diff options
author | Allan Sandfeld Jensen <allan.jensen@theqtcompany.com> | 2016-07-14 17:41:05 +0200 |
---|---|---|
committer | Allan Sandfeld Jensen <allan.jensen@qt.io> | 2016-08-04 12:37:36 +0000 |
commit | 399c965b6064c440ddcf4015f5f8e9d131c7a0a6 (patch) | |
tree | 6b06b60ff365abef0e13b3503d593a0df48d20e8 /chromium/v8/src/code-stub-assembler.cc | |
parent | 7366110654eec46f21b6824f302356426f48cd74 (diff) | |
download | qtwebengine-chromium-399c965b6064c440ddcf4015f5f8e9d131c7a0a6.tar.gz |
BASELINE: Update Chromium to 52.0.2743.76 and Ninja to 1.7.1
Change-Id: I382f51b959689505a60f8b707255ecb344f7d8b4
Reviewed-by: Michael BrĂ¼ning <michael.bruning@qt.io>
Diffstat (limited to 'chromium/v8/src/code-stub-assembler.cc')
-rw-r--r-- | chromium/v8/src/code-stub-assembler.cc | 1572 |
1 files changed, 1572 insertions, 0 deletions
diff --git a/chromium/v8/src/code-stub-assembler.cc b/chromium/v8/src/code-stub-assembler.cc new file mode 100644 index 00000000000..3e26b524362 --- /dev/null +++ b/chromium/v8/src/code-stub-assembler.cc @@ -0,0 +1,1572 @@ +// Copyright 2016 the V8 project authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "src/code-stub-assembler.h" +#include "src/code-factory.h" + +namespace v8 { +namespace internal { + +using compiler::Node; + +CodeStubAssembler::CodeStubAssembler(Isolate* isolate, Zone* zone, + const CallInterfaceDescriptor& descriptor, + Code::Flags flags, const char* name, + size_t result_size) + : compiler::CodeAssembler(isolate, zone, descriptor, flags, name, + result_size) {} + +CodeStubAssembler::CodeStubAssembler(Isolate* isolate, Zone* zone, + int parameter_count, Code::Flags flags, + const char* name) + : compiler::CodeAssembler(isolate, zone, parameter_count, flags, name) {} + +Node* CodeStubAssembler::BooleanMapConstant() { + return HeapConstant(isolate()->factory()->boolean_map()); +} + +Node* CodeStubAssembler::EmptyStringConstant() { + return LoadRoot(Heap::kempty_stringRootIndex); +} + +Node* CodeStubAssembler::HeapNumberMapConstant() { + return HeapConstant(isolate()->factory()->heap_number_map()); +} + +Node* CodeStubAssembler::NoContextConstant() { + return SmiConstant(Smi::FromInt(0)); +} + +Node* CodeStubAssembler::NullConstant() { + return LoadRoot(Heap::kNullValueRootIndex); +} + +Node* CodeStubAssembler::UndefinedConstant() { + return LoadRoot(Heap::kUndefinedValueRootIndex); +} + +Node* CodeStubAssembler::StaleRegisterConstant() { + return LoadRoot(Heap::kStaleRegisterRootIndex); +} + +Node* CodeStubAssembler::Float64Round(Node* x) { + Node* one = Float64Constant(1.0); + Node* one_half = Float64Constant(0.5); + + Variable var_x(this, MachineRepresentation::kFloat64); + Label return_x(this); + + // Round up {x} towards Infinity. + var_x.Bind(Float64Ceil(x)); + + GotoIf(Float64LessThanOrEqual(Float64Sub(var_x.value(), one_half), x), + &return_x); + var_x.Bind(Float64Sub(var_x.value(), one)); + Goto(&return_x); + + Bind(&return_x); + return var_x.value(); +} + +Node* CodeStubAssembler::Float64Ceil(Node* x) { + if (IsFloat64RoundUpSupported()) { + return Float64RoundUp(x); + } + + Node* one = Float64Constant(1.0); + Node* zero = Float64Constant(0.0); + Node* two_52 = Float64Constant(4503599627370496.0E0); + Node* minus_two_52 = Float64Constant(-4503599627370496.0E0); + + Variable var_x(this, MachineRepresentation::kFloat64); + Label return_x(this), return_minus_x(this); + var_x.Bind(x); + + // Check if {x} is greater than zero. + Label if_xgreaterthanzero(this), if_xnotgreaterthanzero(this); + Branch(Float64GreaterThan(x, zero), &if_xgreaterthanzero, + &if_xnotgreaterthanzero); + + Bind(&if_xgreaterthanzero); + { + // Just return {x} unless it's in the range ]0,2^52[. + GotoIf(Float64GreaterThanOrEqual(x, two_52), &return_x); + + // Round positive {x} towards Infinity. + var_x.Bind(Float64Sub(Float64Add(two_52, x), two_52)); + GotoUnless(Float64LessThan(var_x.value(), x), &return_x); + var_x.Bind(Float64Add(var_x.value(), one)); + Goto(&return_x); + } + + Bind(&if_xnotgreaterthanzero); + { + // Just return {x} unless it's in the range ]-2^52,0[ + GotoIf(Float64LessThanOrEqual(x, minus_two_52), &return_x); + GotoUnless(Float64LessThan(x, zero), &return_x); + + // Round negated {x} towards Infinity and return the result negated. + Node* minus_x = Float64Neg(x); + var_x.Bind(Float64Sub(Float64Add(two_52, minus_x), two_52)); + GotoUnless(Float64GreaterThan(var_x.value(), minus_x), &return_minus_x); + var_x.Bind(Float64Sub(var_x.value(), one)); + Goto(&return_minus_x); + } + + Bind(&return_minus_x); + var_x.Bind(Float64Neg(var_x.value())); + Goto(&return_x); + + Bind(&return_x); + return var_x.value(); +} + +Node* CodeStubAssembler::Float64Floor(Node* x) { + if (IsFloat64RoundDownSupported()) { + return Float64RoundDown(x); + } + + Node* one = Float64Constant(1.0); + Node* zero = Float64Constant(0.0); + Node* two_52 = Float64Constant(4503599627370496.0E0); + Node* minus_two_52 = Float64Constant(-4503599627370496.0E0); + + Variable var_x(this, MachineRepresentation::kFloat64); + Label return_x(this), return_minus_x(this); + var_x.Bind(x); + + // Check if {x} is greater than zero. + Label if_xgreaterthanzero(this), if_xnotgreaterthanzero(this); + Branch(Float64GreaterThan(x, zero), &if_xgreaterthanzero, + &if_xnotgreaterthanzero); + + Bind(&if_xgreaterthanzero); + { + // Just return {x} unless it's in the range ]0,2^52[. + GotoIf(Float64GreaterThanOrEqual(x, two_52), &return_x); + + // Round positive {x} towards -Infinity. + var_x.Bind(Float64Sub(Float64Add(two_52, x), two_52)); + GotoUnless(Float64GreaterThan(var_x.value(), x), &return_x); + var_x.Bind(Float64Sub(var_x.value(), one)); + Goto(&return_x); + } + + Bind(&if_xnotgreaterthanzero); + { + // Just return {x} unless it's in the range ]-2^52,0[ + GotoIf(Float64LessThanOrEqual(x, minus_two_52), &return_x); + GotoUnless(Float64LessThan(x, zero), &return_x); + + // Round negated {x} towards -Infinity and return the result negated. + Node* minus_x = Float64Neg(x); + var_x.Bind(Float64Sub(Float64Add(two_52, minus_x), two_52)); + GotoUnless(Float64LessThan(var_x.value(), minus_x), &return_minus_x); + var_x.Bind(Float64Add(var_x.value(), one)); + Goto(&return_minus_x); + } + + Bind(&return_minus_x); + var_x.Bind(Float64Neg(var_x.value())); + Goto(&return_x); + + Bind(&return_x); + return var_x.value(); +} + +Node* CodeStubAssembler::Float64Trunc(Node* x) { + if (IsFloat64RoundTruncateSupported()) { + return Float64RoundTruncate(x); + } + + Node* one = Float64Constant(1.0); + Node* zero = Float64Constant(0.0); + Node* two_52 = Float64Constant(4503599627370496.0E0); + Node* minus_two_52 = Float64Constant(-4503599627370496.0E0); + + Variable var_x(this, MachineRepresentation::kFloat64); + Label return_x(this), return_minus_x(this); + var_x.Bind(x); + + // Check if {x} is greater than 0. + Label if_xgreaterthanzero(this), if_xnotgreaterthanzero(this); + Branch(Float64GreaterThan(x, zero), &if_xgreaterthanzero, + &if_xnotgreaterthanzero); + + Bind(&if_xgreaterthanzero); + { + if (IsFloat64RoundDownSupported()) { + var_x.Bind(Float64RoundDown(x)); + } else { + // Just return {x} unless it's in the range ]0,2^52[. + GotoIf(Float64GreaterThanOrEqual(x, two_52), &return_x); + + // Round positive {x} towards -Infinity. + var_x.Bind(Float64Sub(Float64Add(two_52, x), two_52)); + GotoUnless(Float64GreaterThan(var_x.value(), x), &return_x); + var_x.Bind(Float64Sub(var_x.value(), one)); + } + Goto(&return_x); + } + + Bind(&if_xnotgreaterthanzero); + { + if (IsFloat64RoundUpSupported()) { + var_x.Bind(Float64RoundUp(x)); + Goto(&return_x); + } else { + // Just return {x} unless its in the range ]-2^52,0[. + GotoIf(Float64LessThanOrEqual(x, minus_two_52), &return_x); + GotoUnless(Float64LessThan(x, zero), &return_x); + + // Round negated {x} towards -Infinity and return result negated. + Node* minus_x = Float64Neg(x); + var_x.Bind(Float64Sub(Float64Add(two_52, minus_x), two_52)); + GotoUnless(Float64GreaterThan(var_x.value(), minus_x), &return_minus_x); + var_x.Bind(Float64Sub(var_x.value(), one)); + Goto(&return_minus_x); + } + } + + Bind(&return_minus_x); + var_x.Bind(Float64Neg(var_x.value())); + Goto(&return_x); + + Bind(&return_x); + return var_x.value(); +} + +Node* CodeStubAssembler::SmiFromWord32(Node* value) { + value = ChangeInt32ToIntPtr(value); + return WordShl(value, SmiShiftBitsConstant()); +} + +Node* CodeStubAssembler::SmiTag(Node* value) { + int32_t constant_value; + if (ToInt32Constant(value, constant_value) && Smi::IsValid(constant_value)) { + return SmiConstant(Smi::FromInt(constant_value)); + } + return WordShl(value, SmiShiftBitsConstant()); +} + +Node* CodeStubAssembler::SmiUntag(Node* value) { + return WordSar(value, SmiShiftBitsConstant()); +} + +Node* CodeStubAssembler::SmiToWord32(Node* value) { + Node* result = WordSar(value, SmiShiftBitsConstant()); + if (Is64()) { + result = TruncateInt64ToInt32(result); + } + return result; +} + +Node* CodeStubAssembler::SmiToFloat64(Node* value) { + return ChangeInt32ToFloat64(SmiToWord32(value)); +} + +Node* CodeStubAssembler::SmiAdd(Node* a, Node* b) { return IntPtrAdd(a, b); } + +Node* CodeStubAssembler::SmiAddWithOverflow(Node* a, Node* b) { + return IntPtrAddWithOverflow(a, b); +} + +Node* CodeStubAssembler::SmiSub(Node* a, Node* b) { return IntPtrSub(a, b); } + +Node* CodeStubAssembler::SmiSubWithOverflow(Node* a, Node* b) { + return IntPtrSubWithOverflow(a, b); +} + +Node* CodeStubAssembler::SmiEqual(Node* a, Node* b) { return WordEqual(a, b); } + +Node* CodeStubAssembler::SmiAboveOrEqual(Node* a, Node* b) { + return UintPtrGreaterThanOrEqual(a, b); +} + +Node* CodeStubAssembler::SmiLessThan(Node* a, Node* b) { + return IntPtrLessThan(a, b); +} + +Node* CodeStubAssembler::SmiLessThanOrEqual(Node* a, Node* b) { + return IntPtrLessThanOrEqual(a, b); +} + +Node* CodeStubAssembler::SmiMin(Node* a, Node* b) { + // TODO(bmeurer): Consider using Select once available. + Variable min(this, MachineRepresentation::kTagged); + Label if_a(this), if_b(this), join(this); + BranchIfSmiLessThan(a, b, &if_a, &if_b); + Bind(&if_a); + min.Bind(a); + Goto(&join); + Bind(&if_b); + min.Bind(b); + Goto(&join); + Bind(&join); + return min.value(); +} + +Node* CodeStubAssembler::WordIsSmi(Node* a) { + return WordEqual(WordAnd(a, IntPtrConstant(kSmiTagMask)), IntPtrConstant(0)); +} + +Node* CodeStubAssembler::WordIsPositiveSmi(Node* a) { + return WordEqual(WordAnd(a, IntPtrConstant(kSmiTagMask | kSmiSignMask)), + IntPtrConstant(0)); +} + +Node* CodeStubAssembler::AllocateRawUnaligned(Node* size_in_bytes, + AllocationFlags flags, + Node* top_address, + Node* limit_address) { + Node* top = Load(MachineType::Pointer(), top_address); + Node* limit = Load(MachineType::Pointer(), limit_address); + + // If there's not enough space, call the runtime. + Variable result(this, MachineRepresentation::kTagged); + Label runtime_call(this, Label::kDeferred), no_runtime_call(this); + Label merge_runtime(this, &result); + + Node* new_top = IntPtrAdd(top, size_in_bytes); + Branch(UintPtrGreaterThanOrEqual(new_top, limit), &runtime_call, + &no_runtime_call); + + Bind(&runtime_call); + // AllocateInTargetSpace does not use the context. + Node* context = SmiConstant(Smi::FromInt(0)); + + Node* runtime_result; + if (flags & kPretenured) { + Node* runtime_flags = SmiConstant( + Smi::FromInt(AllocateDoubleAlignFlag::encode(false) | + AllocateTargetSpace::encode(AllocationSpace::OLD_SPACE))); + runtime_result = CallRuntime(Runtime::kAllocateInTargetSpace, context, + SmiTag(size_in_bytes), runtime_flags); + } else { + runtime_result = CallRuntime(Runtime::kAllocateInNewSpace, context, + SmiTag(size_in_bytes)); + } + result.Bind(runtime_result); + Goto(&merge_runtime); + + // When there is enough space, return `top' and bump it up. + Bind(&no_runtime_call); + Node* no_runtime_result = top; + StoreNoWriteBarrier(MachineType::PointerRepresentation(), top_address, + new_top); + no_runtime_result = BitcastWordToTagged( + IntPtrAdd(no_runtime_result, IntPtrConstant(kHeapObjectTag))); + result.Bind(no_runtime_result); + Goto(&merge_runtime); + + Bind(&merge_runtime); + return result.value(); +} + +Node* CodeStubAssembler::AllocateRawAligned(Node* size_in_bytes, + AllocationFlags flags, + Node* top_address, + Node* limit_address) { + Node* top = Load(MachineType::Pointer(), top_address); + Node* limit = Load(MachineType::Pointer(), limit_address); + Variable adjusted_size(this, MachineType::PointerRepresentation()); + adjusted_size.Bind(size_in_bytes); + if (flags & kDoubleAlignment) { + // TODO(epertoso): Simd128 alignment. + Label aligned(this), not_aligned(this), merge(this, &adjusted_size); + Branch(WordAnd(top, IntPtrConstant(kDoubleAlignmentMask)), ¬_aligned, + &aligned); + + Bind(¬_aligned); + Node* not_aligned_size = + IntPtrAdd(size_in_bytes, IntPtrConstant(kPointerSize)); + adjusted_size.Bind(not_aligned_size); + Goto(&merge); + + Bind(&aligned); + Goto(&merge); + + Bind(&merge); + } + + Variable address(this, MachineRepresentation::kTagged); + address.Bind(AllocateRawUnaligned(adjusted_size.value(), kNone, top, limit)); + + Label needs_filler(this), doesnt_need_filler(this), + merge_address(this, &address); + Branch(IntPtrEqual(adjusted_size.value(), size_in_bytes), &doesnt_need_filler, + &needs_filler); + + Bind(&needs_filler); + // Store a filler and increase the address by kPointerSize. + // TODO(epertoso): this code assumes that we only align to kDoubleSize. Change + // it when Simd128 alignment is supported. + StoreNoWriteBarrier(MachineType::PointerRepresentation(), top, + LoadRoot(Heap::kOnePointerFillerMapRootIndex)); + address.Bind(BitcastWordToTagged( + IntPtrAdd(address.value(), IntPtrConstant(kPointerSize)))); + Goto(&merge_address); + + Bind(&doesnt_need_filler); + Goto(&merge_address); + + Bind(&merge_address); + // Update the top. + StoreNoWriteBarrier(MachineType::PointerRepresentation(), top_address, + IntPtrAdd(top, adjusted_size.value())); + return address.value(); +} + +Node* CodeStubAssembler::Allocate(Node* size_in_bytes, AllocationFlags flags) { + bool const new_space = !(flags & kPretenured); + Node* top_address = ExternalConstant( + new_space + ? ExternalReference::new_space_allocation_top_address(isolate()) + : ExternalReference::old_space_allocation_top_address(isolate())); + Node* limit_address = ExternalConstant( + new_space + ? ExternalReference::new_space_allocation_limit_address(isolate()) + : ExternalReference::old_space_allocation_limit_address(isolate())); + +#ifdef V8_HOST_ARCH_32_BIT + if (flags & kDoubleAlignment) { + return AllocateRawAligned(size_in_bytes, flags, top_address, limit_address); + } +#endif + + return AllocateRawUnaligned(size_in_bytes, flags, top_address, limit_address); +} + +Node* CodeStubAssembler::Allocate(int size_in_bytes, AllocationFlags flags) { + return CodeStubAssembler::Allocate(IntPtrConstant(size_in_bytes), flags); +} + +Node* CodeStubAssembler::InnerAllocate(Node* previous, Node* offset) { + return BitcastWordToTagged(IntPtrAdd(previous, offset)); +} + +Node* CodeStubAssembler::InnerAllocate(Node* previous, int offset) { + return InnerAllocate(previous, IntPtrConstant(offset)); +} + +Node* CodeStubAssembler::LoadBufferObject(Node* buffer, int offset, + MachineType rep) { + return Load(rep, buffer, IntPtrConstant(offset)); +} + +Node* CodeStubAssembler::LoadObjectField(Node* object, int offset, + MachineType rep) { + return Load(rep, object, IntPtrConstant(offset - kHeapObjectTag)); +} + +Node* CodeStubAssembler::LoadHeapNumberValue(Node* object) { + return Load(MachineType::Float64(), object, + IntPtrConstant(HeapNumber::kValueOffset - kHeapObjectTag)); +} + +Node* CodeStubAssembler::LoadMap(Node* object) { + return LoadObjectField(object, HeapObject::kMapOffset); +} + +Node* CodeStubAssembler::LoadInstanceType(Node* object) { + return LoadMapInstanceType(LoadMap(object)); +} + +Node* CodeStubAssembler::LoadElements(Node* object) { + return LoadObjectField(object, JSObject::kElementsOffset); +} + +Node* CodeStubAssembler::LoadFixedArrayBaseLength(Node* array) { + return LoadObjectField(array, FixedArrayBase::kLengthOffset); +} + +Node* CodeStubAssembler::LoadMapBitField(Node* map) { + return Load(MachineType::Uint8(), map, + IntPtrConstant(Map::kBitFieldOffset - kHeapObjectTag)); +} + +Node* CodeStubAssembler::LoadMapBitField2(Node* map) { + return Load(MachineType::Uint8(), map, + IntPtrConstant(Map::kBitField2Offset - kHeapObjectTag)); +} + +Node* CodeStubAssembler::LoadMapBitField3(Node* map) { + return Load(MachineType::Uint32(), map, + IntPtrConstant(Map::kBitField3Offset - kHeapObjectTag)); +} + +Node* CodeStubAssembler::LoadMapInstanceType(Node* map) { + return Load(MachineType::Uint8(), map, + IntPtrConstant(Map::kInstanceTypeOffset - kHeapObjectTag)); +} + +Node* CodeStubAssembler::LoadMapDescriptors(Node* map) { + return LoadObjectField(map, Map::kDescriptorsOffset); +} + +Node* CodeStubAssembler::LoadMapPrototype(Node* map) { + return LoadObjectField(map, Map::kPrototypeOffset); +} + +Node* CodeStubAssembler::LoadNameHash(Node* name) { + return Load(MachineType::Uint32(), name, + IntPtrConstant(Name::kHashFieldOffset - kHeapObjectTag)); +} + +Node* CodeStubAssembler::AllocateUninitializedFixedArray(Node* length) { + Node* header_size = IntPtrConstant(FixedArray::kHeaderSize); + Node* data_size = WordShl(length, IntPtrConstant(kPointerSizeLog2)); + Node* total_size = IntPtrAdd(data_size, header_size); + + Node* result = Allocate(total_size, kNone); + StoreMapNoWriteBarrier(result, LoadRoot(Heap::kFixedArrayMapRootIndex)); + StoreObjectFieldNoWriteBarrier(result, FixedArray::kLengthOffset, + SmiTag(length)); + + return result; +} + +Node* CodeStubAssembler::LoadFixedArrayElement(Node* object, Node* index_node, + int additional_offset, + ParameterMode parameter_mode) { + int32_t header_size = + FixedArray::kHeaderSize + additional_offset - kHeapObjectTag; + Node* offset = ElementOffsetFromIndex(index_node, FAST_HOLEY_ELEMENTS, + parameter_mode, header_size); + return Load(MachineType::AnyTagged(), object, offset); +} + +Node* CodeStubAssembler::LoadMapInstanceSize(Node* map) { + return Load(MachineType::Uint8(), map, + IntPtrConstant(Map::kInstanceSizeOffset - kHeapObjectTag)); +} + +Node* CodeStubAssembler::LoadNativeContext(Node* context) { + return LoadFixedArrayElement(context, + Int32Constant(Context::NATIVE_CONTEXT_INDEX)); +} + +Node* CodeStubAssembler::LoadJSArrayElementsMap(ElementsKind kind, + Node* native_context) { + return LoadFixedArrayElement(native_context, + Int32Constant(Context::ArrayMapIndex(kind))); +} + +Node* CodeStubAssembler::StoreHeapNumberValue(Node* object, Node* value) { + return StoreNoWriteBarrier( + MachineRepresentation::kFloat64, object, + IntPtrConstant(HeapNumber::kValueOffset - kHeapObjectTag), value); +} + +Node* CodeStubAssembler::StoreObjectField( + Node* object, int offset, Node* value) { + return Store(MachineRepresentation::kTagged, object, + IntPtrConstant(offset - kHeapObjectTag), value); +} + +Node* CodeStubAssembler::StoreObjectFieldNoWriteBarrier( + Node* object, int offset, Node* value, MachineRepresentation rep) { + return StoreNoWriteBarrier(rep, object, + IntPtrConstant(offset - kHeapObjectTag), value); +} + +Node* CodeStubAssembler::StoreMapNoWriteBarrier(Node* object, Node* map) { + return StoreNoWriteBarrier( + MachineRepresentation::kTagged, object, + IntPtrConstant(HeapNumber::kMapOffset - kHeapObjectTag), map); +} + +Node* CodeStubAssembler::StoreFixedArrayElement(Node* object, Node* index_node, + Node* value, + WriteBarrierMode barrier_mode, + ParameterMode parameter_mode) { + DCHECK(barrier_mode == SKIP_WRITE_BARRIER || + barrier_mode == UPDATE_WRITE_BARRIER); + Node* offset = + ElementOffsetFromIndex(index_node, FAST_HOLEY_ELEMENTS, parameter_mode, + FixedArray::kHeaderSize - kHeapObjectTag); + MachineRepresentation rep = MachineRepresentation::kTagged; + if (barrier_mode == SKIP_WRITE_BARRIER) { + return StoreNoWriteBarrier(rep, object, offset, value); + } else { + return Store(rep, object, offset, value); + } +} + +Node* CodeStubAssembler::StoreFixedDoubleArrayElement( + Node* object, Node* index_node, Node* value, ParameterMode parameter_mode) { + Node* offset = + ElementOffsetFromIndex(index_node, FAST_DOUBLE_ELEMENTS, parameter_mode, + FixedArray::kHeaderSize - kHeapObjectTag); + MachineRepresentation rep = MachineRepresentation::kFloat64; + return StoreNoWriteBarrier(rep, object, offset, value); +} + +Node* CodeStubAssembler::AllocateHeapNumber() { + Node* result = Allocate(HeapNumber::kSize, kNone); + StoreMapNoWriteBarrier(result, HeapNumberMapConstant()); + return result; +} + +Node* CodeStubAssembler::AllocateHeapNumberWithValue(Node* value) { + Node* result = AllocateHeapNumber(); + StoreHeapNumberValue(result, value); + return result; +} + +Node* CodeStubAssembler::AllocateSeqOneByteString(int length) { + Node* result = Allocate(SeqOneByteString::SizeFor(length)); + StoreMapNoWriteBarrier(result, LoadRoot(Heap::kOneByteStringMapRootIndex)); + StoreObjectFieldNoWriteBarrier(result, SeqOneByteString::kLengthOffset, + SmiConstant(Smi::FromInt(length))); + StoreObjectFieldNoWriteBarrier(result, SeqOneByteString::kHashFieldSlot, + IntPtrConstant(String::kEmptyHashField)); + return result; +} + +Node* CodeStubAssembler::AllocateSeqTwoByteString(int length) { + Node* result = Allocate(SeqTwoByteString::SizeFor(length)); + StoreMapNoWriteBarrier(result, LoadRoot(Heap::kStringMapRootIndex)); + StoreObjectFieldNoWriteBarrier(result, SeqTwoByteString::kLengthOffset, + SmiConstant(Smi::FromInt(length))); + StoreObjectFieldNoWriteBarrier(result, SeqTwoByteString::kHashFieldSlot, + IntPtrConstant(String::kEmptyHashField)); + return result; +} + +Node* CodeStubAssembler::AllocateJSArray(ElementsKind kind, Node* array_map, + Node* capacity_node, Node* length_node, + compiler::Node* allocation_site, + ParameterMode mode) { + bool is_double = IsFastDoubleElementsKind(kind); + int base_size = JSArray::kSize + FixedArray::kHeaderSize; + int elements_offset = JSArray::kSize; + + if (allocation_site != nullptr) { + base_size += AllocationMemento::kSize; + elements_offset += AllocationMemento::kSize; + } + + int32_t capacity; + bool constant_capacity = ToInt32Constant(capacity_node, capacity); + Node* total_size = + ElementOffsetFromIndex(capacity_node, kind, mode, base_size); + + // Allocate both array and elements object, and initialize the JSArray. + Heap* heap = isolate()->heap(); + Node* array = Allocate(total_size); + StoreMapNoWriteBarrier(array, array_map); + Node* empty_properties = + HeapConstant(Handle<HeapObject>(heap->empty_fixed_array())); + StoreObjectFieldNoWriteBarrier(array, JSArray::kPropertiesOffset, + empty_properties); + StoreObjectFieldNoWriteBarrier( + array, JSArray::kLengthOffset, + mode == SMI_PARAMETERS ? length_node : SmiTag(length_node)); + + if (allocation_site != nullptr) { + InitializeAllocationMemento(array, JSArray::kSize, allocation_site); + } + + // Setup elements object. + Node* elements = InnerAllocate(array, elements_offset); + StoreObjectFieldNoWriteBarrier(array, JSArray::kElementsOffset, elements); + Handle<Map> elements_map(is_double ? heap->fixed_double_array_map() + : heap->fixed_array_map()); + StoreMapNoWriteBarrier(elements, HeapConstant(elements_map)); + StoreObjectFieldNoWriteBarrier( + elements, FixedArray::kLengthOffset, + mode == SMI_PARAMETERS ? capacity_node : SmiTag(capacity_node)); + + int const first_element_offset = FixedArray::kHeaderSize - kHeapObjectTag; + Node* hole = HeapConstant(Handle<HeapObject>(heap->the_hole_value())); + Node* double_hole = + Is64() ? Int64Constant(kHoleNanInt64) : Int32Constant(kHoleNanLower32); + DCHECK_EQ(kHoleNanLower32, kHoleNanUpper32); + if (constant_capacity && capacity <= kElementLoopUnrollThreshold) { + for (int i = 0; i < capacity; ++i) { + if (is_double) { + Node* offset = ElementOffsetFromIndex(Int32Constant(i), kind, mode, + first_element_offset); + // Don't use doubles to store the hole double, since manipulating the + // signaling NaN used for the hole in C++, e.g. with bit_cast, will + // change its value on ia32 (the x87 stack is used to return values + // and stores to the stack silently clear the signalling bit). + // + // TODO(danno): When we have a Float32/Float64 wrapper class that + // preserves double bits during manipulation, remove this code/change + // this to an indexed Float64 store. + if (Is64()) { + StoreNoWriteBarrier(MachineRepresentation::kWord64, elements, offset, + double_hole); + } else { + StoreNoWriteBarrier(MachineRepresentation::kWord32, elements, offset, + double_hole); + offset = ElementOffsetFromIndex(Int32Constant(i), kind, mode, + first_element_offset + kPointerSize); + StoreNoWriteBarrier(MachineRepresentation::kWord32, elements, offset, + double_hole); + } + } else { + StoreFixedArrayElement(elements, Int32Constant(i), hole, + SKIP_WRITE_BARRIER); + } + } + } else { + // TODO(danno): Add a loop for initialization + UNIMPLEMENTED(); + } + + return array; +} + +void CodeStubAssembler::InitializeAllocationMemento( + compiler::Node* base_allocation, int base_allocation_size, + compiler::Node* allocation_site) { + StoreObjectFieldNoWriteBarrier( + base_allocation, AllocationMemento::kMapOffset + base_allocation_size, + HeapConstant(Handle<Map>(isolate()->heap()->allocation_memento_map()))); + StoreObjectFieldNoWriteBarrier( + base_allocation, + AllocationMemento::kAllocationSiteOffset + base_allocation_size, + allocation_site); + if (FLAG_allocation_site_pretenuring) { + Node* count = LoadObjectField(allocation_site, + AllocationSite::kPretenureCreateCountOffset); + Node* incremented_count = IntPtrAdd(count, SmiConstant(Smi::FromInt(1))); + StoreObjectFieldNoWriteBarrier(allocation_site, + AllocationSite::kPretenureCreateCountOffset, + incremented_count); + } +} + +Node* CodeStubAssembler::TruncateTaggedToFloat64(Node* context, Node* value) { + // We might need to loop once due to ToNumber conversion. + Variable var_value(this, MachineRepresentation::kTagged), + var_result(this, MachineRepresentation::kFloat64); + Label loop(this, &var_value), done_loop(this, &var_result); + var_value.Bind(value); + Goto(&loop); + Bind(&loop); + { + // Load the current {value}. + value = var_value.value(); + + // Check if the {value} is a Smi or a HeapObject. + Label if_valueissmi(this), if_valueisnotsmi(this); + Branch(WordIsSmi(value), &if_valueissmi, &if_valueisnotsmi); + + Bind(&if_valueissmi); + { + // Convert the Smi {value}. + var_result.Bind(SmiToFloat64(value)); + Goto(&done_loop); + } + + Bind(&if_valueisnotsmi); + { + // Check if {value} is a HeapNumber. + Label if_valueisheapnumber(this), + if_valueisnotheapnumber(this, Label::kDeferred); + Branch(WordEqual(LoadMap(value), HeapNumberMapConstant()), + &if_valueisheapnumber, &if_valueisnotheapnumber); + + Bind(&if_valueisheapnumber); + { + // Load the floating point value. + var_result.Bind(LoadHeapNumberValue(value)); + Goto(&done_loop); + } + + Bind(&if_valueisnotheapnumber); + { + // Convert the {value} to a Number first. + Callable callable = CodeFactory::NonNumberToNumber(isolate()); + var_value.Bind(CallStub(callable, context, value)); + Goto(&loop); + } + } + } + Bind(&done_loop); + return var_result.value(); +} + +Node* CodeStubAssembler::TruncateTaggedToWord32(Node* context, Node* value) { + // We might need to loop once due to ToNumber conversion. + Variable var_value(this, MachineRepresentation::kTagged), + var_result(this, MachineRepresentation::kWord32); + Label loop(this, &var_value), done_loop(this, &var_result); + var_value.Bind(value); + Goto(&loop); + Bind(&loop); + { + // Load the current {value}. + value = var_value.value(); + + // Check if the {value} is a Smi or a HeapObject. + Label if_valueissmi(this), if_valueisnotsmi(this); + Branch(WordIsSmi(value), &if_valueissmi, &if_valueisnotsmi); + + Bind(&if_valueissmi); + { + // Convert the Smi {value}. + var_result.Bind(SmiToWord32(value)); + Goto(&done_loop); + } + + Bind(&if_valueisnotsmi); + { + // Check if {value} is a HeapNumber. + Label if_valueisheapnumber(this), + if_valueisnotheapnumber(this, Label::kDeferred); + Branch(WordEqual(LoadMap(value), HeapNumberMapConstant()), + &if_valueisheapnumber, &if_valueisnotheapnumber); + + Bind(&if_valueisheapnumber); + { + // Truncate the floating point value. + var_result.Bind(TruncateHeapNumberValueToWord32(value)); + Goto(&done_loop); + } + + Bind(&if_valueisnotheapnumber); + { + // Convert the {value} to a Number first. + Callable callable = CodeFactory::NonNumberToNumber(isolate()); + var_value.Bind(CallStub(callable, context, value)); + Goto(&loop); + } + } + } + Bind(&done_loop); + return var_result.value(); +} + +Node* CodeStubAssembler::TruncateHeapNumberValueToWord32(Node* object) { + Node* value = LoadHeapNumberValue(object); + return TruncateFloat64ToWord32(value); +} + +Node* CodeStubAssembler::ChangeFloat64ToTagged(Node* value) { + Node* value32 = RoundFloat64ToInt32(value); + Node* value64 = ChangeInt32ToFloat64(value32); + + Label if_valueisint32(this), if_valueisheapnumber(this), if_join(this); + + Label if_valueisequal(this), if_valueisnotequal(this); + Branch(Float64Equal(value, value64), &if_valueisequal, &if_valueisnotequal); + Bind(&if_valueisequal); + { + GotoUnless(Word32Equal(value32, Int32Constant(0)), &if_valueisint32); + BranchIfInt32LessThan(Float64ExtractHighWord32(value), Int32Constant(0), + &if_valueisheapnumber, &if_valueisint32); + } + Bind(&if_valueisnotequal); + Goto(&if_valueisheapnumber); + + Variable var_result(this, MachineRepresentation::kTagged); + Bind(&if_valueisint32); + { + if (Is64()) { + Node* result = SmiTag(ChangeInt32ToInt64(value32)); + var_result.Bind(result); + Goto(&if_join); + } else { + Node* pair = Int32AddWithOverflow(value32, value32); + Node* overflow = Projection(1, pair); + Label if_overflow(this, Label::kDeferred), if_notoverflow(this); + Branch(overflow, &if_overflow, &if_notoverflow); + Bind(&if_overflow); + Goto(&if_valueisheapnumber); + Bind(&if_notoverflow); + { + Node* result = Projection(0, pair); + var_result.Bind(result); + Goto(&if_join); + } + } + } + Bind(&if_valueisheapnumber); + { + Node* result = AllocateHeapNumberWithValue(value); + var_result.Bind(result); + Goto(&if_join); + } + Bind(&if_join); + return var_result.value(); +} + +Node* CodeStubAssembler::ChangeInt32ToTagged(Node* value) { + if (Is64()) { + return SmiTag(ChangeInt32ToInt64(value)); + } + Variable var_result(this, MachineRepresentation::kTagged); + Node* pair = Int32AddWithOverflow(value, value); + Node* overflow = Projection(1, pair); + Label if_overflow(this, Label::kDeferred), if_notoverflow(this), + if_join(this); + Branch(overflow, &if_overflow, &if_notoverflow); + Bind(&if_overflow); + { + Node* value64 = ChangeInt32ToFloat64(value); + Node* result = AllocateHeapNumberWithValue(value64); + var_result.Bind(result); + } + Goto(&if_join); + Bind(&if_notoverflow); + { + Node* result = Projection(0, pair); + var_result.Bind(result); + } + Goto(&if_join); + Bind(&if_join); + return var_result.value(); +} + +Node* CodeStubAssembler::ChangeUint32ToTagged(Node* value) { + Label if_overflow(this, Label::kDeferred), if_not_overflow(this), + if_join(this); + Variable var_result(this, MachineRepresentation::kTagged); + // If {value} > 2^31 - 1, we need to store it in a HeapNumber. + Branch(Int32LessThan(value, Int32Constant(0)), &if_overflow, + &if_not_overflow); + Bind(&if_not_overflow); + { + if (Is64()) { + var_result.Bind(SmiTag(ChangeUint32ToUint64(value))); + } else { + // If tagging {value} results in an overflow, we need to use a HeapNumber + // to represent it. + Node* pair = Int32AddWithOverflow(value, value); + Node* overflow = Projection(1, pair); + GotoIf(overflow, &if_overflow); + + Node* result = Projection(0, pair); + var_result.Bind(result); + } + } + Goto(&if_join); + + Bind(&if_overflow); + { + Node* float64_value = ChangeUint32ToFloat64(value); + var_result.Bind(AllocateHeapNumberWithValue(float64_value)); + } + Goto(&if_join); + + Bind(&if_join); + return var_result.value(); +} + +Node* CodeStubAssembler::ToThisString(Node* context, Node* value, + char const* method_name) { + Variable var_value(this, MachineRepresentation::kTagged); + var_value.Bind(value); + + // Check if the {value} is a Smi or a HeapObject. + Label if_valueissmi(this, Label::kDeferred), if_valueisnotsmi(this), + if_valueisstring(this); + Branch(WordIsSmi(value), &if_valueissmi, &if_valueisnotsmi); + Bind(&if_valueisnotsmi); + { + // Load the instance type of the {value}. + Node* value_instance_type = LoadInstanceType(value); + + // Check if the {value} is already String. + Label if_valueisnotstring(this, Label::kDeferred); + Branch( + Int32LessThan(value_instance_type, Int32Constant(FIRST_NONSTRING_TYPE)), + &if_valueisstring, &if_valueisnotstring); + Bind(&if_valueisnotstring); + { + // Check if the {value} is null. + Label if_valueisnullorundefined(this, Label::kDeferred), + if_valueisnotnullorundefined(this, Label::kDeferred), + if_valueisnotnull(this, Label::kDeferred); + Branch(WordEqual(value, NullConstant()), &if_valueisnullorundefined, + &if_valueisnotnull); + Bind(&if_valueisnotnull); + { + // Check if the {value} is undefined. + Branch(WordEqual(value, UndefinedConstant()), + &if_valueisnullorundefined, &if_valueisnotnullorundefined); + Bind(&if_valueisnotnullorundefined); + { + // Convert the {value} to a String. + Callable callable = CodeFactory::ToString(isolate()); + var_value.Bind(CallStub(callable, context, value)); + Goto(&if_valueisstring); + } + } + + Bind(&if_valueisnullorundefined); + { + // The {value} is either null or undefined. + CallRuntime(Runtime::kThrowCalledOnNullOrUndefined, context, + HeapConstant(factory()->NewStringFromAsciiChecked( + method_name, TENURED))); + Goto(&if_valueisstring); // Never reached. + } + } + } + Bind(&if_valueissmi); + { + // The {value} is a Smi, convert it to a String. + Callable callable = CodeFactory::NumberToString(isolate()); + var_value.Bind(CallStub(callable, context, value)); + Goto(&if_valueisstring); + } + Bind(&if_valueisstring); + return var_value.value(); +} + +Node* CodeStubAssembler::StringCharCodeAt(Node* string, Node* index) { + // Translate the {index} into a Word. + index = SmiToWord(index); + + // We may need to loop in case of cons or sliced strings. + Variable var_index(this, MachineType::PointerRepresentation()); + Variable var_result(this, MachineRepresentation::kWord32); + Variable var_string(this, MachineRepresentation::kTagged); + Variable* loop_vars[] = {&var_index, &var_string}; + Label done_loop(this, &var_result), loop(this, 2, loop_vars); + var_string.Bind(string); + var_index.Bind(index); + Goto(&loop); + Bind(&loop); + { + // Load the current {index}. + index = var_index.value(); + + // Load the current {string}. + string = var_string.value(); + + // Load the instance type of the {string}. + Node* string_instance_type = LoadInstanceType(string); + + // Check if the {string} is a SeqString. + Label if_stringissequential(this), if_stringisnotsequential(this); + Branch(Word32Equal(Word32And(string_instance_type, + Int32Constant(kStringRepresentationMask)), + Int32Constant(kSeqStringTag)), + &if_stringissequential, &if_stringisnotsequential); + + Bind(&if_stringissequential); + { + // Check if the {string} is a TwoByteSeqString or a OneByteSeqString. + Label if_stringistwobyte(this), if_stringisonebyte(this); + Branch(Word32Equal(Word32And(string_instance_type, + Int32Constant(kStringEncodingMask)), + Int32Constant(kTwoByteStringTag)), + &if_stringistwobyte, &if_stringisonebyte); + + Bind(&if_stringisonebyte); + { + var_result.Bind( + Load(MachineType::Uint8(), string, + IntPtrAdd(index, IntPtrConstant(SeqOneByteString::kHeaderSize - + kHeapObjectTag)))); + Goto(&done_loop); + } + + Bind(&if_stringistwobyte); + { + var_result.Bind( + Load(MachineType::Uint16(), string, + IntPtrAdd(WordShl(index, IntPtrConstant(1)), + IntPtrConstant(SeqTwoByteString::kHeaderSize - + kHeapObjectTag)))); + Goto(&done_loop); + } + } + + Bind(&if_stringisnotsequential); + { + // Check if the {string} is a ConsString. + Label if_stringiscons(this), if_stringisnotcons(this); + Branch(Word32Equal(Word32And(string_instance_type, + Int32Constant(kStringRepresentationMask)), + Int32Constant(kConsStringTag)), + &if_stringiscons, &if_stringisnotcons); + + Bind(&if_stringiscons); + { + // Check whether the right hand side is the empty string (i.e. if + // this is really a flat string in a cons string). If that is not + // the case we flatten the string first. + Label if_rhsisempty(this), if_rhsisnotempty(this, Label::kDeferred); + Node* rhs = LoadObjectField(string, ConsString::kSecondOffset); + Branch(WordEqual(rhs, EmptyStringConstant()), &if_rhsisempty, + &if_rhsisnotempty); + + Bind(&if_rhsisempty); + { + // Just operate on the left hand side of the {string}. + var_string.Bind(LoadObjectField(string, ConsString::kFirstOffset)); + Goto(&loop); + } + + Bind(&if_rhsisnotempty); + { + // Flatten the {string} and lookup in the resulting string. + var_string.Bind(CallRuntime(Runtime::kFlattenString, + NoContextConstant(), string)); + Goto(&loop); + } + } + + Bind(&if_stringisnotcons); + { + // Check if the {string} is an ExternalString. + Label if_stringisexternal(this), if_stringisnotexternal(this); + Branch(Word32Equal(Word32And(string_instance_type, + Int32Constant(kStringRepresentationMask)), + Int32Constant(kExternalStringTag)), + &if_stringisexternal, &if_stringisnotexternal); + + Bind(&if_stringisexternal); + { + // Check if the {string} is a short external string. + Label if_stringisshort(this), + if_stringisnotshort(this, Label::kDeferred); + Branch(Word32Equal(Word32And(string_instance_type, + Int32Constant(kShortExternalStringMask)), + Int32Constant(0)), + &if_stringisshort, &if_stringisnotshort); + + Bind(&if_stringisshort); + { + // Load the actual resource data from the {string}. + Node* string_resource_data = + LoadObjectField(string, ExternalString::kResourceDataOffset, + MachineType::Pointer()); + + // Check if the {string} is a TwoByteExternalString or a + // OneByteExternalString. + Label if_stringistwobyte(this), if_stringisonebyte(this); + Branch(Word32Equal(Word32And(string_instance_type, + Int32Constant(kStringEncodingMask)), + Int32Constant(kTwoByteStringTag)), + &if_stringistwobyte, &if_stringisonebyte); + + Bind(&if_stringisonebyte); + { + var_result.Bind( + Load(MachineType::Uint8(), string_resource_data, index)); + Goto(&done_loop); + } + + Bind(&if_stringistwobyte); + { + var_result.Bind(Load(MachineType::Uint16(), string_resource_data, + WordShl(index, IntPtrConstant(1)))); + Goto(&done_loop); + } + } + + Bind(&if_stringisnotshort); + { + // The {string} might be compressed, call the runtime. + var_result.Bind(SmiToWord32( + CallRuntime(Runtime::kExternalStringGetChar, + NoContextConstant(), string, SmiTag(index)))); + Goto(&done_loop); + } + } + + Bind(&if_stringisnotexternal); + { + // The {string} is a SlicedString, continue with its parent. + Node* string_offset = + SmiToWord(LoadObjectField(string, SlicedString::kOffsetOffset)); + Node* string_parent = + LoadObjectField(string, SlicedString::kParentOffset); + var_index.Bind(IntPtrAdd(index, string_offset)); + var_string.Bind(string_parent); + Goto(&loop); + } + } + } + } + + Bind(&done_loop); + return var_result.value(); +} + +Node* CodeStubAssembler::StringFromCharCode(Node* code) { + Variable var_result(this, MachineRepresentation::kTagged); + + // Check if the {code} is a one-byte char code. + Label if_codeisonebyte(this), if_codeistwobyte(this, Label::kDeferred), + if_done(this); + Branch(Int32LessThanOrEqual(code, Int32Constant(String::kMaxOneByteCharCode)), + &if_codeisonebyte, &if_codeistwobyte); + Bind(&if_codeisonebyte); + { + // Load the isolate wide single character string cache. + Node* cache = LoadRoot(Heap::kSingleCharacterStringCacheRootIndex); + + // Check if we have an entry for the {code} in the single character string + // cache already. + Label if_entryisundefined(this, Label::kDeferred), + if_entryisnotundefined(this); + Node* entry = LoadFixedArrayElement(cache, code); + Branch(WordEqual(entry, UndefinedConstant()), &if_entryisundefined, + &if_entryisnotundefined); + + Bind(&if_entryisundefined); + { + // Allocate a new SeqOneByteString for {code} and store it in the {cache}. + Node* result = AllocateSeqOneByteString(1); + StoreNoWriteBarrier( + MachineRepresentation::kWord8, result, + IntPtrConstant(SeqOneByteString::kHeaderSize - kHeapObjectTag), code); + StoreFixedArrayElement(cache, code, result); + var_result.Bind(result); + Goto(&if_done); + } + + Bind(&if_entryisnotundefined); + { + // Return the entry from the {cache}. + var_result.Bind(entry); + Goto(&if_done); + } + } + + Bind(&if_codeistwobyte); + { + // Allocate a new SeqTwoByteString for {code}. + Node* result = AllocateSeqTwoByteString(1); + StoreNoWriteBarrier( + MachineRepresentation::kWord16, result, + IntPtrConstant(SeqTwoByteString::kHeaderSize - kHeapObjectTag), code); + var_result.Bind(result); + Goto(&if_done); + } + + Bind(&if_done); + return var_result.value(); +} + +Node* CodeStubAssembler::BitFieldDecode(Node* word32, uint32_t shift, + uint32_t mask) { + return Word32Shr(Word32And(word32, Int32Constant(mask)), + Int32Constant(shift)); +} + +void CodeStubAssembler::TryToName(Node* key, Label* if_keyisindex, + Variable* var_index, Label* if_keyisunique, + Label* call_runtime) { + DCHECK_EQ(MachineRepresentation::kWord32, var_index->rep()); + + Label if_keyissmi(this), if_keyisnotsmi(this); + Branch(WordIsSmi(key), &if_keyissmi, &if_keyisnotsmi); + Bind(&if_keyissmi); + { + // Negative smi keys are named properties. Handle in the runtime. + Label if_keyispositive(this); + Branch(WordIsPositiveSmi(key), &if_keyispositive, call_runtime); + Bind(&if_keyispositive); + + var_index->Bind(SmiToWord32(key)); + Goto(if_keyisindex); + } + + Bind(&if_keyisnotsmi); + + Node* key_instance_type = LoadInstanceType(key); + Label if_keyisnotsymbol(this); + Branch(Word32Equal(key_instance_type, Int32Constant(SYMBOL_TYPE)), + if_keyisunique, &if_keyisnotsymbol); + Bind(&if_keyisnotsymbol); + { + Label if_keyisinternalized(this); + Node* bits = + WordAnd(key_instance_type, + Int32Constant(kIsNotStringMask | kIsNotInternalizedMask)); + Branch(Word32Equal(bits, Int32Constant(kStringTag | kInternalizedTag)), + &if_keyisinternalized, call_runtime); + Bind(&if_keyisinternalized); + + // Check whether the key is an array index passed in as string. Handle + // uniform with smi keys if so. + // TODO(verwaest): Also support non-internalized strings. + Node* hash = LoadNameHash(key); + Node* bit = + Word32And(hash, Int32Constant(internal::Name::kIsNotArrayIndexMask)); + Label if_isarrayindex(this); + Branch(Word32Equal(bit, Int32Constant(0)), &if_isarrayindex, + if_keyisunique); + Bind(&if_isarrayindex); + var_index->Bind(BitFieldDecode<internal::Name::ArrayIndexValueBits>(hash)); + Goto(if_keyisindex); + } +} + +void CodeStubAssembler::TryLookupProperty(Node* object, Node* map, + Node* instance_type, Node* name, + Label* if_found, Label* if_not_found, + Label* call_runtime) { + { + Label if_objectissimple(this); + Branch(Int32LessThanOrEqual(instance_type, + Int32Constant(LAST_SPECIAL_RECEIVER_TYPE)), + call_runtime, &if_objectissimple); + Bind(&if_objectissimple); + } + + // TODO(verwaest): Perform a dictonary lookup on slow-mode receivers. + Node* bit_field3 = LoadMapBitField3(map); + Node* bit = BitFieldDecode<Map::DictionaryMap>(bit_field3); + Label if_isfastmap(this); + Branch(Word32Equal(bit, Int32Constant(0)), &if_isfastmap, call_runtime); + Bind(&if_isfastmap); + Node* nof = BitFieldDecode<Map::NumberOfOwnDescriptorsBits>(bit_field3); + // Bail out to the runtime for large numbers of own descriptors. The stub only + // does linear search, which becomes too expensive in that case. + { + static const int32_t kMaxLinear = 210; + Label above_max(this), below_max(this); + Branch(Int32LessThanOrEqual(nof, Int32Constant(kMaxLinear)), &below_max, + call_runtime); + Bind(&below_max); + } + Node* descriptors = LoadMapDescriptors(map); + + Variable var_descriptor(this, MachineRepresentation::kWord32); + Label loop(this, &var_descriptor); + var_descriptor.Bind(Int32Constant(0)); + Goto(&loop); + Bind(&loop); + { + Node* index = var_descriptor.value(); + Node* offset = Int32Constant(DescriptorArray::ToKeyIndex(0)); + Node* factor = Int32Constant(DescriptorArray::kDescriptorSize); + Label if_notdone(this); + Branch(Word32Equal(index, nof), if_not_found, &if_notdone); + Bind(&if_notdone); + { + Node* array_index = Int32Add(offset, Int32Mul(index, factor)); + Node* current = LoadFixedArrayElement(descriptors, array_index); + Label if_unequal(this); + Branch(WordEqual(current, name), if_found, &if_unequal); + Bind(&if_unequal); + + var_descriptor.Bind(Int32Add(index, Int32Constant(1))); + Goto(&loop); + } + } +} + +void CodeStubAssembler::TryLookupElement(Node* object, Node* map, + Node* instance_type, Node* index, + Label* if_found, Label* if_not_found, + Label* call_runtime) { + { + Label if_objectissimple(this); + Branch(Int32LessThanOrEqual(instance_type, + Int32Constant(LAST_CUSTOM_ELEMENTS_RECEIVER)), + call_runtime, &if_objectissimple); + Bind(&if_objectissimple); + } + + Node* bit_field2 = LoadMapBitField2(map); + Node* elements_kind = BitFieldDecode<Map::ElementsKindBits>(bit_field2); + + // TODO(verwaest): Support other elements kinds as well. + Label if_isobjectorsmi(this); + Branch( + Int32LessThanOrEqual(elements_kind, Int32Constant(FAST_HOLEY_ELEMENTS)), + &if_isobjectorsmi, call_runtime); + Bind(&if_isobjectorsmi); + { + Node* elements = LoadElements(object); + Node* length = LoadFixedArrayBaseLength(elements); + + Label if_iskeyinrange(this); + Branch(Int32LessThan(index, SmiToWord32(length)), &if_iskeyinrange, + if_not_found); + + Bind(&if_iskeyinrange); + Node* element = LoadFixedArrayElement(elements, index); + Node* the_hole = LoadRoot(Heap::kTheHoleValueRootIndex); + Branch(WordEqual(element, the_hole), if_not_found, if_found); + } +} + +Node* CodeStubAssembler::OrdinaryHasInstance(Node* context, Node* callable, + Node* object) { + Variable var_result(this, MachineRepresentation::kTagged); + Label return_false(this), return_true(this), + return_runtime(this, Label::kDeferred), return_result(this); + + // Goto runtime if {object} is a Smi. + GotoIf(WordIsSmi(object), &return_runtime); + + // Load map of {object}. + Node* object_map = LoadMap(object); + + // Lookup the {callable} and {object} map in the global instanceof cache. + // Note: This is safe because we clear the global instanceof cache whenever + // we change the prototype of any object. + Node* instanceof_cache_function = + LoadRoot(Heap::kInstanceofCacheFunctionRootIndex); + Node* instanceof_cache_map = LoadRoot(Heap::kInstanceofCacheMapRootIndex); + { + Label instanceof_cache_miss(this); + GotoUnless(WordEqual(instanceof_cache_function, callable), + &instanceof_cache_miss); + GotoUnless(WordEqual(instanceof_cache_map, object_map), + &instanceof_cache_miss); + var_result.Bind(LoadRoot(Heap::kInstanceofCacheAnswerRootIndex)); + Goto(&return_result); + Bind(&instanceof_cache_miss); + } + + // Goto runtime if {callable} is a Smi. + GotoIf(WordIsSmi(callable), &return_runtime); + + // Load map of {callable}. + Node* callable_map = LoadMap(callable); + + // Goto runtime if {callable} is not a JSFunction. + Node* callable_instance_type = LoadMapInstanceType(callable_map); + GotoUnless( + Word32Equal(callable_instance_type, Int32Constant(JS_FUNCTION_TYPE)), + &return_runtime); + + // Goto runtime if {callable} is not a constructor or has + // a non-instance "prototype". + Node* callable_bitfield = LoadMapBitField(callable_map); + GotoUnless( + Word32Equal(Word32And(callable_bitfield, + Int32Constant((1 << Map::kHasNonInstancePrototype) | + (1 << Map::kIsConstructor))), + Int32Constant(1 << Map::kIsConstructor)), + &return_runtime); + + // Get the "prototype" (or initial map) of the {callable}. + Node* callable_prototype = + LoadObjectField(callable, JSFunction::kPrototypeOrInitialMapOffset); + { + Variable var_callable_prototype(this, MachineRepresentation::kTagged); + Label callable_prototype_valid(this); + var_callable_prototype.Bind(callable_prototype); + + // Resolve the "prototype" if the {callable} has an initial map. Afterwards + // the {callable_prototype} will be either the JSReceiver prototype object + // or the hole value, which means that no instances of the {callable} were + // created so far and hence we should return false. + Node* callable_prototype_instance_type = + LoadInstanceType(callable_prototype); + GotoUnless( + Word32Equal(callable_prototype_instance_type, Int32Constant(MAP_TYPE)), + &callable_prototype_valid); + var_callable_prototype.Bind( + LoadObjectField(callable_prototype, Map::kPrototypeOffset)); + Goto(&callable_prototype_valid); + Bind(&callable_prototype_valid); + callable_prototype = var_callable_prototype.value(); + } + + // Update the global instanceof cache with the current {object} map and + // {callable}. The cached answer will be set when it is known below. + StoreRoot(Heap::kInstanceofCacheFunctionRootIndex, callable); + StoreRoot(Heap::kInstanceofCacheMapRootIndex, object_map); + + // Loop through the prototype chain looking for the {callable} prototype. + Variable var_object_map(this, MachineRepresentation::kTagged); + var_object_map.Bind(object_map); + Label loop(this, &var_object_map); + Goto(&loop); + Bind(&loop); + { + Node* object_map = var_object_map.value(); + + // Check if the current {object} needs to be access checked. + Node* object_bitfield = LoadMapBitField(object_map); + GotoUnless( + Word32Equal(Word32And(object_bitfield, + Int32Constant(1 << Map::kIsAccessCheckNeeded)), + Int32Constant(0)), + &return_runtime); + + // Check if the current {object} is a proxy. + Node* object_instance_type = LoadMapInstanceType(object_map); + GotoIf(Word32Equal(object_instance_type, Int32Constant(JS_PROXY_TYPE)), + &return_runtime); + + // Check the current {object} prototype. + Node* object_prototype = LoadMapPrototype(object_map); + GotoIf(WordEqual(object_prototype, callable_prototype), &return_true); + GotoIf(WordEqual(object_prototype, NullConstant()), &return_false); + + // Continue with the prototype. + var_object_map.Bind(LoadMap(object_prototype)); + Goto(&loop); + } + + Bind(&return_true); + StoreRoot(Heap::kInstanceofCacheAnswerRootIndex, BooleanConstant(true)); + var_result.Bind(BooleanConstant(true)); + Goto(&return_result); + + Bind(&return_false); + StoreRoot(Heap::kInstanceofCacheAnswerRootIndex, BooleanConstant(false)); + var_result.Bind(BooleanConstant(false)); + Goto(&return_result); + + Bind(&return_runtime); + { + // Invalidate the global instanceof cache. + StoreRoot(Heap::kInstanceofCacheFunctionRootIndex, SmiConstant(0)); + // Fallback to the runtime implementation. + var_result.Bind( + CallRuntime(Runtime::kOrdinaryHasInstance, context, callable, object)); + } + Goto(&return_result); + + Bind(&return_result); + return var_result.value(); +} + +compiler::Node* CodeStubAssembler::ElementOffsetFromIndex(Node* index_node, + ElementsKind kind, + ParameterMode mode, + int base_size) { + bool is_double = IsFastDoubleElementsKind(kind); + int element_size_shift = is_double ? kDoubleSizeLog2 : kPointerSizeLog2; + int element_size = 1 << element_size_shift; + int const kSmiShiftBits = kSmiShiftSize + kSmiTagSize; + int32_t index = 0; + bool constant_index = false; + if (mode == SMI_PARAMETERS) { + element_size_shift -= kSmiShiftBits; + intptr_t temp = 0; + constant_index = ToIntPtrConstant(index_node, temp); + index = temp >> kSmiShiftBits; + } else { + constant_index = ToInt32Constant(index_node, index); + } + if (constant_index) { + return IntPtrConstant(base_size + element_size * index); + } + if (Is64() && mode == INTEGER_PARAMETERS) { + index_node = ChangeInt32ToInt64(index_node); + } + if (base_size == 0) { + return (element_size_shift >= 0) + ? WordShl(index_node, IntPtrConstant(element_size_shift)) + : WordShr(index_node, IntPtrConstant(-element_size_shift)); + } + return IntPtrAdd( + IntPtrConstant(base_size), + (element_size_shift >= 0) + ? WordShl(index_node, IntPtrConstant(element_size_shift)) + : WordShr(index_node, IntPtrConstant(-element_size_shift))); +} + +} // namespace internal +} // namespace v8 |