diff options
Diffstat (limited to 'Source/JavaScriptCore/dfg/DFGOSRExitCompiler64.cpp')
| -rw-r--r-- | Source/JavaScriptCore/dfg/DFGOSRExitCompiler64.cpp | 793 |
1 files changed, 235 insertions, 558 deletions
diff --git a/Source/JavaScriptCore/dfg/DFGOSRExitCompiler64.cpp b/Source/JavaScriptCore/dfg/DFGOSRExitCompiler64.cpp index 5f0ba1a68..6999e5cfb 100644 --- a/Source/JavaScriptCore/dfg/DFGOSRExitCompiler64.cpp +++ b/Source/JavaScriptCore/dfg/DFGOSRExitCompiler64.cpp @@ -1,5 +1,5 @@ /* - * Copyright (C) 2011, 2013 Apple Inc. All rights reserved. + * Copyright (C) 2011, 2013-2015 Apple Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions @@ -29,53 +29,37 @@ #if ENABLE(DFG_JIT) && USE(JSVALUE64) #include "DFGOperations.h" -#include "Operations.h" +#include "DFGOSRExitCompilerCommon.h" +#include "DFGSpeculativeJIT.h" +#include "JSCInlines.h" +#include "VirtualRegister.h" + #include <wtf/DataLog.h> namespace JSC { namespace DFG { void OSRExitCompiler::compileExit(const OSRExit& exit, const Operands<ValueRecovery>& operands, SpeculationRecovery* recovery) { - // 1) Pro-forma stuff. -#if DFG_ENABLE(DEBUG_VERBOSE) - dataLogF("OSR exit for ("); - for (CodeOrigin codeOrigin = exit.m_codeOrigin; ; codeOrigin = codeOrigin.inlineCallFrame->caller) { - dataLogF("bc#%u", codeOrigin.bytecodeIndex); - if (!codeOrigin.inlineCallFrame) - break; - dataLogF(" -> %p ", codeOrigin.inlineCallFrame->executable.get()); - } - dataLogF(") "); - dumpOperands(operands, WTF::dataFile()); -#endif + m_jit.jitAssertTagsInPlace(); + // Pro-forma stuff. if (Options::printEachOSRExit()) { SpeculationFailureDebugInfo* debugInfo = new SpeculationFailureDebugInfo; debugInfo->codeBlock = m_jit.codeBlock(); + debugInfo->kind = exit.m_kind; + debugInfo->bytecodeOffset = exit.m_codeOrigin.bytecodeIndex; m_jit.debugCall(debugOperationPrintSpeculationFailure, debugInfo); } -#if DFG_ENABLE(JIT_BREAK_ON_SPECULATION_FAILURE) - m_jit.breakpoint(); -#endif - -#if DFG_ENABLE(SUCCESS_STATS) - static SamplingCounter counter("SpeculationFailure"); - m_jit.emitCount(counter); -#endif - - // 2) Perform speculation recovery. This only comes into play when an operation - // starts mutating state before verifying the speculation it has already made. - - GPRReg alreadyBoxed = InvalidGPRReg; + // Perform speculation recovery. This only comes into play when an operation + // starts mutating state before verifying the speculation it has already made. if (recovery) { switch (recovery->type()) { case SpeculativeAdd: m_jit.sub32(recovery->src(), recovery->dest()); m_jit.or64(GPRInfo::tagTypeNumberRegister, recovery->dest()); - alreadyBoxed = recovery->dest(); break; case BooleanSpeculationCheck: @@ -87,7 +71,7 @@ void OSRExitCompiler::compileExit(const OSRExit& exit, const Operands<ValueRecov } } - // 3) Refine some array and/or value profile, if appropriate. + // Refine some array and/or value profile, if appropriate. if (!!exit.m_jsValueSource) { if (exit.m_kind == BadCache || exit.m_kind == BadIndexingType) { @@ -113,8 +97,13 @@ void OSRExitCompiler::compileExit(const OSRExit& exit, const Operands<ValueRecov scratch1 = AssemblyHelpers::selectScratchGPR(usedRegister); scratch2 = AssemblyHelpers::selectScratchGPR(usedRegister, scratch1); - m_jit.push(scratch1); - m_jit.push(scratch2); + if (isARM64()) { + m_jit.pushToSave(scratch1); + m_jit.pushToSave(scratch2); + } else { + m_jit.push(scratch1); + m_jit.push(scratch2); + } GPRReg value; if (exit.m_jsValueSource.isAddress()) { @@ -123,15 +112,20 @@ void OSRExitCompiler::compileExit(const OSRExit& exit, const Operands<ValueRecov } else value = exit.m_jsValueSource.gpr(); - m_jit.loadPtr(AssemblyHelpers::Address(value, JSCell::structureOffset()), scratch1); - m_jit.storePtr(scratch1, arrayProfile->addressOfLastSeenStructure()); - m_jit.load8(AssemblyHelpers::Address(scratch1, Structure::indexingTypeOffset()), scratch1); + m_jit.load32(AssemblyHelpers::Address(value, JSCell::structureIDOffset()), scratch1); + m_jit.store32(scratch1, arrayProfile->addressOfLastSeenStructureID()); + m_jit.load8(AssemblyHelpers::Address(value, JSCell::indexingTypeOffset()), scratch1); m_jit.move(AssemblyHelpers::TrustedImm32(1), scratch2); m_jit.lshift32(scratch1, scratch2); m_jit.or32(scratch2, AssemblyHelpers::AbsoluteAddress(arrayProfile->addressOfArrayModes())); - m_jit.pop(scratch2); - m_jit.pop(scratch1); + if (isARM64()) { + m_jit.popToRestore(scratch2); + m_jit.popToRestore(scratch1); + } else { + m_jit.pop(scratch2); + m_jit.pop(scratch1); + } } } @@ -148,90 +142,58 @@ void OSRExitCompiler::compileExit(const OSRExit& exit, const Operands<ValueRecov m_jit.store64(exit.m_jsValueSource.gpr(), bucket); } } + + // What follows is an intentionally simple OSR exit implementation that generates + // fairly poor code but is very easy to hack. In particular, it dumps all state that + // needs conversion into a scratch buffer so that in step 6, where we actually do the + // conversions, we know that all temp registers are free to use and the variable is + // definitely in a well-known spot in the scratch buffer regardless of whether it had + // originally been in a register or spilled. This allows us to decouple "where was + // the variable" from "how was it represented". Consider that the + // Int32DisplacedInJSStack recovery: it tells us that the value is in a + // particular place and that that place holds an unboxed int32. We have two different + // places that a value could be (displaced, register) and a bunch of different + // ways of representing a value. The number of recoveries is two * a bunch. The code + // below means that we have to have two + a bunch cases rather than two * a bunch. + // Once we have loaded the value from wherever it was, the reboxing is the same + // regardless of its location. Likewise, before we do the reboxing, the way we get to + // the value (i.e. where we load it from) is the same regardless of its type. Because + // the code below always dumps everything into a scratch buffer first, the two + // questions become orthogonal, which simplifies adding new types and adding new + // locations. + // + // This raises the question: does using such a suboptimal implementation of OSR exit, + // where we always emit code to dump all state into a scratch buffer only to then + // dump it right back into the stack, hurt us in any way? The asnwer is that OSR exits + // are rare. Our tiering strategy ensures this. This is because if an OSR exit is + // taken more than ~100 times, we jettison the DFG code block along with all of its + // exits. It is impossible for an OSR exit - i.e. the code we compile below - to + // execute frequently enough for the codegen to matter that much. It probably matters + // enough that we don't want to turn this into some super-slow function call, but so + // long as we're generating straight-line code, that code can be pretty bad. Also + // because we tend to exit only along one OSR exit from any DFG code block - that's an + // empirical result that we're extremely confident about - the code size of this + // doesn't matter much. Hence any attempt to optimize the codegen here is just purely + // harmful to the system: it probably won't reduce either net memory usage or net + // execution time. It will only prevent us from cleanly decoupling "where was the + // variable" from "how was it represented", which will make it more difficult to add + // features in the future and it will make it harder to reason about bugs. - // 4) Figure out how many scratch slots we'll need. We need one for every GPR/FPR - // whose destination is now occupied by a DFG virtual register, and we need - // one for every displaced virtual register if there are more than - // GPRInfo::numberOfRegisters of them. Also see if there are any constants, - // any undefined slots, any FPR slots, and any unboxed ints. - - Vector<bool> poisonedVirtualRegisters(operands.numberOfLocals()); - for (unsigned i = 0; i < poisonedVirtualRegisters.size(); ++i) - poisonedVirtualRegisters[i] = false; - - unsigned numberOfPoisonedVirtualRegisters = 0; - unsigned numberOfDisplacedVirtualRegisters = 0; + // Save all state from GPRs into the scratch buffer. - // Booleans for fast checks. We expect that most OSR exits do not have to rebox - // Int32s, have no FPRs, and have no constants. If there are constants, we - // expect most of them to be jsUndefined(); if that's true then we handle that - // specially to minimize code size and execution time. - bool haveUnboxedInt32s = false; - bool haveUnboxedDoubles = false; - bool haveFPRs = false; - bool haveConstants = false; - bool haveUndefined = false; - bool haveUInt32s = false; - bool haveArguments = false; + ScratchBuffer* scratchBuffer = m_jit.vm()->scratchBufferForSize(sizeof(EncodedJSValue) * operands.size()); + EncodedJSValue* scratch = scratchBuffer ? static_cast<EncodedJSValue*>(scratchBuffer->dataBuffer()) : 0; for (size_t index = 0; index < operands.size(); ++index) { const ValueRecovery& recovery = operands[index]; + switch (recovery.technique()) { - case Int32DisplacedInJSStack: - case DoubleDisplacedInJSStack: - case DisplacedInJSStack: - numberOfDisplacedVirtualRegisters++; - ASSERT((int)recovery.virtualRegister() >= 0); - - // See if we might like to store to this virtual register before doing - // virtual register shuffling. If so, we say that the virtual register - // is poisoned: it cannot be stored to until after displaced virtual - // registers are handled. We track poisoned virtual register carefully - // to ensure this happens efficiently. Note that we expect this case - // to be rare, so the handling of it is optimized for the cases in - // which it does not happen. - if (recovery.virtualRegister() < (int)operands.numberOfLocals()) { - switch (operands.local(recovery.virtualRegister()).technique()) { - case InGPR: - case UnboxedInt32InGPR: - case UInt32InGPR: - case InFPR: - if (!poisonedVirtualRegisters[recovery.virtualRegister()]) { - poisonedVirtualRegisters[recovery.virtualRegister()] = true; - numberOfPoisonedVirtualRegisters++; - } - break; - default: - break; - } - } - break; - + case InGPR: case UnboxedInt32InGPR: - case AlreadyInJSStackAsUnboxedInt32: - haveUnboxedInt32s = true; - break; - - case AlreadyInJSStackAsUnboxedDouble: - haveUnboxedDoubles = true; - break; - - case UInt32InGPR: - haveUInt32s = true; - break; - - case InFPR: - haveFPRs = true; - break; - - case Constant: - haveConstants = true; - if (recovery.constant().isUndefined()) - haveUndefined = true; - break; - - case ArgumentsThatWereNotCreated: - haveArguments = true; + case UnboxedInt52InGPR: + case UnboxedStrictInt52InGPR: + case UnboxedCellInGPR: + m_jit.store64(recovery.gpr(), scratch + index); break; default: @@ -239,478 +201,193 @@ void OSRExitCompiler::compileExit(const OSRExit& exit, const Operands<ValueRecov } } -#if DFG_ENABLE(DEBUG_VERBOSE) - dataLogF(" "); - if (numberOfPoisonedVirtualRegisters) - dataLogF("Poisoned=%u ", numberOfPoisonedVirtualRegisters); - if (numberOfDisplacedVirtualRegisters) - dataLogF("Displaced=%u ", numberOfDisplacedVirtualRegisters); - if (haveUnboxedInt32s) - dataLogF("UnboxedInt32 "); - if (haveUnboxedDoubles) - dataLogF("UnboxedDoubles "); - if (haveUInt32s) - dataLogF("UInt32 "); - if (haveFPRs) - dataLogF("FPR "); - if (haveConstants) - dataLogF("Constants "); - if (haveUndefined) - dataLogF("Undefined "); - dataLogF(" "); -#endif - - ScratchBuffer* scratchBuffer = m_jit.vm()->scratchBufferForSize(sizeof(EncodedJSValue) * std::max(haveUInt32s ? 2u : 0u, numberOfPoisonedVirtualRegisters + (numberOfDisplacedVirtualRegisters <= GPRInfo::numberOfRegisters ? 0 : numberOfDisplacedVirtualRegisters))); - EncodedJSValue* scratchDataBuffer = scratchBuffer ? static_cast<EncodedJSValue*>(scratchBuffer->dataBuffer()) : 0; + // And voila, all GPRs are free to reuse. - // From here on, the code assumes that it is profitable to maximize the distance - // between when something is computed and when it is stored. - - // 5) Perform all reboxing of integers. - - if (haveUnboxedInt32s || haveUInt32s) { - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - switch (recovery.technique()) { - case UnboxedInt32InGPR: - if (recovery.gpr() != alreadyBoxed) - m_jit.or64(GPRInfo::tagTypeNumberRegister, recovery.gpr()); - break; - - case AlreadyInJSStackAsUnboxedInt32: - m_jit.store32(AssemblyHelpers::TrustedImm32(static_cast<uint32_t>(TagTypeNumber >> 32)), AssemblyHelpers::tagFor(static_cast<VirtualRegister>(operands.operandForIndex(index)))); - break; - - case UInt32InGPR: { - // This occurs when the speculative JIT left an unsigned 32-bit integer - // in a GPR. If it's positive, we can just box the int. Otherwise we - // need to turn it into a boxed double. - - // We don't try to be clever with register allocation here; we assume - // that the program is using FPRs and we don't try to figure out which - // ones it is using. Instead just temporarily save fpRegT0 and then - // restore it. This makes sense because this path is not cheap to begin - // with, and should happen very rarely. - - GPRReg addressGPR = GPRInfo::regT0; - if (addressGPR == recovery.gpr()) - addressGPR = GPRInfo::regT1; - - m_jit.store64(addressGPR, scratchDataBuffer); - m_jit.move(AssemblyHelpers::TrustedImmPtr(scratchDataBuffer + 1), addressGPR); - m_jit.storeDouble(FPRInfo::fpRegT0, addressGPR); - - AssemblyHelpers::Jump positive = m_jit.branch32(AssemblyHelpers::GreaterThanOrEqual, recovery.gpr(), AssemblyHelpers::TrustedImm32(0)); - - m_jit.convertInt32ToDouble(recovery.gpr(), FPRInfo::fpRegT0); - m_jit.addDouble(AssemblyHelpers::AbsoluteAddress(&AssemblyHelpers::twoToThe32), FPRInfo::fpRegT0); - m_jit.boxDouble(FPRInfo::fpRegT0, recovery.gpr()); - - AssemblyHelpers::Jump done = m_jit.jump(); - - positive.link(&m_jit); - - m_jit.or64(GPRInfo::tagTypeNumberRegister, recovery.gpr()); - - done.link(&m_jit); - - m_jit.loadDouble(addressGPR, FPRInfo::fpRegT0); - m_jit.load64(scratchDataBuffer, addressGPR); - break; - } - - default: - break; - } - } - } - - // 6) Dump all non-poisoned GPRs. For poisoned GPRs, save them into the scratch storage. - // Note that GPRs do not have a fast change (like haveFPRs) because we expect that - // most OSR failure points will have at least one GPR that needs to be dumped. - - initializePoisoned(operands.numberOfLocals()); - unsigned currentPoisonIndex = 0; + // Save all state from FPRs into the scratch buffer. for (size_t index = 0; index < operands.size(); ++index) { const ValueRecovery& recovery = operands[index]; - int operand = operands.operandForIndex(index); + switch (recovery.technique()) { - case InGPR: - case UnboxedInt32InGPR: - case UInt32InGPR: - if (operands.isVariable(index) && poisonedVirtualRegisters[operands.variableForIndex(index)]) { - m_jit.store64(recovery.gpr(), scratchDataBuffer + currentPoisonIndex); - m_poisonScratchIndices[operands.variableForIndex(index)] = currentPoisonIndex; - currentPoisonIndex++; - } else - m_jit.store64(recovery.gpr(), AssemblyHelpers::addressFor((VirtualRegister)operand)); + case UnboxedDoubleInFPR: + case InFPR: + m_jit.move(AssemblyHelpers::TrustedImmPtr(scratch + index), GPRInfo::regT0); + m_jit.storeDouble(recovery.fpr(), MacroAssembler::Address(GPRInfo::regT0)); break; + default: break; } } - // At this point all GPRs are available for scratch use. + // Now, all FPRs are also free. - if (haveFPRs) { - // 7) Box all doubles (relies on there being more GPRs than FPRs) - - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - if (recovery.technique() != InFPR) - continue; - FPRReg fpr = recovery.fpr(); - GPRReg gpr = GPRInfo::toRegister(FPRInfo::toIndex(fpr)); - m_jit.boxDouble(fpr, gpr); - } - - // 8) Dump all doubles into the stack, or to the scratch storage if - // the destination virtual register is poisoned. + // Save all state from the stack into the scratch buffer. For simplicity we + // do this even for state that's already in the right place on the stack. + // It makes things simpler later. + + for (size_t index = 0; index < operands.size(); ++index) { + const ValueRecovery& recovery = operands[index]; - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - if (recovery.technique() != InFPR) - continue; - GPRReg gpr = GPRInfo::toRegister(FPRInfo::toIndex(recovery.fpr())); - if (operands.isVariable(index) && poisonedVirtualRegisters[operands.variableForIndex(index)]) { - m_jit.store64(gpr, scratchDataBuffer + currentPoisonIndex); - m_poisonScratchIndices[operands.variableForIndex(index)] = currentPoisonIndex; - currentPoisonIndex++; - } else - m_jit.store64(gpr, AssemblyHelpers::addressFor((VirtualRegister)operands.operandForIndex(index))); + switch (recovery.technique()) { + case DisplacedInJSStack: + case CellDisplacedInJSStack: + case BooleanDisplacedInJSStack: + case Int32DisplacedInJSStack: + case DoubleDisplacedInJSStack: + case Int52DisplacedInJSStack: + case StrictInt52DisplacedInJSStack: + m_jit.load64(AssemblyHelpers::addressFor(recovery.virtualRegister()), GPRInfo::regT0); + m_jit.store64(GPRInfo::regT0, scratch + index); + break; + + default: + break; } } + + // Need to ensure that the stack pointer accounts for the worst-case stack usage at exit. This + // could toast some stack that the DFG used. We need to do it before storing to stack offsets + // used by baseline. + m_jit.addPtr( + CCallHelpers::TrustedImm32( + -m_jit.codeBlock()->jitCode()->dfgCommon()->requiredRegisterCountForExit * sizeof(Register)), + CCallHelpers::framePointerRegister, CCallHelpers::stackPointerRegister); + + // Restore the DFG callee saves and then save the ones the baseline JIT uses. + m_jit.emitRestoreCalleeSaves(); + m_jit.emitSaveCalleeSavesFor(m_jit.baselineCodeBlock()); + + // The tag registers are needed to materialize recoveries below. + m_jit.emitMaterializeTagCheckRegisters(); + + if (exit.isExceptionHandler()) + m_jit.copyCalleeSavesToVMCalleeSavesBuffer(); + + // Do all data format conversions and store the results into the stack. - // At this point all GPRs and FPRs are available for scratch use. - - // 9) Box all unboxed doubles in the stack. - if (haveUnboxedDoubles) { - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - if (recovery.technique() != AlreadyInJSStackAsUnboxedDouble) - continue; - m_jit.loadDouble(AssemblyHelpers::addressFor((VirtualRegister)operands.operandForIndex(index)), FPRInfo::fpRegT0); + for (size_t index = 0; index < operands.size(); ++index) { + const ValueRecovery& recovery = operands[index]; + VirtualRegister reg = operands.virtualRegisterForIndex(index); + + if (reg.isLocal() && reg.toLocal() < static_cast<int>(m_jit.baselineCodeBlock()->calleeSaveSpaceAsVirtualRegisters())) + continue; + + int operand = reg.offset(); + + switch (recovery.technique()) { + case InGPR: + case UnboxedCellInGPR: + case DisplacedInJSStack: + case CellDisplacedInJSStack: + case BooleanDisplacedInJSStack: + case InFPR: + m_jit.load64(scratch + index, GPRInfo::regT0); + m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor(operand)); + break; + + case UnboxedInt32InGPR: + case Int32DisplacedInJSStack: + m_jit.load64(scratch + index, GPRInfo::regT0); + m_jit.zeroExtend32ToPtr(GPRInfo::regT0, GPRInfo::regT0); + m_jit.or64(GPRInfo::tagTypeNumberRegister, GPRInfo::regT0); + m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor(operand)); + break; + + case UnboxedInt52InGPR: + case Int52DisplacedInJSStack: + m_jit.load64(scratch + index, GPRInfo::regT0); + m_jit.rshift64( + AssemblyHelpers::TrustedImm32(JSValue::int52ShiftAmount), GPRInfo::regT0); + m_jit.boxInt52(GPRInfo::regT0, GPRInfo::regT0, GPRInfo::regT1, FPRInfo::fpRegT0); + m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor(operand)); + break; + + case UnboxedStrictInt52InGPR: + case StrictInt52DisplacedInJSStack: + m_jit.load64(scratch + index, GPRInfo::regT0); + m_jit.boxInt52(GPRInfo::regT0, GPRInfo::regT0, GPRInfo::regT1, FPRInfo::fpRegT0); + m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor(operand)); + break; + + case UnboxedDoubleInFPR: + case DoubleDisplacedInJSStack: + m_jit.move(AssemblyHelpers::TrustedImmPtr(scratch + index), GPRInfo::regT0); + m_jit.loadDouble(MacroAssembler::Address(GPRInfo::regT0), FPRInfo::fpRegT0); + m_jit.purifyNaN(FPRInfo::fpRegT0); m_jit.boxDouble(FPRInfo::fpRegT0, GPRInfo::regT0); - m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor((VirtualRegister)operands.operandForIndex(index))); - } - } - - ASSERT(currentPoisonIndex == numberOfPoisonedVirtualRegisters); - - // 10) Reshuffle displaced virtual registers. Optimize for the case that - // the number of displaced virtual registers is not more than the number - // of available physical registers. - - if (numberOfDisplacedVirtualRegisters) { - if (numberOfDisplacedVirtualRegisters <= GPRInfo::numberOfRegisters) { - // So far this appears to be the case that triggers all the time, but - // that is far from guaranteed. - - unsigned displacementIndex = 0; - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - switch (recovery.technique()) { - case DisplacedInJSStack: - m_jit.load64(AssemblyHelpers::addressFor(recovery.virtualRegister()), GPRInfo::toRegister(displacementIndex++)); - break; - - case Int32DisplacedInJSStack: { - GPRReg gpr = GPRInfo::toRegister(displacementIndex++); - m_jit.load32(AssemblyHelpers::addressFor(recovery.virtualRegister()), gpr); - m_jit.or64(GPRInfo::tagTypeNumberRegister, gpr); - break; - } - - case DoubleDisplacedInJSStack: { - GPRReg gpr = GPRInfo::toRegister(displacementIndex++); - m_jit.load64(AssemblyHelpers::addressFor(recovery.virtualRegister()), gpr); - m_jit.sub64(GPRInfo::tagTypeNumberRegister, gpr); - break; - } - - default: - break; - } - } - - displacementIndex = 0; - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - switch (recovery.technique()) { - case DisplacedInJSStack: - case Int32DisplacedInJSStack: - case DoubleDisplacedInJSStack: - m_jit.store64(GPRInfo::toRegister(displacementIndex++), AssemblyHelpers::addressFor((VirtualRegister)operands.operandForIndex(index))); - break; - - default: - break; - } - } - } else { - // FIXME: This should use the shuffling algorithm that we use - // for speculative->non-speculative jumps, if we ever discover that - // some hot code with lots of live values that get displaced and - // spilled really enjoys frequently failing speculation. - - // For now this code is engineered to be correct but probably not - // super. In particular, it correctly handles cases where for example - // the displacements are a permutation of the destination values, like - // - // 1 -> 2 - // 2 -> 1 - // - // It accomplishes this by simply lifting all of the virtual registers - // from their old (DFG JIT) locations and dropping them in a scratch - // location in memory, and then transferring from that scratch location - // to their new (old JIT) locations. - - unsigned scratchIndex = numberOfPoisonedVirtualRegisters; - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - - switch (recovery.technique()) { - case DisplacedInJSStack: - m_jit.load64(AssemblyHelpers::addressFor(recovery.virtualRegister()), GPRInfo::regT0); - m_jit.store64(GPRInfo::regT0, scratchDataBuffer + scratchIndex++); - break; - - case Int32DisplacedInJSStack: { - m_jit.load32(AssemblyHelpers::addressFor(recovery.virtualRegister()), GPRInfo::regT0); - m_jit.or64(GPRInfo::tagTypeNumberRegister, GPRInfo::regT0); - m_jit.store64(GPRInfo::regT0, scratchDataBuffer + scratchIndex++); - break; - } - - case DoubleDisplacedInJSStack: { - m_jit.load64(AssemblyHelpers::addressFor(recovery.virtualRegister()), GPRInfo::regT0); - m_jit.sub64(GPRInfo::tagTypeNumberRegister, GPRInfo::regT0); - m_jit.store64(GPRInfo::regT0, scratchDataBuffer + scratchIndex++); - break; - } - - default: - break; - } - } - - scratchIndex = numberOfPoisonedVirtualRegisters; - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - switch (recovery.technique()) { - case DisplacedInJSStack: - case Int32DisplacedInJSStack: - case DoubleDisplacedInJSStack: - m_jit.load64(scratchDataBuffer + scratchIndex++, GPRInfo::regT0); - m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor((VirtualRegister)operands.operandForIndex(index))); - break; - - default: - break; - } - } - - ASSERT(scratchIndex == numberOfPoisonedVirtualRegisters + numberOfDisplacedVirtualRegisters); - } - } - - // 11) Dump all poisoned virtual registers. - - if (numberOfPoisonedVirtualRegisters) { - for (int virtualRegister = 0; virtualRegister < (int)operands.numberOfLocals(); ++virtualRegister) { - if (!poisonedVirtualRegisters[virtualRegister]) - continue; + m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor(operand)); + break; - const ValueRecovery& recovery = operands.local(virtualRegister); - switch (recovery.technique()) { - case InGPR: - case UnboxedInt32InGPR: - case UInt32InGPR: - case InFPR: - m_jit.load64(scratchDataBuffer + poisonIndex(virtualRegister), GPRInfo::regT0); - m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor((VirtualRegister)virtualRegister)); - break; - - default: - break; - } + case Constant: + m_jit.store64( + AssemblyHelpers::TrustedImm64(JSValue::encode(recovery.constant())), + AssemblyHelpers::addressFor(operand)); + break; + + case DirectArgumentsThatWereNotCreated: + case ClonedArgumentsThatWereNotCreated: + // Don't do this, yet. + break; + + default: + RELEASE_ASSERT_NOT_REACHED(); + break; } } - - // 12) Dump all constants. Optimize for Undefined, since that's a constant we see - // often. - if (haveConstants) { - if (haveUndefined) - m_jit.move(AssemblyHelpers::TrustedImm64(JSValue::encode(jsUndefined())), GPRInfo::regT0); - - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - if (recovery.technique() != Constant) - continue; - if (recovery.constant().isUndefined()) - m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor((VirtualRegister)operands.operandForIndex(index))); - else - m_jit.store64(AssemblyHelpers::TrustedImm64(JSValue::encode(recovery.constant())), AssemblyHelpers::addressFor((VirtualRegister)operands.operandForIndex(index))); - } - } - - // 13) Adjust the old JIT's execute counter. Since we are exiting OSR, we know - // that all new calls into this code will go to the new JIT, so the execute - // counter only affects call frames that performed OSR exit and call frames - // that were still executing the old JIT at the time of another call frame's - // OSR exit. We want to ensure that the following is true: + // Now that things on the stack are recovered, do the arguments recovery. We assume that arguments + // recoveries don't recursively refer to each other. But, we don't try to assume that they only + // refer to certain ranges of locals. Hence why we need to do this here, once the stack is sensible. + // Note that we also roughly assume that the arguments might still be materialized outside of its + // inline call frame scope - but for now the DFG wouldn't do that. + + emitRestoreArguments(operands); + + // Adjust the old JIT's execute counter. Since we are exiting OSR, we know + // that all new calls into this code will go to the new JIT, so the execute + // counter only affects call frames that performed OSR exit and call frames + // that were still executing the old JIT at the time of another call frame's + // OSR exit. We want to ensure that the following is true: // - // (a) Code the performs an OSR exit gets a chance to reenter optimized - // code eventually, since optimized code is faster. But we don't - // want to do such reentery too aggressively (see (c) below). + // (a) Code the performs an OSR exit gets a chance to reenter optimized + // code eventually, since optimized code is faster. But we don't + // want to do such reentery too aggressively (see (c) below). // - // (b) If there is code on the call stack that is still running the old - // JIT's code and has never OSR'd, then it should get a chance to - // perform OSR entry despite the fact that we've exited. + // (b) If there is code on the call stack that is still running the old + // JIT's code and has never OSR'd, then it should get a chance to + // perform OSR entry despite the fact that we've exited. // - // (c) Code the performs an OSR exit should not immediately retry OSR - // entry, since both forms of OSR are expensive. OSR entry is - // particularly expensive. + // (c) Code the performs an OSR exit should not immediately retry OSR + // entry, since both forms of OSR are expensive. OSR entry is + // particularly expensive. // - // (d) Frequent OSR failures, even those that do not result in the code - // running in a hot loop, result in recompilation getting triggered. + // (d) Frequent OSR failures, even those that do not result in the code + // running in a hot loop, result in recompilation getting triggered. // - // To ensure (c), we'd like to set the execute counter to - // counterValueForOptimizeAfterWarmUp(). This seems like it would endanger - // (a) and (b), since then every OSR exit would delay the opportunity for - // every call frame to perform OSR entry. Essentially, if OSR exit happens - // frequently and the function has few loops, then the counter will never - // become non-negative and OSR entry will never be triggered. OSR entry - // will only happen if a loop gets hot in the old JIT, which does a pretty - // good job of ensuring (a) and (b). But that doesn't take care of (d), - // since each speculation failure would reset the execute counter. - // So we check here if the number of speculation failures is significantly - // larger than the number of successes (we want 90% success rate), and if - // there have been a large enough number of failures. If so, we set the - // counter to 0; otherwise we set the counter to - // counterValueForOptimizeAfterWarmUp(). - - handleExitCounts(exit); - - // 14) Reify inlined call frames. - - ASSERT(m_jit.baselineCodeBlock()->getJITType() == JITCode::BaselineJIT); - m_jit.storePtr(AssemblyHelpers::TrustedImmPtr(m_jit.baselineCodeBlock()), AssemblyHelpers::addressFor((VirtualRegister)JSStack::CodeBlock)); - - for (CodeOrigin codeOrigin = exit.m_codeOrigin; codeOrigin.inlineCallFrame; codeOrigin = codeOrigin.inlineCallFrame->caller) { - InlineCallFrame* inlineCallFrame = codeOrigin.inlineCallFrame; - CodeBlock* baselineCodeBlock = m_jit.baselineCodeBlockFor(codeOrigin); - CodeBlock* baselineCodeBlockForCaller = m_jit.baselineCodeBlockFor(inlineCallFrame->caller); - Vector<BytecodeAndMachineOffset>& decodedCodeMap = m_jit.decodedCodeMapFor(baselineCodeBlockForCaller); - unsigned returnBytecodeIndex = inlineCallFrame->caller.bytecodeIndex + OPCODE_LENGTH(op_call); - BytecodeAndMachineOffset* mapping = binarySearch<BytecodeAndMachineOffset, unsigned>(decodedCodeMap, decodedCodeMap.size(), returnBytecodeIndex, BytecodeAndMachineOffset::getBytecodeIndex); - - ASSERT(mapping); - ASSERT(mapping->m_bytecodeIndex == returnBytecodeIndex); - - void* jumpTarget = baselineCodeBlockForCaller->getJITCode().executableAddressAtOffset(mapping->m_machineCodeOffset); - - GPRReg callerFrameGPR; - if (inlineCallFrame->caller.inlineCallFrame) { - m_jit.addPtr(AssemblyHelpers::TrustedImm32(inlineCallFrame->caller.inlineCallFrame->stackOffset * sizeof(EncodedJSValue)), GPRInfo::callFrameRegister, GPRInfo::regT3); - callerFrameGPR = GPRInfo::regT3; - } else - callerFrameGPR = GPRInfo::callFrameRegister; - - m_jit.storePtr(AssemblyHelpers::TrustedImmPtr(baselineCodeBlock), AssemblyHelpers::addressFor((VirtualRegister)(inlineCallFrame->stackOffset + JSStack::CodeBlock))); - if (!inlineCallFrame->isClosureCall()) - m_jit.store64(AssemblyHelpers::TrustedImm64(JSValue::encode(JSValue(inlineCallFrame->callee->scope()))), AssemblyHelpers::addressFor((VirtualRegister)(inlineCallFrame->stackOffset + JSStack::ScopeChain))); - m_jit.store64(callerFrameGPR, AssemblyHelpers::addressFor((VirtualRegister)(inlineCallFrame->stackOffset + JSStack::CallerFrame))); - m_jit.storePtr(AssemblyHelpers::TrustedImmPtr(jumpTarget), AssemblyHelpers::addressFor((VirtualRegister)(inlineCallFrame->stackOffset + JSStack::ReturnPC))); - m_jit.store32(AssemblyHelpers::TrustedImm32(inlineCallFrame->arguments.size()), AssemblyHelpers::payloadFor((VirtualRegister)(inlineCallFrame->stackOffset + JSStack::ArgumentCount))); - if (!inlineCallFrame->isClosureCall()) - m_jit.store64(AssemblyHelpers::TrustedImm64(JSValue::encode(JSValue(inlineCallFrame->callee.get()))), AssemblyHelpers::addressFor((VirtualRegister)(inlineCallFrame->stackOffset + JSStack::Callee))); - } - - // 15) Create arguments if necessary and place them into the appropriate aliased - // registers. - - if (haveArguments) { - HashSet<InlineCallFrame*, DefaultHash<InlineCallFrame*>::Hash, - NullableHashTraits<InlineCallFrame*> > didCreateArgumentsObject; - - for (size_t index = 0; index < operands.size(); ++index) { - const ValueRecovery& recovery = operands[index]; - if (recovery.technique() != ArgumentsThatWereNotCreated) - continue; - int operand = operands.operandForIndex(index); - // Find the right inline call frame. - InlineCallFrame* inlineCallFrame = 0; - for (InlineCallFrame* current = exit.m_codeOrigin.inlineCallFrame; - current; - current = current->caller.inlineCallFrame) { - if (current->stackOffset <= operand) { - inlineCallFrame = current; - break; - } - } - - if (!m_jit.baselineCodeBlockFor(inlineCallFrame)->usesArguments()) - continue; - int argumentsRegister = m_jit.argumentsRegisterFor(inlineCallFrame); - if (didCreateArgumentsObject.add(inlineCallFrame).isNewEntry) { - // We know this call frame optimized out an arguments object that - // the baseline JIT would have created. Do that creation now. - if (inlineCallFrame) { - m_jit.addPtr(AssemblyHelpers::TrustedImm32(inlineCallFrame->stackOffset * sizeof(EncodedJSValue)), GPRInfo::callFrameRegister, GPRInfo::regT0); - m_jit.setupArguments(GPRInfo::regT0); - } else - m_jit.setupArgumentsExecState(); - m_jit.move( - AssemblyHelpers::TrustedImmPtr( - bitwise_cast<void*>(operationCreateArguments)), - GPRInfo::nonArgGPR0); - m_jit.call(GPRInfo::nonArgGPR0); - m_jit.store64(GPRInfo::returnValueGPR, AssemblyHelpers::addressFor(argumentsRegister)); - m_jit.store64( - GPRInfo::returnValueGPR, - AssemblyHelpers::addressFor(unmodifiedArgumentsRegister(argumentsRegister))); - m_jit.move(GPRInfo::returnValueGPR, GPRInfo::regT0); // no-op move on almost all platforms. - } - - m_jit.load64(AssemblyHelpers::addressFor(argumentsRegister), GPRInfo::regT0); - m_jit.store64(GPRInfo::regT0, AssemblyHelpers::addressFor(operand)); - } - } - - // 16) Load the result of the last bytecode operation into regT0. - - if (exit.m_lastSetOperand != std::numeric_limits<int>::max()) - m_jit.load64(AssemblyHelpers::addressFor((VirtualRegister)exit.m_lastSetOperand), GPRInfo::cachedResultRegister); - - // 17) Adjust the call frame pointer. - - if (exit.m_codeOrigin.inlineCallFrame) - m_jit.addPtr(AssemblyHelpers::TrustedImm32(exit.m_codeOrigin.inlineCallFrame->stackOffset * sizeof(EncodedJSValue)), GPRInfo::callFrameRegister); - - // 18) Jump into the corresponding baseline JIT code. - - CodeBlock* baselineCodeBlock = m_jit.baselineCodeBlockFor(exit.m_codeOrigin); - Vector<BytecodeAndMachineOffset>& decodedCodeMap = m_jit.decodedCodeMapFor(baselineCodeBlock); - - BytecodeAndMachineOffset* mapping = binarySearch<BytecodeAndMachineOffset, unsigned>(decodedCodeMap, decodedCodeMap.size(), exit.m_codeOrigin.bytecodeIndex, BytecodeAndMachineOffset::getBytecodeIndex); - - ASSERT(mapping); - ASSERT(mapping->m_bytecodeIndex == exit.m_codeOrigin.bytecodeIndex); - - void* jumpTarget = baselineCodeBlock->getJITCode().executableAddressAtOffset(mapping->m_machineCodeOffset); - - ASSERT(GPRInfo::regT1 != GPRInfo::cachedResultRegister); - - m_jit.move(AssemblyHelpers::TrustedImmPtr(jumpTarget), GPRInfo::regT1); - - m_jit.jump(GPRInfo::regT1); + // To ensure (c), we'd like to set the execute counter to + // counterValueForOptimizeAfterWarmUp(). This seems like it would endanger + // (a) and (b), since then every OSR exit would delay the opportunity for + // every call frame to perform OSR entry. Essentially, if OSR exit happens + // frequently and the function has few loops, then the counter will never + // become non-negative and OSR entry will never be triggered. OSR entry + // will only happen if a loop gets hot in the old JIT, which does a pretty + // good job of ensuring (a) and (b). But that doesn't take care of (d), + // since each speculation failure would reset the execute counter. + // So we check here if the number of speculation failures is significantly + // larger than the number of successes (we want 90% success rate), and if + // there have been a large enough number of failures. If so, we set the + // counter to 0; otherwise we set the counter to + // counterValueForOptimizeAfterWarmUp(). + + handleExitCounts(m_jit, exit); + + // Reify inlined call frames. + + reifyInlinedCallFrames(m_jit, exit); -#if DFG_ENABLE(DEBUG_VERBOSE) - dataLogF("-> %p\n", jumpTarget); -#endif + // And finish. + adjustAndJumpToTarget(m_jit, exit); } } } // namespace JSC::DFG |