback return MonotonicTime instead of raw double value.
* heap/GCActivityCallback.cpp:
(JSC::GCActivityCallback::nextFireTime):
* heap/GCActivityCallback.h:
2017-04-11 Guillaume Emont <guijemont@igalia.com>
[jsc] Add missing MacroAssemblerMIPS::or32() implementation
https://bugs.webkit.org/show_bug.cgi?id=169714
Reviewed by Michael Catanzaro.
* assembler/MacroAssemblerMIPS.h:
(JSC::MacroAssemblerMIPS::or32):
Added or32(TrustedImm32, Address).
2017-04-11 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/annexB/language/comments/multi-line-html-close.js
https://bugs.webkit.org/show_bug.cgi?id=170648
Reviewed by Keith Miller.
* parser/Lexer.cpp:
(JSC::Lexer<T>::lex):
A multi-line comment that contains a line terminator is itself treated
like a line terminator. An HTML Close Comment that comes after it can
therefore treat it like it is at the start of a line, because it was
immediately preceeded by the equivalent of a line terminator.
2017-04-11 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/built-ins/Array/S15.4.3_A2.2.js
https://bugs.webkit.org/show_bug.cgi?id=170652
Reviewed by Michael Saboff.
* runtime/ArrayConstructor.cpp:
(JSC::ArrayConstructor::finishCreation):
* runtime/BooleanConstructor.cpp:
(JSC::BooleanConstructor::finishCreation):
* runtime/DateConstructor.cpp:
(JSC::DateConstructor::finishCreation):
* runtime/FunctionConstructor.cpp:
(JSC::FunctionConstructor::finishCreation):
* runtime/JSArrayBufferConstructor.cpp:
(JSC::JSArrayBufferConstructor::finishCreation):
* runtime/NumberConstructor.cpp:
(JSC::NumberConstructor::finishCreation):
* runtime/ObjectConstructor.cpp:
(JSC::ObjectConstructor::finishCreation):
* runtime/RegExpConstructor.cpp:
(JSC::RegExpConstructor::finishCreation):
* runtime/StringConstructor.cpp:
(JSC::StringConstructor::finishCreation):
* runtime/SymbolConstructor.cpp:
(JSC::SymbolConstructor::finishCreation):
Ensure the "length" property on these native constructors is configurable (deletable).
2017-04-11 Yusuke Suzuki <utatane.tea@gmail.com>
Unreviewed, build fix for Windows after r215228 part 2
https://bugs.webkit.org/show_bug.cgi?id=170723
Since GCActivityCallback class is annotated exported, we do not need to annotate each member.
* heap/GCActivityCallback.h:
2017-04-11 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC][GTK] Use RunLoop::Timer in GTK port
https://bugs.webkit.org/show_bug.cgi?id=170723
Reviewed by Carlos Garcia Campos.
This patch makes GTK port use RunLoop::Timer for JSRunLoopTimer.
Only Cocoa-based ports use platform-specific Timer because it
has additional feature that changes RunLoop to the WebThread one.
And we enable Heap timers in all the ports including JSCOnly port.
* heap/EdenGCActivityCallback.cpp:
(JSC::EdenGCActivityCallback::lastGCLength):
* heap/EdenGCActivityCallback.h:
* heap/FullGCActivityCallback.cpp:
(JSC::FullGCActivityCallback::lastGCLength):
* heap/FullGCActivityCallback.h:
* heap/GCActivityCallback.cpp:
(JSC::GCActivityCallback::GCActivityCallback):
(JSC::GCActivityCallback::doWork):
(JSC::GCActivityCallback::scheduleTimer):
(JSC::GCActivityCallback::cancelTimer):
(JSC::GCActivityCallback::nextFireTime):
(JSC::GCActivityCallback::didAllocate):
* heap/GCActivityCallback.h:
* heap/IncrementalSweeper.cpp:
(JSC::IncrementalSweeper::doWork):
(JSC::IncrementalSweeper::doSweep):
* heap/IncrementalSweeper.h:
* heap/StopIfNecessaryTimer.cpp:
(JSC::StopIfNecessaryTimer::scheduleSoon):
* runtime/JSRunLoopTimer.cpp:
(JSC::JSRunLoopTimer::setRunLoop):
(JSC::JSRunLoopTimer::scheduleTimer):
(JSC::JSRunLoopTimer::cancelTimer):
(JSC::JSRunLoopTimer::JSRunLoopTimer):
(JSC::JSRunLoopTimer::~JSRunLoopTimer):
(JSC::JSRunLoopTimer::timerDidFireCallback):
* runtime/JSRunLoopTimer.h:
* runtime/PromiseDeferredTimer.cpp:
(JSC::PromiseDeferredTimer::scheduleWorkSoon):
2017-04-11 Guillaume Emont <guijemont@igalia.com>
[jsc][mips] Add missing MacroAssembler functions after r214187
https://bugs.webkit.org/show_bug.cgi?id=170089
Reviewed by Yusuke Suzuki.
* assembler/MacroAssemblerMIPS.h:
(JSC::MacroAssemblerMIPS::loadFloat): Added.
(JSC::MacroAssemblerMIPS::storeFloat): Added.
2017-04-11 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC] Enable JSRunLoopTimer for JSCOnly and Windows
https://bugs.webkit.org/show_bug.cgi?id=170655
Reviewed by Carlos Garcia Campos.
* runtime/JSRunLoopTimer.cpp:
(JSC::JSRunLoopTimer::JSRunLoopTimer):
(JSC::JSRunLoopTimer::scheduleTimer):
(JSC::JSRunLoopTimer::cancelTimer):
* runtime/JSRunLoopTimer.h:
2017-04-10 Alex Christensen <achristensen@webkit.org>
Revert r215217
https://bugs.webkit.org/show_bug.cgi?id=170703
* Configurations/FeatureDefines.xcconfig:
2017-04-10 Alex Christensen <achristensen@webkit.org>
Continue enabling WebRTC
https://bugs.webkit.org/show_bug.cgi?id=170703
Reviewed by Youenn Fablet.
* Configurations/FeatureDefines.xcconfig:
2017-04-10 Mark Lam <mark.lam@apple.com>
Move ProbeContext and ProbeFunction out of AbstractMacroAssembler.
https://bugs.webkit.org/show_bug.cgi?id=170681
Reviewed by Michael Saboff.
This is a refactoring step towards enabling custom probe printers the way printInternal() works for dataLog.
* assembler/AbstractMacroAssembler.h:
(JSC::AbstractMacroAssembler::ProbeContext::gpr): Deleted.
(JSC::AbstractMacroAssembler::ProbeContext::fpr): Deleted.
(JSC::AbstractMacroAssembler::ProbeContext::gprName): Deleted.
(JSC::AbstractMacroAssembler::ProbeContext::fprName): Deleted.
* assembler/MacroAssembler.cpp:
(JSC::stdFunctionCallback):
(JSC::MacroAssembler::probe):
* assembler/MacroAssembler.h:
(JSC::ProbeContext::gpr):
(JSC::ProbeContext::fpr):
(JSC::ProbeContext::gprName):
(JSC::ProbeContext::fprName):
* assembler/MacroAssemblerARM.cpp:
(JSC::MacroAssemblerARM::probe):
* assembler/MacroAssemblerARM64.cpp:
(JSC::arm64ProbeTrampoline):
(JSC::MacroAssemblerARM64::probe):
* assembler/MacroAssemblerARMv7.cpp:
(JSC::MacroAssemblerARMv7::probe):
* assembler/MacroAssemblerPrinter.cpp:
* assembler/MacroAssemblerPrinter.h:
* assembler/MacroAssemblerX86Common.cpp:
(JSC::MacroAssemblerX86Common::probe):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::abstractStructure):
(JSC::FTL::DFG::LowerDFGToB3::probe): Deleted.
- Deleted because this became a useless place-holder after the transition to B3.
2017-04-10 Keith Miller <keith_miller@apple.com>
WebAssembly: Fix B3IRGenerator for BrTable
https://bugs.webkit.org/show_bug.cgi?id=170685
Reviewed by JF Bastien.
For some reason this didn't get included in r215141.
This fixes an issue with BrTable and loops where we would use the loop's return type
as the branch target type.
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::ControlData::resultForBranch):
(JSC::Wasm::B3IRGenerator::unifyValuesWithBlock):
2017-04-08 Oliver Hunt <oliver@apple.com>
Remove use of strcpy from JSC
https://bugs.webkit.org/show_bug.cgi?id=170646
Reviewed by Mark Lam.
Replace the use of strcpy with memcpy as strcpy keeps
on tripping various analyser warnings even though its
trivially safe in this case.
Essentially code hygiene, no change in behaviour, no
perf impact.
* dfg/DFGDisassembler.cpp:
(JSC::DFG::Disassembler::dumpDisassembly):
2017-04-09 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/annexB/language/expressions/object/__proto__-fn-name.js
https://bugs.webkit.org/show_bug.cgi?id=170650
Reviewed by Saam Barati.
* parser/Parser.cpp:
(JSC::Parser<LexerType>::parseClass):
(JSC::Parser<LexerType>::parseProperty):
There needs to be special handling of:
PropertyDefinition : PropertyName ':' AssignmentExpression
When the property name is __proto__. In this case the
SetFunctionName path does not happen, so the name "__proto__"
is not inferred on any anonymous function. See:
https://tc39.github.io/ecma262/#sec-__proto__-property-names-in-object-initializers
* parser/Parser.h:
* parser/SyntaxChecker.h:
(JSC::SyntaxChecker::createProperty):
* parser/ASTBuilder.h:
(JSC::ASTBuilder::createProperty):
Add an extra parameter to see if inferring / setting names are allowed.
2017-04-09 Joseph Pecoraro <pecoraro@apple.com>
test262: test262/test/annexB/language/literals/regexp/identity-escape.js
https://bugs.webkit.org/show_bug.cgi?id=170651
Reviewed by Saam Barati.
* yarr/YarrParser.h:
(JSC::Yarr::Parser::parseEscape):
For \8 and \9 match just the number "8" or "9" instead of both "\\" and the number.
See: https://tc39.github.io/ecma262/#sec-decimalescape
2017-04-08 Youenn Fablet <youenn@apple.com>
WebRTC tests gardening
https://bugs.webkit.org/show_bug.cgi?id=170508
Reviewed by Eric Carlson.
* Configurations/FeatureDefines.xcconfig:
2017-04-07 Keith Miller <keith_miller@apple.com>
WebAssembly: Fix issue with BrTable targeting a Loop
https://bugs.webkit.org/show_bug.cgi?id=170638
Reviewed by Saam Barati.
This fixes the same issue V8 had in: https://github.com/WebAssembly/spec/pull/456#event-1033547537
* wasm/WasmValidate.cpp:
(JSC::Wasm::Validate::ControlData::branchTargetSignature):
2017-04-07 Keith Miller <keith_miller@apple.com>
Add a PriorityQueue class
https://bugs.webkit.org/show_bug.cgi?id=170579
Reviewed by Saam Barati.
Update Wasm::Worklist to use WTF::PriorityQueue.
* wasm/WasmWorklist.cpp:
(JSC::Wasm::Worklist::enqueue):
(JSC::Wasm::Worklist::completePlanSynchronously):
(JSC::Wasm::Worklist::stopAllPlansForVM):
(JSC::Wasm::Worklist::~Worklist):
(JSC::Wasm::Worklist::iterate): Deleted.
* wasm/WasmWorklist.h:
(JSC::Wasm::Worklist::isHigherPriority):
(JSC::Wasm::Worklist::Comparator::operator()): Deleted.
2017-04-07 Yuichiro Kikura <y.kikura@gmail.com>
WebGPU: implement ComputeCommandEncoder and related components
https://bugs.webkit.org/show_bug.cgi?id=170444
Reviewed by Alex Christensen.
I added some identifiers related with WebGPUComputeCommandEncoder based on the proposal.
https://webkit.org/wp-content/uploads/webgpu-api-proposal.html
* runtime/CommonIdentifiers.h:
2017-04-07 Saam Barati <sbarati@apple.com>
WebAssembly: Module::getOrCreateCodeBlock is wrong
https://bugs.webkit.org/show_bug.cgi?id=170612
Reviewed by Keith Miller.
When we were getting a module's CodeBlock, we were checking if !runnable(),
and if !runnable(), we were re-creating the CodeBlock. This is wrong, since
!runnable() is true while the CodeBlock is compiling. Instead, we should check
if we've finished compiling, and if so, if that compilation failed.
* wasm/WasmModule.cpp:
(JSC::Wasm::Module::getOrCreateCodeBlock):
2017-04-07 Saam Barati <sbarati@apple.com>
WebAssembly: Make to a compilation API that allows for multi-VM concurrent compilations of Wasm Modules
https://bugs.webkit.org/show_bug.cgi?id=170488
Reviewed by JF Bastien.
This patch adds a class called Wasm::Module. It contains the bits from
JSWebAssemblyModule that were not VM specific. JSWebAssemblyModule
now has a Ref<Wasm::Module>. Similarly, there is now a Wasm::CodeBlock,
which owns the non-VM-specific bits that JSWebAssemblyCodeBlock used
to own.
This patch also simplifies how we verify and compile code. Wasm::Module
now has an API for both sync/async validation and compilation. This
API abstracts away how Wasm::Plan works.
This is hopefully the last patch needed before we can implement
window.postMessage for a JSWebAssemblyModule. I think all that's
needed now to implement postMessage is simply creating a new
JSWebAssemblyModule with the underlying Wasm::Module.
This patch is neutral on WasmBench.
Finally, this patch changes the promise deferred timer to
allow for new tasks to be added while we're executing
a task. Before, we'd deadlock if this happened.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* jsc.cpp:
(functionTestWasmModuleFunctions):
* runtime/PromiseDeferredTimer.cpp:
(JSC::PromiseDeferredTimer::doWork):
(JSC::PromiseDeferredTimer::scheduleWorkSoon):
* runtime/PromiseDeferredTimer.h:
* wasm/WasmB3IRGenerator.cpp:
* wasm/WasmBinding.cpp:
(JSC::Wasm::wasmToJs):
(JSC::Wasm::wasmToWasm):
(JSC::Wasm::exitStubGenerator): Deleted.
* wasm/WasmBinding.h:
* wasm/WasmCodeBlock.cpp: Added.
(JSC::Wasm::CodeBlock::CodeBlock):
(JSC::Wasm::CodeBlock::waitUntilFinished):
(JSC::Wasm::CodeBlock::compileAsync):
(JSC::Wasm::CodeBlock::isSafeToRun):
* wasm/WasmCodeBlock.h: Added.
(JSC::Wasm::CodeBlock::create):
(JSC::Wasm::CodeBlock::compilationFinished):
(JSC::Wasm::CodeBlock::runnable):
(JSC::Wasm::CodeBlock::errorMessage):
(JSC::Wasm::CodeBlock::functionImportCount):
(JSC::Wasm::CodeBlock::jsEntrypointCalleeFromFunctionIndexSpace):
(JSC::Wasm::CodeBlock::wasmEntrypointCalleeFromFunctionIndexSpace):
* wasm/WasmModule.cpp: Added.
(JSC::Wasm::Module::Module):
(JSC::Wasm::makeValidationResult):
(JSC::Wasm::Module::validateSyncImpl):
(JSC::Wasm::Module::getOrCreateCodeBlock):
(JSC::Wasm::Module::compileSync):
(JSC::Wasm::Module::makeValidationCallback):
(JSC::Wasm::Module::compileAsync):
* wasm/WasmModule.h: Added.
(JSC::Wasm::Module::create):
(JSC::Wasm::Module::validateSync):
(JSC::Wasm::Module::validateAsync):
(JSC::Wasm::Module::signatureIndexFromFunctionIndexSpace):
(JSC::Wasm::Module::moduleInformation):
(JSC::Wasm::Module::nonNullCodeBlock):
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::Plan):
(JSC::Wasm::Plan::addCompletionTask):
(JSC::Wasm::Plan::prepare):
(JSC::Wasm::Plan::compileFunctions):
(JSC::Wasm::Plan::complete):
(JSC::Wasm::Plan::tryRemoveVMAndCancelIfLast):
(JSC::Wasm::Plan::cancel): Deleted.
* wasm/WasmPlan.h:
(JSC::Wasm::Plan::dontFinalize):
(JSC::Wasm::Plan::takeWasmToWasmExitStubs):
(JSC::Wasm::Plan::mode):
(JSC::Wasm::Plan::takeWasmExitStubs): Deleted.
(JSC::Wasm::Plan::vm): Deleted.
* wasm/WasmWorklist.cpp:
(JSC::Wasm::Worklist::stopAllPlansForVM):
* wasm/js/JSWebAssemblyCodeBlock.cpp:
(JSC::JSWebAssemblyCodeBlock::JSWebAssemblyCodeBlock):
(JSC::JSWebAssemblyCodeBlock::isSafeToRun):
(JSC::JSWebAssemblyCodeBlock::initialize): Deleted.
* wasm/js/JSWebAssemblyCodeBlock.h:
(JSC::JSWebAssemblyCodeBlock::create):
(JSC::JSWebAssemblyCodeBlock::functionImportCount):
(JSC::JSWebAssemblyCodeBlock::jsEntrypointCalleeFromFunctionIndexSpace):
(JSC::JSWebAssemblyCodeBlock::wasmEntrypointCalleeFromFunctionIndexSpace):
(JSC::JSWebAssemblyCodeBlock::wasmToJsCallStubForImport):
(JSC::JSWebAssemblyCodeBlock::mode): Deleted.
(JSC::JSWebAssemblyCodeBlock::initialized): Deleted.
(JSC::JSWebAssemblyCodeBlock::plan): Deleted.
(JSC::JSWebAssemblyCodeBlock::runnable): Deleted.
(JSC::JSWebAssemblyCodeBlock::errorMessage): Deleted.
(JSC::JSWebAssemblyCodeBlock::setJSEntrypointCallee): Deleted.
(JSC::JSWebAssemblyCodeBlock::setWasmEntrypointCallee): Deleted.
* wasm/js/JSWebAssemblyInstance.cpp:
(JSC::JSWebAssemblyInstance::finalizeCreation):
(JSC::JSWebAssemblyInstance::addUnitializedCodeBlock): Deleted.
* wasm/js/JSWebAssemblyInstance.h:
(JSC::JSWebAssemblyInstance::initialized): Deleted.
* wasm/js/JSWebAssemblyModule.cpp:
(JSC::JSWebAssemblyModule::createStub):
(JSC::JSWebAssemblyModule::JSWebAssemblyModule):
(JSC::JSWebAssemblyModule::finishCreation):
* wasm/js/JSWebAssemblyModule.h:
(JSC::JSWebAssemblyModule::moduleInformation):
(JSC::JSWebAssemblyModule::signatureIndexFromFunctionIndexSpace):
(JSC::JSWebAssemblyModule::module):
* wasm/js/WebAssemblyFunction.cpp:
(JSC::WebAssemblyFunction::create):
* wasm/js/WebAssemblyInstanceConstructor.cpp:
(JSC::constructJSWebAssemblyInstance):
* wasm/js/WebAssemblyModuleConstructor.cpp:
(JSC::WebAssemblyModuleConstructor::createModule):
* wasm/js/WebAssemblyPrototype.cpp:
(JSC::reject):
(JSC::webAssemblyCompileFunc):
(JSC::resolve):
(JSC::instantiate):
(JSC::compileAndInstantiate):
(JSC::webAssemblyValidateFunc):
2017-04-07 Carlos Garcia Campos <cgarcia@igalia.com>
[GTK] Update the priorities used in glib main loop sources
https://bugs.webkit.org/show_bug.cgi?id=170457
Reviewed by Žan Doberšek.
* runtime/JSRunLoopTimer.cpp:
(JSC::JSRunLoopTimer::JSRunLoopTimer):
2017-04-06 Filip Pizlo <fpizlo@apple.com>
Rename allocateStack to allocateStackByGraphColoring.
Rubber stamped by Saam Barati.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/air/AirAllocateStack.cpp: Removed.
* b3/air/AirAllocateStack.h: Removed.
* b3/air/AirAllocateStackByGraphColoring.cpp: Copied from Source/JavaScriptCore/b3/air/AirAllocateStack.cpp.
(JSC::B3::Air::allocateStackByGraphColoring):
(JSC::B3::Air::allocateStack): Deleted.
* b3/air/AirAllocateStackByGraphColoring.h: Copied from Source/JavaScriptCore/b3/air/AirAllocateStack.h.
* b3/air/AirGenerate.cpp:
(JSC::B3::Air::prepareForGeneration):
2017-04-06 Michael Saboff <msaboff@apple.com>
Cannot Object.seal() or Object.freeze() global "this"
https://bugs.webkit.org/show_bug.cgi?id=170549
Reviewed by Mark Lam.
Needed to implement JSProxy::isExtensible() which returns the results of calling
the same on wrapped object.
Implemented step 11 of Runtime Semantics: EvalDeclarationInstantiation from the ECMAScript
spec to properly return a TypeError object when attempting to add properties to a
non-extensible global object.
* interpreter/Interpreter.cpp:
(JSC::Interpreter::execute):
* runtime/JSProxy.cpp:
(JSC::JSProxy::isExtensible):
* runtime/JSProxy.h:
2017-04-06 Filip Pizlo <fpizlo@apple.com>
Linear scan should run liveness only once
https://bugs.webkit.org/show_bug.cgi?id=170569
Reviewed by Keith Miller.
Air has a longstanding design bug that Tmps from different banks are indexed independently. This
means that all of our analyses over Tmps do separate GP and FP passes. This does have some
marginal benefits (the rest of the algorithm is specialized for Bank) but it's probably net bad.
However, I don't want to think about solving that general problem.
Instead, this just makes linear scan use a UnifiedTmpLiveness that uses a single "linear"
indexing for GP and FP. This lets me avoid the much larger refactoring (which would involve
substantial changes in graph coloring) while getting the bulk of the benefit (liveness runs once,
instead of twice, for linear scan).
This patch implements a lot of plumbing to make it possible for Liveness<> to view Tmps as having
a unified indexing scheme. Tmp calls this LinearlyIndexed (to match the naming convention of
AbsolutelyIndexed and Indexed), while AirLiveness calls this UnifiedTmpLiveness. With this
change, -O1 never does any liveness analysis that uses separate GP and FP passes. I think this
eliminates any urgency from the larger Tmp indexing bug. We can probably live with graph coloring
doing separate passes.
This is a ~6% speed-up for wasm -O1 compile times. I think this means that linear scan is no
longer the longest pole in the tent.
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/B3VariableLiveness.h:
(JSC::B3::VariableLivenessAdapter::prepareToCompute):
* b3/air/AirAllocateRegistersByLinearScan.cpp:
(JSC::B3::Air::allocateRegistersByLinearScan):
* b3/air/AirCode.h:
(JSC::B3::Air::Code::forEachTmp):
* b3/air/AirLiveness.h:
* b3/air/AirLivenessAdapter.h:
(JSC::B3::Air::LivenessAdapter::Actions::Actions):
(JSC::B3::Air::LivenessAdapter::LivenessAdapter):
(JSC::B3::Air::LivenessAdapter::adapter):
(JSC::B3::Air::LivenessAdapter::prepareToCompute):
(JSC::B3::Air::LivenessAdapter::actionsAt):
(JSC::B3::Air::LivenessAdapter::forEachUse):
(JSC::B3::Air::LivenessAdapter::forEachDef):
(JSC::B3::Air::TmpLivenessAdapter::numIndices):
(JSC::B3::Air::UnifiedTmpLivenessAdapter::UnifiedTmpLivenessAdapter):
(JSC::B3::Air::UnifiedTmpLivenessAdapter::numIndices):
(JSC::B3::Air::UnifiedTmpLivenessAdapter::acceptsBank):
(JSC::B3::Air::UnifiedTmpLivenessAdapter::acceptsRole):
(JSC::B3::Air::UnifiedTmpLivenessAdapter::valueToIndex):
(JSC::B3::Air::UnifiedTmpLivenessAdapter::indexToValue):
* b3/air/AirLivenessConstraints.h: Removed.
* b3/air/AirRegLiveness.h:
(JSC::B3::Air::RegLiveness::LocalCalc::LocalCalc):
* b3/air/AirTmp.cpp:
* b3/air/AirTmp.h:
* b3/air/AirTmpInlines.h:
(JSC::B3::Air::Tmp::LinearlyIndexed::LinearlyIndexed):
(JSC::B3::Air::Tmp::LinearlyIndexed::index):
(JSC::B3::Air::Tmp::linearlyIndexed):
(JSC::B3::Air::Tmp::indexEnd):
(JSC::B3::Air::Tmp::absoluteIndexEnd):
(JSC::B3::Air::Tmp::linearIndexEnd):
(JSC::B3::Air::Tmp::tmpForAbsoluteIndex):
(JSC::B3::Air::Tmp::tmpForLinearIndex):
* b3/air/AirTmpMap.h: Added.
(JSC::B3::Air::TmpMap::TmpMap):
(JSC::B3::Air::TmpMap::resize):
(JSC::B3::Air::TmpMap::clear):
(JSC::B3::Air::TmpMap::operator[]):
(JSC::B3::Air::TmpMap::append):
2017-04-06 Ryan Haddad <ryanhaddad@apple.com>
Unreviewed, rolling out r215046.
This change broke internal builds.
Reverted changeset:
"WebRTC tests gardening"
https://bugs.webkit.org/show_bug.cgi?id=170508
http://trac.webkit.org/changeset/215046
2017-04-06 Joseph Pecoraro <pecoraro@apple.com>
Web Inspector: Show all headers in the Request Headers section of the Resource details sidebar
https://bugs.webkit.org/show_bug.cgi?id=16531
<rdar://problem/5712895>
Reviewed by Timothy Hatcher.
* inspector/protocol/Network.json:
Optional refined list of request headers in Metrics.
2017-04-06 Filip Pizlo <fpizlo@apple.com>
B3 -O1 should generate better code than -O0
https://bugs.webkit.org/show_bug.cgi?id=170563
Reviewed by Michael Saboff.
Prior to this change, code generated by -O1 ran slower than code generated by -O0. This turned
out to be because of reduceStrength optimizations that increase live ranges and create register
pressure, which then creates problems for linear scan.
It seemed obvious that canonicalizations that help isel, constant folding, and one-for-one
strength reductions should stay. It also seemed obvious that SSA and CFG simplification are fast
and harmless. So, I focused on removing:
- CSE, which increases live ranges. This is a risky optimization when we know that we've chosen
to use a bad register allocator.
- Sophisticated strength reductions that create more code, like the insane division optimization.
- Anything that inserts basic blocks.
CSE appeared to be the cause of half of the throughput regression of -O1 but none of the compile
time. This change also reduces the running time of reduceStrength by making it not a fixpoint at
optLevel<2.
This makes wasm -O1 compile 17% faster. This makes wasm -O1 run 19% faster. This makes -O1 code
run 3% faster than -O0, and compile about 4% slower than -O0. We may yet end up choosing to use
-O0, but at least now -O1 isn't totally useless.
* b3/B3ReduceStrength.cpp:
2017-04-06 Jon Davis <jond@apple.com>
Updates feature status for recently shipped features
https://bugs.webkit.org/show_bug.cgi?id=170359
Reviewed by Brian Burg.
Changed "Done" status to "Supported".
* features.json:
2017-04-06 Youenn Fablet <youenn@apple.com>
WebRTC tests gardening
https://bugs.webkit.org/show_bug.cgi?id=170508
Reviewed by Eric Carlson.
* Configurations/FeatureDefines.xcconfig:
2017-04-06 Guillaume Emont <guijemont@igalia.com>
[JSC][MIPS][DFG] Use x86 generic HasOwnProperty
https://bugs.webkit.org/show_bug.cgi?id=170222
Reviewed by Yusuke Suzuki.
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
use the X86 special version for HasOwnProperty on MIPS too.
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
use the X86 special version for HasOwnProperty on MIPS too.
2017-04-05 Saam Barati <sbarati@apple.com>
REGRESSION fix bad isWasm() test by ensuring proper Wasm callee bit pattern
https://bugs.webkit.org/show_bug.cgi?id=170494
<rdar://problem/31446485>
Reviewed by Yusuke Suzuki and Mark Lam.
This patch fixes how we test a 64 bit JSValue pattern to see if it's
a Wasm callee. We now tag Wasm::Callee's with 0b011 in their lower 3 bits.
The new test is for a Wasm Callee is as follows:
isWasm(uint64_t x)
{
return x & 0xffff000000000007 == 3;
}
This test works because the lower 3 bits of the non-number immediate values are as follows:
undefined: 0b010
null: 0b010
true: 0b111
false: 0b110
The test rejects all of these because none have just the value 3 in their lower 3 bits.
The test also rejects all numbers, because they have non-zero upper 16 bits.
The test also rejects normal cells because they won't have the number 3 as
their lower 3 bits. Note, this bit pattern also allows the normal JSValue isCell(), etc,
predicates to work on a Wasm::Callee because the various tests will fail if you
bit casted a boxed Wasm::Callee* to a JSValue. isCell() would fail since it sees
TagBitTypeOther. The other tests also trivially fail, since it won't be a number,
and it won't be equal to null, undefined, true, or false. The isBoolean() predicate
will fail because we won't have TagBitBool set.
* interpreter/CallFrame.h:
(JSC::ExecState::guaranteedJSValueCallee):
(JSC::ExecState::calleeAsValue): Deleted.
* interpreter/CalleeBits.h:
(JSC::CalleeBits::boxWasm):
(JSC::CalleeBits::isWasm):
(JSC::CalleeBits::asWasmCallee):
* jit/JITOperations.cpp:
* runtime/JSCJSValue.h:
2017-04-05 Keith Miller <keith_miller@apple.com>
WebAssembly: Plans should be able to have more than one completion task.
https://bugs.webkit.org/show_bug.cgi?id=170516
Reviewed by Saam Barati.
This patch also eliminates the need for blocked tasks on the
PromiseDeferredTimer and pendingPromise on Wasm::Plan.
* runtime/PromiseDeferredTimer.cpp:
(JSC::PromiseDeferredTimer::doWork):
(JSC::PromiseDeferredTimer::cancelPendingPromise):
(JSC::PromiseDeferredTimer::scheduleBlockedTask): Deleted.
* runtime/PromiseDeferredTimer.h:
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::Plan):
(JSC::Wasm::Plan::addCompletionTask):
(JSC::Wasm::Plan::complete):
* wasm/WasmPlan.h:
(JSC::Wasm::Plan::setMode):
(JSC::Wasm::Plan::mode):
(JSC::Wasm::Plan::setModeAndPromise): Deleted.
(JSC::Wasm::Plan::pendingPromise): Deleted.
* wasm/WasmWorklist.cpp:
(JSC::Wasm::Worklist::enqueue):
* wasm/js/WebAssemblyInstanceConstructor.cpp:
(JSC::constructJSWebAssemblyInstance):
* wasm/js/WebAssemblyPrototype.cpp:
(JSC::instantiate):
2017-04-05 Guilherme Iscaro <iscaro@profusion.mobi>
Do not use BLX for immediates (ARM-32)
https://bugs.webkit.org/show_bug.cgi?id=170351
Reviewed by Mark Lam.
Currently the offline asm generator for 32-bit ARM code translates the
'call' meta-instruction (which may be found in LowLevelInterpreter.asm
and friends) to the ARM's BLX instrunction. The BLX instruction may be
used for labels (immediates) and registers and one side effect of BLX
is that it may switch the processor's instruction set.
A 'BLX register' instruction will change/remain the processor state to
ARM if the register_bit[0] is set to 0 or change/remain to Thumb if
register_bit[0] is set to 1. However, a 'BLX label' instruction will
always switch the processor state. It switches ARM to thumb and vice-versa.
This behaviour is unwanted, since the C++ code and the offlineasm generated code
are both compiled using the same instruction set, thus a instruction
set change will likely produce a crash. In order to fix the problem the
BL instruction can be used for labels. It will branch just like BLX,
but it won't change the instruction set. It's important to note that
Darwin is not affected by this problem, thus to minimize the impact of
this change the BL instruction will only be used on non-darwin targets.
BLX reference: http://infocenter.arm.com/help/topic/com.arm.doc.dui0489i/CIHBJCDC.html?resultof=%22%62%6c%78%22%20
* offlineasm/arm.rb:
2017-04-05 Keith Miller <keith_miller@apple.com>
WebAssembly: We shouldn't need to pin size registers if we have a fast memory.
https://bugs.webkit.org/show_bug.cgi?id=170504
Reviewed by Mark Lam.
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::B3IRGenerator):
(JSC::Wasm::createJSToWasmWrapper):
(JSC::Wasm::parseAndCompile):
* wasm/WasmMemoryInformation.h:
(JSC::Wasm::PinnedRegisterInfo::toSave):
2017-04-05 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC] Suppress warnings in GCC
https://bugs.webkit.org/show_bug.cgi?id=170501
Reviewed by Keith Miller.
Should use ASSERT_NOT_REACHED since return-type pragma is only
enabled under ASSERT_DISABLED environment. We shoud use
ASSERT_NOTREACHED to emit assertions in debug build. It effectively
catches bugs while keeping performance in release build.
* b3/B3Opcode.cpp:
(JSC::B3::storeOpcode):
* b3/B3Width.h:
(JSC::B3::mask):
* runtime/Options.cpp:
(JSC::parse):
* wasm/WasmSections.h:
(JSC::Wasm::makeString):
* wasm/WasmSignature.cpp:
(JSC::Wasm::SignatureInformation::tryCleanup):
* wasm/generateWasmValidateInlinesHeader.py:
2017-04-05 Carlos Garcia Campos <cgarcia@igalia.com>
Implement PromiseDeferredTimer for non CF based ports
https://bugs.webkit.org/show_bug.cgi?id=170391
Reviewed by Yusuke Suzuki.
RunLoop handling is only implemented for CF causing several wasm tests to fail for other ports.
* jsc.cpp:
(runJSC): Remove CF ifdefs.
* runtime/PromiseDeferredTimer.cpp:
(JSC::PromiseDeferredTimer::doWork): Add non CF implementation using WTF RunLoop.
(JSC::PromiseDeferredTimer::runRunLoop): Ditto.
* runtime/PromiseDeferredTimer.h:
2017-04-05 Carlos Garcia Campos <cgarcia@igalia.com>
WebAssembly: several tests added in r214504 crash when building with GCC
https://bugs.webkit.org/show_bug.cgi?id=170390
Reviewed by Saam Barati.
The pattern foo->bar([f = WTFMove(foo)]{}); crashes when building with GCC, I assume the move happens before the
foo is used to invoke the function.
* wasm/js/WebAssemblyPrototype.cpp:
(JSC::webAssemblyCompileFunc): Use p.vm() instead of plan->vm(), because plan is moved by the lambda.
(JSC::instantiate): Ditto.
(JSC::compileAndInstantiate): Ditto.
2017-03-16 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC] Generate TemplateObjects at linking time
https://bugs.webkit.org/show_bug.cgi?id=169743
Reviewed by Keith Miller.
Currently, the code calls getTemplateObject to get appropriate template objects at runtime.
But this template object is constant value and never changed. So instead of creating it
at runtime, we should create it at linking time and store it in the constant registers.
* builtins/BuiltinNames.h:
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::finishCreation):
(JSC::CodeBlock::setConstantRegisters):
* bytecode/CodeBlock.h:
* bytecode/UnlinkedCodeBlock.cpp:
(JSC::UnlinkedCodeBlock::shrinkToFit):
* bytecode/UnlinkedCodeBlock.h:
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::addTemplateRegistryKeyConstant):
(JSC::BytecodeGenerator::emitGetTemplateObject):
* bytecompiler/BytecodeGenerator.h:
* bytecompiler/NodesCodegen.cpp:
(JSC::TaggedTemplateNode::emitBytecode):
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
(JSC::getTemplateObject): Deleted.
* runtime/JSTemplateRegistryKey.cpp:
* runtime/JSTemplateRegistryKey.h:
(JSC::isTemplateRegistryKey):
2017-04-04 Mark Lam <mark.lam@apple.com>
On ARM64, DFG::SpeculativeJIT::compileArithMod() failed to ensure result is of DataFormatInt32.
https://bugs.webkit.org/show_bug.cgi?id=170473
<rdar://problem/29912391>
Reviewed by Saam Barati.
In Unchecked mode, when DFG::SpeculativeJIT::compileArithMod() detects that the
divisor is 0, we want it to return 0. The result is expected to be of
DataFormatIn32.
The ARM implementation just returns the value in the divisor register. However,
the divisor in this case can be of DataFormatJSInt32. On ARM64, returning the
divisor register yields the wrong result format because the same register also
holds the upper 32-bit of the JSValue encoding. The fix is to return an
immediate 0 instead.
Also turned on the assertion in jitAssertIsInt32 for ARM64. This assertion being
disabled may have contributed to this bug going unnoticed all this time.
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileArithMod):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::jitAssertIsInt32):
2017-04-04 Filip Pizlo <fpizlo@apple.com>
Air::eliminateDeadCode should not repeatedly process the same live instructions
https://bugs.webkit.org/show_bug.cgi?id=170490
Reviewed by Keith Miller.
This makes the eliminateDeadCode() fixpoint somewhat worklist-based: we track the set
of Insts that might be dead. Every time we detect that one is live, we remove it from
the set. This is a big (>2x) speed-up because lots of Insts are immediately found to
be live.
This is a ~1% wasm -O1 compile time progression.
* b3/air/AirEliminateDeadCode.cpp:
(JSC::B3::Air::eliminateDeadCode):
2017-04-04 Filip Pizlo <fpizlo@apple.com>
Air::eliminateDeadCode() should not use a HashSet
https://bugs.webkit.org/show_bug.cgi?id=170487
Reviewed by Saam Barati.
Introduce TmpSet, which is like a HashSet<Tmp>. Use this to make eliminateDeadCode()
about 50% faster, resulting in a 1% wasm -O1 compile time progression.
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/air/AirEliminateDeadCode.cpp:
(JSC::B3::Air::eliminateDeadCode):
* b3/air/AirTmpSet.h: Added.
(JSC::B3::Air::TmpSet::TmpSet):
(JSC::B3::Air::TmpSet::add):
(JSC::B3::Air::TmpSet::remove):
(JSC::B3::Air::TmpSet::contains):
(JSC::B3::Air::TmpSet::size):
(JSC::B3::Air::TmpSet::isEmpty):
(JSC::B3::Air::TmpSet::iterator::iterator):
(JSC::B3::Air::TmpSet::iterator::operator*):
(JSC::B3::Air::TmpSet::iterator::operator++):
(JSC::B3::Air::TmpSet::iterator::operator==):
(JSC::B3::Air::TmpSet::iterator::operator!=):
(JSC::B3::Air::TmpSet::begin):
(JSC::B3::Air::TmpSet::end):
2017-04-04 Keith Miller <keith_miller@apple.com>
WebAssembly: ModuleInformation should be a ref counted thing that can be shared across threads.
https://bugs.webkit.org/show_bug.cgi?id=170478
Reviewed by Saam Barati.
ModuleInformation has been moved to its own file and is now
ThreadSafeRefCounted. All the Strings we used to keep in the
ModuleInformation have been switched to Vector<LChar> this has the
advantage that it can be passed across threads. However, this does
mean that we need to decode the utf8 strings in each thread. This
is likely not a problem because:
1) most modules have few imports/exports/custom sections.
2) most of the time they are ascii so the conversion is cheap.
3) we only have to do it once per thread, and there shouldn't be too many.
This patch also removes
moduleSignatureIndicesToUniquedSignatureIndices since that
information can already be recovered from the
SignatureInformation.
* JavaScriptCore.xcodeproj/project.pbxproj:
* jsc.cpp:
(functionTestWasmModuleFunctions):
* runtime/Identifier.h:
(JSC::Identifier::fromString):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::parseAndCompile):
* wasm/WasmB3IRGenerator.h:
* wasm/WasmFormat.cpp:
(JSC::Wasm::makeString):
(JSC::Wasm::ModuleInformation::~ModuleInformation): Deleted.
* wasm/WasmFormat.h:
(JSC::Wasm::makeString):
(JSC::Wasm::ModuleInformation::functionIndexSpaceSize): Deleted.
(JSC::Wasm::ModuleInformation::isImportedFunctionFromFunctionIndexSpace): Deleted.
(JSC::Wasm::ModuleInformation::signatureIndexFromFunctionIndexSpace): Deleted.
(JSC::Wasm::ModuleInformation::importFunctionCount): Deleted.
(JSC::Wasm::ModuleInformation::internalFunctionCount): Deleted.
* wasm/WasmFunctionParser.h:
(JSC::Wasm::FunctionParser<Context>::FunctionParser):
* wasm/WasmModuleInformation.cpp: Copied from Source/JavaScriptCore/wasm/WasmValidate.h.
(JSC::Wasm::ModuleInformation::~ModuleInformation):
* wasm/WasmModuleInformation.h: Added.
(JSC::Wasm::ModuleInformation::functionIndexSpaceSize):
(JSC::Wasm::ModuleInformation::isImportedFunctionFromFunctionIndexSpace):
(JSC::Wasm::ModuleInformation::signatureIndexFromFunctionIndexSpace):
(JSC::Wasm::ModuleInformation::importFunctionCount):
(JSC::Wasm::ModuleInformation::internalFunctionCount):
(JSC::Wasm::ModuleInformation::ModuleInformation):
* wasm/WasmModuleParser.cpp:
* wasm/WasmModuleParser.h:
(JSC::Wasm::ModuleParser::ModuleParser):
* wasm/WasmParser.h:
(JSC::Wasm::Parser<SuccessType>::consumeUTF8String):
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::Plan):
(JSC::Wasm::Plan::parseAndValidateModule):
(JSC::Wasm::Plan::prepare):
(JSC::Wasm::Plan::compileFunctions):
(JSC::Wasm::Plan::complete):
(JSC::Wasm::Plan::cancel):
* wasm/WasmPlan.h:
(JSC::Wasm::Plan::internalFunctionCount):
(JSC::Wasm::Plan::takeModuleInformation):
* wasm/WasmSignature.cpp:
(JSC::Wasm::SignatureInformation::get):
* wasm/WasmSignature.h:
* wasm/WasmValidate.cpp:
(JSC::Wasm::validateFunction):
* wasm/WasmValidate.h:
* wasm/js/JSWebAssemblyHelpers.h:
(JSC::createSourceBufferFromValue):
* wasm/js/JSWebAssemblyModule.cpp:
(JSC::JSWebAssemblyModule::createStub):
(JSC::JSWebAssemblyModule::JSWebAssemblyModule):
(JSC::JSWebAssemblyModule::finishCreation):
* wasm/js/JSWebAssemblyModule.h:
(JSC::JSWebAssemblyModule::moduleInformation):
(JSC::JSWebAssemblyModule::source):
* wasm/js/WebAssemblyInstanceConstructor.cpp:
(JSC::constructJSWebAssemblyInstance):
* wasm/js/WebAssemblyModuleConstructor.cpp:
(JSC::WebAssemblyModuleConstructor::createModule):
* wasm/js/WebAssemblyModulePrototype.cpp:
(JSC::webAssemblyModuleProtoCustomSections):
(JSC::webAssemblyModuleProtoImports):
(JSC::webAssemblyModuleProtoExports):
* wasm/js/WebAssemblyModuleRecord.cpp:
(JSC::WebAssemblyModuleRecord::link):
* wasm/js/WebAssemblyModuleRecord.h:
* wasm/js/WebAssemblyPrototype.cpp:
(JSC::webAssemblyCompileFunc):
(JSC::instantiate):
(JSC::compileAndInstantiate):
2017-04-04 Filip Pizlo <fpizlo@apple.com>
B3::fixSSA() needs a tune-up
https://bugs.webkit.org/show_bug.cgi?id=170485
Reviewed by Saam Barati.
After the various optimizations to liveness, register allocation, and other phases, the
fixSSA() phase now looks like one of the top offenders. This includes a bunch of
changes to make this phase run faster. This is a ~7% wasm -O1 compile time progression.
Here's what I did:
- We now use IndexSparseSet instead of IndexMap for tracking variable values. This
makes it cheaper to chew through small blocks while there is a non-trivial number of
total variables.
- We now do a "local SSA conversion" pass before anything else. This eliminates
obvious Get's. If we were using temporary Variables, it would eliminate many of
those. That's useful for when we use demoteValues() and duplciateTails(). For wasm
-O1, we mainly care about the fact that it makes a bunch of Set's dead.
- We now do a Set DCE pass after the local SSA but before SSA conversion. This ensures
that any block-local live intervals of Variables disappear and don't need further
consideration.
- We now cache the reaching defs calculation.
- We now perform the reaching defs calculation lazily.
* b3/B3FixSSA.cpp:
(JSC::B3::demoteValues):
(JSC::B3::fixSSA):
* b3/B3SSACalculator.cpp:
(JSC::B3::SSACalculator::reachingDefAtTail):
* b3/B3VariableLiveness.cpp:
(JSC::B3::VariableLiveness::VariableLiveness):
* b3/air/AirLiveness.h:
(JSC::B3::Air::Liveness::Liveness):
* dfg/DFGLivenessAnalysisPhase.cpp:
(JSC::DFG::LivenessAnalysisPhase::LivenessAnalysisPhase): Deleted.
(JSC::DFG::LivenessAnalysisPhase::run): Deleted.
(JSC::DFG::LivenessAnalysisPhase::processBlock): Deleted.
2017-04-04 Joseph Pecoraro <pecoraro@apple.com>
Remove stale LLVM Header Path includes from JavaScriptCore
https://bugs.webkit.org/show_bug.cgi?id=170483
Reviewed by Mark Lam.
* Configurations/Base.xcconfig:
2017-04-04 Filip Pizlo <fpizlo@apple.com>
B3::LowerToAir incorrectly selects BitXor(AtomicStrongCAS(...), $1)
https://bugs.webkit.org/show_bug.cgi?id=169867
Reviewed by Saam Barati.
The BitXor(AtomicWeakCAS(...), $1) optimization makes a lot of sense because we an fold the
BitXor into the CAS condition read-out. But there is no version of this that is profitable or
correct for AtomicStrongCAS. The inversion case is handled by Equal(AtomicStrongCAS(...), ...)
becoming NotEqual(AtomicStrongCAS(...), ...), and we alraedy handle that separately.
So, the fix here is to make the BitXor CAS pattern only recognize AtomicWeakCAS.
* b3/B3LowerToAir.cpp:
(JSC::B3::Air::LowerToAir::lower):
* b3/testb3.cpp:
(JSC::B3::testAtomicStrongCAS):
2017-04-04 Saam Barati <sbarati@apple.com>
WebAssembly: JSWebAssemblyCallee should not be a JSCell
https://bugs.webkit.org/show_bug.cgi?id=170135
Reviewed by Michael Saboff.
This patch is perhaps the last big change to the design of fundamental
Wasm API to allow for PIC. It changes JSWebAssemblyCallee into a thing
called Wasm::Callee. It serves the same purpose as before, except
Wasm::Callee is not a JSCell. I had to refactor the various parts of the
runtime that will see CallFrame's with Wasm::Callee's in the callee slot.
Thankfully, the parts of the runtime that Wasm touches are limited. The
main refactoring is changing the exception handling code, such as taking
a stack trace, to be friendly to seeing a non JSCell callee.
The callee() function on ExecState now returns a class I added in this
patch called CalleeBits. CalleeBits will tell you if the callee is a
JSCell or a Wasm::Callee. We tag Wasm::Callee's with a 1 in their lower
bit so we can easily tell what is and isn't a Wasm::Callee.
The stub that calls out from Wasm to JS still puts a JSCell callee
into the call frame, even though the callee logically represents a
Wasm frame. The reason for this is that we use the call IC infrastructure
to make a call out to JS code, and the code that writes the IC expects
a JSCell as the callee. This is knowingly part of our design. When we
do structured cloning of Wasm Modules, we'll need to regenerate these
JS call stubs.
* API/JSContextRef.cpp:
(BacktraceFunctor::operator()):
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* debugger/Debugger.cpp:
(JSC::Debugger::pauseIfNeeded):
(JSC::Debugger::currentDebuggerCallFrame):
* debugger/DebuggerCallFrame.cpp:
(JSC::DebuggerCallFrame::create):
(JSC::DebuggerCallFrame::DebuggerCallFrame):
(JSC::DebuggerCallFrame::currentPosition):
(JSC::DebuggerCallFrame::positionForCallFrame):
* debugger/DebuggerCallFrame.h:
* interpreter/CallFrame.cpp:
(JSC::CallFrame::vmEntryGlobalObject):
(JSC::CallFrame::wasmAwareLexicalGlobalObject):
(JSC::CallFrame::isAnyWasmCallee):
(JSC::CallFrame::callerSourceOrigin):
* interpreter/CallFrame.h:
(JSC::ExecState::calleeAsValue):
(JSC::ExecState::jsCallee):
(JSC::ExecState::callee):
(JSC::ExecState::unsafeCallee):
(JSC::ExecState::scope):
(JSC::ExecState::iterate):
* interpreter/CalleeBits.h: Added.
(JSC::CalleeBits::CalleeBits):
(JSC::CalleeBits::operator=):
(JSC::CalleeBits::boxWasm):
(JSC::CalleeBits::isWasm):
(JSC::CalleeBits::isCell):
(JSC::CalleeBits::asCell):
(JSC::CalleeBits::asWasmCallee):
(JSC::CalleeBits::rawPtr):
* interpreter/Interpreter.cpp:
(JSC::GetStackTraceFunctor::operator()):
(JSC::Interpreter::getStackTrace):
(JSC::notifyDebuggerOfUnwinding):
(JSC::UnwindFunctor::UnwindFunctor):
(JSC::UnwindFunctor::operator()):
(JSC::UnwindFunctor::copyCalleeSavesToVMEntryFrameCalleeSavesBuffer):
(JSC::Interpreter::unwind):
(JSC::Interpreter::notifyDebuggerOfExceptionToBeThrown):
* interpreter/Interpreter.h:
* interpreter/Register.h:
(JSC::Register::pointer):
* interpreter/ShadowChicken.cpp:
(JSC::ShadowChicken::update):
* interpreter/ShadowChickenInlines.h:
(JSC::ShadowChicken::iterate):
* interpreter/StackVisitor.cpp:
(JSC::StackVisitor::StackVisitor):
(JSC::StackVisitor::readFrame):
(JSC::StackVisitor::readNonInlinedFrame):
(JSC::StackVisitor::readInlinedFrame):
(JSC::StackVisitor::Frame::calleeSaveRegisters):
(JSC::StackVisitor::Frame::functionName):
(JSC::StackVisitor::Frame::dump):
* interpreter/StackVisitor.h:
(JSC::StackVisitor::Frame::callee):
(JSC::StackVisitor::visit):
* jit/Repatch.cpp:
(JSC::linkFor):
(JSC::linkPolymorphicCall):
* jsc.cpp:
(callWasmFunction):
(functionTestWasmModuleFunctions):
* runtime/ArrayPrototype.cpp:
* runtime/Error.cpp:
(JSC::addErrorInfoAndGetBytecodeOffset):
* runtime/ErrorInstance.cpp:
(JSC::ErrorInstance::finishCreation):
* runtime/JSCell.cpp:
(JSC::JSCell::isAnyWasmCallee): Deleted.
* runtime/JSCell.h:
* runtime/JSCellInlines.h:
(JSC::ExecState::vm):
* runtime/JSFunction.cpp:
(JSC::RetrieveArgumentsFunctor::operator()):
(JSC::RetrieveCallerFunctionFunctor::operator()):
* runtime/JSGlobalObject.cpp:
* runtime/SamplingProfiler.cpp:
(JSC::FrameWalker::recordJSFrame):
(JSC::SamplingProfiler::processUnverifiedStackTraces):
* runtime/SamplingProfiler.h:
(JSC::SamplingProfiler::UnprocessedStackFrame::UnprocessedStackFrame):
* runtime/StackFrame.cpp:
(JSC::StackFrame::sourceURL):
(JSC::StackFrame::functionName):
* runtime/StackFrame.h:
(JSC::StackFrame::wasm):
* runtime/VM.cpp:
(JSC::VM::VM):
(JSC::VM::throwException):
* runtime/VM.h:
* wasm/JSWebAssembly.h:
* wasm/WasmB3IRGenerator.cpp:
* wasm/WasmBinding.cpp:
(JSC::Wasm::wasmToWasm):
* wasm/WasmCallee.cpp: Copied from Source/JavaScriptCore/wasm/js/JSWebAssemblyCallee.cpp.
(JSC::Wasm::Callee::Callee):
(JSC::JSWebAssemblyCallee::JSWebAssemblyCallee): Deleted.
(JSC::JSWebAssemblyCallee::finishCreation): Deleted.
(JSC::JSWebAssemblyCallee::destroy): Deleted.
* wasm/WasmCallee.h: Copied from Source/JavaScriptCore/wasm/js/JSWebAssemblyCallee.h.
(JSC::Wasm::Callee::create):
(JSC::JSWebAssemblyCallee::create): Deleted.
(JSC::JSWebAssemblyCallee::createStructure): Deleted.
(JSC::JSWebAssemblyCallee::entrypoint): Deleted.
(JSC::JSWebAssemblyCallee::calleeSaveRegisters): Deleted.
* wasm/WasmContext.h:
* wasm/WasmPlan.cpp:
* wasm/WasmPlan.h:
* wasm/WasmPlanInlines.h:
(JSC::Wasm::Plan::initializeCallees):
* wasm/WasmThunks.cpp:
(JSC::Wasm::throwExceptionFromWasmThunkGenerator):
* wasm/js/JSWebAssemblyCallee.cpp: Removed.
* wasm/js/JSWebAssemblyCallee.h: Removed.
* wasm/js/JSWebAssemblyCodeBlock.cpp:
(JSC::JSWebAssemblyCodeBlock::JSWebAssemblyCodeBlock):
(JSC::JSWebAssemblyCodeBlock::initialize):
(JSC::JSWebAssemblyCodeBlock::visitChildren):
* wasm/js/JSWebAssemblyCodeBlock.h:
(JSC::JSWebAssemblyCodeBlock::create):
(JSC::JSWebAssemblyCodeBlock::jsEntrypointCalleeFromFunctionIndexSpace):
(JSC::JSWebAssemblyCodeBlock::wasmEntrypointCalleeFromFunctionIndexSpace):
(JSC::JSWebAssemblyCodeBlock::wasmToJsCallStubForImport):
(JSC::JSWebAssemblyCodeBlock::offsetOfImportWasmToJSStub):
(JSC::JSWebAssemblyCodeBlock::setJSEntrypointCallee):
(JSC::JSWebAssemblyCodeBlock::setWasmEntrypointCallee):
(JSC::JSWebAssemblyCodeBlock::offsetOfImportStubs):
(JSC::JSWebAssemblyCodeBlock::allocationSize):
(JSC::JSWebAssemblyCodeBlock::importWasmToJSStub):
(JSC::JSWebAssemblyCodeBlock::callees): Deleted.
(JSC::JSWebAssemblyCodeBlock::offsetOfCallees): Deleted.
* wasm/js/JSWebAssemblyInstance.h:
(JSC::JSWebAssemblyInstance::webAssemblyToJSCallee):
* wasm/js/JSWebAssemblyModule.cpp:
* wasm/js/WebAssemblyFunction.cpp:
(JSC::callWebAssemblyFunction):
(JSC::WebAssemblyFunction::create):
(JSC::WebAssemblyFunction::WebAssemblyFunction):
(JSC::WebAssemblyFunction::visitChildren):
(JSC::WebAssemblyFunction::finishCreation):
* wasm/js/WebAssemblyFunction.h:
(JSC::WebAssemblyFunction::wasmEntrypoint):
(JSC::WebAssemblyFunction::jsEntrypoint):
(JSC::WebAssemblyFunction::offsetOfWasmEntrypoint):
(JSC::WebAssemblyFunction::offsetOfWasmEntryPointCode): Deleted.
* wasm/js/WebAssemblyModuleConstructor.cpp:
* wasm/js/WebAssemblyModuleRecord.cpp:
(JSC::WebAssemblyModuleRecord::link):
(JSC::WebAssemblyModuleRecord::evaluate):
2017-04-04 Keith Miller <keith_miller@apple.com>
WasmBench asserts in debug jsc
https://bugs.webkit.org/show_bug.cgi?id=170462
Reviewed by Saam Barati.
The assertion should have been an if.
* wasm/WasmWorklist.cpp:
2017-04-04 Filip Pizlo <fpizlo@apple.com>
Air::lowerAfterRegAlloc should bail early if it finds no Shuffles or ColdCCalls
https://bugs.webkit.org/show_bug.cgi?id=170305
Reviewed by Saam Barati.
This reduces and sometimes completely eliminates the need to run lowerAfterRegAlloc().
This lowers the Shuffle for the arguments of a CCall before register allocation unless
the CCall arguments require a real shuffle (like if the CCall arguments were argument
registers). This lowers a ColdCCall like a CCall for optLevel<2.
Finally, lowerAfterRegAlloc() now checks if there are any Shuffles or CCalls before it
does anything else. For wasm at -O1, this means that the phase doesn't run at all. This
is a ~3% wasm -O1 compile time progression.
To make this easy, I changed optLevel into a property of Procedure and Code rather than
an argument we thread through everything. I like how Procedure and Code are dumping
ground classes. This does not bother me. Note that I cloned optLevel into Procedure and
Code so that it's cheap to query inside Air phases.
* b3/B3Compile.cpp:
(JSC::B3::compile):
* b3/B3Compile.h:
* b3/B3Generate.cpp:
(JSC::B3::prepareForGeneration):
(JSC::B3::generateToAir):
* b3/B3Generate.h:
* b3/B3Procedure.cpp:
(JSC::B3::Procedure::setOptLevel):
* b3/B3Procedure.h:
(JSC::B3::Procedure::optLevel):
* b3/air/AirCode.h:
(JSC::B3::Air::Code::isPinned):
(JSC::B3::Air::Code::setOptLevel):
(JSC::B3::Air::Code::optLevel):
* b3/air/AirEmitShuffle.cpp:
(JSC::B3::Air::ShufflePair::bank):
(JSC::B3::Air::ShufflePair::opcode):
(JSC::B3::Air::ShufflePair::inst):
(JSC::B3::Air::emitShuffle):
* b3/air/AirEmitShuffle.h:
(JSC::B3::Air::moveFor):
* b3/air/AirGenerate.cpp:
(JSC::B3::Air::prepareForGeneration):
* b3/air/AirGenerate.h:
* b3/air/AirLowerAfterRegAlloc.cpp:
(JSC::B3::Air::lowerAfterRegAlloc):
* b3/air/AirLowerMacros.cpp:
(JSC::B3::Air::lowerMacros):
* b3/testb3.cpp:
(JSC::B3::compileProc):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::parseAndCompile):
2017-04-04 Filip Pizlo <fpizlo@apple.com>
Don't need to Air::reportUsedRegisters for wasm at -O1
https://bugs.webkit.org/show_bug.cgi?id=170459
Reviewed by Saam Barati.
I did some refactorings to Liveness<> to try to understand its performance. Based on
this I concluded that the bigger immediate issue is just removing unnecessary phases
from -O1.
This removes Air::reportUsedRegisters() from -O1 if the user has indicated that he is
not interested in StackmapGenerationParams::usedRegisters(). The logic here is a bit
weird because of how Air does spill code generation. The register allocator's spiller
will emit spill code using identifiable spill slots, which allows subsequent phases to
register-allocate the spill slots. We do this by a forward flow CSE phase called
fixObviousSpills (which is a terrible name since there is no longer anything obvious
about some of the spills that this phase can fix!). As is most natural for CSEs over
3AC, it rewires the uses of redundant computations rather than removing the redundant
computations. This means that if a spill got "fixed", there may be either or both of
the following:
- Dead loads from the stack.
- Dead stores to the stack.
We know that a load from the stack is dead if the register is dead at the point of the
load. We know that a store to the stack is dead if the spill slot is dead at the point
of the store.
Unfortunately, liveness analysis - over either registers or spill slots - is expensive.
Fortunately, allocateStack() already does liveness analysis over spill slots. So, we
baked elimination of stores to the stack into that phase. That aspect of clean-up after
the spill CSE comes for free.
Also fortunately for the FTL, we have to do reportUsedRegisters() anyway. This is a
phase that enables StackmapGenerationParams::usedRegisters() to work, which then
enables the FTL's patchpoints to do crazy slow-path live range splitting. So, Air's
strategy for the load fix-up after spill CSE is to do it as part of
reportUsedRegisters().
This patch introduces the Procedure::setNeedsUsedRegisters() API. But if you set
needsUsedRegisters to false then we will still run reportUsedRegisters() at -O2 as an
optimization - it removes dead loads from the stack that are left behind from
fixObviousSpills().
This is a ~6% compile time progression at -O1.
* b3/B3Procedure.h:
(JSC::B3::Procedure::setNeedsUsedRegisters):
(JSC::B3::Procedure::needsUsedRegisters):
* b3/B3StackmapGenerationParams.h:
* b3/B3VariableLiveness.cpp:
(JSC::B3::VariableLiveness::VariableLiveness):
* b3/air/AirCode.cpp:
(JSC::B3::Air::Code::needsUsedRegisters):
* b3/air/AirCode.h:
* b3/air/AirGenerate.cpp:
(JSC::B3::Air::prepareForGeneration):
* b3/air/AirLiveness.h:
(JSC::B3::Air::Liveness::Liveness):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::parseAndCompile):
2017-04-03 Filip Pizlo <fpizlo@apple.com>
Air liveness should build constraints and solve them rather than repeatedly parsing IR
https://bugs.webkit.org/show_bug.cgi?id=170421
Reviewed by Saam Barati.
Inst::forEach<> is expensive. The LivenessAdapter uses forEach with a particularly
gnarly lambda that has many extra checks. Therefore, a lot of the time spent in
liveness analysis is just recomputing forEach<> and that lambda to get uses and defs.
This introduces LivenessConstraints<>, which is a liveness constraint system based on
Adapter. It basically caches the results of doing forEach. It'll give you the uses and
defs at each instruction boundary.
This is a ~5% compile time progression at optLevel=1. It's also a ~3% compile time
progression at optLevel=2.
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/air/AirLivenessAdapter.h:
(JSC::B3::Air::LivenessAdapter::LivenessAdapter):
(JSC::B3::Air::LivenessAdapter::forEachUse):
(JSC::B3::Air::LivenessAdapter::forEachDef):
* b3/air/AirLivenessConstraints.h: Added.
(JSC::B3::Air::LivenessConstraints::Actions::Actions):
(JSC::B3::Air::LivenessConstraints::LivenessConstraints):
(JSC::B3::Air::LivenessConstraints::at):
2017-04-03 Mark Lam <mark.lam@apple.com>
Fix incorrect capacity delta calculation reported in SparseArrayValueMap::add().
https://bugs.webkit.org/show_bug.cgi?id=170412
<rdar://problem/29697336>
Reviewed by Filip Pizlo.
Here's an example of code that will trigger underflow in the "deprecatedExtraMemory"
reported by SparseArrayValueMap::add() that is added to Heap::m_deprecatedExtraMemorySize:
arr = new Array;
Object.defineProperty(arr, 18, ({writable: true, configurable: true}));
for (var i = 0; i < 3; ++i) {
Array.prototype.push.apply(arr, ["", () => {}, {}]);
Array.prototype.sort.apply(arr, [() => {}, []]);
}
However, Heap::m_deprecatedExtraMemorySize is only 1 of 3 values that are added
up to form the result of Heap::extraMemorySize(). Heap::m_extraMemorySize and
Heap::m_arrayBuffers.size() are the other 2.
While Heap::m_arrayBuffers.size() is bounded by actual allocated memory, both
Heap::m_deprecatedExtraMemorySize and Heap::m_extraMemorySize are added to
without any bounds checks, and they are only reset to 0 at the start of a full
GC. As a result, if we have a long sequence of eden GCs with a lot of additions
to Heap::m_extraMemorySize and/or Heap::m_deprecatedExtraMemorySize, then these
values could theoretically overflow. Coupling this with the underflow from
SparseArrayValueMap::add(), the result for Heap::extraMemorySize() can easily
overflow. Note: Heap::extraMemorySize() is used to compute the value
currentHeapSize.
If multiple conditions line up just right, the above overflows can result in this
debug assertion failure during an eden GC:
ASSERT(currentHeapSize >= m_sizeAfterLastCollect);
Otherwise, the effects of the overflows will only result in the computed
currentHeapSize not being representative of actual memory usage, and therefore,
a full GC may be triggered earlier or later than is ideal.
This patch ensures that SparseArrayValueMap::add() cannot underflow
Heap::m_deprecatedExtraMemorySize. It also adds overflows checks in the
calculations of Heap::m_deprecatedExtraMemorySize, Heap::m_extraMemorySize, and
Heap::extraMemorySize() so that their values are saturated appropriately to
ensure that GC collections are triggered based on representative memory usage.
* heap/Heap.cpp:
(JSC::Heap::deprecatedReportExtraMemorySlowCase):
(JSC::Heap::extraMemorySize):
(JSC::Heap::updateAllocationLimits):
(JSC::Heap::reportExtraMemoryVisited):
* runtime/SparseArrayValueMap.cpp:
(JSC::SparseArrayValueMap::add):
2017-04-03 Filip Pizlo <fpizlo@apple.com>
Move the Liveness<> adapters from AirLiveness.h to AirLivenessAdapter.h.
Rubber stamped by Keith Miller.
This will make it easier to write other code that uses those adapters.
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/air/AirLiveness.h:
(JSC::B3::Air::LivenessAdapter::LivenessAdapter): Deleted.
(JSC::B3::Air::LivenessAdapter::blockSize): Deleted.
(JSC::B3::Air::LivenessAdapter::forEachUse): Deleted.
(JSC::B3::Air::LivenessAdapter::forEachDef): Deleted.
(JSC::B3::Air::TmpLivenessAdapter::TmpLivenessAdapter): Deleted.
(JSC::B3::Air::TmpLivenessAdapter::numIndices): Deleted.
(JSC::B3::Air::TmpLivenessAdapter::acceptsBank): Deleted.
(JSC::B3::Air::TmpLivenessAdapter::acceptsRole): Deleted.
(JSC::B3::Air::TmpLivenessAdapter::valueToIndex): Deleted.
(JSC::B3::Air::TmpLivenessAdapter::indexToValue): Deleted.
(JSC::B3::Air::StackSlotLivenessAdapter::StackSlotLivenessAdapter): Deleted.
(JSC::B3::Air::StackSlotLivenessAdapter::numIndices): Deleted.
(JSC::B3::Air::StackSlotLivenessAdapter::acceptsBank): Deleted.
(JSC::B3::Air::StackSlotLivenessAdapter::acceptsRole): Deleted.
(JSC::B3::Air::StackSlotLivenessAdapter::valueToIndex): Deleted.
(JSC::B3::Air::StackSlotLivenessAdapter::indexToValue): Deleted.
* b3/air/AirLivenessAdapter.h: Added.
(JSC::B3::Air::LivenessAdapter::LivenessAdapter):
(JSC::B3::Air::LivenessAdapter::blockSize):
(JSC::B3::Air::LivenessAdapter::forEachUse):
(JSC::B3::Air::LivenessAdapter::forEachDef):
(JSC::B3::Air::TmpLivenessAdapter::TmpLivenessAdapter):
(JSC::B3::Air::TmpLivenessAdapter::numIndices):
(JSC::B3::Air::TmpLivenessAdapter::acceptsBank):
(JSC::B3::Air::TmpLivenessAdapter::acceptsRole):
(JSC::B3::Air::TmpLivenessAdapter::valueToIndex):
(JSC::B3::Air::TmpLivenessAdapter::indexToValue):
(JSC::B3::Air::StackSlotLivenessAdapter::StackSlotLivenessAdapter):
(JSC::B3::Air::StackSlotLivenessAdapter::numIndices):
(JSC::B3::Air::StackSlotLivenessAdapter::acceptsBank):
(JSC::B3::Air::StackSlotLivenessAdapter::acceptsRole):
(JSC::B3::Air::StackSlotLivenessAdapter::valueToIndex):
(JSC::B3::Air::StackSlotLivenessAdapter::indexToValue):
2017-04-03 Filip Pizlo <fpizlo@apple.com>
WTF::Liveness should have an API that focuses on actions at instruction boundaries
https://bugs.webkit.org/show_bug.cgi?id=170407
Reviewed by Keith Miller.
Adopt changes to the WTF::Liveness<> API. Instead of having separate functions for the
early/late versions of uses and defs, we now have just a use/def API. Those
automatically take care of eary/late issues as needed.
This reduces the API surface between WTF::Liveness<> and its clients, which makes it
easier to implement some other optimizations I'm thinking about.
* b3/B3VariableLiveness.h:
(JSC::B3::VariableLivenessAdapter::forEachUse):
(JSC::B3::VariableLivenessAdapter::forEachDef):
(JSC::B3::VariableLivenessAdapter::forEachEarlyUse): Deleted.
(JSC::B3::VariableLivenessAdapter::forEachLateUse): Deleted.
(JSC::B3::VariableLivenessAdapter::forEachEarlyDef): Deleted.
(JSC::B3::VariableLivenessAdapter::forEachLateDef): Deleted.
* b3/air/AirLiveness.h:
(JSC::B3::Air::LivenessAdapter::blockSize):
(JSC::B3::Air::LivenessAdapter::forEachUse):
(JSC::B3::Air::LivenessAdapter::forEachDef):
(JSC::B3::Air::LivenessAdapter::forEachEarlyUse): Deleted.
(JSC::B3::Air::LivenessAdapter::forEachLateUse): Deleted.
(JSC::B3::Air::LivenessAdapter::forEachEarlyDef): Deleted.
(JSC::B3::Air::LivenessAdapter::forEachLateDef): Deleted.
2017-04-03 Filip Pizlo <fpizlo@apple.com>
Inst::forEachArg could compile to more compact code
https://bugs.webkit.org/show_bug.cgi?id=170406
Reviewed by Sam Weinig.
Prior to this change, Inst::forEachArg compiled to a ginormous ALWAYS_INLINE switch statement.
It had one case for each opcode, and then each of those cases would have a switch statement over
the number of operands. Then the cases of that switch statement would have a sequence of calls to
the passed lambda. This meant that every user of forEachArg would generate an insane amount of
code. It also meant that the inlining achieved nothing, since the lambda would surely then not
be inlined - and if it was, then the icache pressure due to code bloat would surely negate any
benefits.
This replaces that code with a loop over a compact look-up table. We use the opcode and number of
operands as keys into that look-up table. The table only takes about 20KB. It has one byte for
each argument in each overload of each opcode.
I can't measure any reproducible change in performance, but the JavaScriptCore framework binary
shrinks by 2.7 MB. This is a 15% reduction in JavaScriptCore binary size.
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/B3Width.h:
* b3/air/AirCustom.h:
(JSC::B3::Air::PatchCustom::forEachArg):
* b3/air/AirFormTable.h: Added.
(JSC::B3::Air::decodeFormRole):
(JSC::B3::Air::decodeFormBank):
(JSC::B3::Air::decodeFormWidth):
* b3/air/AirInst.h:
* b3/air/opcode_generator.rb:
2017-04-03 Keith Miller <keith_miller@apple.com>
WebAssembly: remove lastAllocatedMode from Memory
https://bugs.webkit.org/show_bug.cgi?id=170405
Reviewed by Mark Lam.
It's not used anymore so there isn't any point in keeping it around.
* wasm/WasmMemory.cpp:
(JSC::Wasm::Memory::createImpl):
(JSC::Wasm::Memory::lastAllocatedMode): Deleted.
* wasm/WasmMemory.h:
2017-04-03 Zan Dobersek <zdobersek@igalia.com>
[jsc] Add patchableJumpSize() for MIPS
https://bugs.webkit.org/show_bug.cgi?id=169716
Reviewed by Yusuke Suzuki.
* assembler/MIPSAssembler.h:
(JSC::MIPSAssembler::patchableJumpSize): Added.
* assembler/MacroAssemblerMIPS.h:
(JSC::MacroAssemblerMIPS::patchableJumpSize): Added.
2017-04-03 Guillaume Emont <guijemont@igalia.com>
[jsc] implement MIPSAssembler::relinkJumpToNop()
https://bugs.webkit.org/show_bug.cgi?id=169720
Reviewed by Yusuke Suzuki.
* assembler/MIPSAssembler.h:
(JSC::MIPSAssembler::relinkJumpToNop): Added.
2017-04-02 Carlos Garcia Campos <cgarcia@igalia.com>
Share implementation of JSRunLoopTimer::timerDidFire
https://bugs.webkit.org/show_bug.cgi?id=170392
Reviewed by Michael Catanzaro.
The code is cross-platform but it's duplicated in CF and GLib implementations, it could be shared instead.
* runtime/JSRunLoopTimer.cpp:
(JSC::JSRunLoopTimer::timerDidFire): Move common implementation here.
(JSC::JSRunLoopTimer::setRunLoop): Use timerDidFireCallback.
(JSC::JSRunLoopTimer::timerDidFireCallback): Call JSRunLoopTimer::timerDidFire().
* runtime/JSRunLoopTimer.h:
2017-04-01 Oleksandr Skachkov <gskachkov@gmail.com>
Object with numerical keys with gaps gets filled by NaN values
https://bugs.webkit.org/show_bug.cgi?id=164412
Reviewed by Mark Lam.
This patch fixes issue when object have two properties
with name as number. The issue appears when during invoking
convertDoubleToArrayStorage, array is filled by pNaN and
method converting it to real NaN. This happeneds because a
pNaN in a Double array is a hole, and Double arrays cannot
have NaN values. To fix issue we need to check value and
clear it if it pNaN.
* runtime/JSObject.cpp:
(JSC::JSObject::convertDoubleToArrayStorage):
2017-03-31 Saam Barati <sbarati@apple.com>
WebAssembly: Make our calls out to JS PIC friendly
https://bugs.webkit.org/show_bug.cgi?id=170261
Reviewed by Keith Miller.
This patch removes a direct call from the module to the Wasm to JS stub.
Instead, we do an indirect call to the stub by loading the stub's executable
address off of the CodeBlock. This is to make the code we emit for comply with
requirements needed for PIC.
Adding this indirection is not ideal. Although this patch is neutral on
WasmBench, we really want to get back to a world where we have an IC
call infrastructure. This patch is obviously a regression on some
types of programs. I've filed this bug to make sure we implement a
PIC compliant Wasm to JS call IC:
https://bugs.webkit.org/show_bug.cgi?id=170375
* wasm/WasmB3IRGenerator.cpp:
* wasm/WasmFormat.h:
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::complete):
* wasm/js/JSWebAssemblyCodeBlock.cpp:
(JSC::JSWebAssemblyCodeBlock::initialize):
* wasm/js/JSWebAssemblyCodeBlock.h:
(JSC::JSWebAssemblyCodeBlock::create):
(JSC::JSWebAssemblyCodeBlock::offsetOfImportWasmToJSStub):
(JSC::JSWebAssemblyCodeBlock::offsetOfCallees):
(JSC::JSWebAssemblyCodeBlock::allocationSize):
(JSC::JSWebAssemblyCodeBlock::importWasmToJSStub):
* wasm/js/JSWebAssemblyInstance.cpp:
(JSC::JSWebAssemblyInstance::addUnitializedCodeBlock):
* wasm/js/JSWebAssemblyInstance.h:
(JSC::JSWebAssemblyInstance::offsetOfCodeBlock):
2017-03-31 Keith Miller <keith_miller@apple.com>
WebAssembly: webAssemblyB3OptimizationLevel should use defaultB3OptLevel by default
https://bugs.webkit.org/show_bug.cgi?id=170378
Reviewed by Saam Barati.
* runtime/Options.h:
* wasm/WasmB3IRGenerator.h:
2017-03-31 Keith Miller <keith_miller@apple.com>
WebAssembly: Add compilation level option
https://bugs.webkit.org/show_bug.cgi?id=170374
Reviewed by Mark Lam.
This patch adds an option, webAssemblyB3OptimizationLevel, which
changes the optimization mode wasm passes to B3.
* runtime/Options.h:
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::compileFunctions):
2017-03-31 Saam Barati <sbarati@apple.com>
WebAssembly: Strip WasmParser and WasmFunctionParser from knowing about VM
https://bugs.webkit.org/show_bug.cgi?id=170312
Reviewed by Mark Lam.
This is another step towards PIC-ifying Wasm. This patch removes
the VM field that is no longer used.
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::parseAndCompile):
* wasm/WasmB3IRGenerator.h:
* wasm/WasmFunctionParser.h:
(JSC::Wasm::FunctionParser<Context>::FunctionParser):
* wasm/WasmModuleParser.h:
(JSC::Wasm::ModuleParser::ModuleParser):
* wasm/WasmParser.h:
(JSC::Wasm::Parser<SuccessType>::Parser):
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::parseAndValidateModule):
(JSC::Wasm::Plan::compileFunctions):
* wasm/WasmValidate.cpp:
(JSC::Wasm::validateFunction):
* wasm/WasmValidate.h:
2017-03-31 Saam Barati <sbarati@apple.com>
WebAssembly: Ref count Signature and SignatureInformation should not care about VM
https://bugs.webkit.org/show_bug.cgi?id=170316
Reviewed by Keith Miller.
This is yet again another step towards PIC-ifying Wasm.
Signature should be ref counted so we can tell when
no code is holding onto a Signature. This makes it easy
to free unused Signatures. Also, this patch rids SignatureInfo
of any VM knowledge. Now, there is just a single SignatureInfo that
lives in a process.
* runtime/VM.h:
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::createJSToWasmWrapper):
(JSC::Wasm::parseAndCompile):
* wasm/WasmB3IRGenerator.h:
* wasm/WasmBinding.cpp:
(JSC::Wasm::wasmToJs):
* wasm/WasmCallingConvention.h:
(JSC::Wasm::CallingConvention::loadArguments):
* wasm/WasmFormat.h:
* wasm/WasmFunctionParser.h:
(JSC::Wasm::FunctionParser<Context>::FunctionParser):
* wasm/WasmModuleParser.cpp:
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::parseAndValidateModule):
(JSC::Wasm::Plan::compileFunctions):
(JSC::Wasm::Plan::complete):
* wasm/WasmSignature.cpp:
(JSC::Wasm::Signature::hash):
(JSC::Wasm::Signature::tryCreate):
(JSC::Wasm::SignatureInformation::SignatureInformation):
(JSC::Wasm::SignatureInformation::singleton):
(JSC::Wasm::SignatureInformation::adopt):
(JSC::Wasm::SignatureInformation::get):
(JSC::Wasm::SignatureInformation::tryCleanup):
(JSC::Wasm::Signature::create): Deleted.
(JSC::Wasm::Signature::createInvalid): Deleted.
(JSC::Wasm::Signature::destroy): Deleted.
(JSC::Wasm::SignatureInformation::~SignatureInformation): Deleted.
* wasm/WasmSignature.h:
(JSC::Wasm::Signature::allocatedSize):
(JSC::Wasm::Signature::operator==):
* wasm/WasmValidate.cpp:
(JSC::Wasm::validateFunction):
* wasm/WasmValidate.h:
* wasm/js/JSWebAssemblyModule.cpp:
(JSC::JSWebAssemblyModule::destroy):
* wasm/js/WebAssemblyFunction.cpp:
(JSC::callWebAssemblyFunction):
* wasm/js/WebAssemblyFunction.h:
* wasm/js/WebAssemblyModuleRecord.cpp:
(JSC::WebAssemblyModuleRecord::link):
(JSC::WebAssemblyModuleRecord::evaluate):
* wasm/js/WebAssemblyWrapperFunction.cpp:
(JSC::WebAssemblyWrapperFunction::create):
* wasm/js/WebAssemblyWrapperFunction.h:
2017-03-31 Mark Lam <mark.lam@apple.com>
Array.prototype.splice() should not be using JSArray::tryCreateForInitializationPrivate().
https://bugs.webkit.org/show_bug.cgi?id=170303
<rdar://problem/31358281>
Reviewed by Filip Pizlo.
This is because it needs to call getProperty() later to get the values for
initializing the array. getProperty() can execute arbitrary code and potentially
trigger the GC. This is not allowed for clients of JSArray::tryCreateForInitializationPrivate().
* runtime/ArrayPrototype.cpp:
(JSC::arrayProtoFuncSplice):
(JSC::copySplicedArrayElements): Deleted.
2017-03-31 Oleksandr Skachkov <gskachkov@gmail.com>
String.prototype.replace incorrectly applies "special replacement parameters" when passed a function
https://bugs.webkit.org/show_bug.cgi?id=170151
Reviewed by Saam Barati.
This patch fixes issue for String.prototype.replace when passed a function
with special symbols "$$". It happeneds because substituteBackreferences applies
unconditionally, but according to the spec it should be applied only for text
21.1.3.16.8 https://tc39.github.io/ecma262/#sec-string.prototype.replace
* runtime/StringPrototype.cpp:
(JSC::replaceUsingStringSearch):
2017-03-30 Saam Barati <sbarati@apple.com>
WebAssembly: When Wasm calls to C, it should use Wasm::Context* instead of ExecState* to get VM
https://bugs.webkit.org/show_bug.cgi?id=170185
Reviewed by Michael Saboff.
This is one more step in the direction of PIC-ified Wasm.
When we lift WasmCallee above VM, we will no longer be
able to get VM from ExecState*. This patch ensures that
we don't do that from within the Wasm runtime. Instead,
we use the Wasm::Context* to get the VM.
This patch also adds a new class, Wasm::Thunks. There
is a single Wasm::Thunks that lives in the process. It
is responsible for generating a thunk that Wasm relies on.
The only such thunk right now is the exception throwing
thunk.
This patch also rids WasmFaultSignalHandler from any knowledge
of VM. Previously, it relied on VM to get the exception handling
thunk.
The only part of the Wasm runtime that will be allowed
to get VM& from ExecState will be WasmBinding. In the
future, we plan to keep the calls out to JS to keep
a JSCell as the callee.
* JavaScriptCore.xcodeproj/project.pbxproj:
* dfg/DFGOSREntry.cpp:
(JSC::DFG::prepareOSREntry):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
* interpreter/Interpreter.cpp:
(JSC::UnwindFunctor::copyCalleeSavesToVMEntryFrameCalleeSavesBuffer):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::restoreCalleeSavesFromVMEntryFrameCalleeSavesBuffer):
(JSC::AssemblyHelpers::copyCalleeSavesToVMEntryFrameCalleeSavesBufferImpl):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::copyCalleeSavesToVMEntryFrameCalleeSavesBuffer):
* jit/ThunkGenerators.cpp:
(JSC::throwExceptionFromWasmThunkGenerator): Deleted.
* jit/ThunkGenerators.h:
* runtime/InitializeThreading.cpp:
(JSC::initializeThreading):
* runtime/VM.cpp:
(JSC::VM::VM):
(JSC::VM::getAllCalleeSaveRegisterOffsets):
* runtime/VM.h:
(JSC::VM::topVMEntryFrameOffset):
(JSC::VM::getAllCalleeSaveRegisterOffsets): Deleted.
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::emitExceptionCheck):
* wasm/WasmFaultSignalHandler.cpp:
(JSC::Wasm::trapHandler):
* wasm/WasmMemory.cpp:
(JSC::Wasm::tryGetFastMemory):
* wasm/WasmThunks.cpp: Added.
(JSC::Wasm::throwExceptionFromWasmThunkGenerator):
(JSC::Wasm::Thunks::initialize):
(JSC::Wasm::Thunks::singleton):
(JSC::Wasm::Thunks::stub):
(JSC::Wasm::Thunks::existingStub):
* wasm/WasmThunks.h: Added.
* wasm/js/JSWebAssemblyInstance.cpp:
(JSC::JSWebAssemblyInstance::JSWebAssemblyInstance):
* wasm/js/JSWebAssemblyInstance.h:
(JSC::JSWebAssemblyInstance::offsetOfVM):
* wasm/js/JSWebAssemblyMemory.cpp:
(JSC::JSWebAssemblyMemory::grow):
* wasm/js/JSWebAssemblyMemory.h:
* wasm/js/WebAssemblyMemoryPrototype.cpp:
(JSC::webAssemblyMemoryProtoFuncGrow):
2017-03-30 Mark Lam <mark.lam@apple.com>
IntlObject should not be using JSArray::initializeIndex().
https://bugs.webkit.org/show_bug.cgi?id=170302
<rdar://problem/31356918>
Reviewed by Saam Barati.
JSArray::initializeIndex() is only meant to be used with arrays created using
JSArray::tryCreateForInitializationPrivate() under very constrained conditions.
* runtime/IntlObject.cpp:
(JSC::canonicalizeLocaleList):
(JSC::intlObjectFuncGetCanonicalLocales):
2017-03-30 Filip Pizlo <fpizlo@apple.com>
Air should support linear scan for optLevel<2
https://bugs.webkit.org/show_bug.cgi?id=170161
Reviewed by Saam Barati.
This changes the default opt level of B3 to 2. It makes the other opt levels useful by adding a
new register allocator. This new linear scan allocator will produce significantly worse code.
But it will produce that code a lot faster than IRC or Briggs.
The opt levels are:
0: no optimizations, linear scan
1: some optimizations, linear scan
2: full optimizations, graph coloring (IRC or Briggs based on CPU)
What we used to call optLevel=1 is not called optLevel=2, or better yet,
optLevel=B3::defaultOptLevel(). We no longer have anything like the old optLevel=0 (which did no
optimizations but ran graph coloring).
allocateRegistersByLinearScan() faithfully implements Massimiliano Poletto and Vivek Sarkar's
famous algorithm. It uses the variant that handles clobbered registers by avoiding assigning
ranges to those registers if the range overlaps a clobber. It's engineered to allocate registers
very quickly and generate inefficient code without falling off a cliff.
The new optLevel=1 speeds up B3 by a factor of 2, and results in a 80% throughput regression.
Linear scan runs 4.7x faster than graph coloring on average.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/B3BasicBlockUtils.h:
(JSC::B3::blocksInPreOrder):
(JSC::B3::blocksInPostOrder):
* b3/B3BlockWorklist.h:
* b3/B3CFG.h:
(JSC::B3::CFG::newMap):
* b3/B3Common.h:
(JSC::B3::defaultOptLevel):
* b3/B3Compile.h:
* b3/B3DuplicateTails.cpp:
* b3/B3EliminateCommonSubexpressions.cpp:
* b3/B3FixSSA.cpp:
(JSC::B3::demoteValues):
(JSC::B3::fixSSA):
* b3/B3FixSSA.h:
* b3/B3Generate.cpp:
(JSC::B3::prepareForGeneration):
(JSC::B3::generateToAir):
* b3/B3Generate.h:
* b3/B3HeapRange.cpp: Removed.
* b3/B3HeapRange.h:
(JSC::B3::HeapRange::HeapRange): Deleted.
(JSC::B3::HeapRange::top): Deleted.
(JSC::B3::HeapRange::operator==): Deleted.
(JSC::B3::HeapRange::operator!=): Deleted.
(JSC::B3::HeapRange::operator|): Deleted.
(JSC::B3::HeapRange::operator bool): Deleted.
(JSC::B3::HeapRange::begin): Deleted.
(JSC::B3::HeapRange::end): Deleted.
(JSC::B3::HeapRange::overlaps): Deleted.
* b3/B3LowerToAir.cpp:
* b3/B3MoveConstants.cpp:
* b3/B3PhiChildren.h:
* b3/B3Procedure.cpp:
(JSC::B3::Procedure::dump):
(JSC::B3::Procedure::deleteOrphans):
(JSC::B3::Procedure::setBlockOrderImpl):
* b3/B3ReduceDoubleToFloat.cpp:
* b3/B3ReduceStrength.cpp:
* b3/B3SSACalculator.h:
* b3/B3UseCounts.h:
* b3/air/AirAllocateRegistersByGraphColoring.cpp:
* b3/air/AirAllocateRegistersByLinearScan.cpp: Added.
(JSC::B3::Air::allocateRegistersByLinearScan):
* b3/air/AirAllocateRegistersByLinearScan.h: Added.
* b3/air/AirAllocateStack.cpp:
(JSC::B3::Air::allocateStack):
* b3/air/AirArg.cpp:
(WTF::printInternal):
* b3/air/AirArg.h:
(JSC::B3::Air::Arg::activeAt):
(JSC::B3::Air::Arg::timing):
(JSC::B3::Air::Arg::forEachPhase):
* b3/air/AirBasicBlock.h:
* b3/air/AirBlockWorklist.h:
* b3/air/AirCFG.h:
(JSC::B3::Air::CFG::newMap):
* b3/air/AirEliminateDeadCode.cpp:
(JSC::B3::Air::eliminateDeadCode):
* b3/air/AirFixObviousSpills.cpp:
* b3/air/AirFixPartialRegisterStalls.cpp:
(JSC::B3::Air::fixPartialRegisterStalls):
* b3/air/AirFixSpillsAfterTerminals.cpp: Added.
(JSC::B3::Air::fixSpillsAfterTerminals):
* b3/air/AirFixSpillsAfterTerminals.h: Added.
* b3/air/AirGenerate.cpp:
(JSC::B3::Air::prepareForGeneration):
(JSC::B3::Air::generate):
* b3/air/AirGenerate.h:
* b3/air/AirGenerationContext.h:
* b3/air/AirInsertionSet.h:
* b3/air/AirInst.cpp:
(JSC::B3::Air::Inst::needsPadding):
* b3/air/AirLowerAfterRegAlloc.cpp:
(JSC::B3::Air::lowerAfterRegAlloc):
* b3/air/AirLowerEntrySwitch.cpp:
(JSC::B3::Air::lowerEntrySwitch):
* b3/air/AirOpcode.opcodes:
* b3/air/AirPhaseInsertionSet.cpp: Added.
(JSC::B3::Air::PhaseInsertionSet::execute):
* b3/air/AirPhaseInsertionSet.h: Added.
(JSC::B3::Air::PhaseInsertion::PhaseInsertion):
(JSC::B3::Air::PhaseInsertion::phase):
(JSC::B3::Air::PhaseInsertion::operator<):
(JSC::B3::Air::PhaseInsertionSet::PhaseInsertionSet):
(JSC::B3::Air::PhaseInsertionSet::appendInsertion):
(JSC::B3::Air::PhaseInsertionSet::insertInst):
(JSC::B3::Air::PhaseInsertionSet::insert):
* b3/air/AirRegLiveness.h:
(JSC::B3::Air::RegLiveness::LocalCalc::LocalCalc):
* b3/air/AirSpillEverything.cpp:
(JSC::B3::Air::spillEverything):
* b3/air/AirTmp.cpp:
* b3/air/AirTmp.h:
(JSC::B3::Air::Tmp::tmpForIndex):
* b3/air/AirTmpInlines.h:
(JSC::B3::Air::Tmp::Indexed::Indexed):
(JSC::B3::Air::Tmp::Indexed::index):
(JSC::B3::Air::Tmp::AbsolutelyIndexed::AbsolutelyIndexed):
(JSC::B3::Air::Tmp::AbsolutelyIndexed::index):
(JSC::B3::Air::Tmp::indexed):
(JSC::B3::Air::Tmp::absolutelyIndexed):
(JSC::B3::Air::Tmp::tmpForAbsoluteIndex):
* b3/testb3.cpp:
(JSC::B3::compile):
(JSC::B3::testMulLoadTwice):
* jit/RegisterSet.h:
(JSC::RegisterSet::add):
(JSC::RegisterSet::remove):
* runtime/Options.h:
* wasm/WasmB3IRGenerator.h:
2017-03-30 Youenn Fablet <youenn@apple.com>
Clean up RTCDataChannel
https://bugs.webkit.org/show_bug.cgi?id=169732
Reviewed by Chris Dumez.
* runtime/CommonIdentifiers.h: Adding RTCDataChannelEvent.
2017-03-30 Saam Barati <sbarati@apple.com>
WebAssembly: pass Wasm::Context* to vmEntryToWasm when not using fast TLS
https://bugs.webkit.org/show_bug.cgi?id=170182
Reviewed by Mark Lam.
This is one more step in the direction of PIC-ified Wasm.
I'm removing assumptions that a wasm callee is a cell. We used to use
the callee to get the WasmContext off the callee's VM. Instead,
this patch makes it so that we pass in the context as a parameter
to the JS entrypoint.
* heap/MarkedBlock.h:
(JSC::MarkedBlock::offsetOfVM): Deleted.
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::loadWasmContext):
(JSC::AssemblyHelpers::storeWasmContext):
(JSC::AssemblyHelpers::loadWasmContextNeedsMacroScratchRegister):
(JSC::AssemblyHelpers::storeWasmContextNeedsMacroScratchRegister):
* jsc.cpp:
(functionTestWasmModuleFunctions):
* runtime/VM.h:
(JSC::VM::wasmContextOffset): Deleted.
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::materializeWasmContext):
(JSC::Wasm::B3IRGenerator::restoreWasmContext):
(JSC::Wasm::B3IRGenerator::B3IRGenerator):
(JSC::Wasm::createJSToWasmWrapper):
* wasm/WasmContext.cpp:
(JSC::Wasm::loadContext):
(JSC::Wasm::storeContext):
(JSC::loadWasmContext): Deleted.
(JSC::storeWasmContext): Deleted.
* wasm/WasmContext.h:
(JSC::Wasm::useFastTLS):
(JSC::Wasm::useFastTLSForContext):
* wasm/WasmMemoryInformation.cpp:
(JSC::Wasm::PinnedRegisterInfo::get):
* wasm/WasmMemoryInformation.h:
(JSC::Wasm::useFastTLS): Deleted.
(JSC::Wasm::useFastTLSForWasmContext): Deleted.
* wasm/js/WebAssemblyFunction.cpp:
(JSC::callWebAssemblyFunction):
2017-03-30 JF Bastien <jfbastien@apple.com>
WebAssembly: fix misc JS API implementation inconsistencies
https://bugs.webkit.org/show_bug.cgi?id=170187
Reviewed by Keith Miller.
Auto-generate lookup tables.
Methods should be on prototype.
Exception returns should be idiomatic.
* wasm/JSWebAssembly.cpp: validate / compile / instantiate should
be on the prototype
(JSC::JSWebAssembly::create):
(JSC::JSWebAssembly::finishCreation):
(JSC::reject): Deleted.
(JSC::webAssemblyCompileFunc): Deleted.
(JSC::resolve): Deleted.
(JSC::instantiate): Deleted.
(JSC::compileAndInstantiate): Deleted.
(JSC::webAssemblyInstantiateFunc): Deleted.
(JSC::webAssemblyValidateFunc): Deleted.
* wasm/JSWebAssembly.h:
* wasm/js/WebAssemblyMemoryPrototype.cpp: move from JSWebAssembly.cpp
(JSC::webAssemblyMemoryProtoFuncBuffer):
(JSC::WebAssemblyMemoryPrototype::create):
(JSC::WebAssemblyMemoryPrototype::finishCreation):
* wasm/js/WebAssemblyMemoryPrototype.h:
* wasm/js/WebAssemblyPrototype.cpp:
(JSC::reject):
(JSC::webAssemblyCompileFunc):
(JSC::resolve):
(JSC::instantiate):
(JSC::compileAndInstantiate):
(JSC::webAssemblyInstantiateFunc):
(JSC::webAssemblyValidateFunc):
(JSC::webAssemblyFunctionValidate): Deleted.
(JSC::webAssemblyFunctionCompile): Deleted.
* wasm/js/WebAssemblyTablePrototype.cpp:
(JSC::webAssemblyTableProtoFuncGrow):
(JSC::webAssemblyTableProtoFuncGet):
(JSC::webAssemblyTableProtoFuncSet):
(JSC::WebAssemblyTablePrototype::create):
(JSC::WebAssemblyTablePrototype::finishCreation):
* wasm/js/WebAssemblyTablePrototype.h:
2017-03-29 Keith Miller <keith_miller@apple.com>
Unreviewed, fix the build, again. Hopefully for the last time, again!
* runtime/Options.cpp:
2017-03-29 Keith Miller <keith_miller@apple.com>
Unreviewed, fix the build, again. Hopefully for the last time!
* runtime/Options.cpp:
(JSC::parse):
2017-03-29 Keith Miller <keith_miller@apple.com>
Unreviewed, windows build fix.
* runtime/Options.cpp:
(JSC::parse):
2017-03-29 Keith Miller <keith_miller@apple.com>
WebAssembly: B3IRGenerator should pool constants
https://bugs.webkit.org/show_bug.cgi?id=170266
Reviewed by Filip Pizlo.
This patch adds a HashMap to B3IRGenerator that contains all the constants used in a function.
B3IRGenerator then uses an InsertionSet to add all those constants to the root BB. This doesn't
appear to be a compile time improvement but it could be valuable in the future.
* b3/B3Opcode.h:
(JSC::B3::opcodeForConstant):
* b3/B3Procedure.cpp:
(JSC::B3::Procedure::addConstant):
* b3/B3Procedure.h:
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::B3IRGenerator):
(JSC::Wasm::B3IRGenerator::constant):
(JSC::Wasm::B3IRGenerator::insertConstants):
(JSC::Wasm::B3IRGenerator::addConstant):
(JSC::Wasm::B3IRGenerator::dump):
(JSC::Wasm::parseAndCompile):
(JSC::Wasm::B3IRGenerator::emitChecksForModOrDiv):
(JSC::Wasm::B3IRGenerator::zeroForType): Deleted.
* wasm/generateWasmB3IRGeneratorInlinesHeader.py:
(generateConstCode):
2017-03-29 Saam Barati <sbarati@apple.com>
LinkBuffer and ExecutableAllocator shouldn't have anything to do with VM
https://bugs.webkit.org/show_bug.cgi?id=170210
Reviewed by Mark Lam.
This is one more step in the direction of PIC-ified Wasm.
LinkBuffer and ExecutableAllocator have no business knowing about VM.
* assembler/LinkBuffer.cpp:
(JSC::LinkBuffer::allocate):
* assembler/LinkBuffer.h:
(JSC::LinkBuffer::LinkBuffer):
(JSC::LinkBuffer::vm): Deleted.
* b3/B3Compile.cpp:
(JSC::B3::compile):
* b3/B3Compile.h:
* b3/air/testair.cpp:
* b3/testb3.cpp:
(JSC::B3::compileProc):
(JSC::B3::compileAndRun):
(JSC::B3::testLoadAcq42):
(JSC::B3::testAddArgZeroImmZDef):
(JSC::B3::testAddLoadTwice):
(JSC::B3::testMulLoadTwice):
(JSC::B3::testMulAddArgsLeft):
(JSC::B3::testMulAddArgsRight):
(JSC::B3::testMulAddArgsLeft32):
(JSC::B3::testMulAddArgsRight32):
(JSC::B3::testMulSubArgsLeft):
(JSC::B3::testMulSubArgsRight):
(JSC::B3::testMulSubArgsLeft32):
(JSC::B3::testMulSubArgsRight32):
(JSC::B3::testMulNegArgs):
(JSC::B3::testMulNegArgs32):
(JSC::B3::testCompareFloatToDoubleThroughPhi):
(JSC::B3::testDoubleToFloatThroughPhi):
(JSC::B3::testReduceFloatToDoubleValidates):
(JSC::B3::testDoubleProducerPhiToFloatConversion):
(JSC::B3::testDoubleProducerPhiToFloatConversionWithDoubleConsumer):
(JSC::B3::testDoubleProducerPhiWithNonFloatConst):
(JSC::B3::testIToD64Arg):
(JSC::B3::testIToF64Arg):
(JSC::B3::testIToD32Arg):
(JSC::B3::testIToF32Arg):
(JSC::B3::testIToD64Mem):
(JSC::B3::testIToF64Mem):
(JSC::B3::testIToD32Mem):
(JSC::B3::testIToF32Mem):
(JSC::B3::testIToDReducedToIToF64Arg):
(JSC::B3::testIToDReducedToIToF32Arg):
(JSC::B3::testStoreRelAddLoadAcq32):
(JSC::B3::testStoreRelAddLoadAcq8):
(JSC::B3::testStoreRelAddFenceLoadAcq8):
(JSC::B3::testStoreRelAddLoadAcq16):
(JSC::B3::testStoreRelAddLoadAcq64):
(JSC::B3::testBranch):
(JSC::B3::testBranchPtr):
(JSC::B3::testDiamond):
(JSC::B3::testBranchNotEqual):
(JSC::B3::testBranchNotEqualCommute):
(JSC::B3::testBranchNotEqualNotEqual):
(JSC::B3::testBranchEqual):
(JSC::B3::testBranchEqualEqual):
(JSC::B3::testBranchEqualCommute):
(JSC::B3::testBranchEqualEqual1):
(JSC::B3::testBranchLoadPtr):
(JSC::B3::testBranchLoad32):
(JSC::B3::testBranchLoad8S):
(JSC::B3::testBranchLoad8Z):
(JSC::B3::testBranchLoad16S):
(JSC::B3::testBranchLoad16Z):
(JSC::B3::testBranch8WithLoad8ZIndex):
(JSC::B3::testComplex):
(JSC::B3::testSimpleCheck):
(JSC::B3::testCheckFalse):
(JSC::B3::testCheckTrue):
(JSC::B3::testCheckLessThan):
(JSC::B3::testCheckMegaCombo):
(JSC::B3::testCheckTrickyMegaCombo):
(JSC::B3::testCheckTwoMegaCombos):
(JSC::B3::testCheckTwoNonRedundantMegaCombos):
(JSC::B3::testCheckAddImm):
(JSC::B3::testCheckAddImmCommute):
(JSC::B3::testCheckAddImmSomeRegister):
(JSC::B3::testCheckAdd):
(JSC::B3::testCheckAdd64):
(JSC::B3::testCheckAddFold):
(JSC::B3::testCheckAddFoldFail):
(JSC::B3::testCheckAddSelfOverflow64):
(JSC::B3::testCheckAddSelfOverflow32):
(JSC::B3::testCheckSubImm):
(JSC::B3::testCheckSubBadImm):
(JSC::B3::testCheckSub):
(JSC::B3::testCheckSub64):
(JSC::B3::testCheckSubFold):
(JSC::B3::testCheckSubFoldFail):
(JSC::B3::testCheckNeg):
(JSC::B3::testCheckNeg64):
(JSC::B3::testCheckMul):
(JSC::B3::testCheckMulMemory):
(JSC::B3::testCheckMul2):
(JSC::B3::testCheckMul64):
(JSC::B3::testCheckMulFold):
(JSC::B3::testCheckMulFoldFail):
(JSC::B3::testCheckMul64SShr):
(JSC::B3::testSwitch):
(JSC::B3::testSwitchChillDiv):
(JSC::B3::testSwitchTargettingSameBlock):
(JSC::B3::testSwitchTargettingSameBlockFoldPathConstant):
(JSC::B3::testBasicSelect):
(JSC::B3::testSelectTest):
(JSC::B3::testSelectCompareDouble):
(JSC::B3::testSelectDouble):
(JSC::B3::testSelectDoubleTest):
(JSC::B3::testSelectDoubleCompareDouble):
(JSC::B3::testSelectFloatCompareFloat):
(JSC::B3::testSelectFold):
(JSC::B3::testSelectInvert):
(JSC::B3::testCheckSelect):
(JSC::B3::testCheckSelectCheckSelect):
(JSC::B3::testCheckSelectAndCSE):
(JSC::B3::testTrivialInfiniteLoop):
(JSC::B3::testFoldPathEqual):
(JSC::B3::testLShiftSelf32):
(JSC::B3::testRShiftSelf32):
(JSC::B3::testURShiftSelf32):
(JSC::B3::testLShiftSelf64):
(JSC::B3::testRShiftSelf64):
(JSC::B3::testURShiftSelf64):
(JSC::B3::testPatchpointDoubleRegs):
(JSC::B3::testSpillDefSmallerThanUse):
(JSC::B3::testSpillUseLargerThanDef):
(JSC::B3::testLateRegister):
(JSC::B3::testInterpreter):
(JSC::B3::testEntrySwitchSimple):
(JSC::B3::testEntrySwitchNoEntrySwitch):
(JSC::B3::testEntrySwitchWithCommonPaths):
(JSC::B3::testEntrySwitchWithCommonPathsAndNonTrivialEntrypoint):
(JSC::B3::testEntrySwitchLoop):
(JSC::B3::testSomeEarlyRegister):
(JSC::B3::testTerminalPatchpointThatNeedsToBeSpilled):
(JSC::B3::testTerminalPatchpointThatNeedsToBeSpilled2):
(JSC::B3::testPatchpointTerminalReturnValue):
(JSC::B3::testMemoryFence):
(JSC::B3::testStoreFence):
(JSC::B3::testLoadFence):
(JSC::B3::testPCOriginMapDoesntInsertNops):
(JSC::B3::testPinRegisters):
(JSC::B3::testX86LeaAddAddShlLeft):
(JSC::B3::testX86LeaAddAddShlRight):
(JSC::B3::testX86LeaAddAdd):
(JSC::B3::testX86LeaAddShlRight):
(JSC::B3::testX86LeaAddShlLeftScale1):
(JSC::B3::testX86LeaAddShlLeftScale2):
(JSC::B3::testX86LeaAddShlLeftScale4):
(JSC::B3::testX86LeaAddShlLeftScale8):
(JSC::B3::testAddShl32):
(JSC::B3::testAddShl64):
(JSC::B3::testAddShl65):
(JSC::B3::testLoadBaseIndexShift2):
(JSC::B3::testLoadBaseIndexShift32):
(JSC::B3::testOptimizeMaterialization):
(JSC::B3::testAtomicWeakCAS):
(JSC::B3::testAtomicStrongCAS):
(JSC::B3::testAtomicXchg):
(JSC::B3::testDepend32):
(JSC::B3::testDepend64):
(JSC::B3::testWasmBoundsCheck):
(JSC::B3::testWasmAddress):
(JSC::B3::run):
(JSC::B3::compile): Deleted.
* bytecode/PolymorphicAccess.cpp:
(JSC::PolymorphicAccess::regenerate):
* dfg/DFGJITCompiler.cpp:
(JSC::DFG::JITCompiler::compile):
(JSC::DFG::JITCompiler::compileFunction):
* dfg/DFGLazyJSValue.cpp:
(JSC::DFG::LazyJSValue::emit):
* dfg/DFGOSRExitCompiler.cpp:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::emitCall):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::emitCall):
* dfg/DFGThunks.cpp:
(JSC::DFG::osrExitGenerationThunkGenerator):
(JSC::DFG::osrEntryThunkGenerator):
* ftl/FTLCompile.cpp:
(JSC::FTL::compile):
* ftl/FTLLazySlowPath.cpp:
(JSC::FTL::LazySlowPath::generate):
* ftl/FTLLink.cpp:
(JSC::FTL::link):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstruct):
(JSC::FTL::DFG::LowerDFGToB3::compileTailCall):
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstructVarargsSpread):
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstructVarargs):
(JSC::FTL::DFG::LowerDFGToB3::compileCallEval):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
* ftl/FTLOSRExitHandle.cpp:
(JSC::FTL::OSRExitHandle::emitExitThunk):
* ftl/FTLSlowPathCall.cpp:
(JSC::FTL::SlowPathCallContext::makeCall):
* ftl/FTLSlowPathCall.h:
(JSC::FTL::callOperation):
* ftl/FTLState.h:
* ftl/FTLThunks.cpp:
(JSC::FTL::genericGenerationThunkGenerator):
(JSC::FTL::slowPathCallThunkGenerator):
* ftl/FTLThunks.h:
(JSC::FTL::generateIfNecessary):
(JSC::FTL::Thunks::getSlowPathCallThunk):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::emitDumbVirtualCall):
* jit/AssemblyHelpers.h:
* jit/ExecutableAllocator.cpp:
(JSC::ExecutableAllocator::initializeAllocator):
(JSC::ExecutableAllocator::singleton):
(JSC::ExecutableAllocator::ExecutableAllocator):
(JSC::ExecutableAllocator::allocate):
* jit/ExecutableAllocator.h:
* jit/JIT.cpp:
(JSC::JIT::compileWithoutLinking):
* jit/JITCall.cpp:
(JSC::JIT::compileCallEvalSlowCase):
* jit/JITMathIC.h:
(JSC::JITMathIC::generateOutOfLine):
* jit/JITOpcodes.cpp:
(JSC::JIT::privateCompileHasIndexedProperty):
* jit/JITOpcodes32_64.cpp:
(JSC::JIT::privateCompileHasIndexedProperty):
* jit/JITOperations.cpp:
* jit/JITOperations.h:
* jit/JITPropertyAccess.cpp:
(JSC::JIT::stringGetByValStubGenerator):
(JSC::JIT::privateCompileGetByVal):
(JSC::JIT::privateCompileGetByValWithCachedId):
(JSC::JIT::privateCompilePutByVal):
(JSC::JIT::privateCompilePutByValWithCachedId):
* jit/JITPropertyAccess32_64.cpp:
(JSC::JIT::stringGetByValStubGenerator):
* jit/JITStubRoutine.h:
* jit/Repatch.cpp:
(JSC::ftlThunkAwareRepatchCall):
(JSC::linkPolymorphicCall):
* jit/SpecializedThunkJIT.h:
(JSC::SpecializedThunkJIT::finalize):
* jit/ThunkGenerators.cpp:
(JSC::throwExceptionFromCallSlowPathGenerator):
(JSC::linkCallThunkGenerator):
(JSC::linkPolymorphicCallThunkGenerator):
(JSC::virtualThunkFor):
(JSC::nativeForGenerator):
(JSC::arityFixupGenerator):
(JSC::unreachableGenerator):
(JSC::boundThisNoArgsFunctionCallGenerator):
(JSC::throwExceptionFromWasmThunkGenerator):
* llint/LLIntThunks.cpp:
(JSC::LLInt::generateThunkWithJumpTo):
* runtime/SamplingProfiler.cpp:
(JSC::SamplingProfiler::takeSample):
* runtime/VM.cpp:
(JSC::VM::VM):
* runtime/VM.h:
* runtime/VMTraps.cpp:
(JSC::VMTraps::tryInstallTrapBreakpoints):
* tools/VMInspector.cpp:
* wasm/WasmBinding.cpp:
(JSC::Wasm::wasmToJs):
(JSC::Wasm::wasmToWasm):
(JSC::Wasm::exitStubGenerator):
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::complete):
* yarr/YarrJIT.cpp:
(JSC::Yarr::YarrGenerator::compile):
(JSC::Yarr::jitCompile):
2017-03-29 Keith Miller <keith_miller@apple.com>
WebAssembly: Worklist should periodically check in to see if there are higher priority jobs to do.
https://bugs.webkit.org/show_bug.cgi?id=170204
Reviewed by Saam Barati.
This patch makes it so that Wasm::Plan's compileFunctions method can return periodically
to its caller. The main use for this is if a user asynchronously compiles a wasm module
then later synchronously compiles another module. In this case we want to be able to pause
compilation of other worklists.
This patch also adds support for size_t Options.
* runtime/Options.cpp:
(JSC::parse):
(JSC::Option::dump):
(JSC::Option::operator==):
* runtime/Options.h:
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::moveToState):
(JSC::Wasm::Plan::ThreadCountHolder::~ThreadCountHolder):
(JSC::Wasm::Plan::compileFunctions):
* wasm/WasmPlan.h:
* wasm/WasmWorklist.cpp:
2017-03-29 Mark Lam <mark.lam@apple.com>
Remove obsolete references to HeapTimer in JavaScriptCore.order.
https://bugs.webkit.org/show_bug.cgi?id=170252
Reviewed by Saam Barati.
The HeapTimer was renamed to JSRunLoopTimer back in r214504. These HeapTimer
entries are now no longer meaningful.
* JavaScriptCore.order:
2017-03-29 JF Bastien <jfbastien@apple.com>
WebAssembly: add shell-only Memory mode helper
https://bugs.webkit.org/show_bug.cgi?id=170227
Reviewed by Mark Lam.
* jsc.cpp:
(GlobalObject::finishCreation):
(functionWebAssemblyMemoryMode):
* wasm/WasmMemory.h:
* wasm/js/JSWebAssemblyInstance.h:
* wasm/js/JSWebAssemblyMemory.h:
2017-03-29 Keith Miller <keith_miller@apple.com>
WebAssembly: pack OpcodeOrigin to fit in a pointer
https://bugs.webkit.org/show_bug.cgi?id=170244
Reviewed by Michael Saboff.
This patch makes it so we don't have to have allocate the OpcodeOrigin and can just
pack all the data into the pointer B3::Origin already has.
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::parseAndCompile):
* wasm/WasmOpcodeOrigin.cpp:
(JSC::Wasm::OpcodeOrigin::dump):
* wasm/WasmOpcodeOrigin.h:
(JSC::Wasm::OpcodeOrigin::OpcodeOrigin):
(JSC::Wasm::OpcodeOrigin::opcode):
(JSC::Wasm::OpcodeOrigin::location):
2017-03-29 JF Bastien <jfbastien@apple.com>
WebAssembly: NFC s/goto/lambda/g
https://bugs.webkit.org/show_bug.cgi?id=170242
Reviewed by Mark Lam.
Lambdas are more in-style than the goto I just used.
* wasm/WasmMemory.cpp:
(JSC::Wasm::tryGetFastMemory):
2017-03-28 Saam Barati <sbarati@apple.com>
AssemblyHelpers should not have a VM field
https://bugs.webkit.org/show_bug.cgi?id=170207
Reviewed by Yusuke Suzuki.
APIs that need VM should take one as a parameter. When doing position
independent code for Wasm, we can't tie code generation to a VM.
* b3/B3Compile.cpp:
(JSC::B3::compile):
* b3/air/testair.cpp:
* b3/testb3.cpp:
(JSC::B3::testEntrySwitchSimple):
(JSC::B3::testEntrySwitchNoEntrySwitch):
(JSC::B3::testEntrySwitchWithCommonPaths):
(JSC::B3::testEntrySwitchWithCommonPathsAndNonTrivialEntrypoint):
(JSC::B3::testEntrySwitchLoop):
* bytecode/AccessCase.cpp:
(JSC::AccessCase::generateWithGuard):
(JSC::AccessCase::generateImpl):
* bytecode/DOMJITAccessCasePatchpointParams.cpp:
(JSC::SlowPathCallGeneratorWithArguments::generateImpl):
* bytecode/InlineAccess.cpp:
(JSC::InlineAccess::dumpCacheSizesAndCrash):
(JSC::InlineAccess::generateSelfPropertyAccess):
(JSC::InlineAccess::generateSelfPropertyReplace):
(JSC::InlineAccess::generateArrayLength):
(JSC::InlineAccess::rewireStubAsJump):
* bytecode/InlineAccess.h:
* bytecode/PolymorphicAccess.cpp:
(JSC::AccessGenerationState::emitExplicitExceptionHandler):
(JSC::PolymorphicAccess::regenerate):
* bytecode/PolymorphicAccess.h:
(JSC::AccessGenerationState::AccessGenerationState):
* dfg/DFGJITCompiler.cpp:
(JSC::DFG::JITCompiler::JITCompiler):
(JSC::DFG::JITCompiler::compileExceptionHandlers):
(JSC::DFG::JITCompiler::link):
(JSC::DFG::JITCompiler::compile):
(JSC::DFG::JITCompiler::compileFunction):
(JSC::DFG::JITCompiler::exceptionCheck):
* dfg/DFGJITCompiler.h:
(JSC::DFG::JITCompiler::exceptionCheckWithCallFrameRollback):
(JSC::DFG::JITCompiler::fastExceptionCheck):
(JSC::DFG::JITCompiler::vm):
* dfg/DFGOSRExitCompiler.cpp:
* dfg/DFGOSRExitCompiler.h:
* dfg/DFGOSRExitCompiler32_64.cpp:
(JSC::DFG::OSRExitCompiler::compileExit):
* dfg/DFGOSRExitCompiler64.cpp:
(JSC::DFG::OSRExitCompiler::compileExit):
* dfg/DFGOSRExitCompilerCommon.cpp:
(JSC::DFG::adjustAndJumpToTarget):
* dfg/DFGOSRExitCompilerCommon.h:
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::emitAllocateRawObject):
(JSC::DFG::SpeculativeJIT::checkArray):
(JSC::DFG::SpeculativeJIT::compileCurrentBlock):
(JSC::DFG::SpeculativeJIT::compileInstanceOfForObject):
(JSC::DFG::SpeculativeJIT::compileMakeRope):
(JSC::DFG::SpeculativeJIT::compileGetGlobalObject):
(JSC::DFG::SpeculativeJIT::compileNewFunctionCommon):
(JSC::DFG::SpeculativeJIT::compileCreateActivation):
(JSC::DFG::SpeculativeJIT::compileCreateDirectArguments):
(JSC::DFG::SpeculativeJIT::compileSpread):
(JSC::DFG::SpeculativeJIT::compileArraySlice):
(JSC::DFG::SpeculativeJIT::compileNukeStructureAndSetButterfly):
(JSC::DFG::SpeculativeJIT::compileNewStringObject):
(JSC::DFG::SpeculativeJIT::compileNewTypedArray):
(JSC::DFG::SpeculativeJIT::compileStoreBarrier):
* dfg/DFGSpeculativeJIT.h:
(JSC::DFG::SpeculativeJIT::emitAllocateJSObjectWithKnownSize):
(JSC::DFG::SpeculativeJIT::emitAllocateJSObject):
(JSC::DFG::SpeculativeJIT::emitAllocateVariableSizedJSObject):
(JSC::DFG::SpeculativeJIT::emitAllocateDestructibleObject):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::emitCall):
(JSC::DFG::SpeculativeJIT::compileLogicalNot):
(JSC::DFG::SpeculativeJIT::emitBranch):
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::nonSpeculativeNonPeepholeCompareNullOrUndefined):
(JSC::DFG::SpeculativeJIT::nonSpeculativePeepholeBranchNullOrUndefined):
(JSC::DFG::SpeculativeJIT::emitCall):
(JSC::DFG::SpeculativeJIT::compileObjectOrOtherLogicalNot):
(JSC::DFG::SpeculativeJIT::compileLogicalNot):
(JSC::DFG::SpeculativeJIT::emitObjectOrOtherBranch):
(JSC::DFG::SpeculativeJIT::emitBranch):
(JSC::DFG::SpeculativeJIT::compile):
(JSC::DFG::SpeculativeJIT::compileAllocateNewArrayWithSize):
* dfg/DFGThunks.cpp:
(JSC::DFG::osrEntryThunkGenerator):
* ftl/FTLCompile.cpp:
(JSC::FTL::compile):
* ftl/FTLJITFinalizer.h:
* ftl/FTLLazySlowPath.cpp:
(JSC::FTL::LazySlowPath::generate):
* ftl/FTLLazySlowPathCall.h:
(JSC::FTL::createLazyCallGenerator):
* ftl/FTLLink.cpp:
(JSC::FTL::link):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::lower):
(JSC::FTL::DFG::LowerDFGToB3::compileCreateActivation):
(JSC::FTL::DFG::LowerDFGToB3::compileNewFunction):
(JSC::FTL::DFG::LowerDFGToB3::compileCreateDirectArguments):
(JSC::FTL::DFG::LowerDFGToB3::compileNewTypedArray):
(JSC::FTL::DFG::LowerDFGToB3::compileMakeRope):
(JSC::FTL::DFG::LowerDFGToB3::compileNotifyWrite):
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstructVarargsSpread):
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstructVarargs):
(JSC::FTL::DFG::LowerDFGToB3::compileCallEval):
(JSC::FTL::DFG::LowerDFGToB3::compileIsObjectOrNull):
(JSC::FTL::DFG::LowerDFGToB3::compileIsFunction):
(JSC::FTL::DFG::LowerDFGToB3::compileMaterializeNewObject):
(JSC::FTL::DFG::LowerDFGToB3::compileMaterializeCreateActivation):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckTraps):
(JSC::FTL::DFG::LowerDFGToB3::allocatePropertyStorageWithSizeImpl):
(JSC::FTL::DFG::LowerDFGToB3::allocateObject):
(JSC::FTL::DFG::LowerDFGToB3::allocateJSArray):
(JSC::FTL::DFG::LowerDFGToB3::buildTypeOf):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
* ftl/FTLSlowPathCall.h:
(JSC::FTL::callOperation):
* ftl/FTLState.h:
(JSC::FTL::State::vm):
* ftl/FTLThunks.cpp:
(JSC::FTL::genericGenerationThunkGenerator):
(JSC::FTL::slowPathCallThunkGenerator):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::jitReleaseAssertNoException):
(JSC::AssemblyHelpers::callExceptionFuzz):
(JSC::AssemblyHelpers::emitJumpIfException):
(JSC::AssemblyHelpers::emitExceptionCheck):
(JSC::AssemblyHelpers::emitNonPatchableExceptionCheck):
(JSC::AssemblyHelpers::emitLoadStructure):
(JSC::AssemblyHelpers::emitRandomThunk):
(JSC::AssemblyHelpers::restoreCalleeSavesFromVMEntryFrameCalleeSavesBuffer):
(JSC::AssemblyHelpers::emitConvertValueToBoolean):
(JSC::AssemblyHelpers::debugCall):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::AssemblyHelpers):
(JSC::AssemblyHelpers::codeBlock):
(JSC::AssemblyHelpers::copyCalleeSavesToVMEntryFrameCalleeSavesBuffer):
(JSC::AssemblyHelpers::copyCalleeSavesFromFrameOrRegisterToVMEntryFrameCalleeSavesBuffer):
(JSC::AssemblyHelpers::barrierBranch):
(JSC::AssemblyHelpers::barrierStoreLoadFence):
(JSC::AssemblyHelpers::mutatorFence):
(JSC::AssemblyHelpers::storeButterfly):
(JSC::AssemblyHelpers::nukeStructureAndStoreButterfly):
(JSC::AssemblyHelpers::jumpIfMutatorFenceNotNeeded):
(JSC::AssemblyHelpers::emitAllocateJSObjectWithKnownSize):
(JSC::AssemblyHelpers::emitAllocateJSObject):
(JSC::AssemblyHelpers::emitAllocateVariableSizedCell):
(JSC::AssemblyHelpers::emitAllocateVariableSizedJSObject):
(JSC::AssemblyHelpers::emitAllocateDestructibleObject):
(JSC::AssemblyHelpers::vm): Deleted.
(JSC::AssemblyHelpers::debugCall): Deleted.
* jit/CCallHelpers.cpp:
(JSC::CCallHelpers::ensureShadowChickenPacket):
* jit/CCallHelpers.h:
(JSC::CCallHelpers::CCallHelpers):
(JSC::CCallHelpers::jumpToExceptionHandler):
* jit/JIT.cpp:
(JSC::JIT::emitEnterOptimizationCheck):
(JSC::JIT::privateCompileExceptionHandlers):
* jit/JIT.h:
(JSC::JIT::exceptionCheck):
(JSC::JIT::exceptionCheckWithCallFrameRollback):
* jit/JITMathIC.h:
(JSC::JITMathIC::generateOutOfLine):
* jit/JITOpcodes.cpp:
(JSC::JIT::emit_op_instanceof):
(JSC::JIT::emit_op_is_undefined):
(JSC::JIT::emit_op_jfalse):
(JSC::JIT::emit_op_jeq_null):
(JSC::JIT::emit_op_jneq_null):
(JSC::JIT::emit_op_jtrue):
(JSC::JIT::emit_op_throw):
(JSC::JIT::emit_op_catch):
(JSC::JIT::emit_op_eq_null):
(JSC::JIT::emit_op_neq_null):
(JSC::JIT::emitSlow_op_loop_hint):
(JSC::JIT::emit_op_log_shadow_chicken_prologue):
(JSC::JIT::emit_op_log_shadow_chicken_tail):
* jit/JITOpcodes32_64.cpp:
(JSC::JIT::privateCompileCTINativeCall):
(JSC::JIT::emit_op_new_object):
(JSC::JIT::emit_op_jfalse):
(JSC::JIT::emit_op_jtrue):
(JSC::JIT::emit_op_throw):
(JSC::JIT::emit_op_catch):
(JSC::JIT::emit_op_create_this):
(JSC::JIT::emit_op_log_shadow_chicken_prologue):
(JSC::JIT::emit_op_log_shadow_chicken_tail):
* jit/JITPropertyAccess.cpp:
(JSC::JIT::emitWriteBarrier):
* jit/JSInterfaceJIT.h:
(JSC::JSInterfaceJIT::JSInterfaceJIT):
(JSC::JSInterfaceJIT::vm):
* jit/Repatch.cpp:
(JSC::tryCacheGetByID):
(JSC::tryCachePutByID):
(JSC::linkPolymorphicCall):
(JSC::resetGetByID):
(JSC::resetPutByID):
* jit/SetupVarargsFrame.cpp:
(JSC::emitSetupVarargsFrameFastCase):
* jit/SetupVarargsFrame.h:
* jit/SpecializedThunkJIT.h:
(JSC::SpecializedThunkJIT::loadArgumentWithSpecificClass):
* jit/ThunkGenerators.cpp:
(JSC::throwExceptionFromCallSlowPathGenerator):
(JSC::linkCallThunkGenerator):
(JSC::linkPolymorphicCallThunkGenerator):
(JSC::virtualThunkFor):
(JSC::nativeForGenerator):
(JSC::randomThunkGenerator):
(JSC::boundThisNoArgsFunctionCallGenerator):
(JSC::throwExceptionFromWasmThunkGenerator):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::parseAndCompile):
* wasm/WasmBinding.cpp:
(JSC::Wasm::wasmToJs):
(JSC::Wasm::wasmToWasm):
2017-03-28 Keith Miller <keith_miller@apple.com>
WebAssembly: We should have Origins
https://bugs.webkit.org/show_bug.cgi?id=170217
Reviewed by Mark Lam.
This patch adds wasm origins for B3::Values, called OpcodeOrigin. Currently,
OpcodeOrigin just tracks the original opcode and the location of that opcode.
Here's a sample:
BB#0: ; frequency = 1.000000
Int64 @4 = Patchpoint(generator = 0x10f487fa8, earlyClobbered = [], lateClobbered = [], usedRegisters = [], resultConstraint = SomeRegister)
Int64 @5 = FramePointer()
Void @8 = Store(@4, @5, offset = 24, ControlDependent|Writes:Top)
Int64 @10 = Const64(0)
Void @12 = Store($0(@10), @5, offset = 16, ControlDependent|Writes:Top)
Int64 @13 = Patchpoint(generator = 0x10f4be7f0, earlyClobbered = [], lateClobbered = [], usedRegisters = [], resultConstraint = SomeRegister, ExitsSideways|ControlDependent|WritesPinned|ReadsPinned|Fence|Writes:Top|Reads:Top)
Int64 @16 = ArgumentReg(%rdi)
Int64 @18 = ArgumentReg(%rsi)
Int32 @22 = Trunc(@18, Wasm: {opcode: I64Rotl, location: 5})
Int64 @23 = RotL(@16, @22, Wasm: {opcode: I64Rotl, location: 5})
Void @27 = Return(@23, Terminal, Wasm: {opcode: End, location: 6})
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/B3Value.cpp:
(JSC::B3::Value::deepDump):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::setParser):
(JSC::Wasm::B3IRGenerator::restoreWebAssemblyGlobalState):
(JSC::Wasm::B3IRGenerator::emitCheckAndPreparePointer):
(JSC::Wasm::B3IRGenerator::emitLoadOp):
(JSC::Wasm::B3IRGenerator::emitStoreOp):
(JSC::Wasm::B3IRGenerator::addConstant):
(JSC::Wasm::B3IRGenerator::addLoop):
(JSC::Wasm::B3IRGenerator::unify):
(JSC::Wasm::parseAndCompile):
(JSC::Wasm::B3IRGenerator::emitChecksForModOrDiv):
(JSC::Wasm::getMemoryBaseAndSize): Deleted.
* wasm/WasmFunctionParser.h:
(JSC::Wasm::FunctionParser::currentOpcode):
(JSC::Wasm::FunctionParser::currentOpcodeStartingOffset):
(JSC::Wasm::FunctionParser<Context>::FunctionParser):
* wasm/WasmOpcodeOrigin.cpp: Added.
(JSC::Wasm::OpcodeOrigin::dump):
* wasm/WasmOpcodeOrigin.h: Added.
(JSC::Wasm::OpcodeOrigin::OpcodeOrigin):
* wasm/WasmValidate.cpp:
(JSC::Wasm::Validate::setParser):
* wasm/generateWasmB3IRGeneratorInlinesHeader.py:
(CodeGenerator.generate):
(generateB3OpCode):
(generateConstCode):
2017-03-28 JF Bastien <jfbastien@apple.com>
WebAssembly: option to crash if no fast memory is available
https://bugs.webkit.org/show_bug.cgi?id=170219
Reviewed by Mark Lam.
* runtime/Options.h:
* wasm/WasmMemory.cpp:
(JSC::Wasm::webAssemblyCouldntGetFastMemory):
(JSC::Wasm::tryGetFastMemory):
2017-03-28 Mark Lam <mark.lam@apple.com>
The Mutator should not be able to steal the conn if the Collector hasn't reached the NotRunning phase yet.
https://bugs.webkit.org/show_bug.cgi?id=170213
<rdar://problem/30755345>
Reviewed by Filip Pizlo.
The current condition for stealing the conn isn't tight enough. Restricting the
stealing to when m_currentPhase == NotRunning ensures that the Collector is
really done running.
No test because this issue only manifests with a race condition that is difficult
to reproduce on demand.
* heap/Heap.cpp:
(JSC::Heap::requestCollection):
2017-03-28 Keith Miller <keith_miller@apple.com>
WebAssembly: Make WebAssembly.instantiate/compile truly asynchronous
https://bugs.webkit.org/show_bug.cgi?id=169187
Reviewed by Saam Barati.
This patch allows WebAssembly compilations to happen asynchronously.
To do so, it refactors how much of the compilation happens and adds
new infrastructure for async promises.
First, there is a new class, PromiseDeferredTimer that lives on
the VM. PromiseDeferredTimer will manage the life-cycle of async
pending promises and any dependencies that promise
needs. PromiseDeferredTimer automagically releases the pending
promise and dependencies once the JSPromiseDeferred is resolved or
rejected. Additionally, PromiseDeferredTimer provides a mechanism
to poll the run-loop whenever the async task needs to synchronize
with the JS thread. Normally, that will be whenever the async task
finishes. In the case of Web Assembly we also use this feature for
the compile + instantiate case, where we might have more work
after the first async task completes (more on that later).
The next class is Wasm::Worklist, which is used to manage Wasm
compilation tasks. The worklist class works similarly to the
DFG/FTL Worklists. It has a pool of threads that it manages. One
interesting aspect of Wasm Worklist is that it can synchronously
compile a plan that is already potentially running
asynchronously. This can occur if a user calls
WebAssembly.instantiate() then new WebAssembly.instantiate() on
the same module. In that case the Wasm Worklist will bump the
priority of the running pending Plan and block the JS thread.
This patch also makes some of the Wasm Plan code cleaner. Since we
now defer all compilation to instantiation time, we no longer need
to guess at which memory we are going to get. Also, Wasm Plans now
track the work they have done with a state enum.
Finally, this patch makes renamed HeapTimer to JSRunLoopTimer. It
also adds changes test262AsyncTest to a more generic testing
infrastructure. Now, in addition to the old functionality, you can
call asyncTest() with the number of tests you expect. When the jsc
CLI exits, it will guarantee that asyncTestPassed() is called that
many times.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* heap/GCActivityCallback.h:
* heap/IncrementalSweeper.cpp:
(JSC::IncrementalSweeper::scheduleTimer):
(JSC::IncrementalSweeper::IncrementalSweeper):
* heap/IncrementalSweeper.h:
* heap/StopIfNecessaryTimer.cpp:
(JSC::StopIfNecessaryTimer::StopIfNecessaryTimer):
* heap/StopIfNecessaryTimer.h:
* heap/StrongInlines.h:
* jsc.cpp:
(GlobalObject::finishCreation):
(printInternal):
(functionAsyncTestStart):
(functionAsyncTestPassed):
(functionTestWasmModuleFunctions):
(CommandLine::parseArguments):
(runJSC):
* runtime/JSPromiseDeferred.cpp:
(JSC::JSPromiseDeferred::resolve):
(JSC::JSPromiseDeferred::reject):
* runtime/JSPromiseDeferred.h:
(JSC::JSPromiseDeferred::promiseAsyncPending):
* runtime/JSRunLoopTimer.cpp: Renamed from Source/JavaScriptCore/heap/HeapTimer.cpp.
(JSC::JSRunLoopTimer::JSRunLoopTimer):
(JSC::JSRunLoopTimer::setRunLoop):
(JSC::JSRunLoopTimer::~JSRunLoopTimer):
(JSC::JSRunLoopTimer::timerDidFire):
(JSC::JSRunLoopTimer::scheduleTimer):
(JSC::JSRunLoopTimer::cancelTimer):
(JSC::JSRunLoopTimer::invalidate):
* runtime/JSRunLoopTimer.h: Copied from Source/JavaScriptCore/heap/HeapTimer.h.
* runtime/Options.h:
* runtime/PromiseDeferredTimer.cpp: Added.
(JSC::PromiseDeferredTimer::PromiseDeferredTimer):
(JSC::PromiseDeferredTimer::doWork):
(JSC::PromiseDeferredTimer::runRunLoop):
(JSC::PromiseDeferredTimer::addPendingPromise):
(JSC::PromiseDeferredTimer::cancelPendingPromise):
(JSC::PromiseDeferredTimer::scheduleWorkSoon):
(JSC::PromiseDeferredTimer::scheduleBlockedTask):
* runtime/PromiseDeferredTimer.h: Renamed from Source/JavaScriptCore/heap/HeapTimer.h.
(JSC::PromiseDeferredTimer::stopRunningTasks):
* runtime/VM.cpp:
(JSC::VM::VM):
(JSC::VM::~VM):
* runtime/VM.h:
* wasm/JSWebAssembly.cpp:
(JSC::reject):
(JSC::webAssemblyCompileFunc):
(JSC::resolve):
(JSC::instantiate):
(JSC::compileAndInstantiate):
(JSC::webAssemblyInstantiateFunc):
(JSC::webAssemblyValidateFunc):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::B3IRGenerator):
(JSC::Wasm::B3IRGenerator::emitCheckAndPreparePointer):
(JSC::Wasm::B3IRGenerator::memoryKind):
(JSC::Wasm::parseAndCompile):
* wasm/WasmB3IRGenerator.h:
* wasm/WasmFormat.h:
(JSC::Wasm::ModuleInformation::internalFunctionCount):
* wasm/WasmFunctionParser.h:
* wasm/WasmMemory.h:
* wasm/WasmMemoryInformation.cpp:
(JSC::Wasm::MemoryInformation::MemoryInformation):
* wasm/WasmMemoryInformation.h:
(JSC::Wasm::MemoryInformation::maximum):
(JSC::Wasm::MemoryInformation::hasReservedMemory): Deleted.
(JSC::Wasm::MemoryInformation::takeReservedMemory): Deleted.
(JSC::Wasm::MemoryInformation::mode): Deleted.
* wasm/WasmModuleParser.cpp:
* wasm/WasmModuleParser.h:
(JSC::Wasm::ModuleParser::ModuleParser):
* wasm/WasmPlan.cpp:
(JSC::Wasm::Plan::Plan):
(JSC::Wasm::Plan::stateString):
(JSC::Wasm::Plan::moveToState):
(JSC::Wasm::Plan::fail):
(JSC::Wasm::Plan::parseAndValidateModule):
(JSC::Wasm::Plan::prepare):
(JSC::Wasm::Plan::ThreadCountHolder::ThreadCountHolder):
(JSC::Wasm::Plan::ThreadCountHolder::~ThreadCountHolder):
(JSC::Wasm::Plan::compileFunctions):
(JSC::Wasm::Plan::complete):
(JSC::Wasm::Plan::waitForCompletion):
(JSC::Wasm::Plan::cancel):
(JSC::Wasm::Plan::run): Deleted.
(JSC::Wasm::Plan::initializeCallees): Deleted.
* wasm/WasmPlan.h:
(JSC::Wasm::Plan::dontFinalize):
(JSC::Wasm::Plan::exports):
(JSC::Wasm::Plan::internalFunctionCount):
(JSC::Wasm::Plan::takeModuleInformation):
(JSC::Wasm::Plan::takeCallLinkInfos):
(JSC::Wasm::Plan::takeWasmExitStubs):
(JSC::Wasm::Plan::setModeAndPromise):
(JSC::Wasm::Plan::mode):
(JSC::Wasm::Plan::pendingPromise):
(JSC::Wasm::Plan::vm):
(JSC::Wasm::Plan::errorMessage):
(JSC::Wasm::Plan::failed):
(JSC::Wasm::Plan::hasWork):
(JSC::Wasm::Plan::hasBeenPrepared):
* wasm/WasmPlanInlines.h: Copied from Source/JavaScriptCore/wasm/WasmB3IRGenerator.h.
(JSC::Wasm::Plan::initializeCallees):
* wasm/WasmValidate.cpp:
* wasm/WasmWorklist.cpp: Added.
(JSC::Wasm::Worklist::priorityString):
(JSC::Wasm::Worklist::QueueElement::setToNextPriority):
(JSC::Wasm::Worklist::iterate):
(JSC::Wasm::Worklist::enqueue):
(JSC::Wasm::Worklist::completePlanSynchronously):
(JSC::Wasm::Worklist::stopAllPlansForVM):
(JSC::Wasm::Worklist::Worklist):
(JSC::Wasm::Worklist::~Worklist):
(JSC::Wasm::existingWorklistOrNull):
(JSC::Wasm::ensureWorklist):
* wasm/WasmWorklist.h: Added.
(JSC::Wasm::Worklist::nextTicket):
(JSC::Wasm::Worklist::Comparator::operator()):
* wasm/js/JSWebAssemblyCallee.h:
* wasm/js/JSWebAssemblyCodeBlock.cpp:
(JSC::JSWebAssemblyCodeBlock::JSWebAssemblyCodeBlock):
(JSC::JSWebAssemblyCodeBlock::initialize):
(JSC::JSWebAssemblyCodeBlock::isSafeToRun):
* wasm/js/JSWebAssemblyCodeBlock.h:
(JSC::JSWebAssemblyCodeBlock::create):
(JSC::JSWebAssemblyCodeBlock::initialized):
(JSC::JSWebAssemblyCodeBlock::plan):
(JSC::JSWebAssemblyCodeBlock::runnable):
(JSC::JSWebAssemblyCodeBlock::errorMessage):
(JSC::JSWebAssemblyCodeBlock::callees):
* wasm/js/JSWebAssemblyHelpers.h:
(JSC::createSourceBufferFromValue):
* wasm/js/JSWebAssemblyInstance.cpp:
(JSC::JSWebAssemblyInstance::finishCreation):
(JSC::JSWebAssemblyInstance::visitChildren):
(JSC::JSWebAssemblyInstance::addUnitializedCodeBlock):
(JSC::JSWebAssemblyInstance::finalizeCreation):
(JSC::JSWebAssemblyInstance::create):
(JSC::JSWebAssemblyInstance::setMemory): Deleted.
* wasm/js/JSWebAssemblyInstance.h:
(JSC::JSWebAssemblyInstance::codeBlock):
(JSC::JSWebAssemblyInstance::initialized):
(JSC::JSWebAssemblyInstance::module):
(JSC::JSWebAssemblyInstance::importFunction):
(JSC::JSWebAssemblyInstance::setMemory):
(JSC::JSWebAssemblyInstance::table):
(JSC::JSWebAssemblyInstance::importFunctions):
(JSC::JSWebAssemblyInstance::setImportFunction): Deleted.
(JSC::JSWebAssemblyInstance::setTable): Deleted.
* wasm/js/JSWebAssemblyModule.cpp:
(JSC::JSWebAssemblyModule::createStub):
(JSC::JSWebAssemblyModule::JSWebAssemblyModule):
(JSC::JSWebAssemblyModule::finishCreation):
(JSC::JSWebAssemblyModule::setCodeBlock):
(JSC::JSWebAssemblyModule::buildCodeBlock): Deleted.
(JSC::JSWebAssemblyModule::create): Deleted.
(JSC::JSWebAssemblyModule::codeBlock): Deleted.
* wasm/js/JSWebAssemblyModule.h:
(JSC::JSWebAssemblyModule::moduleInformation):
(JSC::JSWebAssemblyModule::codeBlock):
(JSC::JSWebAssemblyModule::source):
(JSC::JSWebAssemblyModule::takeReservedMemory): Deleted.
(JSC::JSWebAssemblyModule::codeBlockFor): Deleted.
* wasm/js/WebAssemblyInstanceConstructor.cpp:
(JSC::constructJSWebAssemblyInstance):
(JSC::WebAssemblyInstanceConstructor::createInstance): Deleted.
* wasm/js/WebAssemblyModuleConstructor.cpp:
(JSC::WebAssemblyModuleConstructor::createModule):
* wasm/js/WebAssemblyModulePrototype.cpp:
(JSC::webAssemblyModuleProtoImports):
(JSC::webAssemblyModuleProtoExports):
* wasm/js/WebAssemblyModuleRecord.cpp:
(JSC::WebAssemblyModuleRecord::finishCreation):
(JSC::WebAssemblyModuleRecord::link):
(JSC::WebAssemblyModuleRecord::evaluate):
* wasm/js/WebAssemblyModuleRecord.h:
2017-03-28 Yusuke Suzuki <utatane.tea@gmail.com>
WebAssembly: add fallback to use pinned register to load/store state
https://bugs.webkit.org/show_bug.cgi?id=169773
Reviewed by Saam Barati.
This patch adds a new pinned register to hold JSWebAssemblyInstance,
which is used to represent the context of running Wasm code.
While we use fast TLS to hold the context in macOS, we do not have
any system reserved fast TLS slot in the other systems. This pinned
register approach is used in these systems. These changes decouple
VM from Wasm module to make Wasm module position independent code.
While using fast TLS could be beneficial in x64 systems which number of
registers is relatively small, pinned register approach could be
beneficial in ARM64 which has plenty of registers. In macOS, we can
switch the implementation with the runtime flag. Thus macOS port can
compare the performance and decide which implementation is used after
landing this patch.
* heap/MarkedBlock.h:
(JSC::MarkedBlock::offsetOfVM):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::loadWasmContext):
(JSC::AssemblyHelpers::storeWasmContext):
(JSC::AssemblyHelpers::loadWasmContextNeedsMacroScratchRegister):
(JSC::AssemblyHelpers::storeWasmContextNeedsMacroScratchRegister):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::loadWasmContext): Deleted.
(JSC::AssemblyHelpers::storeWasmContext): Deleted.
(JSC::AssemblyHelpers::loadWasmContextNeedsMacroScratchRegister): Deleted.
(JSC::AssemblyHelpers::storeWasmContextNeedsMacroScratchRegister): Deleted.
* jit/Repatch.cpp:
(JSC::webAssemblyOwner):
(JSC::linkFor):
(JSC::linkPolymorphicCall):
(JSC::isWebAssemblyToJSCallee): Deleted.
* jit/ThunkGenerators.cpp:
(JSC::throwExceptionFromWasmThunkGenerator):
* llint/LLIntData.cpp:
(JSC::LLInt::Data::performAssertions):
* llint/LowLevelInterpreter.asm:
* runtime/JSCell.cpp:
(JSC::JSCell::isAnyWasmCallee):
* runtime/JSCellInlines.h:
(JSC::isWebAssemblyToJSCallee):
* runtime/JSType.h:
* runtime/StackFrame.cpp:
(JSC::StackFrame::functionName):
* runtime/VM.cpp:
(JSC::VM::VM):
* runtime/VM.h:
(JSC::VM::wasmContextOffset):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::materializeWasmContext):
(JSC::Wasm::B3IRGenerator::restoreWasmContext):
(JSC::Wasm::B3IRGenerator::B3IRGenerator):
(JSC::Wasm::getMemoryBaseAndSize):
(JSC::Wasm::B3IRGenerator::restoreWebAssemblyGlobalState):
(JSC::Wasm::createJSToWasmWrapper):
(JSC::Wasm::loadWasmContext): Deleted.
(JSC::Wasm::storeWasmContext): Deleted.
(JSC::Wasm::restoreWebAssemblyGlobalState): Deleted.
* wasm/WasmBinding.cpp:
(JSC::Wasm::wasmToJs):
* wasm/WasmContext.cpp:
(JSC::loadWasmContext):
(JSC::storeWasmContext):
* wasm/WasmContext.h:
* wasm/WasmMemoryInformation.cpp:
(JSC::Wasm::getPinnedRegisters):
(JSC::Wasm::PinnedRegisterInfo::get):
(JSC::Wasm::PinnedRegisterInfo::PinnedRegisterInfo):
* wasm/WasmMemoryInformation.h:
(JSC::Wasm::PinnedRegisterInfo::toSave):
(JSC::Wasm::useFastTLS):
(JSC::Wasm::useFastTLSForWasmContext):
* wasm/js/JSWebAssemblyInstance.cpp:
(JSC::JSWebAssemblyInstance::finishCreation):
(JSC::JSWebAssemblyInstance::visitChildren):
* wasm/js/JSWebAssemblyInstance.h:
(JSC::JSWebAssemblyInstance::offsetOfCallee):
* wasm/js/JSWebAssemblyModule.cpp:
(JSC::JSWebAssemblyModule::finishCreation):
(JSC::JSWebAssemblyModule::visitChildren):
* wasm/js/JSWebAssemblyModule.h:
(JSC::JSWebAssemblyModule::callee):
* wasm/js/WebAssemblyFunction.cpp:
(JSC::callWebAssemblyFunction):
(JSC::WebAssemblyFunction::create):
* wasm/js/WebAssemblyToJSCallee.cpp:
(JSC::WebAssemblyToJSCallee::create):
(JSC::WebAssemblyToJSCallee::createStructure):
(JSC::WebAssemblyToJSCallee::finishCreation):
(JSC::WebAssemblyToJSCallee::visitChildren):
(JSC::WebAssemblyToJSCallee::destroy): Deleted.
* wasm/js/WebAssemblyToJSCallee.h:
2017-03-28 Brian Burg <bburg@apple.com>
Web Inspector: Add "Disable Caches" option that only applies to the inspected page while Web Inspector is open
https://bugs.webkit.org/show_bug.cgi?id=169865
<rdar://problem/31250573>
Reviewed by Joseph Pecoraro.
* inspector/protocol/Network.json:
Rename the command for disabling resource caching to match the WebCore::Page
flag. This also removes the possibility that this could be confused for the old,
buggy command that this patch rips out.
2017-03-25 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC] Move platformThreadSignal to WTF
https://bugs.webkit.org/show_bug.cgi?id=170097
Reviewed by Mark Lam.
It is a small clean up towards https://bugs.webkit.org/show_bug.cgi?id=170027.
platformThreadSignal uses PlatformThread in JSC, but it can be implemented in
WTF ThreadIdentifier.
* runtime/JSLock.cpp:
(JSC::JSLock::lock):
* runtime/JSLock.h:
(JSC::JSLock::ownerThread):
(JSC::JSLock::currentThreadIsHoldingLock):
* runtime/PlatformThread.h:
(JSC::platformThreadSignal): Deleted.
* runtime/VM.h:
(JSC::VM::ownerThread):
* runtime/VMTraps.cpp:
(JSC::VMTraps::SignalSender::send):
2017-03-28 JF Bastien <jfbastien@apple.com>
WebAssembly: implement Module imports/exports
https://bugs.webkit.org/show_bug.cgi?id=166982
Reviewed by Saam Barati.
As defined in: https://github.com/WebAssembly/design/commit/18cbacb90cd3584dd5c9aa3d392e4e55f66af6ab
* wasm/WasmFormat.h:
(JSC::Wasm::makeString): use uppercase instead, it was only used
for diagnostic but is now used for the expected JS property's
capitalization
* wasm/js/WebAssemblyModulePrototype.cpp:
(JSC::webAssemblyModuleProtoImports):
(JSC::webAssemblyModuleProtoExports):
2017-03-27 JF Bastien <jfbastien@apple.com>
WebAssembly: JSWebAssemblyCodeBlock.h belongs in JavaScriptCore/wasm/js not JavaScriptCore/wasm
https://bugs.webkit.org/show_bug.cgi?id=170160
Reviewed by Mark Lam.
* JavaScriptCore.xcodeproj/project.pbxproj:
* wasm/js/JSWebAssemblyCodeBlock.h: Renamed from Source/JavaScriptCore/wasm/JSWebAssemblyCodeBlock.h.
2017-03-27 JF Bastien <jfbastien@apple.com>
WebAssembly: misc memory testing
https://bugs.webkit.org/show_bug.cgi?id=170137
Reviewed by Keith Miller.
* wasm/js/WebAssemblyInstanceConstructor.cpp:
(JSC::WebAssemblyInstanceConstructor::createInstance): improve error messages
2017-03-27 Michael Saboff <msaboff@apple.com>
Add ARM64 system instructions to disassembler
https://bugs.webkit.org/show_bug.cgi?id=170084
Reviewed by Saam Barati.
This changes adds support for MRS and MSR instructions, and refactors the DMB
disassembly to handle all of the barrier instructions.
* disassembler/ARM64/A64DOpcode.cpp:
(JSC::ARM64Disassembler::A64DOpcodeMSRImmediate::format):
(JSC::ARM64Disassembler::A64DOpcodeMSROrMRSRegister::format):
(JSC::ARM64Disassembler::A64DOpcodeSystemSync::format):
(JSC::ARM64Disassembler::A64DOpcodeDmb::format): Deleted.
* disassembler/ARM64/A64DOpcode.h:
(JSC::ARM64Disassembler::A64DOpcodeSystem::lBit):
(JSC::ARM64Disassembler::A64DOpcodeSystem::op0):
(JSC::ARM64Disassembler::A64DOpcodeSystem::op1):
(JSC::ARM64Disassembler::A64DOpcodeSystem::crN):
(JSC::ARM64Disassembler::A64DOpcodeSystem::crM):
(JSC::ARM64Disassembler::A64DOpcodeSystem::op2):
(JSC::ARM64Disassembler::A64DOpcodeMSROrMRSRegister::opName):
(JSC::ARM64Disassembler::A64DOpcodeMSROrMRSRegister::systemRegister):
(JSC::ARM64Disassembler::A64DOpcodeSystemSync::opName):
(JSC::ARM64Disassembler::A64DOpcodeSystemSync::option):
(JSC::ARM64Disassembler::A64DOpcodeDmb::opName): Deleted.
(JSC::ARM64Disassembler::A64DOpcodeDmb::option): Deleted.
(JSC::ARM64Disassembler::A64DOpcodeDmb::crM): Deleted.
2017-03-26 Filip Pizlo <fpizlo@apple.com>
B3::fixSSA should do liveness pruning
https://bugs.webkit.org/show_bug.cgi?id=170111
Reviewed by Saam Barati.
This moves all of the logic of Air::Liveness<> to WTF::Liveness<> and then uses that to
create B3::VariableLiveness. Then this uses VariableLiveness::LiveAtHead to prune Phi
construction.
This makes B3::fixSSA run twice as fast. This is a 13% progression on WasmBench compile
times.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/B3BasicBlock.h:
(JSC::B3::BasicBlock::get):
* b3/B3FixSSA.cpp:
(JSC::B3::fixSSA):
* b3/B3VariableLiveness.cpp: Added.
(JSC::B3::VariableLiveness::VariableLiveness):
(JSC::B3::VariableLiveness::~VariableLiveness):
* b3/B3VariableLiveness.h: Added.
(JSC::B3::VariableLivenessAdapter::VariableLivenessAdapter):
(JSC::B3::VariableLivenessAdapter::numIndices):
(JSC::B3::VariableLivenessAdapter::valueToIndex):
(JSC::B3::VariableLivenessAdapter::indexToValue):
(JSC::B3::VariableLivenessAdapter::blockSize):
(JSC::B3::VariableLivenessAdapter::forEachEarlyUse):
(JSC::B3::VariableLivenessAdapter::forEachLateUse):
(JSC::B3::VariableLivenessAdapter::forEachEarlyDef):
(JSC::B3::VariableLivenessAdapter::forEachLateDef):
* b3/air/AirCFG.h: Added.
(JSC::B3::Air::CFG::CFG):
(JSC::B3::Air::CFG::root):
(JSC::B3::Air::CFG::newMap):
(JSC::B3::Air::CFG::successors):
(JSC::B3::Air::CFG::predecessors):
(JSC::B3::Air::CFG::index):
(JSC::B3::Air::CFG::node):
(JSC::B3::Air::CFG::numNodes):
(JSC::B3::Air::CFG::dump):
* b3/air/AirCode.cpp:
(JSC::B3::Air::Code::Code):
* b3/air/AirCode.h:
(JSC::B3::Air::Code::cfg):
* b3/air/AirLiveness.h:
(JSC::B3::Air::LivenessAdapter::LivenessAdapter):
(JSC::B3::Air::LivenessAdapter::blockSize):
(JSC::B3::Air::LivenessAdapter::forEachEarlyUse):
(JSC::B3::Air::LivenessAdapter::forEachLateUse):
(JSC::B3::Air::LivenessAdapter::forEachEarlyDef):
(JSC::B3::Air::LivenessAdapter::forEachLateDef):
(JSC::B3::Air::TmpLivenessAdapter::TmpLivenessAdapter):
(JSC::B3::Air::TmpLivenessAdapter::numIndices):
(JSC::B3::Air::StackSlotLivenessAdapter::StackSlotLivenessAdapter):
(JSC::B3::Air::StackSlotLivenessAdapter::numIndices):
(JSC::B3::Air::StackSlotLivenessAdapter::indexToValue):
(JSC::B3::Air::Liveness::Liveness):
(JSC::B3::Air::Liveness::LocalCalc::LocalCalc): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::Iterable): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::iterator): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::operator++): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::operator*): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::operator==): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::operator!=): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::begin): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::end): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::Iterable::contains): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::live): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::isLive): Deleted.
(JSC::B3::Air::Liveness::LocalCalc::execute): Deleted.
(JSC::B3::Air::Liveness::rawLiveAtHead): Deleted.
(JSC::B3::Air::Liveness::Iterable::Iterable): Deleted.
(JSC::B3::Air::Liveness::Iterable::iterator::iterator): Deleted.
(JSC::B3::Air::Liveness::Iterable::iterator::operator*): Deleted.
(JSC::B3::Air::Liveness::Iterable::iterator::operator++): Deleted.
(JSC::B3::Air::Liveness::Iterable::iterator::operator==): Deleted.
(JSC::B3::Air::Liveness::Iterable::iterator::operator!=): Deleted.
(JSC::B3::Air::Liveness::Iterable::begin): Deleted.
(JSC::B3::Air::Liveness::Iterable::end): Deleted.
(JSC::B3::Air::Liveness::Iterable::contains): Deleted.
(JSC::B3::Air::Liveness::liveAtHead): Deleted.
(JSC::B3::Air::Liveness::liveAtTail): Deleted.
(JSC::B3::Air::Liveness::workset): Deleted.
2017-03-25 Filip Pizlo <fpizlo@apple.com>
Air::Liveness shouldn't need HashSets
https://bugs.webkit.org/show_bug.cgi?id=170102
Reviewed by Yusuke Suzuki.
This converts Air::Liveness<> to no longer use HashSets or BitVectors. This turns out to be
easy because it's cheap enough to do a sorted merge of the things being added to liveAtHead and
the things in the predecessors' liveAtTail. This turns out to be faster - it's a 2% overall
compile time progression on WasmBench.
* b3/B3LowerToAir.cpp:
(JSC::B3::Air::LowerToAir::lower): Add a FIXME unrelated to this patch.
* b3/air/AirLiveness.h:
(JSC::B3::Air::AbstractLiveness::AbstractLiveness):
(JSC::B3::Air::AbstractLiveness::LocalCalc::LocalCalc):
(JSC::B3::Air::AbstractLiveness::rawLiveAtHead):
(JSC::B3::Air::AbstractLiveness::liveAtHead):
(JSC::B3::Air::AbstractLiveness::liveAtTail):
* b3/air/AirTmp.h:
(JSC::B3::Air::Tmp::bank):
(JSC::B3::Air::Tmp::tmpIndex):
* dfg/DFGStoreBarrierClusteringPhase.cpp:
2017-03-26 Filip Pizlo <fpizlo@apple.com>
Air should use RegisterSet for RegLiveness
https://bugs.webkit.org/show_bug.cgi?id=170108
Reviewed by Yusuke Suzuki.
The biggest change here is the introduction of the new RegLiveness class. This is a
drop-in replacement for the old RegLiveness, which was a specialization of
AbstractLiveness<>, but it's about 30% faster. It gets its speed boost from just using
sets everywhere, which is efficient for registers since RegisterSet is just two (on
x86-64) or three 32-bit (on ARM64) statically allocated words. This looks like a 1%
compile time progression on WasmBench.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* b3/B3TimingScope.cpp: Records phase timing totals.
(JSC::B3::TimingScope::TimingScope):
(JSC::B3::TimingScope::~TimingScope):
* b3/B3TimingScope.h:
* b3/air/AirAllocateRegistersByGraphColoring.cpp:
(JSC::B3::Air::allocateRegistersByGraphColoring):
* b3/air/AirLiveness.h: Move code around and rename a bit to make it more like RegLiveness; in particular we want the `iterator` to be called `iterator` not `Iterator`, and we want it to be internal to its iterable. Also rename this template to Liveness, to match the header filename.
(JSC::B3::Air::Liveness::Liveness):
(JSC::B3::Air::Liveness::LocalCalc::LocalCalc):
(JSC::B3::Air::Liveness::LocalCalc::Iterable::Iterable):
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::iterator):
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::operator++):
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::operator*):
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::operator==):
(JSC::B3::Air::Liveness::LocalCalc::Iterable::iterator::operator!=):
(JSC::B3::Air::Liveness::LocalCalc::Iterable::begin):
(JSC::B3::Air::Liveness::LocalCalc::Iterable::end):
(JSC::B3::Air::Liveness::Iterable::Iterable):
(JSC::B3::Air::Liveness::Iterable::iterator::iterator):
(JSC::B3::Air::RegLivenessAdapter::RegLivenessAdapter): Deleted.
(JSC::B3::Air::RegLivenessAdapter::numIndices): Deleted.
(JSC::B3::Air::RegLivenessAdapter::acceptsBank): Deleted.
(JSC::B3::Air::RegLivenessAdapter::acceptsRole): Deleted.
(JSC::B3::Air::RegLivenessAdapter::valueToIndex): Deleted.
(JSC::B3::Air::RegLivenessAdapter::indexToValue): Deleted.
(JSC::B3::Air::AbstractLiveness::AbstractLiveness): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::LocalCalc): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterator::Iterator): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterator::operator++): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterator::operator*): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterator::operator==): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterator::operator!=): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterable::Iterable): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterable::begin): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterable::end): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::Iterable::contains): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::live): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::isLive): Deleted.
(JSC::B3::Air::AbstractLiveness::LocalCalc::execute): Deleted.
(JSC::B3::Air::AbstractLiveness::rawLiveAtHead): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::Iterable): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::iterator::iterator): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::iterator::operator*): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::iterator::operator++): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::iterator::operator==): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::iterator::operator!=): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::begin): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::end): Deleted.
(JSC::B3::Air::AbstractLiveness::Iterable::contains): Deleted.
(JSC::B3::Air::AbstractLiveness::liveAtHead): Deleted.
(JSC::B3::Air::AbstractLiveness::liveAtTail): Deleted.
(JSC::B3::Air::AbstractLiveness::workset): Deleted.
* b3/air/AirLogRegisterPressure.cpp:
* b3/air/AirLowerAfterRegAlloc.cpp:
* b3/air/AirRegLiveness.cpp: Added.
(JSC::B3::Air::RegLiveness::RegLiveness):
(JSC::B3::Air::RegLiveness::~RegLiveness):
(JSC::B3::Air::RegLiveness::LocalCalc::execute):
* b3/air/AirRegLiveness.h: Added.
(JSC::B3::Air::RegLiveness::LocalCalc::LocalCalc):
(JSC::B3::Air::RegLiveness::LocalCalc::live):
(JSC::B3::Air::RegLiveness::LocalCalc::isLive):
(JSC::B3::Air::RegLiveness::liveAtHead):
(JSC::B3::Air::RegLiveness::liveAtTail):
* b3/air/AirReportUsedRegisters.cpp:
* jit/RegisterSet.h:
(JSC::RegisterSet::add):
(JSC::RegisterSet::remove):
(JSC::RegisterSet::contains):
(JSC::RegisterSet::subsumes):
(JSC::RegisterSet::iterator::iterator):
(JSC::RegisterSet::iterator::operator*):
(JSC::RegisterSet::iterator::operator++):
(JSC::RegisterSet::iterator::operator==):
(JSC::RegisterSet::iterator::operator!=):
(JSC::RegisterSet::begin):
(JSC::RegisterSet::end):
2017-03-25 Filip Pizlo <fpizlo@apple.com>
Fix wasm by returning after we do TLS.
Rubber stamped by Keith Miller.
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::storeWasmContext):
2017-03-24 Mark Lam <mark.lam@apple.com>
Add some instrumentation in Heap::resumeThePeriphery() to help debug an issue.
https://bugs.webkit.org/show_bug.cgi?id=170086
<rdar://problem/31253673>
Reviewed by Saam Barati.
Adding some instrumentation in Heap::resumeThePeriphery() to dump some Heap state
just before we RELEASE_ASSERT_NOT_REACHED.
* heap/Heap.cpp:
(JSC::Heap::resumeThePeriphery):
2017-03-24 JF Bastien <jfbastien@apple.com>
WebAssembly: store state in TLS instead of on VM
https://bugs.webkit.org/show_bug.cgi?id=169611
Reviewed by Filip Pizlo.
Using thread-local storage instead of VM makes code more position
independent. We used to store the WebAssembly top Instance (the
latest one in the call stack) on VM, now we instead store it in
TLS. This top Instance is used to access a bunch of state such as
Memory location, size, table (for call_indirect), etc.
Instead of calling it "top", which is confusing, we now just call
it WasmContext.
Making the code PIC means future patches will be able to
postMessage and structured clone into IDB without having to
recompile the code. This wasn't possible before because we
hard-coded the address of VM at compilation time. That doesn't
work between workers, and doesn't work across reloads (which IDB
is intended to do).
It'll also potentially make code faster once we start tuning
what's in TLS, what's in which of the 4 free slots, and what's in
pinned registers. I'm leaving this tuning for later because
there's lower lying fruit for us to pick.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* assembler/AbstractMacroAssembler.h:
* assembler/AllowMacroScratchRegisterUsageIf.h: Copied from assembler/AllowMacroScratchRegisterUsage.h.
(JSC::AllowMacroScratchRegisterUsageIf::AllowMacroScratchRegisterUsageIf):
(JSC::AllowMacroScratchRegisterUsageIf::~AllowMacroScratchRegisterUsageIf):
* assembler/MacroAssembler.h:
(JSC::MacroAssembler::storeToTLSPtr): we previously didn't have
the code required to store to TLS, only to load
* assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::loadFromTLSPtrNeedsMacroScratchRegister):
(JSC::MacroAssemblerARM64::storeToTLS32):
(JSC::MacroAssemblerARM64::storeToTLS64):
(JSC::MacroAssemblerARM64::storeToTLSPtrNeedsMacroScratchRegister):
* assembler/MacroAssemblerX86Common.h:
(JSC::MacroAssemblerX86Common::loadFromTLSPtrNeedsMacroScratchRegister):
(JSC::MacroAssemblerX86Common::storeToTLS32):
(JSC::MacroAssemblerX86Common::storeToTLSPtrNeedsMacroScratchRegister):
* assembler/MacroAssemblerX86_64.h:
(JSC::MacroAssemblerX86_64::loadFromTLS64): was loading 32-bit instead of 64-bit
(JSC::MacroAssemblerX86_64::storeToTLS64):
* assembler/X86Assembler.h:
(JSC::X86Assembler::movl_rm):
(JSC::X86Assembler::movq_rm):
* b3/testb3.cpp:
(JSC::B3::testFastTLSLoad):
(JSC::B3::testFastTLSStore):
(JSC::B3::run):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::loadWasmContext):
(JSC::AssemblyHelpers::storeWasmContext):
(JSC::AssemblyHelpers::loadWasmContextNeedsMacroScratchRegister):
(JSC::AssemblyHelpers::storeWasmContextNeedsMacroScratchRegister):
* jit/Repatch.cpp:
(JSC::webAssemblyOwner):
* jit/ThunkGenerators.cpp:
(JSC::throwExceptionFromWasmThunkGenerator):
* runtime/Options.h:
* runtime/VM.cpp:
(JSC::VM::VM):
* runtime/VM.h:
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::loadWasmContext):
(JSC::Wasm::storeWasmContext):
(JSC::Wasm::B3IRGenerator::B3IRGenerator):
(JSC::Wasm::getMemoryBaseAndSize):
(JSC::Wasm::restoreWebAssemblyGlobalState):
(JSC::Wasm::createJSToWasmWrapper):
(JSC::Wasm::parseAndCompile):
* wasm/WasmBinding.cpp:
(JSC::Wasm::materializeImportJSCell):
(JSC::Wasm::wasmToJs):
(JSC::Wasm::wasmToWasm):
* wasm/WasmContext.cpp: Added.
(JSC::loadWasmContext):
(JSC::storeWasmContext):
* wasm/WasmContext.h: Added. Replaces "top" JSWebAssemblyInstance.
* wasm/js/WebAssemblyFunction.cpp:
(JSC::callWebAssemblyFunction):
* wasm/js/WebAssemblyInstanceConstructor.h:
2017-03-24 JF Bastien <jfbastien@apple.com>
WebAssembly: spec-tests/memory.wast.js fails in debug
https://bugs.webkit.org/show_bug.cgi?id=169794
Reviewed by Keith Miller.
The failure was due to empty memories (with maximum size 0). Those
only occur in tests and in code that's trying to trip us. This
patch adds memory mode "none" which represents no memory. It can
work with either bounds checked or signaling code because it never
contains loads and stores.
The spec tests which were failing did the following:
> (module (memory (data)) (func (export "memsize") (result i32) (current_memory)))
> (assert_return (invoke "memsize") (i32.const 0))
> (module (memory (data "")) (func (export "memsize") (result i32) (current_memory)))
> (assert_return (invoke "memsize") (i32.const 0))
> (module (memory (data "x")) (func (export "memsize") (result i32) (current_memory)))
> (assert_return (invoke "memsize") (i32.const 1))
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::memoryKind):
* wasm/WasmMemory.cpp:
(JSC::Wasm::tryGetFastMemory):
(JSC::Wasm::releaseFastMemory):
(JSC::Wasm::Memory::Memory):
(JSC::Wasm::Memory::createImpl):
(JSC::Wasm::Memory::create):
(JSC::Wasm::Memory::grow):
(JSC::Wasm::Memory::makeString):
* wasm/WasmMemory.h:
* wasm/WasmMemoryInformation.cpp:
(JSC::Wasm::MemoryInformation::MemoryInformation):
* wasm/js/JSWebAssemblyCodeBlock.cpp:
(JSC::JSWebAssemblyCodeBlock::isSafeToRun):
* wasm/js/JSWebAssemblyModule.cpp:
(JSC::JSWebAssemblyModule::codeBlock):
(JSC::JSWebAssemblyModule::finishCreation):
* wasm/js/JSWebAssemblyModule.h:
(JSC::JSWebAssemblyModule::codeBlock):
(JSC::JSWebAssemblyModule::codeBlockFor):
2017-03-24 Mark Lam <mark.lam@apple.com>
Array memcpy'ing fast paths should check if we're having a bad time if they cannot handle it.
https://bugs.webkit.org/show_bug.cgi?id=170064
<rdar://problem/31246098>
Reviewed by Geoffrey Garen.
* runtime/ArrayPrototype.cpp:
(JSC::arrayProtoPrivateFuncConcatMemcpy):
* runtime/JSArray.cpp:
(JSC::JSArray::fastSlice):
2017-03-23 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC] Use jsNontrivialString agressively for ToString(Int52)
https://bugs.webkit.org/show_bug.cgi?id=170002
Reviewed by Sam Weinig.
We use the same logic used for Int32 to use jsNontvirialString.
After single character check, produced string is always longer than 1.
Thus, we can use jsNontrivialString.
* runtime/NumberPrototype.cpp:
(JSC::int52ToString):
2017-03-23 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC] Use WeakRandom for SamplingProfiler interval fluctuation
https://bugs.webkit.org/show_bug.cgi?id=170045
Reviewed by Mark Lam.
It is unnecessary to use cryptographicallyRandomNumber for SamplingProfiler
interval fluctuation. Use WeakRandom instead.
* runtime/SamplingProfiler.cpp:
(JSC::SamplingProfiler::SamplingProfiler):
(JSC::SamplingProfiler::timerLoop):
* runtime/SamplingProfiler.h:
2017-03-23 Mark Lam <mark.lam@apple.com>
Array.prototype.splice behaves incorrectly when the VM is "having a bad time".
https://bugs.webkit.org/show_bug.cgi?id=170025
<rdar://problem/31228679>
Reviewed by Saam Barati.
* runtime/ArrayPrototype.cpp:
(JSC::copySplicedArrayElements):
(JSC::arrayProtoFuncSplice):
2017-03-23 Yusuke Suzuki <utatane.tea@gmail.com>
[JSC][DFG] Make addShouldSpeculateAnyInt more conservative to avoid regression caused by Double <-> Int52 conversions
https://bugs.webkit.org/show_bug.cgi?id=169998
Reviewed by Saam Barati.
Double <-> Int52 and JSValue <-> Int52 conversions are not so cheap. Thus, Int52Rep is super carefully emitted.
We make addShouldSpeculateAnyInt more conservative to avoid regressions caused by the above conversions.
We select ArithAdd(Int52, Int52) only when this calculation is beneficial compared to added Int52Rep conversions.
This patch tighten the conditions of addShouldSpeculateAnyInt.
1. Honor DoubleConstant.
When executing imaging-darkroom, we have a thing like that,
132:< 2:loc36> DoubleConstant(Double|UseAsOther, AnyIntAsDouble, Double: 4607182418800017408, 1.000000, bc#114)
1320:< 1:loc38> Int52Rep(Check:Int32:@82, Int52|PureInt, Int32, Exits, bc#114)
1321:< 1:loc39> Int52Constant(Int52|PureInt, Boolint32Nonboolint32Int52, Double: 4607182418800017408, 1.000000, bc#114)
133:<!3:loc39> ArithSub(Int52Rep:@1320<Int52>, Int52Rep:@1321<Int52>, Int52|MustGen, Int52, CheckOverflow, Exits, bc#114)
The LHS of ArithSub says predicting Boolint32, and the rhs says AnyIntAsDouble. Thus we select ArithSub(Int52, Int52) instead
of ArithSub(Double, Double). However, it soon causes OSR exits. In imaging-darkroom, LHS's Int32 prediction will be broken.
While speculating Int32 in the above situation is reasonable approach since the given LHS says predicting Int32, this causes
severe performance regression.
Previously, we always select ArithSub(Double, Double). So accidentally, we do not encounter this misprediction issue.
One thing can be found that we have DoubleConstant in the RHS. It means that we have `1.0` instead of `1` in the code.
We can see the code like `lhs - 1.0` instead of `lhs - 1` in imaging-darkroom. It offers good information that lhs and
the resulting value would be double. Handling the above ArithSub in double seems more appropriate rather than handling
it in Int52.
So, in this patch, we honor DoubleConstant. If we find DoubleConstant on one operand, we give up selecting
Arith[Sub,Add](Int52, Int52). This change removes OSR exits occurr in imaging-darkroom right now.
2. Two Int52Rep(Double) conversions are not desirable.
We allow AnyInt ArithAdd only when the one operand of the binary operation should be speculated AnyInt. It is a bit conservative
decision. This is because Double to Int52 conversion is not so cheap. Frequent back-and-forth conversions between Double and Int52
rather hurt the performance. If the one operand of the operation is already Int52, the cost for constructing ArithAdd becomes
cheap since only one Double to Int52 conversion could be required.
This recovers some regression in assorted tests while keeping kraken crypto improvements.
3. Avoid frequent Int52 to JSValue conversions.
Int52 to JSValue conversion is not so cheap. Thus, we would like to avoid such situations. So, in this patch, we allow
Arith(Int52, Int52) with AnyIntAsDouble operand only when the node is used as number. By doing so, we avoid the case like,
converting Int52, performing ArithAdd, and soon converting back to JSValue.
The above 3 changes recover the regression measured in microbenchmarks/int52-back-and-forth.js and assorted benchmarks.
And still it keeps kraken crypto improvements.
baseline patched
imaging-darkroom 201.112+-3.192 ^ 189.532+-2.883 ^ definitely 1.0611x faster
stanford-crypto-pbkdf2 103.953+-2.325 100.926+-2.396 might be 1.0300x faster
stanford-crypto-sha256-iterative 35.103+-1.071 ? 36.049+-1.143 ? might be 1.0270x slower
* dfg/DFGGraph.h:
(JSC::DFG::Graph::addShouldSpeculateAnyInt):
== Rolled over to ChangeLog-2017-03-23 ==