| /* |
| * Copyright (C) 2009-2019 Apple Inc. All rights reserved. |
| * Copyright (C) 2010 Patrick Gansterer <paroga@paroga.com> |
| * |
| * 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. |
| */ |
| |
| #include "config.h" |
| |
| #if ENABLE(JIT) |
| #if USE(JSVALUE32_64) |
| #include "JIT.h" |
| |
| #include "BytecodeStructs.h" |
| #include "CCallHelpers.h" |
| #include "Exception.h" |
| #include "JITInlines.h" |
| #include "JSArray.h" |
| #include "JSCast.h" |
| #include "JSFunction.h" |
| #include "JSPropertyNameEnumerator.h" |
| #include "LinkBuffer.h" |
| #include "MaxFrameExtentForSlowPathCall.h" |
| #include "OpcodeInlines.h" |
| #include "SlowPathCall.h" |
| #include "TypeProfilerLog.h" |
| #include "VirtualRegister.h" |
| |
| namespace JSC { |
| |
| void JIT::emit_op_mov(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpMov>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_src.offset(); |
| |
| if (m_codeBlock->isConstantRegisterIndex(src)) |
| emitStore(dst, getConstantOperand(src)); |
| else { |
| emitLoad(src, regT1, regT0); |
| emitStore(dst, regT1, regT0); |
| } |
| } |
| |
| void JIT::emit_op_end(const Instruction* currentInstruction) |
| { |
| ASSERT(returnValueGPR != callFrameRegister); |
| auto bytecode = currentInstruction->as<OpEnd>(); |
| emitLoad(bytecode.m_value.offset(), regT1, returnValueGPR); |
| emitRestoreCalleeSaves(); |
| emitFunctionEpilogue(); |
| ret(); |
| } |
| |
| void JIT::emit_op_jmp(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJmp>(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| addJump(jump(), target); |
| } |
| |
| void JIT::emit_op_new_object(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpNewObject>(); |
| auto& metadata = bytecode.metadata(m_codeBlock); |
| Structure* structure = metadata.m_objectAllocationProfile.structure(); |
| size_t allocationSize = JSFinalObject::allocationSize(structure->inlineCapacity()); |
| Allocator allocator = allocatorForNonVirtualConcurrently<JSFinalObject>(*m_vm, allocationSize, AllocatorForMode::AllocatorIfExists); |
| |
| RegisterID resultReg = returnValueGPR; |
| RegisterID allocatorReg = regT1; |
| RegisterID scratchReg = regT3; |
| |
| if (!allocator) |
| addSlowCase(jump()); |
| else { |
| JumpList slowCases; |
| auto butterfly = TrustedImmPtr(nullptr); |
| emitAllocateJSObject(resultReg, JITAllocator::constant(allocator), allocatorReg, TrustedImmPtr(structure), butterfly, scratchReg, slowCases); |
| emitInitializeInlineStorage(resultReg, structure->inlineCapacity()); |
| addSlowCase(slowCases); |
| emitStoreCell(bytecode.m_dst.offset(), resultReg); |
| } |
| } |
| |
| void JIT::emitSlow_op_new_object(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| linkAllSlowCases(iter); |
| |
| auto bytecode = currentInstruction->as<OpNewObject>(); |
| auto& metadata = bytecode.metadata(m_codeBlock); |
| int dst = bytecode.m_dst.offset(); |
| Structure* structure = metadata.m_objectAllocationProfile.structure(); |
| callOperation(operationNewObject, TrustedImmPtr(&vm()), structure); |
| emitStoreCell(dst, returnValueGPR); |
| } |
| |
| void JIT::emit_op_overrides_has_instance(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpOverridesHasInstance>(); |
| int dst = bytecode.m_dst.offset(); |
| int constructor = bytecode.m_constructor.offset(); |
| int hasInstanceValue = bytecode.m_hasInstanceValue.offset(); |
| |
| emitLoadPayload(hasInstanceValue, regT0); |
| // We don't jump if we know what Symbol.hasInstance would do. |
| Jump hasInstanceValueNotCell = emitJumpIfNotJSCell(hasInstanceValue); |
| Jump customhasInstanceValue = branchPtr(NotEqual, regT0, TrustedImmPtr(m_codeBlock->globalObject()->functionProtoHasInstanceSymbolFunction())); |
| |
| // We know that constructor is an object from the way bytecode is emitted for instanceof expressions. |
| emitLoadPayload(constructor, regT0); |
| |
| // Check that constructor 'ImplementsDefaultHasInstance' i.e. the object is not a C-API user nor a bound function. |
| test8(Zero, Address(regT0, JSCell::typeInfoFlagsOffset()), TrustedImm32(ImplementsDefaultHasInstance), regT0); |
| Jump done = jump(); |
| |
| hasInstanceValueNotCell.link(this); |
| customhasInstanceValue.link(this); |
| move(TrustedImm32(1), regT0); |
| |
| done.link(this); |
| emitStoreBool(dst, regT0); |
| |
| } |
| |
| void JIT::emit_op_instanceof(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpInstanceof>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_value.offset(); |
| int proto = bytecode.m_prototype.offset(); |
| |
| // Load the operands into registers. |
| // We use regT0 for baseVal since we will be done with this first, and we can then use it for the result. |
| emitLoadPayload(value, regT2); |
| emitLoadPayload(proto, regT1); |
| |
| // Check that proto are cells. baseVal must be a cell - this is checked by the get_by_id for Symbol.hasInstance. |
| emitJumpSlowCaseIfNotJSCell(value); |
| emitJumpSlowCaseIfNotJSCell(proto); |
| |
| JITInstanceOfGenerator gen( |
| m_codeBlock, CodeOrigin(m_bytecodeIndex), CallSiteIndex(m_bytecodeIndex), |
| RegisterSet::stubUnavailableRegisters(), |
| regT0, // result |
| regT2, // value |
| regT1, // proto |
| regT3, regT4); // scratch |
| gen.generateFastPath(*this); |
| m_instanceOfs.append(gen); |
| |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_instanceof_custom(const Instruction*) |
| { |
| // This always goes to slow path since we expect it to be rare. |
| addSlowCase(jump()); |
| } |
| |
| void JIT::emitSlow_op_instanceof(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| linkAllSlowCases(iter); |
| |
| auto bytecode = currentInstruction->as<OpInstanceof>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_value.offset(); |
| int proto = bytecode.m_prototype.offset(); |
| |
| JITInstanceOfGenerator& gen = m_instanceOfs[m_instanceOfIndex++]; |
| |
| Label coldPathBegin = label(); |
| emitLoadTag(value, regT0); |
| emitLoadTag(proto, regT3); |
| Call call = callOperation(operationInstanceOfOptimize, dst, m_codeBlock->globalObject(), gen.stubInfo(), JSValueRegs(regT0, regT2), JSValueRegs(regT3, regT1)); |
| gen.reportSlowPathCall(coldPathBegin, call); |
| } |
| |
| void JIT::emitSlow_op_instanceof_custom(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| linkAllSlowCases(iter); |
| |
| auto bytecode = currentInstruction->as<OpInstanceofCustom>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_value.offset(); |
| int constructor = bytecode.m_constructor.offset(); |
| int hasInstanceValue = bytecode.m_hasInstanceValue.offset(); |
| |
| emitLoad(value, regT1, regT0); |
| emitLoadPayload(constructor, regT2); |
| emitLoad(hasInstanceValue, regT4, regT3); |
| callOperation(operationInstanceOfCustom, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), regT2, JSValueRegs(regT4, regT3)); |
| emitStoreBool(dst, returnValueGPR); |
| } |
| |
| void JIT::emit_op_is_empty(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpIsEmpty>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_operand.offset(); |
| |
| emitLoad(value, regT1, regT0); |
| compare32(Equal, regT1, TrustedImm32(JSValue::EmptyValueTag), regT0); |
| |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_is_undefined(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpIsUndefined>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_operand.offset(); |
| |
| emitLoad(value, regT1, regT0); |
| Jump isCell = branchIfCell(regT1); |
| |
| compare32(Equal, regT1, TrustedImm32(JSValue::UndefinedTag), regT0); |
| Jump done = jump(); |
| |
| isCell.link(this); |
| Jump isMasqueradesAsUndefined = branchTest8(NonZero, Address(regT0, JSCell::typeInfoFlagsOffset()), TrustedImm32(MasqueradesAsUndefined)); |
| move(TrustedImm32(0), regT0); |
| Jump notMasqueradesAsUndefined = jump(); |
| |
| isMasqueradesAsUndefined.link(this); |
| loadPtr(Address(regT0, JSCell::structureIDOffset()), regT1); |
| move(TrustedImmPtr(m_codeBlock->globalObject()), regT0); |
| loadPtr(Address(regT1, Structure::globalObjectOffset()), regT1); |
| compare32(Equal, regT0, regT1, regT0); |
| |
| notMasqueradesAsUndefined.link(this); |
| done.link(this); |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_is_undefined_or_null(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpIsUndefinedOrNull>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_operand.offset(); |
| |
| emitLoadTag(value, regT0); |
| static_assert((JSValue::UndefinedTag + 1 == JSValue::NullTag) && (JSValue::NullTag & 0x1), ""); |
| or32(TrustedImm32(1), regT0); |
| compare32(Equal, regT0, TrustedImm32(JSValue::NullTag), regT0); |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_is_boolean(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpIsBoolean>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_operand.offset(); |
| |
| emitLoadTag(value, regT0); |
| compare32(Equal, regT0, TrustedImm32(JSValue::BooleanTag), regT0); |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_is_number(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpIsNumber>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_operand.offset(); |
| |
| emitLoadTag(value, regT0); |
| add32(TrustedImm32(1), regT0); |
| compare32(Below, regT0, TrustedImm32(JSValue::LowestTag + 1), regT0); |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_is_cell_with_type(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpIsCellWithType>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_operand.offset(); |
| int type = bytecode.m_type; |
| |
| emitLoad(value, regT1, regT0); |
| Jump isNotCell = branchIfNotCell(regT1); |
| |
| compare8(Equal, Address(regT0, JSCell::typeInfoTypeOffset()), TrustedImm32(type), regT0); |
| Jump done = jump(); |
| |
| isNotCell.link(this); |
| move(TrustedImm32(0), regT0); |
| |
| done.link(this); |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_is_object(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpIsObject>(); |
| int dst = bytecode.m_dst.offset(); |
| int value = bytecode.m_operand.offset(); |
| |
| emitLoad(value, regT1, regT0); |
| Jump isNotCell = branchIfNotCell(regT1); |
| |
| compare8(AboveOrEqual, Address(regT0, JSCell::typeInfoTypeOffset()), TrustedImm32(ObjectType), regT0); |
| Jump done = jump(); |
| |
| isNotCell.link(this); |
| move(TrustedImm32(0), regT0); |
| |
| done.link(this); |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_to_primitive(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpToPrimitive>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_src.offset(); |
| |
| emitLoad(src, regT1, regT0); |
| |
| Jump isImm = branchIfNotCell(regT1); |
| addSlowCase(branchIfObject(regT0)); |
| isImm.link(this); |
| |
| if (dst != src) |
| emitStore(dst, regT1, regT0); |
| } |
| |
| void JIT::emit_op_set_function_name(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpSetFunctionName>(); |
| int func = bytecode.m_function.offset(); |
| int name = bytecode.m_name.offset(); |
| emitLoadPayload(func, regT1); |
| emitLoad(name, regT3, regT2); |
| callOperation(operationSetFunctionName, m_codeBlock->globalObject(), regT1, JSValueRegs(regT3, regT2)); |
| } |
| |
| void JIT::emit_op_not(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpNot>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_operand.offset(); |
| |
| emitLoadTag(src, regT0); |
| |
| emitLoad(src, regT1, regT0); |
| addSlowCase(branchIfNotBoolean(regT1, InvalidGPRReg)); |
| xor32(TrustedImm32(1), regT0); |
| |
| emitStoreBool(dst, regT0, (dst == src)); |
| } |
| |
| void JIT::emit_op_jfalse(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJfalse>(); |
| int cond = bytecode.m_condition.offset(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| |
| emitLoad(cond, regT1, regT0); |
| |
| JSValueRegs value(regT1, regT0); |
| GPRReg scratch1 = regT2; |
| GPRReg scratch2 = regT3; |
| bool shouldCheckMasqueradesAsUndefined = true; |
| addJump(branchIfFalsey(vm(), value, scratch1, scratch2, fpRegT0, fpRegT1, shouldCheckMasqueradesAsUndefined, m_codeBlock->globalObject()), target); |
| } |
| |
| void JIT::emit_op_jtrue(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJtrue>(); |
| int cond = bytecode.m_condition.offset(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| |
| emitLoad(cond, regT1, regT0); |
| bool shouldCheckMasqueradesAsUndefined = true; |
| JSValueRegs value(regT1, regT0); |
| GPRReg scratch1 = regT2; |
| GPRReg scratch2 = regT3; |
| addJump(branchIfTruthy(vm(), value, scratch1, scratch2, fpRegT0, fpRegT1, shouldCheckMasqueradesAsUndefined, m_codeBlock->globalObject()), target); |
| } |
| |
| void JIT::emit_op_jeq_null(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJeqNull>(); |
| int src = bytecode.m_value.offset(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| |
| emitLoad(src, regT1, regT0); |
| |
| Jump isImmediate = branchIfNotCell(regT1); |
| |
| Jump isNotMasqueradesAsUndefined = branchTest8(Zero, Address(regT0, JSCell::typeInfoFlagsOffset()), TrustedImm32(MasqueradesAsUndefined)); |
| loadPtr(Address(regT0, JSCell::structureIDOffset()), regT2); |
| move(TrustedImmPtr(m_codeBlock->globalObject()), regT0); |
| addJump(branchPtr(Equal, Address(regT2, Structure::globalObjectOffset()), regT0), target); |
| Jump masqueradesGlobalObjectIsForeign = jump(); |
| |
| // Now handle the immediate cases - undefined & null |
| isImmediate.link(this); |
| static_assert((JSValue::UndefinedTag + 1 == JSValue::NullTag) && (JSValue::NullTag & 0x1), ""); |
| or32(TrustedImm32(1), regT1); |
| addJump(branchIfNull(regT1), target); |
| |
| isNotMasqueradesAsUndefined.link(this); |
| masqueradesGlobalObjectIsForeign.link(this); |
| } |
| |
| void JIT::emit_op_jneq_null(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJneqNull>(); |
| int src = bytecode.m_value.offset(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| |
| emitLoad(src, regT1, regT0); |
| |
| Jump isImmediate = branchIfNotCell(regT1); |
| |
| addJump(branchTest8(Zero, Address(regT0, JSCell::typeInfoFlagsOffset()), TrustedImm32(MasqueradesAsUndefined)), target); |
| loadPtr(Address(regT0, JSCell::structureIDOffset()), regT2); |
| move(TrustedImmPtr(m_codeBlock->globalObject()), regT0); |
| addJump(branchPtr(NotEqual, Address(regT2, Structure::globalObjectOffset()), regT0), target); |
| Jump wasNotImmediate = jump(); |
| |
| // Now handle the immediate cases - undefined & null |
| isImmediate.link(this); |
| |
| static_assert((JSValue::UndefinedTag + 1 == JSValue::NullTag) && (JSValue::NullTag & 0x1), ""); |
| or32(TrustedImm32(1), regT1); |
| addJump(branchIfNotNull(regT1), target); |
| |
| wasNotImmediate.link(this); |
| } |
| |
| void JIT::emit_op_jundefined_or_null(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJundefinedOrNull>(); |
| int value = bytecode.m_value.offset(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| |
| emitLoadTag(value, regT0); |
| static_assert((JSValue::UndefinedTag + 1 == JSValue::NullTag) && (JSValue::NullTag & 0x1), ""); |
| or32(TrustedImm32(1), regT0); |
| addJump(branchIfNull(regT0), target); |
| } |
| |
| void JIT::emit_op_jnundefined_or_null(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJnundefinedOrNull>(); |
| int value = bytecode.m_value.offset(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| |
| emitLoadTag(value, regT0); |
| static_assert((JSValue::UndefinedTag + 1 == JSValue::NullTag) && (JSValue::NullTag & 0x1), ""); |
| or32(TrustedImm32(1), regT0); |
| addJump(branchIfNotNull(regT0), target); |
| } |
| |
| void JIT::emit_op_jneq_ptr(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJneqPtr>(); |
| auto& metadata = bytecode.metadata(m_codeBlock); |
| int src = bytecode.m_value.offset(); |
| JSValue specialPointer = getConstantOperand(bytecode.m_specialPointer.offset()); |
| ASSERT(specialPointer.isCell()); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| |
| emitLoad(src, regT1, regT0); |
| Jump notCell = branchIfNotCell(regT1); |
| Jump equal = branchPtr(Equal, regT0, TrustedImmPtr(specialPointer.asCell())); |
| notCell.link(this); |
| store8(TrustedImm32(1), &metadata.m_hasJumped); |
| addJump(jump(), target); |
| equal.link(this); |
| } |
| |
| void JIT::emit_op_eq(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpEq>(); |
| |
| int dst = bytecode.m_dst.offset(); |
| int src1 = bytecode.m_lhs.offset(); |
| int src2 = bytecode.m_rhs.offset(); |
| |
| emitLoad2(src1, regT1, regT0, src2, regT3, regT2); |
| addSlowCase(branch32(NotEqual, regT1, regT3)); |
| addSlowCase(branchIfCell(regT1)); |
| addSlowCase(branch32(Below, regT1, TrustedImm32(JSValue::LowestTag))); |
| |
| compare32(Equal, regT0, regT2, regT0); |
| |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emitSlow_op_eq(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| auto bytecode = currentInstruction->as<OpEq>(); |
| int dst = bytecode.m_dst.offset(); |
| |
| JumpList storeResult; |
| JumpList genericCase; |
| |
| genericCase.append(getSlowCase(iter)); // tags not equal |
| |
| linkSlowCase(iter); // tags equal and JSCell |
| genericCase.append(branchIfNotString(regT0)); |
| genericCase.append(branchIfNotString(regT2)); |
| |
| // String case. |
| callOperation(operationCompareStringEq, m_codeBlock->globalObject(), regT0, regT2); |
| storeResult.append(jump()); |
| |
| // Generic case. |
| genericCase.append(getSlowCase(iter)); // doubles |
| genericCase.link(this); |
| callOperation(operationCompareEq, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), JSValueRegs(regT3, regT2)); |
| |
| storeResult.link(this); |
| emitStoreBool(dst, returnValueGPR); |
| } |
| |
| void JIT::emit_op_jeq(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJeq>(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| int src1 = bytecode.m_lhs.offset(); |
| int src2 = bytecode.m_rhs.offset(); |
| |
| emitLoad2(src1, regT1, regT0, src2, regT3, regT2); |
| addSlowCase(branch32(NotEqual, regT1, regT3)); |
| addSlowCase(branchIfCell(regT1)); |
| addSlowCase(branch32(Below, regT1, TrustedImm32(JSValue::LowestTag))); |
| |
| addJump(branch32(Equal, regT0, regT2), target); |
| } |
| |
| void JIT::compileOpEqJumpSlow(Vector<SlowCaseEntry>::iterator& iter, CompileOpEqType type, int jumpTarget) |
| { |
| JumpList done; |
| JumpList genericCase; |
| |
| genericCase.append(getSlowCase(iter)); // tags not equal |
| |
| linkSlowCase(iter); // tags equal and JSCell |
| genericCase.append(branchIfNotString(regT0)); |
| genericCase.append(branchIfNotString(regT2)); |
| |
| // String case. |
| callOperation(operationCompareStringEq, m_codeBlock->globalObject(), regT0, regT2); |
| emitJumpSlowToHot(branchTest32(type == CompileOpEqType::Eq ? NonZero : Zero, returnValueGPR), jumpTarget); |
| done.append(jump()); |
| |
| // Generic case. |
| genericCase.append(getSlowCase(iter)); // doubles |
| genericCase.link(this); |
| callOperation(operationCompareEq, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), JSValueRegs(regT3, regT2)); |
| emitJumpSlowToHot(branchTest32(type == CompileOpEqType::Eq ? NonZero : Zero, returnValueGPR), jumpTarget); |
| |
| done.link(this); |
| } |
| |
| void JIT::emitSlow_op_jeq(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| auto bytecode = currentInstruction->as<OpJeq>(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| compileOpEqJumpSlow(iter, CompileOpEqType::Eq, target); |
| } |
| |
| void JIT::emit_op_neq(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpNeq>(); |
| int dst = bytecode.m_dst.offset(); |
| int src1 = bytecode.m_lhs.offset(); |
| int src2 = bytecode.m_rhs.offset(); |
| |
| emitLoad2(src1, regT1, regT0, src2, regT3, regT2); |
| addSlowCase(branch32(NotEqual, regT1, regT3)); |
| addSlowCase(branchIfCell(regT1)); |
| addSlowCase(branch32(Below, regT1, TrustedImm32(JSValue::LowestTag))); |
| |
| compare32(NotEqual, regT0, regT2, regT0); |
| |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emitSlow_op_neq(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| auto bytecode = currentInstruction->as<OpNeq>(); |
| int dst = bytecode.m_dst.offset(); |
| |
| JumpList storeResult; |
| JumpList genericCase; |
| |
| genericCase.append(getSlowCase(iter)); // tags not equal |
| |
| linkSlowCase(iter); // tags equal and JSCell |
| genericCase.append(branchIfNotString(regT0)); |
| genericCase.append(branchIfNotString(regT2)); |
| |
| // String case. |
| callOperation(operationCompareStringEq, m_codeBlock->globalObject(), regT0, regT2); |
| storeResult.append(jump()); |
| |
| // Generic case. |
| genericCase.append(getSlowCase(iter)); // doubles |
| genericCase.link(this); |
| callOperation(operationCompareEq, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), JSValueRegs(regT3, regT2)); |
| |
| storeResult.link(this); |
| xor32(TrustedImm32(0x1), returnValueGPR); |
| emitStoreBool(dst, returnValueGPR); |
| } |
| |
| void JIT::emit_op_jneq(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpJneq>(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| int src1 = bytecode.m_lhs.offset(); |
| int src2 = bytecode.m_rhs.offset(); |
| |
| emitLoad2(src1, regT1, regT0, src2, regT3, regT2); |
| addSlowCase(branch32(NotEqual, regT1, regT3)); |
| addSlowCase(branchIfCell(regT1)); |
| addSlowCase(branch32(Below, regT1, TrustedImm32(JSValue::LowestTag))); |
| |
| addJump(branch32(NotEqual, regT0, regT2), target); |
| } |
| |
| void JIT::emitSlow_op_jneq(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| auto bytecode = currentInstruction->as<OpJneq>(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| compileOpEqJumpSlow(iter, CompileOpEqType::NEq, target); |
| } |
| |
| template <typename Op> |
| void JIT::compileOpStrictEq(const Instruction* currentInstruction, CompileOpStrictEqType type) |
| { |
| auto bytecode = currentInstruction->as<Op>(); |
| int dst = bytecode.m_dst.offset(); |
| int src1 = bytecode.m_lhs.offset(); |
| int src2 = bytecode.m_rhs.offset(); |
| |
| emitLoad2(src1, regT1, regT0, src2, regT3, regT2); |
| |
| // Bail if the tags differ, or are double. |
| addSlowCase(branch32(NotEqual, regT1, regT3)); |
| addSlowCase(branch32(Below, regT1, TrustedImm32(JSValue::LowestTag))); |
| |
| // Jump to a slow case if both are strings or symbols (non object). |
| Jump notCell = branchIfNotCell(regT1); |
| Jump firstIsObject = branchIfObject(regT0); |
| addSlowCase(branchIfNotObject(regT2)); |
| notCell.link(this); |
| firstIsObject.link(this); |
| |
| // Simply compare the payloads. |
| if (type == CompileOpStrictEqType::StrictEq) |
| compare32(Equal, regT0, regT2, regT0); |
| else |
| compare32(NotEqual, regT0, regT2, regT0); |
| |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::emit_op_stricteq(const Instruction* currentInstruction) |
| { |
| compileOpStrictEq<OpStricteq>(currentInstruction, CompileOpStrictEqType::StrictEq); |
| } |
| |
| void JIT::emit_op_nstricteq(const Instruction* currentInstruction) |
| { |
| compileOpStrictEq<OpNstricteq>(currentInstruction, CompileOpStrictEqType::NStrictEq); |
| } |
| |
| template<typename Op> |
| void JIT::compileOpStrictEqJump(const Instruction* currentInstruction, CompileOpStrictEqType type) |
| { |
| auto bytecode = currentInstruction->as<Op>(); |
| int target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| int src1 = bytecode.m_lhs.offset(); |
| int src2 = bytecode.m_rhs.offset(); |
| |
| emitLoad2(src1, regT1, regT0, src2, regT3, regT2); |
| |
| // Bail if the tags differ, or are double. |
| addSlowCase(branch32(NotEqual, regT1, regT3)); |
| addSlowCase(branch32(Below, regT1, TrustedImm32(JSValue::LowestTag))); |
| |
| // Jump to a slow case if both are strings or symbols (non object). |
| Jump notCell = branchIfNotCell(regT1); |
| Jump firstIsObject = branchIfObject(regT0); |
| addSlowCase(branchIfNotObject(regT2)); |
| notCell.link(this); |
| firstIsObject.link(this); |
| |
| // Simply compare the payloads. |
| if (type == CompileOpStrictEqType::StrictEq) |
| addJump(branch32(Equal, regT0, regT2), target); |
| else |
| addJump(branch32(NotEqual, regT0, regT2), target); |
| } |
| |
| void JIT::emit_op_jstricteq(const Instruction* currentInstruction) |
| { |
| compileOpStrictEqJump<OpJstricteq>(currentInstruction, CompileOpStrictEqType::StrictEq); |
| } |
| |
| void JIT::emit_op_jnstricteq(const Instruction* currentInstruction) |
| { |
| compileOpStrictEqJump<OpJnstricteq>(currentInstruction, CompileOpStrictEqType::NStrictEq); |
| } |
| |
| void JIT::emitSlow_op_jstricteq(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| linkAllSlowCases(iter); |
| |
| auto bytecode = currentInstruction->as<OpJstricteq>(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| callOperation(operationCompareStrictEq, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), JSValueRegs(regT3, regT2)); |
| emitJumpSlowToHot(branchTest32(NonZero, returnValueGPR), target); |
| } |
| |
| void JIT::emitSlow_op_jnstricteq(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| linkAllSlowCases(iter); |
| |
| auto bytecode = currentInstruction->as<OpJnstricteq>(); |
| unsigned target = jumpTarget(currentInstruction, bytecode.m_targetLabel); |
| callOperation(operationCompareStrictEq, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), JSValueRegs(regT3, regT2)); |
| emitJumpSlowToHot(branchTest32(Zero, returnValueGPR), target); |
| } |
| |
| void JIT::emit_op_eq_null(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpEqNull>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_operand.offset(); |
| |
| emitLoad(src, regT1, regT0); |
| Jump isImmediate = branchIfNotCell(regT1); |
| |
| Jump isMasqueradesAsUndefined = branchTest8(NonZero, Address(regT0, JSCell::typeInfoFlagsOffset()), TrustedImm32(MasqueradesAsUndefined)); |
| move(TrustedImm32(0), regT1); |
| Jump wasNotMasqueradesAsUndefined = jump(); |
| |
| isMasqueradesAsUndefined.link(this); |
| loadPtr(Address(regT0, JSCell::structureIDOffset()), regT2); |
| move(TrustedImmPtr(m_codeBlock->globalObject()), regT0); |
| loadPtr(Address(regT2, Structure::globalObjectOffset()), regT2); |
| compare32(Equal, regT0, regT2, regT1); |
| Jump wasNotImmediate = jump(); |
| |
| isImmediate.link(this); |
| |
| compare32(Equal, regT1, TrustedImm32(JSValue::NullTag), regT2); |
| compare32(Equal, regT1, TrustedImm32(JSValue::UndefinedTag), regT1); |
| or32(regT2, regT1); |
| |
| wasNotImmediate.link(this); |
| wasNotMasqueradesAsUndefined.link(this); |
| |
| emitStoreBool(dst, regT1); |
| } |
| |
| void JIT::emit_op_neq_null(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpNeqNull>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_operand.offset(); |
| |
| emitLoad(src, regT1, regT0); |
| Jump isImmediate = branchIfNotCell(regT1); |
| |
| Jump isMasqueradesAsUndefined = branchTest8(NonZero, Address(regT0, JSCell::typeInfoFlagsOffset()), TrustedImm32(MasqueradesAsUndefined)); |
| move(TrustedImm32(1), regT1); |
| Jump wasNotMasqueradesAsUndefined = jump(); |
| |
| isMasqueradesAsUndefined.link(this); |
| loadPtr(Address(regT0, JSCell::structureIDOffset()), regT2); |
| move(TrustedImmPtr(m_codeBlock->globalObject()), regT0); |
| loadPtr(Address(regT2, Structure::globalObjectOffset()), regT2); |
| compare32(NotEqual, regT0, regT2, regT1); |
| Jump wasNotImmediate = jump(); |
| |
| isImmediate.link(this); |
| |
| compare32(NotEqual, regT1, TrustedImm32(JSValue::NullTag), regT2); |
| compare32(NotEqual, regT1, TrustedImm32(JSValue::UndefinedTag), regT1); |
| and32(regT2, regT1); |
| |
| wasNotImmediate.link(this); |
| wasNotMasqueradesAsUndefined.link(this); |
| |
| emitStoreBool(dst, regT1); |
| } |
| |
| void JIT::emit_op_throw(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpThrow>(); |
| ASSERT(regT0 == returnValueGPR); |
| copyCalleeSavesToEntryFrameCalleeSavesBuffer(vm().topEntryFrame); |
| emitLoad(bytecode.m_value.offset(), regT1, regT0); |
| callOperationNoExceptionCheck(operationThrow, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0)); |
| jumpToExceptionHandler(vm()); |
| } |
| |
| void JIT::emit_op_to_number(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpToNumber>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_operand.offset(); |
| |
| emitLoad(src, regT1, regT0); |
| |
| Jump isInt32 = branchIfInt32(regT1); |
| addSlowCase(branch32(AboveOrEqual, regT1, TrustedImm32(JSValue::LowestTag))); |
| isInt32.link(this); |
| |
| emitValueProfilingSite(bytecode.metadata(m_codeBlock)); |
| if (src != dst) |
| emitStore(dst, regT1, regT0); |
| } |
| |
| void JIT::emit_op_to_numeric(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpToNumeric>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_operand.offset(); |
| JSValueRegs argumentValueRegs(regT1, regT0); |
| |
| emitLoad(src, regT1, regT0); |
| |
| Jump isNotCell = branchIfNotCell(regT1); |
| addSlowCase(branchIfNotBigInt(regT0)); |
| Jump isBigInt = jump(); |
| |
| isNotCell.link(this); |
| addSlowCase(branchIfNotNumber(argumentValueRegs, regT2)); |
| isBigInt.link(this); |
| |
| emitValueProfilingSite(bytecode.metadata(m_codeBlock)); |
| if (src != dst) |
| emitStore(dst, regT1, regT0); |
| } |
| |
| void JIT::emit_op_to_string(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpToString>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_operand.offset(); |
| |
| emitLoad(src, regT1, regT0); |
| |
| addSlowCase(branchIfNotCell(regT1)); |
| addSlowCase(branchIfNotString(regT0)); |
| |
| if (src != dst) |
| emitStore(dst, regT1, regT0); |
| } |
| |
| void JIT::emit_op_to_object(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpToObject>(); |
| int dst = bytecode.m_dst.offset(); |
| int src = bytecode.m_operand.offset(); |
| |
| emitLoad(src, regT1, regT0); |
| |
| addSlowCase(branchIfNotCell(regT1)); |
| addSlowCase(branchIfNotObject(regT0)); |
| |
| emitValueProfilingSite(bytecode.metadata(m_codeBlock)); |
| if (src != dst) |
| emitStore(dst, regT1, regT0); |
| } |
| |
| void JIT::emit_op_catch(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpCatch>(); |
| |
| restoreCalleeSavesFromEntryFrameCalleeSavesBuffer(vm().topEntryFrame); |
| |
| move(TrustedImmPtr(m_vm), regT3); |
| // operationThrow returns the callFrame for the handler. |
| load32(Address(regT3, VM::callFrameForCatchOffset()), callFrameRegister); |
| storePtr(TrustedImmPtr(nullptr), Address(regT3, VM::callFrameForCatchOffset())); |
| |
| addPtr(TrustedImm32(stackPointerOffsetFor(codeBlock()) * sizeof(Register)), callFrameRegister, stackPointerRegister); |
| |
| callOperationNoExceptionCheck(operationCheckIfExceptionIsUncatchableAndNotifyProfiler, TrustedImmPtr(&vm())); |
| Jump isCatchableException = branchTest32(Zero, returnValueGPR); |
| jumpToExceptionHandler(vm()); |
| isCatchableException.link(this); |
| |
| move(TrustedImmPtr(m_vm), regT3); |
| |
| // Now store the exception returned by operationThrow. |
| load32(Address(regT3, VM::exceptionOffset()), regT2); |
| move(TrustedImm32(JSValue::CellTag), regT1); |
| |
| store32(TrustedImm32(0), Address(regT3, VM::exceptionOffset())); |
| |
| unsigned exception = bytecode.m_exception.offset(); |
| emitStore(exception, regT1, regT2); |
| |
| load32(Address(regT2, Exception::valueOffset() + OBJECT_OFFSETOF(JSValue, u.asBits.payload)), regT0); |
| load32(Address(regT2, Exception::valueOffset() + OBJECT_OFFSETOF(JSValue, u.asBits.tag)), regT1); |
| |
| unsigned thrownValue = bytecode.m_thrownValue.offset(); |
| emitStore(thrownValue, regT1, regT0); |
| |
| #if ENABLE(DFG_JIT) |
| // FIXME: consider inline caching the process of doing OSR entry, including |
| // argument type proofs, storing locals to the buffer, etc |
| // https://bugs.webkit.org/show_bug.cgi?id=175598 |
| |
| auto& metadata = bytecode.metadata(m_codeBlock); |
| ValueProfileAndOperandBuffer* buffer = metadata.m_buffer; |
| if (buffer || !shouldEmitProfiling()) |
| callOperation(operationTryOSREnterAtCatch, &vm(), m_bytecodeIndex.asBits()); |
| else |
| callOperation(operationTryOSREnterAtCatchAndValueProfile, &vm(), m_bytecodeIndex.asBits()); |
| auto skipOSREntry = branchTestPtr(Zero, returnValueGPR); |
| emitRestoreCalleeSaves(); |
| farJump(returnValueGPR, NoPtrTag); |
| skipOSREntry.link(this); |
| if (buffer && shouldEmitProfiling()) { |
| buffer->forEach([&] (ValueProfileAndOperand& profile) { |
| JSValueRegs regs(regT1, regT0); |
| emitGetVirtualRegister(profile.m_operand, regs); |
| emitValueProfilingSite(static_cast<ValueProfile&>(profile)); |
| }); |
| } |
| #endif // ENABLE(DFG_JIT) |
| } |
| |
| void JIT::emit_op_identity_with_profile(const Instruction*) |
| { |
| // We don't need to do anything here... |
| } |
| |
| void JIT::emit_op_get_parent_scope(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpGetParentScope>(); |
| int currentScope = bytecode.m_scope.offset(); |
| emitLoadPayload(currentScope, regT0); |
| loadPtr(Address(regT0, JSScope::offsetOfNext()), regT0); |
| emitStoreCell(bytecode.m_dst.offset(), regT0); |
| } |
| |
| void JIT::emit_op_switch_imm(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpSwitchImm>(); |
| size_t tableIndex = bytecode.m_tableIndex; |
| unsigned defaultOffset = jumpTarget(currentInstruction, bytecode.m_defaultOffset); |
| unsigned scrutinee = bytecode.m_scrutinee.offset(); |
| |
| // create jump table for switch destinations, track this switch statement. |
| SimpleJumpTable* jumpTable = &m_codeBlock->switchJumpTable(tableIndex); |
| m_switches.append(SwitchRecord(jumpTable, m_bytecodeIndex, defaultOffset, SwitchRecord::Immediate)); |
| jumpTable->ensureCTITable(); |
| |
| emitLoad(scrutinee, regT1, regT0); |
| callOperation(operationSwitchImmWithUnknownKeyType, TrustedImmPtr(&vm()), JSValueRegs(regT1, regT0), tableIndex); |
| farJump(returnValueGPR, NoPtrTag); |
| } |
| |
| void JIT::emit_op_switch_char(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpSwitchChar>(); |
| size_t tableIndex = bytecode.m_tableIndex; |
| unsigned defaultOffset = jumpTarget(currentInstruction, bytecode.m_defaultOffset); |
| unsigned scrutinee = bytecode.m_scrutinee.offset(); |
| |
| // create jump table for switch destinations, track this switch statement. |
| SimpleJumpTable* jumpTable = &m_codeBlock->switchJumpTable(tableIndex); |
| m_switches.append(SwitchRecord(jumpTable, m_bytecodeIndex, defaultOffset, SwitchRecord::Character)); |
| jumpTable->ensureCTITable(); |
| |
| emitLoad(scrutinee, regT1, regT0); |
| callOperation(operationSwitchCharWithUnknownKeyType, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), tableIndex); |
| farJump(returnValueGPR, NoPtrTag); |
| } |
| |
| void JIT::emit_op_switch_string(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpSwitchString>(); |
| size_t tableIndex = bytecode.m_tableIndex; |
| unsigned defaultOffset = jumpTarget(currentInstruction, bytecode.m_defaultOffset); |
| unsigned scrutinee = bytecode.m_scrutinee.offset(); |
| |
| // create jump table for switch destinations, track this switch statement. |
| StringJumpTable* jumpTable = &m_codeBlock->stringSwitchJumpTable(tableIndex); |
| m_switches.append(SwitchRecord(jumpTable, m_bytecodeIndex, defaultOffset)); |
| |
| emitLoad(scrutinee, regT1, regT0); |
| callOperation(operationSwitchStringWithUnknownKeyType, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), tableIndex); |
| farJump(returnValueGPR, NoPtrTag); |
| } |
| |
| void JIT::emit_op_debug(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpDebug>(); |
| load32(codeBlock()->debuggerRequestsAddress(), regT0); |
| Jump noDebuggerRequests = branchTest32(Zero, regT0); |
| callOperation(operationDebug, &vm(), static_cast<int>(bytecode.m_debugHookType)); |
| noDebuggerRequests.link(this); |
| } |
| |
| |
| void JIT::emit_op_enter(const Instruction* currentInstruction) |
| { |
| emitEnterOptimizationCheck(); |
| |
| // Even though JIT code doesn't use them, we initialize our constant |
| // registers to zap stale pointers, to avoid unnecessarily prolonging |
| // object lifetime and increasing GC pressure. |
| for (int i = CodeBlock::llintBaselineCalleeSaveSpaceAsVirtualRegisters(); i < m_codeBlock->numVars(); ++i) |
| emitStore(virtualRegisterForLocal(i).offset(), jsUndefined()); |
| |
| JITSlowPathCall slowPathCall(this, currentInstruction, slow_path_enter); |
| slowPathCall.call(); |
| } |
| |
| void JIT::emit_op_get_scope(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpGetScope>(); |
| int dst = bytecode.m_dst.offset(); |
| emitGetFromCallFrameHeaderPtr(CallFrameSlot::callee, regT0); |
| loadPtr(Address(regT0, JSFunction::offsetOfScopeChain()), regT0); |
| emitStoreCell(dst, regT0); |
| } |
| |
| void JIT::emit_op_create_this(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpCreateThis>(); |
| auto& metadata = bytecode.metadata(m_codeBlock); |
| int callee = bytecode.m_callee.offset(); |
| WriteBarrierBase<JSCell>* cachedFunction = &metadata.m_cachedCallee; |
| RegisterID calleeReg = regT0; |
| RegisterID rareDataReg = regT4; |
| RegisterID resultReg = regT0; |
| RegisterID allocatorReg = regT1; |
| RegisterID structureReg = regT2; |
| RegisterID cachedFunctionReg = regT4; |
| RegisterID scratchReg = regT3; |
| |
| emitLoadPayload(callee, calleeReg); |
| addSlowCase(branchIfNotFunction(calleeReg)); |
| loadPtr(Address(calleeReg, JSFunction::offsetOfRareData()), rareDataReg); |
| addSlowCase(branchTestPtr(Zero, rareDataReg)); |
| load32(Address(rareDataReg, FunctionRareData::offsetOfObjectAllocationProfile() + ObjectAllocationProfileWithPrototype::offsetOfAllocator()), allocatorReg); |
| loadPtr(Address(rareDataReg, FunctionRareData::offsetOfObjectAllocationProfile() + ObjectAllocationProfileWithPrototype::offsetOfStructure()), structureReg); |
| |
| loadPtr(cachedFunction, cachedFunctionReg); |
| Jump hasSeenMultipleCallees = branchPtr(Equal, cachedFunctionReg, TrustedImmPtr(JSCell::seenMultipleCalleeObjects())); |
| addSlowCase(branchPtr(NotEqual, calleeReg, cachedFunctionReg)); |
| hasSeenMultipleCallees.link(this); |
| |
| JumpList slowCases; |
| auto butterfly = TrustedImmPtr(nullptr); |
| emitAllocateJSObject(resultReg, JITAllocator::variable(), allocatorReg, structureReg, butterfly, scratchReg, slowCases); |
| load8(Address(structureReg, Structure::inlineCapacityOffset()), scratchReg); |
| emitInitializeInlineStorage(resultReg, scratchReg); |
| addSlowCase(slowCases); |
| emitStoreCell(bytecode.m_dst.offset(), resultReg); |
| } |
| |
| void JIT::emit_op_to_this(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpToThis>(); |
| auto& metadata = bytecode.metadata(m_codeBlock); |
| StructureID* cachedStructureID = &metadata.m_cachedStructureID; |
| int thisRegister = bytecode.m_srcDst.offset(); |
| |
| emitLoad(thisRegister, regT3, regT2); |
| |
| addSlowCase(branchIfNotCell(regT3)); |
| addSlowCase(branchIfNotType(regT2, FinalObjectType)); |
| loadPtr(Address(regT2, JSCell::structureIDOffset()), regT0); |
| load32(cachedStructureID, regT2); |
| addSlowCase(branchPtr(NotEqual, regT0, regT2)); |
| } |
| |
| void JIT::emit_op_check_tdz(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpCheckTdz>(); |
| emitLoadTag(bytecode.m_targetVirtualRegister.offset(), regT0); |
| addSlowCase(branchIfEmpty(regT0)); |
| } |
| |
| void JIT::emit_op_has_structure_property(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpHasStructureProperty>(); |
| int dst = bytecode.m_dst.offset(); |
| int base = bytecode.m_base.offset(); |
| int enumerator = bytecode.m_enumerator.offset(); |
| |
| emitLoadPayload(base, regT0); |
| emitJumpSlowCaseIfNotJSCell(base); |
| |
| emitLoadPayload(enumerator, regT1); |
| |
| load32(Address(regT0, JSCell::structureIDOffset()), regT0); |
| addSlowCase(branch32(NotEqual, regT0, Address(regT1, JSPropertyNameEnumerator::cachedStructureIDOffset()))); |
| |
| move(TrustedImm32(1), regT0); |
| emitStoreBool(dst, regT0); |
| } |
| |
| void JIT::privateCompileHasIndexedProperty(ByValInfo* byValInfo, ReturnAddressPtr returnAddress, JITArrayMode arrayMode) |
| { |
| const Instruction* currentInstruction = m_codeBlock->instructions().at(byValInfo->bytecodeIndex).ptr(); |
| |
| PatchableJump badType; |
| |
| // FIXME: Add support for other types like TypedArrays and Arguments. |
| // See https://bugs.webkit.org/show_bug.cgi?id=135033 and https://bugs.webkit.org/show_bug.cgi?id=135034. |
| JumpList slowCases = emitLoadForArrayMode(currentInstruction, arrayMode, badType); |
| move(TrustedImm32(1), regT0); |
| Jump done = jump(); |
| |
| LinkBuffer patchBuffer(*this, m_codeBlock); |
| |
| patchBuffer.link(badType, byValInfo->slowPathTarget); |
| patchBuffer.link(slowCases, byValInfo->slowPathTarget); |
| |
| patchBuffer.link(done, byValInfo->badTypeDoneTarget); |
| |
| byValInfo->stubRoutine = FINALIZE_CODE_FOR_STUB( |
| m_codeBlock, patchBuffer, JITStubRoutinePtrTag, |
| "Baseline has_indexed_property stub for %s, return point %p", toCString(*m_codeBlock).data(), returnAddress.value()); |
| |
| MacroAssembler::repatchJump(byValInfo->badTypeJump, CodeLocationLabel<JITStubRoutinePtrTag>(byValInfo->stubRoutine->code().code())); |
| MacroAssembler::repatchCall(CodeLocationCall<NoPtrTag>(MacroAssemblerCodePtr<NoPtrTag>(returnAddress)), FunctionPtr<OperationPtrTag>(operationHasIndexedPropertyGeneric)); |
| } |
| |
| void JIT::emit_op_has_indexed_property(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpHasIndexedProperty>(); |
| auto& metadata = bytecode.metadata(m_codeBlock); |
| int dst = bytecode.m_dst.offset(); |
| int base = bytecode.m_base.offset(); |
| int property = bytecode.m_property.offset(); |
| ArrayProfile* profile = &metadata.m_arrayProfile; |
| ByValInfo* byValInfo = m_codeBlock->addByValInfo(); |
| |
| emitLoadPayload(base, regT0); |
| emitJumpSlowCaseIfNotJSCell(base); |
| |
| emitLoad(property, regT3, regT1); |
| addSlowCase(branchIfNotInt32(regT3)); |
| |
| // This is technically incorrect - we're zero-extending an int32. On the hot path this doesn't matter. |
| // We check the value as if it was a uint32 against the m_vectorLength - which will always fail if |
| // number was signed since m_vectorLength is always less than intmax (since the total allocation |
| // size is always less than 4Gb). As such zero extending will have been correct (and extending the value |
| // to 64-bits is necessary since it's used in the address calculation. We zero extend rather than sign |
| // extending since it makes it easier to re-tag the value in the slow case. |
| zeroExtend32ToPtr(regT1, regT1); |
| |
| emitArrayProfilingSiteWithCell(regT0, regT2, profile); |
| and32(TrustedImm32(IndexingShapeMask), regT2); |
| |
| JITArrayMode mode = chooseArrayMode(profile); |
| PatchableJump badType; |
| |
| // FIXME: Add support for other types like TypedArrays and Arguments. |
| // See https://bugs.webkit.org/show_bug.cgi?id=135033 and https://bugs.webkit.org/show_bug.cgi?id=135034. |
| JumpList slowCases = emitLoadForArrayMode(currentInstruction, mode, badType); |
| move(TrustedImm32(1), regT0); |
| |
| addSlowCase(badType); |
| addSlowCase(slowCases); |
| |
| Label done = label(); |
| |
| emitStoreBool(dst, regT0); |
| |
| Label nextHotPath = label(); |
| |
| m_byValCompilationInfo.append(ByValCompilationInfo(byValInfo, m_bytecodeIndex, PatchableJump(), badType, mode, profile, done, nextHotPath)); |
| } |
| |
| void JIT::emitSlow_op_has_indexed_property(const Instruction* currentInstruction, Vector<SlowCaseEntry>::iterator& iter) |
| { |
| linkAllSlowCases(iter); |
| |
| auto bytecode = currentInstruction->as<OpHasIndexedProperty>(); |
| int dst = bytecode.m_dst.offset(); |
| int base = bytecode.m_base.offset(); |
| int property = bytecode.m_property.offset(); |
| ByValInfo* byValInfo = m_byValCompilationInfo[m_byValInstructionIndex].byValInfo; |
| |
| Label slowPath = label(); |
| |
| emitLoad(base, regT1, regT0); |
| emitLoad(property, regT3, regT2); |
| Call call = callOperation(operationHasIndexedPropertyDefault, dst, m_codeBlock->globalObject(), JSValueRegs(regT1, regT0), JSValueRegs(regT3, regT2), byValInfo); |
| |
| m_byValCompilationInfo[m_byValInstructionIndex].slowPathTarget = slowPath; |
| m_byValCompilationInfo[m_byValInstructionIndex].returnAddress = call; |
| m_byValInstructionIndex++; |
| } |
| |
| void JIT::emit_op_get_direct_pname(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpGetDirectPname>(); |
| int dst = bytecode.m_dst.offset(); |
| int base = bytecode.m_base.offset(); |
| int index = bytecode.m_index.offset(); |
| int enumerator = bytecode.m_enumerator.offset(); |
| |
| // Check that base is a cell |
| emitLoadPayload(base, regT0); |
| emitJumpSlowCaseIfNotJSCell(base); |
| |
| // Check the structure |
| emitLoadPayload(enumerator, regT1); |
| load32(Address(regT0, JSCell::structureIDOffset()), regT2); |
| addSlowCase(branch32(NotEqual, regT2, Address(regT1, JSPropertyNameEnumerator::cachedStructureIDOffset()))); |
| |
| // Compute the offset |
| emitLoadPayload(index, regT2); |
| // If index is less than the enumerator's cached inline storage, then it's an inline access |
| Jump outOfLineAccess = branch32(AboveOrEqual, regT2, Address(regT1, JSPropertyNameEnumerator::cachedInlineCapacityOffset())); |
| addPtr(TrustedImm32(JSObject::offsetOfInlineStorage()), regT0); |
| load32(BaseIndex(regT0, regT2, TimesEight, OBJECT_OFFSETOF(JSValue, u.asBits.tag)), regT1); |
| load32(BaseIndex(regT0, regT2, TimesEight, OBJECT_OFFSETOF(JSValue, u.asBits.payload)), regT0); |
| |
| Jump done = jump(); |
| |
| // Otherwise it's out of line |
| outOfLineAccess.link(this); |
| loadPtr(Address(regT0, JSObject::butterflyOffset()), regT0); |
| sub32(Address(regT1, JSPropertyNameEnumerator::cachedInlineCapacityOffset()), regT2); |
| neg32(regT2); |
| int32_t offsetOfFirstProperty = static_cast<int32_t>(offsetInButterfly(firstOutOfLineOffset)) * sizeof(EncodedJSValue); |
| load32(BaseIndex(regT0, regT2, TimesEight, offsetOfFirstProperty + OBJECT_OFFSETOF(JSValue, u.asBits.tag)), regT1); |
| load32(BaseIndex(regT0, regT2, TimesEight, offsetOfFirstProperty + OBJECT_OFFSETOF(JSValue, u.asBits.payload)), regT0); |
| |
| done.link(this); |
| emitValueProfilingSite(bytecode.metadata(m_codeBlock)); |
| emitStore(dst, regT1, regT0); |
| } |
| |
| void JIT::emit_op_enumerator_structure_pname(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpEnumeratorStructurePname>(); |
| int dst = bytecode.m_dst.offset(); |
| int enumerator = bytecode.m_enumerator.offset(); |
| int index = bytecode.m_index.offset(); |
| |
| emitLoadPayload(index, regT0); |
| emitLoadPayload(enumerator, regT1); |
| Jump inBounds = branch32(Below, regT0, Address(regT1, JSPropertyNameEnumerator::endStructurePropertyIndexOffset())); |
| |
| move(TrustedImm32(JSValue::NullTag), regT2); |
| move(TrustedImm32(0), regT0); |
| |
| Jump done = jump(); |
| inBounds.link(this); |
| |
| loadPtr(Address(regT1, JSPropertyNameEnumerator::cachedPropertyNamesVectorOffset()), regT1); |
| loadPtr(BaseIndex(regT1, regT0, timesPtr()), regT0); |
| move(TrustedImm32(JSValue::CellTag), regT2); |
| |
| done.link(this); |
| emitStore(dst, regT2, regT0); |
| } |
| |
| void JIT::emit_op_enumerator_generic_pname(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpEnumeratorGenericPname>(); |
| int dst = bytecode.m_dst.offset(); |
| int enumerator = bytecode.m_enumerator.offset(); |
| int index = bytecode.m_index.offset(); |
| |
| emitLoadPayload(index, regT0); |
| emitLoadPayload(enumerator, regT1); |
| Jump inBounds = branch32(Below, regT0, Address(regT1, JSPropertyNameEnumerator::endGenericPropertyIndexOffset())); |
| |
| move(TrustedImm32(JSValue::NullTag), regT2); |
| move(TrustedImm32(0), regT0); |
| |
| Jump done = jump(); |
| inBounds.link(this); |
| |
| loadPtr(Address(regT1, JSPropertyNameEnumerator::cachedPropertyNamesVectorOffset()), regT1); |
| loadPtr(BaseIndex(regT1, regT0, timesPtr()), regT0); |
| move(TrustedImm32(JSValue::CellTag), regT2); |
| |
| done.link(this); |
| emitStore(dst, regT2, regT0); |
| } |
| |
| void JIT::emit_op_profile_type(const Instruction* currentInstruction) |
| { |
| auto bytecode = currentInstruction->as<OpProfileType>(); |
| auto& metadata = bytecode.metadata(m_codeBlock); |
| TypeLocation* cachedTypeLocation = metadata.m_typeLocation; |
| int valueToProfile = bytecode.m_targetVirtualRegister.offset(); |
| |
| // Load payload in T0. Load tag in T3. |
| emitLoadPayload(valueToProfile, regT0); |
| emitLoadTag(valueToProfile, regT3); |
| |
| JumpList jumpToEnd; |
| |
| jumpToEnd.append(branchIfEmpty(regT3)); |
| |
| // Compile in a predictive type check, if possible, to see if we can skip writing to the log. |
| // These typechecks are inlined to match those of the 32-bit JSValue type checks. |
| if (cachedTypeLocation->m_lastSeenType == TypeUndefined) |
| jumpToEnd.append(branchIfUndefined(regT3)); |
| else if (cachedTypeLocation->m_lastSeenType == TypeNull) |
| jumpToEnd.append(branchIfNull(regT3)); |
| else if (cachedTypeLocation->m_lastSeenType == TypeBoolean) |
| jumpToEnd.append(branchIfBoolean(regT3, InvalidGPRReg)); |
| else if (cachedTypeLocation->m_lastSeenType == TypeAnyInt) |
| jumpToEnd.append(branchIfInt32(regT3)); |
| else if (cachedTypeLocation->m_lastSeenType == TypeNumber) { |
| jumpToEnd.append(branchIfNumber(JSValueRegs(regT3, regT0), regT1)); |
| } else if (cachedTypeLocation->m_lastSeenType == TypeString) { |
| Jump isNotCell = branchIfNotCell(regT3); |
| jumpToEnd.append(branchIfString(regT0)); |
| isNotCell.link(this); |
| } |
| |
| // Load the type profiling log into T2. |
| TypeProfilerLog* cachedTypeProfilerLog = m_vm->typeProfilerLog(); |
| move(TrustedImmPtr(cachedTypeProfilerLog), regT2); |
| |
| // Load the next log entry into T1. |
| loadPtr(Address(regT2, TypeProfilerLog::currentLogEntryOffset()), regT1); |
| |
| // Store the JSValue onto the log entry. |
| store32(regT0, Address(regT1, TypeProfilerLog::LogEntry::valueOffset() + OBJECT_OFFSETOF(JSValue, u.asBits.payload))); |
| store32(regT3, Address(regT1, TypeProfilerLog::LogEntry::valueOffset() + OBJECT_OFFSETOF(JSValue, u.asBits.tag))); |
| |
| // Store the structureID of the cell if argument is a cell, otherwise, store 0 on the log entry. |
| Jump notCell = branchIfNotCell(regT3); |
| load32(Address(regT0, JSCell::structureIDOffset()), regT0); |
| store32(regT0, Address(regT1, TypeProfilerLog::LogEntry::structureIDOffset())); |
| Jump skipNotCell = jump(); |
| notCell.link(this); |
| store32(TrustedImm32(0), Address(regT1, TypeProfilerLog::LogEntry::structureIDOffset())); |
| skipNotCell.link(this); |
| |
| // Store the typeLocation on the log entry. |
| move(TrustedImmPtr(cachedTypeLocation), regT0); |
| store32(regT0, Address(regT1, TypeProfilerLog::LogEntry::locationOffset())); |
| |
| // Increment the current log entry. |
| addPtr(TrustedImm32(sizeof(TypeProfilerLog::LogEntry)), regT1); |
| store32(regT1, Address(regT2, TypeProfilerLog::currentLogEntryOffset())); |
| jumpToEnd.append(branchPtr(NotEqual, regT1, TrustedImmPtr(cachedTypeProfilerLog->logEndPtr()))); |
| // Clear the log if we're at the end of the log. |
| callOperation(operationProcessTypeProfilerLog, &vm()); |
| |
| jumpToEnd.link(this); |
| } |
| |
| void JIT::emit_op_log_shadow_chicken_prologue(const Instruction* currentInstruction) |
| { |
| RELEASE_ASSERT(vm().shadowChicken()); |
| updateTopCallFrame(); |
| static_assert(nonArgGPR0 != regT0 && nonArgGPR0 != regT2, "we will have problems if this is true."); |
| auto bytecode = currentInstruction->as<OpLogShadowChickenPrologue>(); |
| GPRReg shadowPacketReg = regT0; |
| GPRReg scratch1Reg = nonArgGPR0; // This must be a non-argument register. |
| GPRReg scratch2Reg = regT2; |
| ensureShadowChickenPacket(vm(), shadowPacketReg, scratch1Reg, scratch2Reg); |
| |
| scratch1Reg = regT4; |
| emitLoadPayload(bytecode.m_scope.offset(), regT3); |
| logShadowChickenProloguePacket(shadowPacketReg, scratch1Reg, regT3); |
| } |
| |
| void JIT::emit_op_log_shadow_chicken_tail(const Instruction* currentInstruction) |
| { |
| RELEASE_ASSERT(vm().shadowChicken()); |
| updateTopCallFrame(); |
| static_assert(nonArgGPR0 != regT0 && nonArgGPR0 != regT2, "we will have problems if this is true."); |
| auto bytecode = currentInstruction->as<OpLogShadowChickenTail>(); |
| GPRReg shadowPacketReg = regT0; |
| GPRReg scratch1Reg = nonArgGPR0; // This must be a non-argument register. |
| GPRReg scratch2Reg = regT2; |
| ensureShadowChickenPacket(vm(), shadowPacketReg, scratch1Reg, scratch2Reg); |
| emitLoadPayload(bytecode.m_thisValue.offset(), regT2); |
| emitLoadTag(bytecode.m_thisValue.offset(), regT1); |
| JSValueRegs thisRegs(regT1, regT2); |
| emitLoadPayload(bytecode.m_scope.offset(), regT3); |
| logShadowChickenTailPacket(shadowPacketReg, thisRegs, regT3, m_codeBlock, CallSiteIndex(BytecodeIndex(bitwise_cast<uint32_t>(currentInstruction)))); |
| } |
| |
| } // namespace JSC |
| |
| #endif // USE(JSVALUE32_64) |
| #endif // ENABLE(JIT) |