mirror of
https://github.com/darlinghq/darling-JavaScriptCore.git
synced 2024-12-04 18:16:39 +00:00
3292 lines
117 KiB
C++
3292 lines
117 KiB
C++
/*
|
|
* Copyright (C) 2013-2017 Apple Inc. All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY
|
|
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
|
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR
|
|
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
|
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
|
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
|
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
|
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
|
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#if ENABLE(DFG_JIT)
|
|
|
|
#include "ArrayConstructor.h"
|
|
#include "DFGAbstractInterpreter.h"
|
|
#include "DOMJITGetterSetter.h"
|
|
#include "DOMJITSignature.h"
|
|
#include "GetByIdStatus.h"
|
|
#include "GetterSetter.h"
|
|
#include "HashMapImpl.h"
|
|
#include "JITOperations.h"
|
|
#include "MathCommon.h"
|
|
#include "Operations.h"
|
|
#include "PutByIdStatus.h"
|
|
#include "StringObject.h"
|
|
|
|
namespace JSC { namespace DFG {
|
|
|
|
template<typename AbstractStateType>
|
|
AbstractInterpreter<AbstractStateType>::AbstractInterpreter(Graph& graph, AbstractStateType& state)
|
|
: m_codeBlock(graph.m_codeBlock)
|
|
, m_graph(graph)
|
|
, m_vm(m_graph.m_vm)
|
|
, m_state(state)
|
|
{
|
|
if (m_graph.m_form == SSA)
|
|
m_phiChildren = std::make_unique<PhiChildren>(m_graph);
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
AbstractInterpreter<AbstractStateType>::~AbstractInterpreter()
|
|
{
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
typename AbstractInterpreter<AbstractStateType>::BooleanResult
|
|
AbstractInterpreter<AbstractStateType>::booleanResult(
|
|
Node* node, AbstractValue& value)
|
|
{
|
|
JSValue childConst = value.value();
|
|
if (childConst) {
|
|
if (childConst.toBoolean(m_codeBlock->globalObjectFor(node->origin.semantic)->globalExec()))
|
|
return DefinitelyTrue;
|
|
return DefinitelyFalse;
|
|
}
|
|
|
|
// Next check if we can fold because we know that the source is an object or string and does not equal undefined.
|
|
if (isCellSpeculation(value.m_type) && !value.m_structure.isTop()) {
|
|
bool allTrue = true;
|
|
for (unsigned i = value.m_structure.size(); i--;) {
|
|
RegisteredStructure structure = value.m_structure[i];
|
|
if (structure->masqueradesAsUndefined(m_codeBlock->globalObjectFor(node->origin.semantic))
|
|
|| structure->typeInfo().type() == StringType) {
|
|
allTrue = false;
|
|
break;
|
|
}
|
|
}
|
|
if (allTrue)
|
|
return DefinitelyTrue;
|
|
}
|
|
|
|
return UnknownBooleanResult;
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::startExecuting()
|
|
{
|
|
ASSERT(m_state.block());
|
|
ASSERT(m_state.isValid());
|
|
|
|
m_state.setDidClobber(false);
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::executeEdges(Node* node)
|
|
{
|
|
m_graph.doToChildren(
|
|
node,
|
|
[&] (Edge& edge) {
|
|
filterEdgeByUse(edge);
|
|
});
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::executeKnownEdgeTypes(Node* node)
|
|
{
|
|
// Some use kinds are required to not have checks, because we know somehow that the incoming
|
|
// value will already have the type we want. In those cases, AI may not be smart enough to
|
|
// prove that this is indeed the case. But the existance of the edge is enough to prove that
|
|
// it is indeed the case. Taking advantage of this is not optional, since otherwise the DFG
|
|
// and FTL backends may emit checks in a node that lacks a valid exit origin.
|
|
m_graph.doToChildren(
|
|
node,
|
|
[&] (Edge& edge) {
|
|
if (mayHaveTypeCheck(edge.useKind()))
|
|
return;
|
|
|
|
filterEdgeByUse(edge);
|
|
});
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::verifyEdge(Node* node, Edge edge)
|
|
{
|
|
if (!(forNode(edge).m_type & ~typeFilterFor(edge.useKind())))
|
|
return;
|
|
|
|
DFG_CRASH(m_graph, node, toCString("Edge verification error: ", node, "->", edge, " was expected to have type ", SpeculationDump(typeFilterFor(edge.useKind())), " but has type ", SpeculationDump(forNode(edge).m_type), " (", forNode(edge).m_type, ")").data(), AbstractInterpreterInvalidType, node->op(), edge->op(), edge.useKind(), forNode(edge).m_type);
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::verifyEdges(Node* node)
|
|
{
|
|
DFG_NODE_DO_TO_CHILDREN(m_graph, node, verifyEdge);
|
|
}
|
|
|
|
inline bool isToThisAnIdentity(bool isStrictMode, AbstractValue& valueForNode)
|
|
{
|
|
// We look at the type first since that will cover most cases and does not require iterating all the structures.
|
|
if (isStrictMode) {
|
|
if (valueForNode.m_type && !(valueForNode.m_type & SpecObjectOther))
|
|
return true;
|
|
} else {
|
|
if (valueForNode.m_type && !(valueForNode.m_type & (~SpecObject | SpecObjectOther)))
|
|
return true;
|
|
}
|
|
|
|
if ((isStrictMode || (valueForNode.m_type && !(valueForNode.m_type & ~SpecObject))) && valueForNode.m_structure.isFinite()) {
|
|
bool overridesToThis = false;
|
|
valueForNode.m_structure.forEach([&](RegisteredStructure structure) {
|
|
TypeInfo type = structure->typeInfo();
|
|
ASSERT(type.isObject() || type.type() == StringType || type.type() == SymbolType);
|
|
if (!isStrictMode)
|
|
ASSERT(type.isObject());
|
|
// We don't need to worry about strings/symbols here since either:
|
|
// 1) We are in strict mode and strings/symbols are not wrapped
|
|
// 2) The AI has proven that the type of this is a subtype of object
|
|
if (type.isObject() && type.overridesToThis())
|
|
overridesToThis = true;
|
|
});
|
|
return !overridesToThis;
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
bool AbstractInterpreter<AbstractStateType>::executeEffects(unsigned clobberLimit, Node* node)
|
|
{
|
|
verifyEdges(node);
|
|
|
|
m_state.createValueForNode(node);
|
|
|
|
switch (node->op()) {
|
|
case JSConstant:
|
|
case DoubleConstant:
|
|
case Int52Constant: {
|
|
setBuiltInConstant(node, *node->constant());
|
|
break;
|
|
}
|
|
|
|
case LazyJSConstant: {
|
|
LazyJSValue value = node->lazyJSValue();
|
|
switch (value.kind()) {
|
|
case LazyJSValue::KnownValue:
|
|
setConstant(node, value.value()->value());
|
|
break;
|
|
case LazyJSValue::SingleCharacterString:
|
|
case LazyJSValue::KnownStringImpl:
|
|
case LazyJSValue::NewStringImpl:
|
|
forNode(node).setType(m_graph, SpecString);
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case Identity: {
|
|
forNode(node) = forNode(node->child1());
|
|
if (forNode(node).value())
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
|
|
case ExtractOSREntryLocal: {
|
|
forNode(node).makeBytecodeTop();
|
|
break;
|
|
}
|
|
|
|
case GetLocal: {
|
|
VariableAccessData* variableAccessData = node->variableAccessData();
|
|
AbstractValue value = m_state.variables().operand(variableAccessData->local().offset());
|
|
// The value in the local should already be checked.
|
|
DFG_ASSERT(m_graph, node, value.isType(typeFilterFor(variableAccessData->flushFormat())));
|
|
if (value.value())
|
|
m_state.setFoundConstants(true);
|
|
forNode(node) = value;
|
|
break;
|
|
}
|
|
|
|
case GetStack: {
|
|
StackAccessData* data = node->stackAccessData();
|
|
AbstractValue value = m_state.variables().operand(data->local);
|
|
// The value in the local should already be checked.
|
|
DFG_ASSERT(m_graph, node, value.isType(typeFilterFor(data->format)));
|
|
if (value.value())
|
|
m_state.setFoundConstants(true);
|
|
forNode(node) = value;
|
|
break;
|
|
}
|
|
|
|
case GetLocalUnlinked: {
|
|
AbstractValue value = m_state.variables().operand(node->unlinkedLocal().offset());
|
|
if (value.value())
|
|
m_state.setFoundConstants(true);
|
|
forNode(node) = value;
|
|
break;
|
|
}
|
|
|
|
case SetLocal: {
|
|
m_state.variables().operand(node->local()) = forNode(node->child1());
|
|
break;
|
|
}
|
|
|
|
case PutStack: {
|
|
m_state.variables().operand(node->stackAccessData()->local) = forNode(node->child1());
|
|
break;
|
|
}
|
|
|
|
case MovHint: {
|
|
// Don't need to do anything. A MovHint only informs us about what would have happened
|
|
// in bytecode, but this code is just concerned with what is actually happening during
|
|
// DFG execution.
|
|
break;
|
|
}
|
|
|
|
case KillStack: {
|
|
// This is just a hint telling us that the OSR state of the local is no longer inside the
|
|
// flushed data.
|
|
break;
|
|
}
|
|
|
|
case SetArgument:
|
|
// Assert that the state of arguments has been set. SetArgument means that someone set
|
|
// the argument values out-of-band, and currently this always means setting to a
|
|
// non-clear value.
|
|
ASSERT(!m_state.variables().operand(node->local()).isClear());
|
|
break;
|
|
|
|
case LoadVarargs:
|
|
case ForwardVarargs: {
|
|
// FIXME: ForwardVarargs should check if the count becomes known, and if it does, it should turn
|
|
// itself into a straight-line sequence of GetStack/PutStack.
|
|
// https://bugs.webkit.org/show_bug.cgi?id=143071
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
LoadVarargsData* data = node->loadVarargsData();
|
|
m_state.variables().operand(data->count).setType(SpecInt32Only);
|
|
for (unsigned i = data->limit - 1; i--;)
|
|
m_state.variables().operand(data->start.offset() + i).makeHeapTop();
|
|
break;
|
|
}
|
|
|
|
case BitAnd:
|
|
case BitOr:
|
|
case BitXor:
|
|
case BitRShift:
|
|
case BitLShift:
|
|
case BitURShift: {
|
|
if (node->child1().useKind() == UntypedUse || node->child2().useKind() == UntypedUse) {
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
if (left && right && left.isInt32() && right.isInt32()) {
|
|
int32_t a = left.asInt32();
|
|
int32_t b = right.asInt32();
|
|
switch (node->op()) {
|
|
case BitAnd:
|
|
setConstant(node, JSValue(a & b));
|
|
break;
|
|
case BitOr:
|
|
setConstant(node, JSValue(a | b));
|
|
break;
|
|
case BitXor:
|
|
setConstant(node, JSValue(a ^ b));
|
|
break;
|
|
case BitRShift:
|
|
setConstant(node, JSValue(a >> static_cast<uint32_t>(b)));
|
|
break;
|
|
case BitLShift:
|
|
setConstant(node, JSValue(a << static_cast<uint32_t>(b)));
|
|
break;
|
|
case BitURShift:
|
|
setConstant(node, JSValue(static_cast<uint32_t>(a) >> static_cast<uint32_t>(b)));
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
if (node->op() == BitAnd
|
|
&& (isBoolInt32Speculation(forNode(node->child1()).m_type) ||
|
|
isBoolInt32Speculation(forNode(node->child2()).m_type))) {
|
|
forNode(node).setType(SpecBoolInt32);
|
|
break;
|
|
}
|
|
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case UInt32ToNumber: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
if (doesOverflow(node->arithMode())) {
|
|
if (enableInt52()) {
|
|
if (child && child.isAnyInt()) {
|
|
int64_t machineInt = child.asAnyInt();
|
|
setConstant(node, jsNumber(static_cast<uint32_t>(machineInt)));
|
|
break;
|
|
}
|
|
forNode(node).setType(SpecAnyInt);
|
|
break;
|
|
}
|
|
if (child && child.isInt32()) {
|
|
uint32_t value = child.asInt32();
|
|
setConstant(node, jsNumber(value));
|
|
break;
|
|
}
|
|
forNode(node).setType(SpecAnyIntAsDouble);
|
|
break;
|
|
}
|
|
if (child && child.isInt32()) {
|
|
int32_t value = child.asInt32();
|
|
if (value >= 0) {
|
|
setConstant(node, jsNumber(value));
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case BooleanToNumber: {
|
|
JSValue concreteValue = forNode(node->child1()).value();
|
|
if (concreteValue) {
|
|
if (concreteValue.isBoolean())
|
|
setConstant(node, jsNumber(concreteValue.asBoolean()));
|
|
else
|
|
setConstant(node, *m_graph.freeze(concreteValue));
|
|
break;
|
|
}
|
|
AbstractValue& value = forNode(node);
|
|
value = forNode(node->child1());
|
|
if (node->child1().useKind() == UntypedUse && !(value.m_type & ~SpecBoolean))
|
|
m_state.setFoundConstants(true);
|
|
if (value.m_type & SpecBoolean) {
|
|
value.merge(SpecBoolInt32);
|
|
value.filter(~SpecBoolean);
|
|
}
|
|
break;
|
|
}
|
|
|
|
case DoubleAsInt32: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
if (child && child.isNumber()) {
|
|
double asDouble = child.asNumber();
|
|
int32_t asInt = JSC::toInt32(asDouble);
|
|
if (bitwise_cast<int64_t>(static_cast<double>(asInt)) == bitwise_cast<int64_t>(asDouble)) {
|
|
setConstant(node, JSValue(asInt));
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case ValueToInt32: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
if (child) {
|
|
if (child.isNumber()) {
|
|
if (child.isInt32())
|
|
setConstant(node, child);
|
|
else
|
|
setConstant(node, JSValue(JSC::toInt32(child.asDouble())));
|
|
break;
|
|
}
|
|
if (child.isBoolean()) {
|
|
setConstant(node, jsNumber(child.asBoolean()));
|
|
break;
|
|
}
|
|
if (child.isUndefinedOrNull()) {
|
|
setConstant(node, jsNumber(0));
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (isBooleanSpeculation(forNode(node->child1()).m_type)) {
|
|
forNode(node).setType(SpecBoolInt32);
|
|
break;
|
|
}
|
|
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case DoubleRep: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
if (std::optional<double> number = child.toNumberFromPrimitive()) {
|
|
setConstant(node, jsDoubleNumber(*number));
|
|
break;
|
|
}
|
|
|
|
SpeculatedType type = forNode(node->child1()).m_type;
|
|
switch (node->child1().useKind()) {
|
|
case NotCellUse: {
|
|
if (type & SpecOther) {
|
|
type &= ~SpecOther;
|
|
type |= SpecDoublePureNaN | SpecBoolInt32; // Null becomes zero, undefined becomes NaN.
|
|
}
|
|
if (type & SpecBoolean) {
|
|
type &= ~SpecBoolean;
|
|
type |= SpecBoolInt32; // True becomes 1, false becomes 0.
|
|
}
|
|
type &= SpecBytecodeNumber;
|
|
break;
|
|
}
|
|
|
|
case Int52RepUse:
|
|
case NumberUse:
|
|
case RealNumberUse:
|
|
break;
|
|
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
}
|
|
forNode(node).setType(type);
|
|
forNode(node).fixTypeForRepresentation(m_graph, node);
|
|
break;
|
|
}
|
|
|
|
case Int52Rep: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
if (child && child.isAnyInt()) {
|
|
setConstant(node, child);
|
|
break;
|
|
}
|
|
|
|
forNode(node).setType(SpecAnyInt);
|
|
break;
|
|
}
|
|
|
|
case ValueRep: {
|
|
JSValue value = forNode(node->child1()).value();
|
|
if (value) {
|
|
setConstant(node, value);
|
|
break;
|
|
}
|
|
|
|
forNode(node).setType(m_graph, forNode(node->child1()).m_type & ~SpecDoubleImpureNaN);
|
|
forNode(node).fixTypeForRepresentation(m_graph, node);
|
|
break;
|
|
}
|
|
|
|
case ValueAdd: {
|
|
ASSERT(node->binaryUseKind() == UntypedUse);
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecString | SpecBytecodeNumber);
|
|
break;
|
|
}
|
|
|
|
case StrCat: {
|
|
forNode(node).setType(m_graph, SpecString);
|
|
break;
|
|
}
|
|
|
|
case ArithAdd: {
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
switch (node->binaryUseKind()) {
|
|
case Int32Use:
|
|
if (left && right && left.isInt32() && right.isInt32()) {
|
|
if (!shouldCheckOverflow(node->arithMode())) {
|
|
setConstant(node, jsNumber(left.asInt32() + right.asInt32()));
|
|
break;
|
|
}
|
|
JSValue result = jsNumber(left.asNumber() + right.asNumber());
|
|
if (result.isInt32()) {
|
|
setConstant(node, result);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Int52RepUse:
|
|
if (left && right && left.isAnyInt() && right.isAnyInt()) {
|
|
JSValue result = jsNumber(left.asAnyInt() + right.asAnyInt());
|
|
if (result.isAnyInt()) {
|
|
setConstant(node, result);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecAnyInt);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (left && right && left.isNumber() && right.isNumber()) {
|
|
setConstant(node, jsDoubleNumber(left.asNumber() + right.asNumber()));
|
|
break;
|
|
}
|
|
forNode(node).setType(
|
|
typeOfDoubleSum(
|
|
forNode(node->child1()).m_type, forNode(node->child2()).m_type));
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case AtomicsIsLockFree: {
|
|
if (node->child1().useKind() != Int32Use)
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(SpecBoolInt32);
|
|
break;
|
|
}
|
|
|
|
case ArithClz32: {
|
|
JSValue operand = forNode(node->child1()).value();
|
|
if (std::optional<double> number = operand.toNumberFromPrimitive()) {
|
|
uint32_t value = toUInt32(*number);
|
|
setConstant(node, jsNumber(clz32(value)));
|
|
break;
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case MakeRope: {
|
|
unsigned numberOfChildren = 0;
|
|
unsigned numberOfRemovedChildren = 0;
|
|
std::optional<unsigned> nonEmptyIndex;
|
|
for (unsigned i = 0; i < AdjacencyList::Size; ++i) {
|
|
Edge& edge = node->children.child(i);
|
|
if (!edge)
|
|
break;
|
|
++numberOfChildren;
|
|
|
|
JSValue childConstant = m_state.forNode(edge).value();
|
|
if (!childConstant) {
|
|
nonEmptyIndex = i;
|
|
continue;
|
|
}
|
|
if (!childConstant.isString()) {
|
|
nonEmptyIndex = i;
|
|
continue;
|
|
}
|
|
if (asString(childConstant)->length()) {
|
|
nonEmptyIndex = i;
|
|
continue;
|
|
}
|
|
|
|
++numberOfRemovedChildren;
|
|
}
|
|
|
|
if (numberOfRemovedChildren) {
|
|
m_state.setFoundConstants(true);
|
|
if (numberOfRemovedChildren == numberOfChildren) {
|
|
// Propagate the last child. This is the way taken in the constant folding phase.
|
|
forNode(node) = forNode(node->children.child(numberOfChildren - 1));
|
|
break;
|
|
}
|
|
if ((numberOfRemovedChildren + 1) == numberOfChildren) {
|
|
ASSERT(nonEmptyIndex);
|
|
forNode(node) = forNode(node->children.child(nonEmptyIndex.value()));
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).set(m_graph, m_vm.stringStructure.get());
|
|
break;
|
|
}
|
|
|
|
case ArithSub: {
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
switch (node->binaryUseKind()) {
|
|
case Int32Use:
|
|
if (left && right && left.isInt32() && right.isInt32()) {
|
|
if (!shouldCheckOverflow(node->arithMode())) {
|
|
setConstant(node, jsNumber(left.asInt32() - right.asInt32()));
|
|
break;
|
|
}
|
|
JSValue result = jsNumber(left.asNumber() - right.asNumber());
|
|
if (result.isInt32()) {
|
|
setConstant(node, result);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Int52RepUse:
|
|
if (left && right && left.isAnyInt() && right.isAnyInt()) {
|
|
JSValue result = jsNumber(left.asAnyInt() - right.asAnyInt());
|
|
if (result.isAnyInt() || !shouldCheckOverflow(node->arithMode())) {
|
|
setConstant(node, result);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecAnyInt);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (left && right && left.isNumber() && right.isNumber()) {
|
|
setConstant(node, jsDoubleNumber(left.asNumber() - right.asNumber()));
|
|
break;
|
|
}
|
|
forNode(node).setType(
|
|
typeOfDoubleDifference(
|
|
forNode(node->child1()).m_type, forNode(node->child2()).m_type));
|
|
break;
|
|
case UntypedUse:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecBytecodeNumber);
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithNegate: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
switch (node->child1().useKind()) {
|
|
case Int32Use:
|
|
if (child && child.isInt32()) {
|
|
if (!shouldCheckOverflow(node->arithMode())) {
|
|
setConstant(node, jsNumber(-child.asInt32()));
|
|
break;
|
|
}
|
|
double doubleResult;
|
|
if (shouldCheckNegativeZero(node->arithMode()))
|
|
doubleResult = -child.asNumber();
|
|
else
|
|
doubleResult = 0 - child.asNumber();
|
|
JSValue valueResult = jsNumber(doubleResult);
|
|
if (valueResult.isInt32()) {
|
|
setConstant(node, valueResult);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Int52RepUse:
|
|
if (child && child.isAnyInt()) {
|
|
double doubleResult;
|
|
if (shouldCheckNegativeZero(node->arithMode()))
|
|
doubleResult = -child.asNumber();
|
|
else
|
|
doubleResult = 0 - child.asNumber();
|
|
JSValue valueResult = jsNumber(doubleResult);
|
|
if (valueResult.isAnyInt()) {
|
|
setConstant(node, valueResult);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecAnyInt);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (child && child.isNumber()) {
|
|
setConstant(node, jsDoubleNumber(-child.asNumber()));
|
|
break;
|
|
}
|
|
forNode(node).setType(
|
|
typeOfDoubleNegation(
|
|
forNode(node->child1()).m_type));
|
|
break;
|
|
default:
|
|
DFG_ASSERT(m_graph, node, node->child1().useKind() == UntypedUse);
|
|
forNode(node).setType(SpecBytecodeNumber);
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithMul: {
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
switch (node->binaryUseKind()) {
|
|
case Int32Use:
|
|
if (left && right && left.isInt32() && right.isInt32()) {
|
|
if (!shouldCheckOverflow(node->arithMode())) {
|
|
setConstant(node, jsNumber(left.asInt32() * right.asInt32()));
|
|
break;
|
|
}
|
|
double doubleResult = left.asNumber() * right.asNumber();
|
|
if (!shouldCheckNegativeZero(node->arithMode()))
|
|
doubleResult += 0; // Sanitizes zero.
|
|
JSValue valueResult = jsNumber(doubleResult);
|
|
if (valueResult.isInt32()) {
|
|
setConstant(node, valueResult);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Int52RepUse:
|
|
if (left && right && left.isAnyInt() && right.isAnyInt()) {
|
|
double doubleResult = left.asNumber() * right.asNumber();
|
|
if (!shouldCheckNegativeZero(node->arithMode()))
|
|
doubleResult += 0;
|
|
JSValue valueResult = jsNumber(doubleResult);
|
|
if (valueResult.isAnyInt()) {
|
|
setConstant(node, valueResult);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecAnyInt);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (left && right && left.isNumber() && right.isNumber()) {
|
|
setConstant(node, jsDoubleNumber(left.asNumber() * right.asNumber()));
|
|
break;
|
|
}
|
|
forNode(node).setType(
|
|
typeOfDoubleProduct(
|
|
forNode(node->child1()).m_type, forNode(node->child2()).m_type));
|
|
break;
|
|
case UntypedUse:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecBytecodeNumber);
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithDiv: {
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
switch (node->binaryUseKind()) {
|
|
case Int32Use:
|
|
if (left && right && left.isInt32() && right.isInt32()) {
|
|
double doubleResult = left.asNumber() / right.asNumber();
|
|
if (!shouldCheckOverflow(node->arithMode()))
|
|
doubleResult = toInt32(doubleResult);
|
|
else if (!shouldCheckNegativeZero(node->arithMode()))
|
|
doubleResult += 0; // Sanitizes zero.
|
|
JSValue valueResult = jsNumber(doubleResult);
|
|
if (valueResult.isInt32()) {
|
|
setConstant(node, valueResult);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (left && right && left.isNumber() && right.isNumber()) {
|
|
setConstant(node, jsDoubleNumber(left.asNumber() / right.asNumber()));
|
|
break;
|
|
}
|
|
forNode(node).setType(
|
|
typeOfDoubleQuotient(
|
|
forNode(node->child1()).m_type, forNode(node->child2()).m_type));
|
|
break;
|
|
case UntypedUse:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecBytecodeNumber);
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithMod: {
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
switch (node->binaryUseKind()) {
|
|
case Int32Use:
|
|
if (left && right && left.isInt32() && right.isInt32()) {
|
|
double doubleResult = fmod(left.asNumber(), right.asNumber());
|
|
if (!shouldCheckOverflow(node->arithMode()))
|
|
doubleResult = toInt32(doubleResult);
|
|
else if (!shouldCheckNegativeZero(node->arithMode()))
|
|
doubleResult += 0; // Sanitizes zero.
|
|
JSValue valueResult = jsNumber(doubleResult);
|
|
if (valueResult.isInt32()) {
|
|
setConstant(node, valueResult);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (left && right && left.isNumber() && right.isNumber()) {
|
|
setConstant(node, jsDoubleNumber(fmod(left.asNumber(), right.asNumber())));
|
|
break;
|
|
}
|
|
forNode(node).setType(
|
|
typeOfDoubleBinaryOp(
|
|
forNode(node->child1()).m_type, forNode(node->child2()).m_type));
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithMin: {
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
switch (node->binaryUseKind()) {
|
|
case Int32Use:
|
|
if (left && right && left.isInt32() && right.isInt32()) {
|
|
setConstant(node, jsNumber(std::min(left.asInt32(), right.asInt32())));
|
|
break;
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (left && right && left.isNumber() && right.isNumber()) {
|
|
double a = left.asNumber();
|
|
double b = right.asNumber();
|
|
setConstant(node, jsDoubleNumber(a < b ? a : (b <= a ? b : a + b)));
|
|
break;
|
|
}
|
|
forNode(node).setType(
|
|
typeOfDoubleMinMax(
|
|
forNode(node->child1()).m_type, forNode(node->child2()).m_type));
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithMax: {
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
switch (node->binaryUseKind()) {
|
|
case Int32Use:
|
|
if (left && right && left.isInt32() && right.isInt32()) {
|
|
setConstant(node, jsNumber(std::max(left.asInt32(), right.asInt32())));
|
|
break;
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (left && right && left.isNumber() && right.isNumber()) {
|
|
double a = left.asNumber();
|
|
double b = right.asNumber();
|
|
setConstant(node, jsDoubleNumber(a > b ? a : (b >= a ? b : a + b)));
|
|
break;
|
|
}
|
|
forNode(node).setType(
|
|
typeOfDoubleMinMax(
|
|
forNode(node->child1()).m_type, forNode(node->child2()).m_type));
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithAbs: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
switch (node->child1().useKind()) {
|
|
case Int32Use:
|
|
if (std::optional<double> number = child.toNumberFromPrimitive()) {
|
|
JSValue result = jsNumber(fabs(*number));
|
|
if (result.isInt32()) {
|
|
setConstant(node, result);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case DoubleRepUse:
|
|
if (std::optional<double> number = child.toNumberFromPrimitive()) {
|
|
setConstant(node, jsDoubleNumber(fabs(*number)));
|
|
break;
|
|
}
|
|
forNode(node).setType(typeOfDoubleAbs(forNode(node->child1()).m_type));
|
|
break;
|
|
default:
|
|
DFG_ASSERT(m_graph, node, node->child1().useKind() == UntypedUse);
|
|
forNode(node).setType(SpecFullNumber);
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithPow: {
|
|
JSValue childY = forNode(node->child2()).value();
|
|
if (childY && childY.isNumber()) {
|
|
if (!childY.asNumber()) {
|
|
setConstant(node, jsDoubleNumber(1));
|
|
break;
|
|
}
|
|
|
|
JSValue childX = forNode(node->child1()).value();
|
|
if (childX && childX.isNumber()) {
|
|
setConstant(node, jsDoubleNumber(operationMathPow(childX.asNumber(), childY.asNumber())));
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(typeOfDoublePow(forNode(node->child1()).m_type, forNode(node->child2()).m_type));
|
|
break;
|
|
}
|
|
|
|
case ArithRandom: {
|
|
forNode(node).setType(m_graph, SpecDoubleReal);
|
|
break;
|
|
}
|
|
|
|
case ArithRound:
|
|
case ArithFloor:
|
|
case ArithCeil:
|
|
case ArithTrunc: {
|
|
JSValue operand = forNode(node->child1()).value();
|
|
if (std::optional<double> number = operand.toNumberFromPrimitive()) {
|
|
double roundedValue = 0;
|
|
if (node->op() == ArithRound)
|
|
roundedValue = jsRound(*number);
|
|
else if (node->op() == ArithFloor)
|
|
roundedValue = floor(*number);
|
|
else if (node->op() == ArithCeil)
|
|
roundedValue = ceil(*number);
|
|
else {
|
|
ASSERT(node->op() == ArithTrunc);
|
|
roundedValue = trunc(*number);
|
|
}
|
|
|
|
if (node->child1().useKind() == UntypedUse) {
|
|
setConstant(node, jsNumber(roundedValue));
|
|
break;
|
|
}
|
|
if (producesInteger(node->arithRoundingMode())) {
|
|
int32_t roundedValueAsInt32 = static_cast<int32_t>(roundedValue);
|
|
if (roundedValueAsInt32 == roundedValue) {
|
|
if (shouldCheckNegativeZero(node->arithRoundingMode())) {
|
|
if (roundedValueAsInt32 || !std::signbit(roundedValue)) {
|
|
setConstant(node, jsNumber(roundedValueAsInt32));
|
|
break;
|
|
}
|
|
} else {
|
|
setConstant(node, jsNumber(roundedValueAsInt32));
|
|
break;
|
|
}
|
|
}
|
|
} else {
|
|
setConstant(node, jsDoubleNumber(roundedValue));
|
|
break;
|
|
}
|
|
}
|
|
if (node->child1().useKind() == DoubleRepUse) {
|
|
if (producesInteger(node->arithRoundingMode()))
|
|
forNode(node).setType(SpecInt32Only);
|
|
else if (node->child1().useKind() == DoubleRepUse)
|
|
forNode(node).setType(typeOfDoubleRounding(forNode(node->child1()).m_type));
|
|
} else {
|
|
DFG_ASSERT(m_graph, node, node->child1().useKind() == UntypedUse);
|
|
forNode(node).setType(SpecFullNumber);
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArithSqrt:
|
|
executeDoubleUnaryOpEffects(node, sqrt);
|
|
break;
|
|
|
|
case ArithFRound:
|
|
executeDoubleUnaryOpEffects(node, [](double value) -> double { return static_cast<float>(value); });
|
|
break;
|
|
|
|
case ArithUnary:
|
|
executeDoubleUnaryOpEffects(node, arithUnaryFunction(node->arithUnaryType()));
|
|
break;
|
|
|
|
case LogicalNot: {
|
|
switch (booleanResult(node, forNode(node->child1()))) {
|
|
case DefinitelyTrue:
|
|
setConstant(node, jsBoolean(false));
|
|
break;
|
|
case DefinitelyFalse:
|
|
setConstant(node, jsBoolean(true));
|
|
break;
|
|
default:
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case MapHash: {
|
|
if (JSValue key = forNode(node->child1()).value()) {
|
|
if (std::optional<uint32_t> hash = concurrentJSMapHash(key)) {
|
|
// Although C++ code uses uint32_t for the hash, the closest type in DFG IR is Int32
|
|
// and that's what MapHash returns. So, we have to cast to int32_t to avoid large
|
|
// unsigned values becoming doubles. This casting between signed and unsigned
|
|
// happens in the assembly code we emit when we don't constant fold this node.
|
|
setConstant(node, jsNumber(static_cast<int32_t>(*hash)));
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case ToLowerCase: {
|
|
forNode(node).setType(m_graph, SpecString);
|
|
break;
|
|
}
|
|
|
|
case LoadFromJSMapBucket:
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case GetMapBucket:
|
|
forNode(node).setType(m_graph, SpecCellOther);
|
|
break;
|
|
|
|
case IsNonEmptyMapBucket:
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
|
|
case IsEmpty:
|
|
case IsUndefined:
|
|
case IsBoolean:
|
|
case IsNumber:
|
|
case IsObject:
|
|
case IsObjectOrNull:
|
|
case IsFunction:
|
|
case IsCellWithType:
|
|
case IsTypedArrayView: {
|
|
AbstractValue child = forNode(node->child1());
|
|
if (child.value()) {
|
|
bool constantWasSet = true;
|
|
switch (node->op()) {
|
|
case IsCellWithType:
|
|
setConstant(node, jsBoolean(child.value().isCell() && child.value().asCell()->type() == node->queriedType()));
|
|
break;
|
|
case IsUndefined:
|
|
setConstant(node, jsBoolean(
|
|
child.value().isCell()
|
|
? child.value().asCell()->structure()->masqueradesAsUndefined(m_codeBlock->globalObjectFor(node->origin.semantic))
|
|
: child.value().isUndefined()));
|
|
break;
|
|
case IsBoolean:
|
|
setConstant(node, jsBoolean(child.value().isBoolean()));
|
|
break;
|
|
case IsNumber:
|
|
setConstant(node, jsBoolean(child.value().isNumber()));
|
|
break;
|
|
case IsObject:
|
|
setConstant(node, jsBoolean(child.value().isObject()));
|
|
break;
|
|
case IsObjectOrNull:
|
|
if (child.value().isObject()) {
|
|
JSObject* object = asObject(child.value());
|
|
if (object->type() == JSFunctionType)
|
|
setConstant(node, jsBoolean(false));
|
|
else if (!(object->inlineTypeFlags() & TypeOfShouldCallGetCallData))
|
|
setConstant(node, jsBoolean(!child.value().asCell()->structure()->masqueradesAsUndefined(m_codeBlock->globalObjectFor(node->origin.semantic))));
|
|
else {
|
|
// FIXME: This could just call getCallData.
|
|
// https://bugs.webkit.org/show_bug.cgi?id=144457
|
|
constantWasSet = false;
|
|
}
|
|
} else
|
|
setConstant(node, jsBoolean(child.value().isNull()));
|
|
break;
|
|
case IsFunction:
|
|
if (child.value().isObject()) {
|
|
JSObject* object = asObject(child.value());
|
|
if (object->type() == JSFunctionType)
|
|
setConstant(node, jsBoolean(true));
|
|
else if (!(object->inlineTypeFlags() & TypeOfShouldCallGetCallData))
|
|
setConstant(node, jsBoolean(false));
|
|
else {
|
|
// FIXME: This could just call getCallData.
|
|
// https://bugs.webkit.org/show_bug.cgi?id=144457
|
|
constantWasSet = false;
|
|
}
|
|
} else
|
|
setConstant(node, jsBoolean(false));
|
|
break;
|
|
case IsEmpty:
|
|
setConstant(node, jsBoolean(child.value().isEmpty()));
|
|
break;
|
|
case IsTypedArrayView:
|
|
setConstant(node, jsBoolean(child.value().isObject() && isTypedView(child.value().getObject()->classInfo(m_vm)->typedArrayStorageType)));
|
|
break;
|
|
default:
|
|
constantWasSet = false;
|
|
break;
|
|
}
|
|
if (constantWasSet)
|
|
break;
|
|
}
|
|
|
|
// FIXME: This code should really use AbstractValue::isType() and
|
|
// AbstractValue::couldBeType().
|
|
// https://bugs.webkit.org/show_bug.cgi?id=146870
|
|
|
|
bool constantWasSet = false;
|
|
switch (node->op()) {
|
|
case IsEmpty: {
|
|
if (child.m_type && !(child.m_type & SpecEmpty)) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
if (child.m_type && !(child.m_type & ~SpecEmpty)) {
|
|
setConstant(node, jsBoolean(true));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
break;
|
|
}
|
|
case IsUndefined:
|
|
// FIXME: Use the masquerades-as-undefined watchpoint thingy.
|
|
// https://bugs.webkit.org/show_bug.cgi?id=144456
|
|
|
|
if (!(child.m_type & (SpecOther | SpecObjectOther))) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
break;
|
|
case IsBoolean:
|
|
if (!(child.m_type & ~SpecBoolean)) {
|
|
setConstant(node, jsBoolean(true));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
if (!(child.m_type & SpecBoolean)) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
break;
|
|
case IsNumber:
|
|
if (!(child.m_type & ~SpecFullNumber)) {
|
|
setConstant(node, jsBoolean(true));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
if (!(child.m_type & SpecFullNumber)) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
break;
|
|
case IsObject:
|
|
if (!(child.m_type & ~SpecObject)) {
|
|
setConstant(node, jsBoolean(true));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
if (!(child.m_type & SpecObject)) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
break;
|
|
case IsObjectOrNull:
|
|
// FIXME: Use the masquerades-as-undefined watchpoint thingy.
|
|
// https://bugs.webkit.org/show_bug.cgi?id=144456
|
|
|
|
// These expressions are complicated to parse. A helpful way to parse this is that
|
|
// "!(T & ~S)" means "T is a subset of S". Conversely, "!(T & S)" means "T is a
|
|
// disjoint set from S". Things like "T - S" means that, provided that S is a
|
|
// subset of T, it's the "set of all things in T but not in S". Things like "T | S"
|
|
// mean the "union of T and S".
|
|
|
|
// Is the child's type an object that isn't an other-object (i.e. object that could
|
|
// have masquaredes-as-undefined traps) and isn't a function? Then: we should fold
|
|
// this to true.
|
|
if (!(child.m_type & ~(SpecObject - SpecObjectOther - SpecFunction))) {
|
|
setConstant(node, jsBoolean(true));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
// Is the child's type definitely not either of: an object that isn't a function,
|
|
// or either undefined or null? Then: we should fold this to false. This means
|
|
// for example that if it's any non-function object, including those that have
|
|
// masquerades-as-undefined traps, then we don't fold. It also means we won't fold
|
|
// if it's undefined-or-null, since the type bits don't distinguish between
|
|
// undefined (which should fold to false) and null (which should fold to true).
|
|
if (!(child.m_type & ((SpecObject - SpecFunction) | SpecOther))) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
break;
|
|
case IsFunction:
|
|
if (!(child.m_type & ~SpecFunction)) {
|
|
setConstant(node, jsBoolean(true));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
|
|
if (!(child.m_type & (SpecFunction | SpecObjectOther | SpecProxyObject))) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
break;
|
|
|
|
case IsCellWithType:
|
|
if (!(child.m_type & ~node->speculatedTypeForQuery())) {
|
|
setConstant(node, jsBoolean(true));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
if (!(child.m_type & node->speculatedTypeForQuery())) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
break;
|
|
|
|
case IsTypedArrayView:
|
|
if (!(child.m_type & ~SpecTypedArrayView)) {
|
|
setConstant(node, jsBoolean(true));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
if (!(child.m_type & SpecTypedArrayView)) {
|
|
setConstant(node, jsBoolean(false));
|
|
constantWasSet = true;
|
|
break;
|
|
}
|
|
break;
|
|
|
|
default:
|
|
break;
|
|
}
|
|
if (constantWasSet)
|
|
break;
|
|
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
|
|
case TypeOf: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
AbstractValue& abstractChild = forNode(node->child1());
|
|
if (child) {
|
|
JSValue typeString = jsTypeStringForValue(m_vm, m_codeBlock->globalObjectFor(node->origin.semantic), child);
|
|
setConstant(node, *m_graph.freeze(typeString));
|
|
break;
|
|
}
|
|
|
|
if (isFullNumberSpeculation(abstractChild.m_type)) {
|
|
setConstant(node, *m_graph.freeze(m_vm.smallStrings.numberString()));
|
|
break;
|
|
}
|
|
|
|
if (isStringSpeculation(abstractChild.m_type)) {
|
|
setConstant(node, *m_graph.freeze(m_vm.smallStrings.stringString()));
|
|
break;
|
|
}
|
|
|
|
// FIXME: We could use the masquerades-as-undefined watchpoint here.
|
|
// https://bugs.webkit.org/show_bug.cgi?id=144456
|
|
if (!(abstractChild.m_type & ~(SpecObject - SpecObjectOther - SpecFunction))) {
|
|
setConstant(node, *m_graph.freeze(m_vm.smallStrings.objectString()));
|
|
break;
|
|
}
|
|
|
|
if (isFunctionSpeculation(abstractChild.m_type)) {
|
|
setConstant(node, *m_graph.freeze(m_vm.smallStrings.functionString()));
|
|
break;
|
|
}
|
|
|
|
if (isBooleanSpeculation(abstractChild.m_type)) {
|
|
setConstant(node, *m_graph.freeze(m_vm.smallStrings.booleanString()));
|
|
break;
|
|
}
|
|
|
|
if (isSymbolSpeculation(abstractChild.m_type)) {
|
|
setConstant(node, *m_graph.freeze(m_vm.smallStrings.symbolString()));
|
|
break;
|
|
}
|
|
|
|
forNode(node).setType(m_graph, SpecStringIdent);
|
|
break;
|
|
}
|
|
|
|
case CompareLess:
|
|
case CompareLessEq:
|
|
case CompareGreater:
|
|
case CompareGreaterEq:
|
|
case CompareEq: {
|
|
JSValue leftConst = forNode(node->child1()).value();
|
|
JSValue rightConst = forNode(node->child2()).value();
|
|
if (leftConst && rightConst) {
|
|
if (leftConst.isNumber() && rightConst.isNumber()) {
|
|
double a = leftConst.asNumber();
|
|
double b = rightConst.asNumber();
|
|
switch (node->op()) {
|
|
case CompareLess:
|
|
setConstant(node, jsBoolean(a < b));
|
|
break;
|
|
case CompareLessEq:
|
|
setConstant(node, jsBoolean(a <= b));
|
|
break;
|
|
case CompareGreater:
|
|
setConstant(node, jsBoolean(a > b));
|
|
break;
|
|
case CompareGreaterEq:
|
|
setConstant(node, jsBoolean(a >= b));
|
|
break;
|
|
case CompareEq:
|
|
setConstant(node, jsBoolean(a == b));
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
if (leftConst.isString() && rightConst.isString()) {
|
|
const StringImpl* a = asString(leftConst)->tryGetValueImpl();
|
|
const StringImpl* b = asString(rightConst)->tryGetValueImpl();
|
|
if (a && b) {
|
|
bool result;
|
|
if (node->op() == CompareEq)
|
|
result = WTF::equal(a, b);
|
|
else if (node->op() == CompareLess)
|
|
result = codePointCompare(a, b) < 0;
|
|
else if (node->op() == CompareLessEq)
|
|
result = codePointCompare(a, b) <= 0;
|
|
else if (node->op() == CompareGreater)
|
|
result = codePointCompare(a, b) > 0;
|
|
else if (node->op() == CompareGreaterEq)
|
|
result = codePointCompare(a, b) >= 0;
|
|
else
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
setConstant(node, jsBoolean(result));
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (node->op() == CompareEq && leftConst.isSymbol() && rightConst.isSymbol()) {
|
|
setConstant(node, jsBoolean(asSymbol(leftConst) == asSymbol(rightConst)));
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (node->op() == CompareEq) {
|
|
SpeculatedType leftType = forNode(node->child1()).m_type;
|
|
SpeculatedType rightType = forNode(node->child2()).m_type;
|
|
if (!valuesCouldBeEqual(leftType, rightType)) {
|
|
setConstant(node, jsBoolean(false));
|
|
break;
|
|
}
|
|
|
|
if (leftType == SpecOther)
|
|
std::swap(leftType, rightType);
|
|
if (rightType == SpecOther) {
|
|
// Undefined and Null are always equal when compared to eachother.
|
|
if (!(leftType & ~SpecOther)) {
|
|
setConstant(node, jsBoolean(true));
|
|
break;
|
|
}
|
|
|
|
// Any other type compared to Null or Undefined is always false
|
|
// as long as the MasqueradesAsUndefined watchpoint is valid.
|
|
//
|
|
// MasqueradesAsUndefined only matters for SpecObjectOther, other
|
|
// cases are always "false".
|
|
if (!(leftType & (SpecObjectOther | SpecOther))) {
|
|
setConstant(node, jsBoolean(false));
|
|
break;
|
|
}
|
|
|
|
if (!(leftType & SpecOther) && m_graph.masqueradesAsUndefinedWatchpointIsStillValid(node->origin.semantic)) {
|
|
JSGlobalObject* globalObject = m_graph.globalObjectFor(node->origin.semantic);
|
|
m_graph.watchpoints().addLazily(globalObject->masqueradesAsUndefinedWatchpoint());
|
|
setConstant(node, jsBoolean(false));
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (node->child1() == node->child2()) {
|
|
if (node->isBinaryUseKind(Int32Use) ||
|
|
node->isBinaryUseKind(Int52RepUse) ||
|
|
node->isBinaryUseKind(StringUse) ||
|
|
node->isBinaryUseKind(BooleanUse) ||
|
|
node->isBinaryUseKind(SymbolUse) ||
|
|
node->isBinaryUseKind(StringIdentUse) ||
|
|
node->isBinaryUseKind(ObjectUse) ||
|
|
node->isBinaryUseKind(ObjectUse, ObjectOrOtherUse) ||
|
|
node->isBinaryUseKind(ObjectOrOtherUse, ObjectUse)) {
|
|
switch (node->op()) {
|
|
case CompareLess:
|
|
case CompareGreater:
|
|
setConstant(node, jsBoolean(false));
|
|
break;
|
|
case CompareLessEq:
|
|
case CompareGreaterEq:
|
|
case CompareEq:
|
|
setConstant(node, jsBoolean(true));
|
|
break;
|
|
default:
|
|
DFG_CRASH(m_graph, node, "Unexpected node type");
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
|
|
case CompareStrictEq: {
|
|
Node* leftNode = node->child1().node();
|
|
Node* rightNode = node->child2().node();
|
|
JSValue left = forNode(leftNode).value();
|
|
JSValue right = forNode(rightNode).value();
|
|
if (left && right) {
|
|
if (left.isString() && right.isString()) {
|
|
// We need this case because JSValue::strictEqual is otherwise too racy for
|
|
// string comparisons.
|
|
const StringImpl* a = asString(left)->tryGetValueImpl();
|
|
const StringImpl* b = asString(right)->tryGetValueImpl();
|
|
if (a && b) {
|
|
setConstant(node, jsBoolean(WTF::equal(a, b)));
|
|
break;
|
|
}
|
|
} else {
|
|
setConstant(node, jsBoolean(JSValue::strictEqual(0, left, right)));
|
|
break;
|
|
}
|
|
}
|
|
|
|
SpeculatedType leftLUB = leastUpperBoundOfStrictlyEquivalentSpeculations(forNode(leftNode).m_type);
|
|
SpeculatedType rightLUB = leastUpperBoundOfStrictlyEquivalentSpeculations(forNode(rightNode).m_type);
|
|
if (!(leftLUB & rightLUB)) {
|
|
setConstant(node, jsBoolean(false));
|
|
break;
|
|
}
|
|
|
|
if (node->child1() == node->child2()) {
|
|
if (node->isBinaryUseKind(BooleanUse) ||
|
|
node->isBinaryUseKind(Int32Use) ||
|
|
node->isBinaryUseKind(Int52RepUse) ||
|
|
node->isBinaryUseKind(StringUse) ||
|
|
node->isBinaryUseKind(StringIdentUse) ||
|
|
node->isBinaryUseKind(SymbolUse) ||
|
|
node->isBinaryUseKind(ObjectUse) ||
|
|
node->isBinaryUseKind(MiscUse, UntypedUse) ||
|
|
node->isBinaryUseKind(UntypedUse, MiscUse) ||
|
|
node->isBinaryUseKind(StringIdentUse, NotStringVarUse) ||
|
|
node->isBinaryUseKind(NotStringVarUse, StringIdentUse) ||
|
|
node->isBinaryUseKind(StringUse, UntypedUse) ||
|
|
node->isBinaryUseKind(UntypedUse, StringUse)) {
|
|
setConstant(node, jsBoolean(true));
|
|
break;
|
|
}
|
|
}
|
|
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
|
|
case CompareEqPtr: {
|
|
Node* childNode = node->child1().node();
|
|
JSValue childValue = forNode(childNode).value();
|
|
if (childValue) {
|
|
setConstant(node, jsBoolean(childValue.isCell() && childValue.asCell() == node->cellOperand()->cell()));
|
|
break;
|
|
}
|
|
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
|
|
case StringCharCodeAt:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
|
|
case StringFromCharCode:
|
|
forNode(node).setType(m_graph, SpecString);
|
|
break;
|
|
|
|
case StringCharAt:
|
|
forNode(node).set(m_graph, m_vm.stringStructure.get());
|
|
break;
|
|
|
|
case GetByVal:
|
|
case AtomicsAdd:
|
|
case AtomicsAnd:
|
|
case AtomicsCompareExchange:
|
|
case AtomicsExchange:
|
|
case AtomicsLoad:
|
|
case AtomicsOr:
|
|
case AtomicsStore:
|
|
case AtomicsSub:
|
|
case AtomicsXor: {
|
|
if (node->op() != GetByVal)
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
switch (node->arrayMode().type()) {
|
|
case Array::SelectUsingPredictions:
|
|
case Array::Unprofiled:
|
|
case Array::SelectUsingArguments:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
case Array::ForceExit:
|
|
m_state.setIsValid(false);
|
|
break;
|
|
case Array::Undecided: {
|
|
JSValue index = forNode(node->child2()).value();
|
|
if (index && index.isInt32() && index.asInt32() >= 0) {
|
|
setConstant(node, jsUndefined());
|
|
break;
|
|
}
|
|
forNode(node).setType(SpecOther);
|
|
break;
|
|
}
|
|
case Array::Generic:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
case Array::String:
|
|
if (node->arrayMode().isOutOfBounds()) {
|
|
// If the watchpoint was still valid we could totally set this to be
|
|
// SpecString | SpecOther. Except that we'd have to be careful. If we
|
|
// tested the watchpoint state here then it could change by the time
|
|
// we got to the backend. So to do this right, we'd have to get the
|
|
// fixup phase to check the watchpoint state and then bake into the
|
|
// GetByVal operation the fact that we're using a watchpoint, using
|
|
// something like Array::SaneChain (except not quite, because that
|
|
// implies an in-bounds access). None of this feels like it's worth it,
|
|
// so we're going with TOP for now. The same thing applies to
|
|
// clobbering the world.
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
} else
|
|
forNode(node).set(m_graph, m_vm.stringStructure.get());
|
|
break;
|
|
case Array::DirectArguments:
|
|
case Array::ScopedArguments:
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
case Array::Int32:
|
|
if (node->arrayMode().isOutOfBounds()) {
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
} else
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Array::Double:
|
|
if (node->arrayMode().isOutOfBounds()) {
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
} else if (node->arrayMode().isSaneChain())
|
|
forNode(node).setType(SpecBytecodeDouble);
|
|
else
|
|
forNode(node).setType(SpecDoubleReal);
|
|
break;
|
|
case Array::Contiguous:
|
|
case Array::ArrayStorage:
|
|
case Array::SlowPutArrayStorage:
|
|
if (node->arrayMode().isOutOfBounds())
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
case Array::Int8Array:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Array::Int16Array:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Array::Int32Array:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Array::Uint8Array:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Array::Uint8ClampedArray:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Array::Uint16Array:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
case Array::Uint32Array:
|
|
if (node->shouldSpeculateInt32())
|
|
forNode(node).setType(SpecInt32Only);
|
|
else if (enableInt52() && node->shouldSpeculateAnyInt())
|
|
forNode(node).setType(SpecAnyInt);
|
|
else
|
|
forNode(node).setType(SpecAnyIntAsDouble);
|
|
break;
|
|
case Array::Float32Array:
|
|
forNode(node).setType(SpecFullDouble);
|
|
break;
|
|
case Array::Float64Array:
|
|
forNode(node).setType(SpecFullDouble);
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case PutByValDirect:
|
|
case PutByVal:
|
|
case PutByValAlias: {
|
|
switch (node->arrayMode().modeForPut().type()) {
|
|
case Array::ForceExit:
|
|
m_state.setIsValid(false);
|
|
break;
|
|
case Array::Generic:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
case Array::Int32:
|
|
if (node->arrayMode().isOutOfBounds())
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
case Array::Double:
|
|
if (node->arrayMode().isOutOfBounds())
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
case Array::Contiguous:
|
|
case Array::ArrayStorage:
|
|
if (node->arrayMode().isOutOfBounds())
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
case Array::SlowPutArrayStorage:
|
|
if (node->arrayMode().mayStoreToHole())
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case ArrayPush:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(SpecBytecodeNumber);
|
|
break;
|
|
|
|
case ArraySlice: {
|
|
JSGlobalObject* globalObject = m_graph.globalObjectFor(node->origin.semantic);
|
|
|
|
// FIXME: We could do better here if we prove that the
|
|
// incoming value has only a single structure.
|
|
RegisteredStructureSet structureSet;
|
|
structureSet.add(m_graph.registerStructure(globalObject->originalArrayStructureForIndexingType(ArrayWithInt32)));
|
|
structureSet.add(m_graph.registerStructure(globalObject->originalArrayStructureForIndexingType(ArrayWithContiguous)));
|
|
structureSet.add(m_graph.registerStructure(globalObject->originalArrayStructureForIndexingType(ArrayWithDouble)));
|
|
|
|
forNode(node).set(m_graph, structureSet);
|
|
break;
|
|
}
|
|
|
|
case ArrayIndexOf: {
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case ArrayPop:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case GetMyArgumentByVal:
|
|
case GetMyArgumentByValOutOfBounds: {
|
|
JSValue index = forNode(node->child2()).m_value;
|
|
InlineCallFrame* inlineCallFrame = node->child1()->origin.semantic.inlineCallFrame;
|
|
|
|
if (index && index.isInt32()) {
|
|
// This pretends to return TOP for accesses that are actually proven out-of-bounds because
|
|
// that's the conservative thing to do. Otherwise we'd need to write more code to mark such
|
|
// paths as unreachable, or to return undefined. We could implement that eventually.
|
|
|
|
unsigned argumentIndex = index.asUInt32() + node->numberOfArgumentsToSkip();
|
|
if (inlineCallFrame) {
|
|
if (argumentIndex < inlineCallFrame->arguments.size() - 1) {
|
|
forNode(node) = m_state.variables().operand(
|
|
virtualRegisterForArgument(argumentIndex + 1) + inlineCallFrame->stackOffset);
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
} else {
|
|
if (argumentIndex < m_state.variables().numberOfArguments() - 1) {
|
|
forNode(node) = m_state.variables().argument(argumentIndex + 1);
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (inlineCallFrame) {
|
|
// We have a bound on the types even though it's random access. Take advantage of this.
|
|
|
|
AbstractValue result;
|
|
for (unsigned i = 1 + node->numberOfArgumentsToSkip(); i < inlineCallFrame->arguments.size(); ++i) {
|
|
result.merge(
|
|
m_state.variables().operand(
|
|
virtualRegisterForArgument(i) + inlineCallFrame->stackOffset));
|
|
}
|
|
|
|
if (node->op() == GetMyArgumentByValOutOfBounds)
|
|
result.merge(SpecOther);
|
|
|
|
if (result.value())
|
|
m_state.setFoundConstants(true);
|
|
|
|
forNode(node) = result;
|
|
break;
|
|
}
|
|
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
}
|
|
|
|
case RegExpExec:
|
|
if (node->child2().useKind() == RegExpObjectUse
|
|
&& node->child3().useKind() == StringUse) {
|
|
// This doesn't clobber the world since there are no conversions to perform.
|
|
} else
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
if (JSValue globalObjectValue = forNode(node->child1()).m_value) {
|
|
if (JSGlobalObject* globalObject = jsDynamicCast<JSGlobalObject*>(m_vm, globalObjectValue)) {
|
|
if (!globalObject->isHavingABadTime()) {
|
|
m_graph.watchpoints().addLazily(globalObject->havingABadTimeWatchpoint());
|
|
Structure* structure = globalObject->regExpMatchesArrayStructure();
|
|
m_graph.registerStructure(structure);
|
|
forNode(node).set(m_graph, structure);
|
|
forNode(node).merge(SpecOther);
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
forNode(node).setType(m_graph, SpecOther | SpecArray);
|
|
break;
|
|
|
|
case RegExpTest:
|
|
if (node->child2().useKind() == RegExpObjectUse
|
|
&& node->child3().useKind() == StringUse) {
|
|
// This doesn't clobber the world since there are no conversions to perform.
|
|
} else
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
|
|
case StringReplace:
|
|
case StringReplaceRegExp:
|
|
if (node->child1().useKind() == StringUse
|
|
&& node->child2().useKind() == RegExpObjectUse
|
|
&& node->child3().useKind() == StringUse) {
|
|
// This doesn't clobber the world. It just reads and writes regexp state.
|
|
} else
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).set(m_graph, m_vm.stringStructure.get());
|
|
break;
|
|
|
|
case Jump:
|
|
break;
|
|
|
|
case Branch: {
|
|
Node* child = node->child1().node();
|
|
BooleanResult result = booleanResult(node, forNode(child));
|
|
if (result == DefinitelyTrue) {
|
|
m_state.setBranchDirection(TakeTrue);
|
|
break;
|
|
}
|
|
if (result == DefinitelyFalse) {
|
|
m_state.setBranchDirection(TakeFalse);
|
|
break;
|
|
}
|
|
// FIXME: The above handles the trivial cases of sparse conditional
|
|
// constant propagation, but we can do better:
|
|
// We can specialize the source variable's value on each direction of
|
|
// the branch.
|
|
m_state.setBranchDirection(TakeBoth);
|
|
break;
|
|
}
|
|
|
|
case Switch: {
|
|
// Nothing to do for now.
|
|
// FIXME: Do sparse conditional things.
|
|
break;
|
|
}
|
|
|
|
case Return:
|
|
m_state.setIsValid(false);
|
|
break;
|
|
|
|
case TailCall:
|
|
case DirectTailCall:
|
|
case TailCallVarargs:
|
|
case TailCallForwardVarargs:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
m_state.setIsValid(false);
|
|
break;
|
|
|
|
case Throw:
|
|
case ThrowStaticError:
|
|
m_state.setIsValid(false);
|
|
break;
|
|
|
|
case ToPrimitive: {
|
|
JSValue childConst = forNode(node->child1()).value();
|
|
if (childConst && childConst.isNumber()) {
|
|
setConstant(node, childConst);
|
|
break;
|
|
}
|
|
|
|
ASSERT(node->child1().useKind() == UntypedUse);
|
|
|
|
if (!(forNode(node->child1()).m_type & ~(SpecFullNumber | SpecBoolean | SpecString | SpecSymbol))) {
|
|
m_state.setFoundConstants(true);
|
|
forNode(node) = forNode(node->child1());
|
|
break;
|
|
}
|
|
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
|
|
forNode(node).setType(m_graph, SpecHeapTop & ~SpecObject);
|
|
break;
|
|
}
|
|
|
|
case ToNumber: {
|
|
JSValue childConst = forNode(node->child1()).value();
|
|
if (childConst && childConst.isNumber()) {
|
|
setConstant(node, childConst);
|
|
break;
|
|
}
|
|
|
|
ASSERT(node->child1().useKind() == UntypedUse);
|
|
|
|
if (!(forNode(node->child1()).m_type & ~SpecBytecodeNumber)) {
|
|
m_state.setFoundConstants(true);
|
|
forNode(node) = forNode(node->child1());
|
|
break;
|
|
}
|
|
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecBytecodeNumber);
|
|
break;
|
|
}
|
|
|
|
case ToString:
|
|
case CallStringConstructor: {
|
|
switch (node->child1().useKind()) {
|
|
case StringObjectUse:
|
|
// This also filters that the StringObject has the primordial StringObject
|
|
// structure.
|
|
filter(
|
|
node->child1(),
|
|
m_graph.registerStructure(m_graph.globalObjectFor(node->origin.semantic)->stringObjectStructure()));
|
|
break;
|
|
case StringOrStringObjectUse:
|
|
case Int32Use:
|
|
case Int52RepUse:
|
|
case DoubleRepUse:
|
|
case NotCellUse:
|
|
break;
|
|
case CellUse:
|
|
case UntypedUse:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
forNode(node).set(m_graph, m_vm.stringStructure.get());
|
|
break;
|
|
}
|
|
|
|
case NumberToStringWithRadix:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).set(m_graph, m_graph.m_vm.stringStructure.get());
|
|
break;
|
|
|
|
case NewStringObject: {
|
|
ASSERT(node->structure()->classInfo() == StringObject::info());
|
|
forNode(node).set(m_graph, node->structure());
|
|
break;
|
|
}
|
|
|
|
case NewArray:
|
|
forNode(node).set(
|
|
m_graph,
|
|
m_graph.globalObjectFor(node->origin.semantic)->arrayStructureForIndexingTypeDuringAllocation(node->indexingType()));
|
|
break;
|
|
|
|
case NewArrayWithSpread:
|
|
if (m_graph.isWatchingHavingABadTimeWatchpoint(node)) {
|
|
// We've compiled assuming we're not having a bad time, so to be consistent
|
|
// with StructureRegisterationPhase we must say we produce an original array
|
|
// allocation structure.
|
|
forNode(node).set(
|
|
m_graph,
|
|
m_graph.globalObjectFor(node->origin.semantic)->originalArrayStructureForIndexingType(ArrayWithContiguous));
|
|
} else {
|
|
forNode(node).set(
|
|
m_graph,
|
|
m_graph.globalObjectFor(node->origin.semantic)->arrayStructureForIndexingTypeDuringAllocation(ArrayWithContiguous));
|
|
}
|
|
|
|
break;
|
|
|
|
case Spread:
|
|
forNode(node).set(
|
|
m_graph, m_vm.fixedArrayStructure.get());
|
|
break;
|
|
|
|
case NewArrayBuffer:
|
|
forNode(node).set(
|
|
m_graph,
|
|
m_graph.globalObjectFor(node->origin.semantic)->arrayStructureForIndexingTypeDuringAllocation(node->indexingType()));
|
|
break;
|
|
|
|
case NewArrayWithSize:
|
|
forNode(node).setType(m_graph, SpecArray);
|
|
break;
|
|
|
|
case NewTypedArray:
|
|
switch (node->child1().useKind()) {
|
|
case Int32Use:
|
|
break;
|
|
case UntypedUse:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
forNode(node).set(
|
|
m_graph,
|
|
m_graph.globalObjectFor(node->origin.semantic)->typedArrayStructureConcurrently(
|
|
node->typedArrayType()));
|
|
break;
|
|
|
|
case NewRegexp:
|
|
forNode(node).set(m_graph, m_graph.globalObjectFor(node->origin.semantic)->regExpStructure());
|
|
break;
|
|
|
|
case ToThis: {
|
|
AbstractValue& source = forNode(node->child1());
|
|
AbstractValue& destination = forNode(node);
|
|
bool strictMode = m_graph.executableFor(node->origin.semantic)->isStrictMode();
|
|
|
|
if (isToThisAnIdentity(strictMode, source)) {
|
|
m_state.setFoundConstants(true);
|
|
destination = source;
|
|
break;
|
|
}
|
|
|
|
if (strictMode)
|
|
destination.makeHeapTop();
|
|
else {
|
|
destination = source;
|
|
destination.merge(SpecObject);
|
|
}
|
|
break;
|
|
}
|
|
|
|
case CreateThis: {
|
|
// FIXME: We can fold this to NewObject if the incoming callee is a constant.
|
|
forNode(node).setType(m_graph, SpecFinalObject);
|
|
break;
|
|
}
|
|
|
|
case NewObject:
|
|
ASSERT(!!node->structure().get());
|
|
forNode(node).set(m_graph, node->structure());
|
|
break;
|
|
|
|
case CallObjectConstructor: {
|
|
AbstractValue& source = forNode(node->child1());
|
|
AbstractValue& destination = forNode(node);
|
|
|
|
if (!(source.m_type & ~SpecObject)) {
|
|
m_state.setFoundConstants(true);
|
|
destination = source;
|
|
break;
|
|
}
|
|
|
|
forNode(node).setType(m_graph, SpecObject);
|
|
break;
|
|
}
|
|
|
|
case PhantomNewObject:
|
|
case PhantomNewFunction:
|
|
case PhantomNewGeneratorFunction:
|
|
case PhantomNewAsyncFunction:
|
|
case PhantomCreateActivation:
|
|
case PhantomDirectArguments:
|
|
case PhantomClonedArguments:
|
|
case PhantomCreateRest:
|
|
case PhantomSpread:
|
|
case PhantomNewArrayWithSpread:
|
|
case BottomValue:
|
|
m_state.setDidClobber(true); // Prevent constant folding.
|
|
// This claims to return bottom.
|
|
break;
|
|
|
|
case PutHint:
|
|
break;
|
|
|
|
case MaterializeNewObject: {
|
|
forNode(node).set(m_graph, node->structureSet());
|
|
break;
|
|
}
|
|
|
|
case CreateActivation:
|
|
case MaterializeCreateActivation:
|
|
forNode(node).set(
|
|
m_graph, m_codeBlock->globalObjectFor(node->origin.semantic)->activationStructure());
|
|
break;
|
|
|
|
case CreateDirectArguments:
|
|
forNode(node).set(m_graph, m_codeBlock->globalObjectFor(node->origin.semantic)->directArgumentsStructure());
|
|
break;
|
|
|
|
case CreateScopedArguments:
|
|
forNode(node).set(m_graph, m_codeBlock->globalObjectFor(node->origin.semantic)->scopedArgumentsStructure());
|
|
break;
|
|
|
|
case CreateClonedArguments:
|
|
if (!m_graph.isWatchingHavingABadTimeWatchpoint(node)) {
|
|
forNode(node).setType(m_graph, SpecObject);
|
|
break;
|
|
}
|
|
forNode(node).set(m_graph, m_codeBlock->globalObjectFor(node->origin.semantic)->clonedArgumentsStructure());
|
|
break;
|
|
|
|
case NewGeneratorFunction:
|
|
forNode(node).set(
|
|
m_graph, m_codeBlock->globalObjectFor(node->origin.semantic)->generatorFunctionStructure());
|
|
break;
|
|
|
|
case NewAsyncFunction:
|
|
forNode(node).set(
|
|
m_graph, m_codeBlock->globalObjectFor(node->origin.semantic)->asyncFunctionStructure());
|
|
break;
|
|
|
|
case NewFunction:
|
|
forNode(node).set(
|
|
m_graph, m_codeBlock->globalObjectFor(node->origin.semantic)->functionStructure());
|
|
break;
|
|
|
|
case GetCallee:
|
|
if (FunctionExecutable* executable = jsDynamicCast<FunctionExecutable*>(m_vm, m_codeBlock->ownerExecutable())) {
|
|
InferredValue* singleton = executable->singletonFunction();
|
|
if (JSValue value = singleton->inferredValue()) {
|
|
m_graph.watchpoints().addLazily(singleton);
|
|
JSFunction* function = jsCast<JSFunction*>(value);
|
|
setConstant(node, *m_graph.freeze(function));
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(m_graph, SpecFunction);
|
|
break;
|
|
|
|
case GetArgumentCountIncludingThis:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
|
|
case GetRestLength:
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
|
|
case GetGetter: {
|
|
JSValue base = forNode(node->child1()).m_value;
|
|
if (base) {
|
|
GetterSetter* getterSetter = jsCast<GetterSetter*>(base);
|
|
if (!getterSetter->isGetterNull()) {
|
|
setConstant(node, *m_graph.freeze(getterSetter->getterConcurrently()));
|
|
break;
|
|
}
|
|
}
|
|
|
|
forNode(node).setType(m_graph, SpecObject);
|
|
break;
|
|
}
|
|
|
|
case GetSetter: {
|
|
JSValue base = forNode(node->child1()).m_value;
|
|
if (base) {
|
|
GetterSetter* getterSetter = jsCast<GetterSetter*>(base);
|
|
if (!getterSetter->isSetterNull()) {
|
|
setConstant(node, *m_graph.freeze(getterSetter->setterConcurrently()));
|
|
break;
|
|
}
|
|
}
|
|
|
|
forNode(node).setType(m_graph, SpecObject);
|
|
break;
|
|
}
|
|
|
|
case GetScope:
|
|
if (JSValue base = forNode(node->child1()).m_value) {
|
|
if (JSFunction* function = jsDynamicCast<JSFunction*>(m_vm, base)) {
|
|
setConstant(node, *m_graph.freeze(function->scope()));
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(m_graph, SpecObjectOther);
|
|
break;
|
|
|
|
case SkipScope: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
if (child) {
|
|
setConstant(node, *m_graph.freeze(JSValue(jsCast<JSScope*>(child.asCell())->next())));
|
|
break;
|
|
}
|
|
forNode(node).setType(m_graph, SpecObjectOther);
|
|
break;
|
|
}
|
|
|
|
case GetGlobalObject: {
|
|
JSValue child = forNode(node->child1()).value();
|
|
if (child) {
|
|
setConstant(node, *m_graph.freeze(JSValue(asObject(child)->globalObject())));
|
|
break;
|
|
}
|
|
|
|
if (forNode(node->child1()).m_structure.isFinite()) {
|
|
JSGlobalObject* globalObject = nullptr;
|
|
bool ok = true;
|
|
forNode(node->child1()).m_structure.forEach(
|
|
[&] (RegisteredStructure structure) {
|
|
if (!globalObject)
|
|
globalObject = structure->globalObject();
|
|
else if (globalObject != structure->globalObject())
|
|
ok = false;
|
|
});
|
|
if (globalObject && ok) {
|
|
setConstant(node, *m_graph.freeze(JSValue(globalObject)));
|
|
break;
|
|
}
|
|
}
|
|
|
|
forNode(node).setType(m_graph, SpecObjectOther);
|
|
break;
|
|
}
|
|
|
|
case GetClosureVar:
|
|
if (JSValue value = m_graph.tryGetConstantClosureVar(forNode(node->child1()), node->scopeOffset())) {
|
|
setConstant(node, *m_graph.freeze(value));
|
|
break;
|
|
}
|
|
forNode(node).makeBytecodeTop();
|
|
break;
|
|
|
|
case PutClosureVar:
|
|
break;
|
|
|
|
case GetRegExpObjectLastIndex:
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case SetRegExpObjectLastIndex:
|
|
case RecordRegExpCachedResult:
|
|
break;
|
|
|
|
case GetFromArguments:
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case PutToArguments:
|
|
break;
|
|
|
|
case GetArgument:
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case TryGetById:
|
|
// FIXME: This should constant fold at least as well as the normal GetById case.
|
|
// https://bugs.webkit.org/show_bug.cgi?id=156422
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case GetById:
|
|
case GetByIdFlush: {
|
|
if (!node->prediction()) {
|
|
m_state.setIsValid(false);
|
|
break;
|
|
}
|
|
|
|
AbstractValue& value = forNode(node->child1());
|
|
if (value.m_structure.isFinite()
|
|
&& (node->child1().useKind() == CellUse || !(value.m_type & ~SpecCell))) {
|
|
UniquedStringImpl* uid = m_graph.identifiers()[node->identifierNumber()];
|
|
GetByIdStatus status = GetByIdStatus::computeFor(value.m_structure.toStructureSet(), uid);
|
|
if (status.isSimple()) {
|
|
// Figure out what the result is going to be - is it TOP, a constant, or maybe
|
|
// something more subtle?
|
|
AbstractValue result;
|
|
for (unsigned i = status.numVariants(); i--;) {
|
|
// This thing won't give us a variant that involves prototypes. If it did, we'd
|
|
// have more work to do here.
|
|
DFG_ASSERT(m_graph, node, status[i].conditionSet().isEmpty());
|
|
|
|
result.merge(
|
|
m_graph.inferredValueForProperty(
|
|
value, uid, status[i].offset(), m_state.structureClobberState()));
|
|
}
|
|
m_state.setFoundConstants(true);
|
|
forNode(node) = result;
|
|
break;
|
|
}
|
|
}
|
|
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
}
|
|
|
|
case GetByValWithThis:
|
|
case GetByIdWithThis:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case GetArrayLength: {
|
|
JSArrayBufferView* view = m_graph.tryGetFoldableView(
|
|
forNode(node->child1()).m_value, node->arrayMode());
|
|
if (view) {
|
|
setConstant(node, jsNumber(view->length()));
|
|
break;
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case GetVectorLength: {
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case DeleteById:
|
|
case DeleteByVal: {
|
|
// FIXME: This could decide if the delete will be successful based on the set of structures that
|
|
// we get from our base value. https://bugs.webkit.org/show_bug.cgi?id=156611
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
|
|
case CheckStructure: {
|
|
AbstractValue& value = forNode(node->child1());
|
|
|
|
const RegisteredStructureSet& set = node->structureSet();
|
|
|
|
// It's interesting that we could have proven that the object has a larger structure set
|
|
// that includes the set we're testing. In that case we could make the structure check
|
|
// more efficient. We currently don't.
|
|
|
|
if (value.m_structure.isSubsetOf(set))
|
|
m_state.setFoundConstants(true);
|
|
|
|
SpeculatedType admittedTypes = SpecNone;
|
|
switch (node->child1().useKind()) {
|
|
case CellUse:
|
|
case KnownCellUse:
|
|
admittedTypes = SpecNone;
|
|
break;
|
|
case CellOrOtherUse:
|
|
admittedTypes = SpecOther;
|
|
break;
|
|
default:
|
|
DFG_CRASH(m_graph, node, "Bad use kind");
|
|
break;
|
|
}
|
|
|
|
filter(value, set, admittedTypes);
|
|
break;
|
|
}
|
|
|
|
case CheckStructureImmediate: {
|
|
// FIXME: This currently can only reason about one structure at a time.
|
|
// https://bugs.webkit.org/show_bug.cgi?id=136988
|
|
|
|
AbstractValue& value = forNode(node->child1());
|
|
const RegisteredStructureSet& set = node->structureSet();
|
|
|
|
if (value.value()) {
|
|
if (Structure* structure = jsDynamicCast<Structure*>(m_vm, value.value())) {
|
|
if (set.contains(m_graph.registerStructure(structure))) {
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
}
|
|
m_state.setIsValid(false);
|
|
break;
|
|
}
|
|
|
|
if (m_phiChildren) {
|
|
bool allGood = true;
|
|
m_phiChildren->forAllTransitiveIncomingValues(
|
|
node,
|
|
[&] (Node* incoming) {
|
|
if (Structure* structure = incoming->dynamicCastConstant<Structure*>(m_vm)) {
|
|
if (set.contains(m_graph.registerStructure(structure)))
|
|
return;
|
|
}
|
|
allGood = false;
|
|
});
|
|
if (allGood) {
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (RegisteredStructure structure = set.onlyStructure()) {
|
|
filterByValue(node->child1(), *m_graph.freeze(structure.get()));
|
|
break;
|
|
}
|
|
|
|
// Aw shucks, we can't do anything!
|
|
break;
|
|
}
|
|
|
|
case PutStructure:
|
|
if (!forNode(node->child1()).m_structure.isClear()) {
|
|
if (forNode(node->child1()).m_structure.onlyStructure() == node->transition()->next)
|
|
m_state.setFoundConstants(true);
|
|
else {
|
|
observeTransition(
|
|
clobberLimit, node->transition()->previous, node->transition()->next);
|
|
forNode(node->child1()).changeStructure(m_graph, node->transition()->next);
|
|
}
|
|
}
|
|
break;
|
|
case GetButterfly:
|
|
case AllocatePropertyStorage:
|
|
case ReallocatePropertyStorage:
|
|
case NukeStructureAndSetButterfly:
|
|
// FIXME: We don't model the fact that the structureID is nuked, simply because currently
|
|
// nobody would currently benefit from having that information. But it's a bug nonetheless.
|
|
forNode(node).clear(); // The result is not a JS value.
|
|
break;
|
|
case CheckSubClass: {
|
|
JSValue constant = forNode(node->child1()).value();
|
|
if (constant) {
|
|
if (constant.isCell() && constant.asCell()->inherits(m_vm, node->classInfo())) {
|
|
m_state.setFoundConstants(true);
|
|
ASSERT(constant);
|
|
break;
|
|
}
|
|
}
|
|
|
|
AbstractValue& value = forNode(node->child1());
|
|
|
|
if (value.m_structure.isSubClassOf(node->classInfo()))
|
|
m_state.setFoundConstants(true);
|
|
|
|
filterClassInfo(value, node->classInfo());
|
|
break;
|
|
}
|
|
case CallDOMGetter: {
|
|
CallDOMGetterData* callDOMGetterData = node->callDOMGetterData();
|
|
DOMJIT::CallDOMGetterSnippet* snippet = callDOMGetterData->snippet;
|
|
if (snippet->effect.writes)
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, callDOMGetterData->domJIT->resultType());
|
|
break;
|
|
}
|
|
case CallDOM: {
|
|
const DOMJIT::Signature* signature = node->signature();
|
|
if (signature->effect.writes)
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, signature->result);
|
|
break;
|
|
}
|
|
case CheckArray: {
|
|
if (node->arrayMode().alreadyChecked(m_graph, node, forNode(node->child1()))) {
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
switch (node->arrayMode().type()) {
|
|
case Array::String:
|
|
filter(node->child1(), SpecString);
|
|
break;
|
|
case Array::Int32:
|
|
case Array::Double:
|
|
case Array::Contiguous:
|
|
case Array::Undecided:
|
|
case Array::ArrayStorage:
|
|
case Array::SlowPutArrayStorage:
|
|
break;
|
|
case Array::DirectArguments:
|
|
filter(node->child1(), SpecDirectArguments);
|
|
break;
|
|
case Array::ScopedArguments:
|
|
filter(node->child1(), SpecScopedArguments);
|
|
break;
|
|
case Array::Int8Array:
|
|
filter(node->child1(), SpecInt8Array);
|
|
break;
|
|
case Array::Int16Array:
|
|
filter(node->child1(), SpecInt16Array);
|
|
break;
|
|
case Array::Int32Array:
|
|
filter(node->child1(), SpecInt32Array);
|
|
break;
|
|
case Array::Uint8Array:
|
|
filter(node->child1(), SpecUint8Array);
|
|
break;
|
|
case Array::Uint8ClampedArray:
|
|
filter(node->child1(), SpecUint8ClampedArray);
|
|
break;
|
|
case Array::Uint16Array:
|
|
filter(node->child1(), SpecUint16Array);
|
|
break;
|
|
case Array::Uint32Array:
|
|
filter(node->child1(), SpecUint32Array);
|
|
break;
|
|
case Array::Float32Array:
|
|
filter(node->child1(), SpecFloat32Array);
|
|
break;
|
|
case Array::Float64Array:
|
|
filter(node->child1(), SpecFloat64Array);
|
|
break;
|
|
case Array::AnyTypedArray:
|
|
filter(node->child1(), SpecTypedArrayView);
|
|
break;
|
|
default:
|
|
RELEASE_ASSERT_NOT_REACHED();
|
|
break;
|
|
}
|
|
filterArrayModes(node->child1(), node->arrayMode().arrayModesThatPassFiltering());
|
|
break;
|
|
}
|
|
case Arrayify: {
|
|
if (node->arrayMode().alreadyChecked(m_graph, node, forNode(node->child1()))) {
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
ASSERT(node->arrayMode().conversion() == Array::Convert);
|
|
clobberStructures(clobberLimit);
|
|
filterArrayModes(node->child1(), node->arrayMode().arrayModesThatPassFiltering());
|
|
break;
|
|
}
|
|
case ArrayifyToStructure: {
|
|
AbstractValue& value = forNode(node->child1());
|
|
if (value.m_structure.isSubsetOf(RegisteredStructureSet(node->structure())))
|
|
m_state.setFoundConstants(true);
|
|
clobberStructures(clobberLimit);
|
|
|
|
// We have a bunch of options of how to express the abstract set at this point. Let set S
|
|
// be the set of structures that the value had before clobbering and assume that all of
|
|
// them are watchable. The new value should be the least expressible upper bound of the
|
|
// intersection of "values that currently have structure = node->structure()" and "values
|
|
// that have structure in S plus any structure transition-reachable from S". Assume that
|
|
// node->structure() is not in S but it is transition-reachable from S. Then we would
|
|
// like to say that the result is "values that have structure = node->structure() until
|
|
// we invalidate", but there is no way to express this using the AbstractValue syntax. So
|
|
// we must choose between:
|
|
//
|
|
// 1) "values that currently have structure = node->structure()". This is a valid
|
|
// superset of the value that we really want, and it's specific enough to satisfy the
|
|
// preconditions of the array access that this is guarding. It's also specific enough
|
|
// to allow relevant optimizations in the case that we didn't have a contradiction
|
|
// like in this example. Notice that in the abscence of any contradiction, this result
|
|
// is precise rather than being a conservative LUB.
|
|
//
|
|
// 2) "values that currently hava structure in S plus any structure transition-reachable
|
|
// from S". This is also a valid superset of the value that we really want, but it's
|
|
// not specific enough to satisfy the preconditions of the array access that this is
|
|
// guarding - so playing such shenanigans would preclude us from having assertions on
|
|
// the typing preconditions of any array accesses. This would also not be a desirable
|
|
// answer in the absence of a contradiction.
|
|
//
|
|
// Note that it's tempting to simply say that the resulting value is BOTTOM because of
|
|
// the contradiction. That would be wrong, since we haven't hit an invalidation point,
|
|
// yet.
|
|
value.set(m_graph, node->structure());
|
|
break;
|
|
}
|
|
case GetIndexedPropertyStorage: {
|
|
JSArrayBufferView* view = m_graph.tryGetFoldableView(
|
|
forNode(node->child1()).m_value, node->arrayMode());
|
|
if (view)
|
|
m_state.setFoundConstants(true);
|
|
forNode(node).clear();
|
|
break;
|
|
}
|
|
case ConstantStoragePointer: {
|
|
forNode(node).clear();
|
|
break;
|
|
}
|
|
|
|
case GetTypedArrayByteOffset: {
|
|
JSArrayBufferView* view = m_graph.tryGetFoldableView(forNode(node->child1()).m_value);
|
|
if (view) {
|
|
setConstant(node, jsNumber(view->byteOffset()));
|
|
break;
|
|
}
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
|
|
case GetByOffset: {
|
|
StorageAccessData& data = node->storageAccessData();
|
|
UniquedStringImpl* uid = m_graph.identifiers()[data.identifierNumber];
|
|
|
|
// FIXME: The part of this that handles inferred property types relies on AI knowing the structure
|
|
// right now. That's probably not optimal. In some cases, we may perform an optimization (usually
|
|
// by something other than AI, maybe by CSE for example) that obscures AI's view of the structure
|
|
// at the point where GetByOffset runs. Currently, when that happens, we'll have to rely entirely
|
|
// on the type that ByteCodeParser was able to prove.
|
|
AbstractValue value = m_graph.inferredValueForProperty(
|
|
forNode(node->child2()), uid, data.offset, m_state.structureClobberState());
|
|
|
|
// It's possible that the type that ByteCodeParser came up with is better.
|
|
AbstractValue typeFromParsing;
|
|
typeFromParsing.set(m_graph, data.inferredType, m_state.structureClobberState());
|
|
value.filter(typeFromParsing);
|
|
|
|
// If we decide that there does not exist any value that this can return, then it's probably
|
|
// because the compilation was already invalidated.
|
|
if (value.isClear())
|
|
m_state.setIsValid(false);
|
|
|
|
forNode(node) = value;
|
|
if (value.m_value)
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
|
|
case GetGetterSetterByOffset: {
|
|
StorageAccessData& data = node->storageAccessData();
|
|
JSValue result = m_graph.tryGetConstantProperty(forNode(node->child2()), data.offset);
|
|
if (result && jsDynamicCast<GetterSetter*>(m_vm, result)) {
|
|
setConstant(node, *m_graph.freeze(result));
|
|
break;
|
|
}
|
|
|
|
forNode(node).set(m_graph, m_graph.globalObjectFor(node->origin.semantic)->getterSetterStructure());
|
|
break;
|
|
}
|
|
|
|
case MultiGetByOffset: {
|
|
// This code will filter the base value in a manner that is possibly different (either more
|
|
// or less precise) than the way it would be filtered if this was strength-reduced to a
|
|
// CheckStructure. This is fine. It's legal for different passes over the code to prove
|
|
// different things about the code, so long as all of them are sound. That even includes
|
|
// one guy proving that code should never execute (due to a contradiction) and another guy
|
|
// not finding that contradiction. If someone ever proved that there would be a
|
|
// contradiction then there must always be a contradiction even if subsequent passes don't
|
|
// realize it. This is the case here.
|
|
|
|
// Ordinarily you have to be careful with calling setFoundConstants()
|
|
// because of the effect on compile times, but this node is FTL-only.
|
|
m_state.setFoundConstants(true);
|
|
|
|
UniquedStringImpl* uid = m_graph.identifiers()[node->multiGetByOffsetData().identifierNumber];
|
|
|
|
AbstractValue base = forNode(node->child1());
|
|
RegisteredStructureSet baseSet;
|
|
AbstractValue result;
|
|
for (const MultiGetByOffsetCase& getCase : node->multiGetByOffsetData().cases) {
|
|
RegisteredStructureSet set = getCase.set();
|
|
set.filter(base);
|
|
if (set.isEmpty())
|
|
continue;
|
|
baseSet.merge(set);
|
|
|
|
switch (getCase.method().kind()) {
|
|
case GetByOffsetMethod::Constant: {
|
|
AbstractValue thisResult;
|
|
thisResult.set(
|
|
m_graph,
|
|
*getCase.method().constant(),
|
|
m_state.structureClobberState());
|
|
result.merge(thisResult);
|
|
break;
|
|
}
|
|
|
|
case GetByOffsetMethod::Load: {
|
|
result.merge(
|
|
m_graph.inferredValueForProperty(
|
|
set, uid, m_state.structureClobberState()));
|
|
break;
|
|
}
|
|
|
|
default: {
|
|
result.makeHeapTop();
|
|
break;
|
|
} }
|
|
}
|
|
|
|
if (forNode(node->child1()).changeStructure(m_graph, baseSet) == Contradiction)
|
|
m_state.setIsValid(false);
|
|
|
|
forNode(node) = result;
|
|
break;
|
|
}
|
|
|
|
case PutByOffset: {
|
|
break;
|
|
}
|
|
|
|
case MultiPutByOffset: {
|
|
RegisteredStructureSet newSet;
|
|
TransitionVector transitions;
|
|
|
|
// Ordinarily you have to be careful with calling setFoundConstants()
|
|
// because of the effect on compile times, but this node is FTL-only.
|
|
m_state.setFoundConstants(true);
|
|
|
|
AbstractValue base = forNode(node->child1());
|
|
AbstractValue originalValue = forNode(node->child2());
|
|
AbstractValue resultingValue;
|
|
|
|
for (unsigned i = node->multiPutByOffsetData().variants.size(); i--;) {
|
|
const PutByIdVariant& variant = node->multiPutByOffsetData().variants[i];
|
|
RegisteredStructureSet thisSet = *m_graph.addStructureSet(variant.oldStructure());
|
|
thisSet.filter(base);
|
|
if (thisSet.isEmpty())
|
|
continue;
|
|
|
|
AbstractValue thisValue = originalValue;
|
|
thisValue.filter(m_graph, variant.requiredType());
|
|
resultingValue.merge(thisValue);
|
|
|
|
if (variant.kind() == PutByIdVariant::Transition) {
|
|
RegisteredStructure newStructure = m_graph.registerStructure(variant.newStructure());
|
|
if (thisSet.onlyStructure() != newStructure) {
|
|
transitions.append(
|
|
Transition(m_graph.registerStructure(variant.oldStructureForTransition()), newStructure));
|
|
} // else this is really a replace.
|
|
newSet.add(newStructure);
|
|
} else {
|
|
ASSERT(variant.kind() == PutByIdVariant::Replace);
|
|
newSet.merge(thisSet);
|
|
}
|
|
}
|
|
|
|
observeTransitions(clobberLimit, transitions);
|
|
if (forNode(node->child1()).changeStructure(m_graph, newSet) == Contradiction)
|
|
m_state.setIsValid(false);
|
|
forNode(node->child2()) = resultingValue;
|
|
if (!!originalValue && !resultingValue)
|
|
m_state.setIsValid(false);
|
|
break;
|
|
}
|
|
|
|
case GetExecutable: {
|
|
JSValue value = forNode(node->child1()).value();
|
|
if (value) {
|
|
JSFunction* function = jsDynamicCast<JSFunction*>(m_vm, value);
|
|
if (function) {
|
|
setConstant(node, *m_graph.freeze(function->executable()));
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(m_graph, SpecCellOther);
|
|
break;
|
|
}
|
|
|
|
case CheckCell: {
|
|
JSValue value = forNode(node->child1()).value();
|
|
if (value == node->cellOperand()->value()) {
|
|
m_state.setFoundConstants(true);
|
|
ASSERT(value);
|
|
break;
|
|
}
|
|
filterByValue(node->child1(), *node->cellOperand());
|
|
break;
|
|
}
|
|
|
|
case CheckNotEmpty: {
|
|
AbstractValue& value = forNode(node->child1());
|
|
if (!(value.m_type & SpecEmpty)) {
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
|
|
filter(value, ~SpecEmpty);
|
|
break;
|
|
}
|
|
|
|
case CheckStringIdent: {
|
|
AbstractValue& value = forNode(node->child1());
|
|
UniquedStringImpl* uid = node->uidOperand();
|
|
ASSERT(!(value.m_type & ~SpecStringIdent)); // Edge filtering should have already ensured this.
|
|
|
|
JSValue childConstant = value.value();
|
|
if (childConstant) {
|
|
ASSERT(childConstant.isString());
|
|
if (asString(childConstant)->tryGetValueImpl() == uid) {
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
}
|
|
|
|
filter(value, SpecStringIdent);
|
|
break;
|
|
}
|
|
|
|
case CheckInBounds: {
|
|
JSValue left = forNode(node->child1()).value();
|
|
JSValue right = forNode(node->child2()).value();
|
|
if (left && right && left.isInt32() && right.isInt32()
|
|
&& static_cast<uint32_t>(left.asInt32()) < static_cast<uint32_t>(right.asInt32())) {
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
case PutById:
|
|
case PutByIdFlush:
|
|
case PutByIdDirect: {
|
|
AbstractValue& value = forNode(node->child1());
|
|
if (value.m_structure.isFinite()) {
|
|
PutByIdStatus status = PutByIdStatus::computeFor(
|
|
m_graph.globalObjectFor(node->origin.semantic),
|
|
value.m_structure.toStructureSet(),
|
|
m_graph.identifiers()[node->identifierNumber()],
|
|
node->op() == PutByIdDirect);
|
|
|
|
if (status.isSimple()) {
|
|
RegisteredStructureSet newSet;
|
|
TransitionVector transitions;
|
|
|
|
for (unsigned i = status.numVariants(); i--;) {
|
|
const PutByIdVariant& variant = status[i];
|
|
if (variant.kind() == PutByIdVariant::Transition) {
|
|
RegisteredStructure newStructure = m_graph.registerStructure(variant.newStructure());
|
|
transitions.append(
|
|
Transition(
|
|
m_graph.registerStructure(variant.oldStructureForTransition()), newStructure));
|
|
newSet.add(newStructure);
|
|
} else {
|
|
ASSERT(variant.kind() == PutByIdVariant::Replace);
|
|
newSet.merge(*m_graph.addStructureSet(variant.oldStructure()));
|
|
}
|
|
}
|
|
|
|
if (status.numVariants() == 1 || isFTL(m_graph.m_plan.mode))
|
|
m_state.setFoundConstants(true);
|
|
|
|
observeTransitions(clobberLimit, transitions);
|
|
if (forNode(node->child1()).changeStructure(m_graph, newSet) == Contradiction)
|
|
m_state.setIsValid(false);
|
|
break;
|
|
}
|
|
}
|
|
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
}
|
|
|
|
case PutByValWithThis:
|
|
case PutByIdWithThis:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
|
|
case PutGetterById:
|
|
case PutSetterById:
|
|
case PutGetterSetterById:
|
|
case PutGetterByVal:
|
|
case PutSetterByVal: {
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
}
|
|
|
|
case DefineDataProperty:
|
|
case DefineAccessorProperty:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
|
|
case In: {
|
|
// FIXME: We can determine when the property definitely exists based on abstract
|
|
// value information.
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
|
|
case HasOwnProperty: {
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
|
|
case GetEnumerableLength: {
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
case HasGenericProperty: {
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
case HasStructureProperty: {
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
case HasIndexedProperty: {
|
|
ArrayMode mode = node->arrayMode();
|
|
switch (mode.type()) {
|
|
case Array::Int32:
|
|
case Array::Double:
|
|
case Array::Contiguous:
|
|
case Array::ArrayStorage: {
|
|
break;
|
|
}
|
|
default: {
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
}
|
|
}
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
}
|
|
case GetDirectPname: {
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
}
|
|
case GetPropertyEnumerator: {
|
|
forNode(node).setType(m_graph, SpecCell);
|
|
break;
|
|
}
|
|
case GetEnumeratorStructurePname: {
|
|
forNode(node).setType(m_graph, SpecString | SpecOther);
|
|
break;
|
|
}
|
|
case GetEnumeratorGenericPname: {
|
|
forNode(node).setType(m_graph, SpecString | SpecOther);
|
|
break;
|
|
}
|
|
case ToIndexString: {
|
|
forNode(node).setType(m_graph, SpecString);
|
|
break;
|
|
}
|
|
|
|
case GetGlobalVar:
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case GetGlobalLexicalVariable:
|
|
forNode(node).makeBytecodeTop();
|
|
break;
|
|
|
|
case GetDynamicVar:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeBytecodeTop();
|
|
break;
|
|
|
|
case PutDynamicVar:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
|
|
case ResolveScope:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecObject);
|
|
break;
|
|
|
|
case ResolveScopeForHoistingFuncDeclInEval:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeBytecodeTop();
|
|
break;
|
|
|
|
case PutGlobalVariable:
|
|
case NotifyWrite:
|
|
break;
|
|
|
|
case OverridesHasInstance:
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
|
|
case InstanceOf:
|
|
// Sadly, we don't propagate the fact that we've done InstanceOf
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
|
|
case InstanceOfCustom:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(SpecBoolean);
|
|
break;
|
|
|
|
case Phi:
|
|
RELEASE_ASSERT(m_graph.m_form == SSA);
|
|
forNode(node) = forNode(NodeFlowProjection(node, NodeFlowProjection::Shadow));
|
|
// The state of this node would have already been decided, but it may have become a
|
|
// constant, in which case we'd like to know.
|
|
if (forNode(node).m_value)
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
|
|
case Upsilon: {
|
|
NodeFlowProjection shadow(node->phi(), NodeFlowProjection::Shadow);
|
|
if (shadow.isStillValid()) {
|
|
m_state.createValueForNode(shadow);
|
|
forNode(shadow) = forNode(node->child1());
|
|
}
|
|
break;
|
|
}
|
|
|
|
case Flush:
|
|
case PhantomLocal:
|
|
break;
|
|
|
|
case Call:
|
|
case TailCallInlinedCaller:
|
|
case Construct:
|
|
case CallVarargs:
|
|
case CallForwardVarargs:
|
|
case TailCallVarargsInlinedCaller:
|
|
case ConstructVarargs:
|
|
case ConstructForwardVarargs:
|
|
case TailCallForwardVarargsInlinedCaller:
|
|
case CallEval:
|
|
case DirectCall:
|
|
case DirectConstruct:
|
|
case DirectTailCallInlinedCaller:
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).makeHeapTop();
|
|
break;
|
|
|
|
case ForceOSRExit:
|
|
case CheckBadCell:
|
|
m_state.setIsValid(false);
|
|
break;
|
|
|
|
case InvalidationPoint:
|
|
forAllValues(clobberLimit, AbstractValue::observeInvalidationPointFor);
|
|
m_state.setStructureClobberState(StructuresAreWatched);
|
|
break;
|
|
|
|
case CheckTraps:
|
|
case LogShadowChickenPrologue:
|
|
case LogShadowChickenTail:
|
|
break;
|
|
|
|
case ProfileType:
|
|
case ProfileControlFlow:
|
|
case Phantom:
|
|
case CountExecution:
|
|
case CheckTierUpInLoop:
|
|
case CheckTierUpAtReturn:
|
|
case CheckTypeInfoFlags:
|
|
break;
|
|
|
|
case ParseInt: {
|
|
AbstractValue value = forNode(node->child1());
|
|
if (value.m_type && !(value.m_type & ~SpecInt32Only)) {
|
|
JSValue radix;
|
|
if (!node->child2())
|
|
radix = jsNumber(0);
|
|
else
|
|
radix = forNode(node->child2()).m_value;
|
|
|
|
if (radix.isNumber()
|
|
&& (radix.asNumber() == 0 || radix.asNumber() == 10)) {
|
|
m_state.setFoundConstants(true);
|
|
forNode(node).setType(SpecInt32Only);
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (node->child1().useKind() == UntypedUse)
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecBytecodeNumber);
|
|
break;
|
|
}
|
|
|
|
case CreateRest:
|
|
if (!m_graph.isWatchingHavingABadTimeWatchpoint(node)) {
|
|
// This means we're already having a bad time.
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
forNode(node).setType(m_graph, SpecArray);
|
|
break;
|
|
}
|
|
forNode(node).set(
|
|
m_graph,
|
|
m_graph.globalObjectFor(node->origin.semantic)->restParameterStructure());
|
|
break;
|
|
|
|
case Check: {
|
|
// Simplify out checks that don't actually do checking.
|
|
for (unsigned i = 0; i < AdjacencyList::Size; ++i) {
|
|
Edge edge = node->children.child(i);
|
|
if (!edge)
|
|
break;
|
|
if (edge.isProved() || edge.willNotHaveCheck()) {
|
|
m_state.setFoundConstants(true);
|
|
break;
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
|
|
case SetFunctionName: {
|
|
clobberWorld(node->origin.semantic, clobberLimit);
|
|
break;
|
|
}
|
|
|
|
case StoreBarrier:
|
|
case FencedStoreBarrier: {
|
|
filter(node->child1(), SpecCell);
|
|
break;
|
|
}
|
|
|
|
case CheckTierUpAndOSREnter:
|
|
case LoopHint:
|
|
case ZombieHint:
|
|
case ExitOK:
|
|
break;
|
|
|
|
case Unreachable:
|
|
// It may be that during a previous run of AI we proved that something was unreachable, but
|
|
// during this run of AI we forget that it's unreachable. AI's proofs don't have to get
|
|
// monotonically stronger over time. So, we don't assert that AI doesn't reach the
|
|
// Unreachable. We have no choice but to take our past proof at face value. Otherwise we'll
|
|
// crash whenever AI fails to be as powerful on run K as it was on run K-1.
|
|
m_state.setIsValid(false);
|
|
break;
|
|
|
|
case LastNodeType:
|
|
case ArithIMul:
|
|
case FiatInt52:
|
|
DFG_CRASH(m_graph, node, "Unexpected node type");
|
|
break;
|
|
}
|
|
|
|
return m_state.isValid();
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
bool AbstractInterpreter<AbstractStateType>::executeEffects(unsigned indexInBlock)
|
|
{
|
|
return executeEffects(indexInBlock, m_state.block()->at(indexInBlock));
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
bool AbstractInterpreter<AbstractStateType>::execute(unsigned indexInBlock)
|
|
{
|
|
Node* node = m_state.block()->at(indexInBlock);
|
|
|
|
startExecuting();
|
|
executeEdges(node);
|
|
return executeEffects(indexInBlock, node);
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
bool AbstractInterpreter<AbstractStateType>::execute(Node* node)
|
|
{
|
|
startExecuting();
|
|
executeEdges(node);
|
|
return executeEffects(UINT_MAX, node);
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::clobberWorld(
|
|
const CodeOrigin&, unsigned clobberLimit)
|
|
{
|
|
clobberStructures(clobberLimit);
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
template<typename Functor>
|
|
void AbstractInterpreter<AbstractStateType>::forAllValues(
|
|
unsigned clobberLimit, Functor& functor)
|
|
{
|
|
if (clobberLimit >= m_state.block()->size())
|
|
clobberLimit = m_state.block()->size();
|
|
else
|
|
clobberLimit++;
|
|
ASSERT(clobberLimit <= m_state.block()->size());
|
|
for (size_t i = clobberLimit; i--;) {
|
|
NodeFlowProjection::forEach(
|
|
m_state.block()->at(i),
|
|
[&] (NodeFlowProjection nodeProjection) {
|
|
functor(forNode(nodeProjection));
|
|
});
|
|
}
|
|
if (m_graph.m_form == SSA) {
|
|
for (NodeFlowProjection node : m_state.block()->ssa->liveAtHead) {
|
|
if (node.isStillValid())
|
|
functor(forNode(node));
|
|
}
|
|
}
|
|
for (size_t i = m_state.variables().numberOfArguments(); i--;)
|
|
functor(m_state.variables().argument(i));
|
|
for (size_t i = m_state.variables().numberOfLocals(); i--;)
|
|
functor(m_state.variables().local(i));
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::clobberStructures(unsigned clobberLimit)
|
|
{
|
|
forAllValues(clobberLimit, AbstractValue::clobberStructuresFor);
|
|
setDidClobber();
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::observeTransition(
|
|
unsigned clobberLimit, RegisteredStructure from, RegisteredStructure to)
|
|
{
|
|
AbstractValue::TransitionObserver transitionObserver(from, to);
|
|
forAllValues(clobberLimit, transitionObserver);
|
|
|
|
ASSERT(!from->dfgShouldWatch()); // We don't need to claim to be in a clobbered state because 'from' was never watchable (during the time we were compiling), hence no constants ever introduced into the DFG IR that ever had a watchable structure would ever have the same structure as from.
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::observeTransitions(
|
|
unsigned clobberLimit, const TransitionVector& vector)
|
|
{
|
|
AbstractValue::TransitionsObserver transitionsObserver(vector);
|
|
forAllValues(clobberLimit, transitionsObserver);
|
|
|
|
if (!ASSERT_DISABLED) {
|
|
// We don't need to claim to be in a clobbered state because none of the Transition::previous structures are watchable.
|
|
for (unsigned i = vector.size(); i--;)
|
|
ASSERT(!vector[i].previous->dfgShouldWatch());
|
|
}
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::setDidClobber()
|
|
{
|
|
m_state.setDidClobber(true);
|
|
m_state.setStructureClobberState(StructuresAreClobbered);
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::dump(PrintStream& out) const
|
|
{
|
|
const_cast<AbstractInterpreter<AbstractStateType>*>(this)->dump(out);
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::dump(PrintStream& out)
|
|
{
|
|
CommaPrinter comma(" ");
|
|
HashSet<NodeFlowProjection> seen;
|
|
if (m_graph.m_form == SSA) {
|
|
for (NodeFlowProjection node : m_state.block()->ssa->liveAtHead) {
|
|
seen.add(node);
|
|
AbstractValue& value = forNode(node);
|
|
if (value.isClear())
|
|
continue;
|
|
out.print(comma, node, ":", value);
|
|
}
|
|
}
|
|
for (size_t i = 0; i < m_state.block()->size(); ++i) {
|
|
NodeFlowProjection::forEach(
|
|
m_state.block()->at(i), [&] (NodeFlowProjection nodeProjection) {
|
|
seen.add(nodeProjection);
|
|
AbstractValue& value = forNode(nodeProjection);
|
|
if (value.isClear())
|
|
return;
|
|
out.print(comma, nodeProjection, ":", value);
|
|
});
|
|
}
|
|
if (m_graph.m_form == SSA) {
|
|
for (NodeFlowProjection node : m_state.block()->ssa->liveAtTail) {
|
|
if (seen.contains(node))
|
|
continue;
|
|
AbstractValue& value = forNode(node);
|
|
if (value.isClear())
|
|
continue;
|
|
out.print(comma, node, ":", value);
|
|
}
|
|
}
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
FiltrationResult AbstractInterpreter<AbstractStateType>::filter(
|
|
AbstractValue& value, const RegisteredStructureSet& set, SpeculatedType admittedTypes)
|
|
{
|
|
if (value.filter(m_graph, set, admittedTypes) == FiltrationOK)
|
|
return FiltrationOK;
|
|
m_state.setIsValid(false);
|
|
return Contradiction;
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
FiltrationResult AbstractInterpreter<AbstractStateType>::filterArrayModes(
|
|
AbstractValue& value, ArrayModes arrayModes)
|
|
{
|
|
if (value.filterArrayModes(arrayModes) == FiltrationOK)
|
|
return FiltrationOK;
|
|
m_state.setIsValid(false);
|
|
return Contradiction;
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
FiltrationResult AbstractInterpreter<AbstractStateType>::filter(
|
|
AbstractValue& value, SpeculatedType type)
|
|
{
|
|
if (value.filter(type) == FiltrationOK)
|
|
return FiltrationOK;
|
|
m_state.setIsValid(false);
|
|
return Contradiction;
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
FiltrationResult AbstractInterpreter<AbstractStateType>::filterByValue(
|
|
AbstractValue& abstractValue, FrozenValue concreteValue)
|
|
{
|
|
if (abstractValue.filterByValue(concreteValue) == FiltrationOK)
|
|
return FiltrationOK;
|
|
m_state.setIsValid(false);
|
|
return Contradiction;
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
FiltrationResult AbstractInterpreter<AbstractStateType>::filterClassInfo(
|
|
AbstractValue& value, const ClassInfo* classInfo)
|
|
{
|
|
if (value.filterClassInfo(m_graph, classInfo) == FiltrationOK)
|
|
return FiltrationOK;
|
|
m_state.setIsValid(false);
|
|
return Contradiction;
|
|
}
|
|
|
|
template<typename AbstractStateType>
|
|
void AbstractInterpreter<AbstractStateType>::executeDoubleUnaryOpEffects(Node* node, double(*equivalentFunction)(double))
|
|
{
|
|
JSValue child = forNode(node->child1()).value();
|
|
if (std::optional<double> number = child.toNumberFromPrimitive()) {
|
|
setConstant(node, jsDoubleNumber(equivalentFunction(*number)));
|
|
return;
|
|
}
|
|
SpeculatedType type = SpecFullNumber;
|
|
if (node->child1().useKind() == DoubleRepUse)
|
|
type = typeOfDoubleUnaryOp(forNode(node->child1()).m_type);
|
|
forNode(node).setType(type);
|
|
}
|
|
|
|
} } // namespace JSC::DFG
|
|
|
|
#endif // ENABLE(DFG_JIT)
|