X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FCodeGen%2FMachineTraceMetrics.cpp;h=76f2aa7bdba00f63d8a9804b07b5966017af4972;hb=25529b337f75a4b9b174592e2c95136e781bd824;hp=ff7f94ea9423613be26fcd42254736104ba57ad0;hpb=8677f2ff9acf317461987b439ede693f01baa5ec;p=oota-llvm.git diff --git a/lib/CodeGen/MachineTraceMetrics.cpp b/lib/CodeGen/MachineTraceMetrics.cpp index ff7f94ea942..76f2aa7bdba 100644 --- a/lib/CodeGen/MachineTraceMetrics.cpp +++ b/lib/CodeGen/MachineTraceMetrics.cpp @@ -52,8 +52,8 @@ void MachineTraceMetrics::getAnalysisUsage(AnalysisUsage &AU) const { bool MachineTraceMetrics::runOnMachineFunction(MachineFunction &Func) { MF = &Func; - TII = MF->getTarget().getInstrInfo(); - TRI = MF->getTarget().getRegisterInfo(); + TII = MF->getSubtarget().getInstrInfo(); + TRI = MF->getSubtarget().getRegisterInfo(); MRI = &MF->getRegInfo(); Loops = &getAnalysis(); const TargetSubtargetInfo &ST = @@ -97,19 +97,17 @@ MachineTraceMetrics::getResources(const MachineBasicBlock *MBB) { unsigned PRKinds = SchedModel.getNumProcResourceKinds(); SmallVector PRCycles(PRKinds); - for (MachineBasicBlock::const_iterator I = MBB->begin(), E = MBB->end(); - I != E; ++I) { - const MachineInstr *MI = I; - if (MI->isTransient()) + for (const auto &MI : *MBB) { + if (MI.isTransient()) continue; ++InstrCount; - if (MI->isCall()) + if (MI.isCall()) FBI->HasCalls = true; // Count processor resources used. if (!SchedModel.hasInstrSchedModel()) continue; - const MCSchedClassDesc *SC = SchedModel.resolveSchedClass(MI); + const MCSchedClassDesc *SC = SchedModel.resolveSchedClass(&MI); if (!SC->isValid()) continue; @@ -570,9 +568,8 @@ MachineTraceMetrics::Ensemble::invalidate(const MachineBasicBlock *BadMBB) { // invalidated, but their instructions will stay the same, so there is no // need to erase the Cycle entries. They will be overwritten when we // recompute. - for (MachineBasicBlock::const_iterator I = BadMBB->begin(), E = BadMBB->end(); - I != E; ++I) - Cycles.erase(I); + for (const auto &I : *BadMBB) + Cycles.erase(&I); } void MachineTraceMetrics::Ensemble::verify() const { @@ -830,16 +827,13 @@ computeInstrDepths(const MachineBasicBlock *MBB) { if (TBI.HasValidInstrHeights) TBI.CriticalPath = computeCrossBlockCriticalPath(TBI); - for (MachineBasicBlock::const_iterator I = MBB->begin(), E = MBB->end(); - I != E; ++I) { - const MachineInstr *UseMI = I; - + for (const auto &UseMI : *MBB) { // Collect all data dependencies. Deps.clear(); - if (UseMI->isPHI()) - getPHIDeps(UseMI, Deps, TBI.Pred, MTM.MRI); - else if (getDataDeps(UseMI, Deps, MTM.MRI)) - updatePhysDepsDownwards(UseMI, Deps, RegUnits, MTM.TRI); + if (UseMI.isPHI()) + getPHIDeps(&UseMI, Deps, TBI.Pred, MTM.MRI); + else if (getDataDeps(&UseMI, Deps, MTM.MRI)) + updatePhysDepsDownwards(&UseMI, Deps, RegUnits, MTM.TRI); // Filter and process dependencies, computing the earliest issue cycle. unsigned Cycle = 0; @@ -855,20 +849,20 @@ computeInstrDepths(const MachineBasicBlock *MBB) { // Add latency if DefMI is a real instruction. Transients get latency 0. if (!Dep.DefMI->isTransient()) DepCycle += MTM.SchedModel - .computeOperandLatency(Dep.DefMI, Dep.DefOp, UseMI, Dep.UseOp); + .computeOperandLatency(Dep.DefMI, Dep.DefOp, &UseMI, Dep.UseOp); Cycle = std::max(Cycle, DepCycle); } // Remember the instruction depth. - InstrCycles &MICycles = Cycles[UseMI]; + InstrCycles &MICycles = Cycles[&UseMI]; MICycles.Depth = Cycle; if (!TBI.HasValidInstrHeights) { - DEBUG(dbgs() << Cycle << '\t' << *UseMI); + DEBUG(dbgs() << Cycle << '\t' << UseMI); continue; } // Update critical path length. TBI.CriticalPath = std::max(TBI.CriticalPath, Cycle + MICycles.Height); - DEBUG(dbgs() << TBI.CriticalPath << '\t' << Cycle << '\t' << *UseMI); + DEBUG(dbgs() << TBI.CriticalPath << '\t' << Cycle << '\t' << UseMI); } } } @@ -1057,16 +1051,16 @@ computeInstrHeights(const MachineBasicBlock *MBB) { Succ = Loop->getHeader(); if (Succ) { - for (MachineBasicBlock::const_iterator I = Succ->begin(), E = Succ->end(); - I != E && I->isPHI(); ++I) { - const MachineInstr *PHI = I; + for (const auto &PHI : *Succ) { + if (!PHI.isPHI()) + break; Deps.clear(); - getPHIDeps(PHI, Deps, MBB, MTM.MRI); + getPHIDeps(&PHI, Deps, MBB, MTM.MRI); if (!Deps.empty()) { // Loop header PHI heights are all 0. - unsigned Height = TBI.Succ ? Cycles.lookup(PHI).Height : 0; - DEBUG(dbgs() << "pred\t" << Height << '\t' << *PHI); - if (pushDepHeight(Deps.front(), PHI, Height, + unsigned Height = TBI.Succ ? Cycles.lookup(&PHI).Height : 0; + DEBUG(dbgs() << "pred\t" << Height << '\t' << PHI); + if (pushDepHeight(Deps.front(), &PHI, Height, Heights, MTM.SchedModel, MTM.TII)) addLiveIns(Deps.front().DefMI, Deps.front().DefOp, Stack); } @@ -1175,6 +1169,7 @@ MachineTraceMetrics::Trace::getPHIDepth(const MachineInstr *PHI) const { return DepCycle; } +/// When bottom is set include instructions in current block in estimate. unsigned MachineTraceMetrics::Trace::getResourceDepth(bool Bottom) const { // Find the limiting processor resource. // Numbers have been pre-scaled to be comparable. @@ -1191,7 +1186,9 @@ unsigned MachineTraceMetrics::Trace::getResourceDepth(bool Bottom) const { // Convert to cycle count. PRMax = TE.MTM.getCycles(PRMax); + /// All instructions before current block unsigned Instrs = TBI.InstrDepth; + // plus instructions in current block if (Bottom) Instrs += TE.MTM.BlockInfo[getBlockNum()].InstrCount; if (unsigned IW = TE.MTM.SchedModel.getIssueWidth()) @@ -1200,44 +1197,72 @@ unsigned MachineTraceMetrics::Trace::getResourceDepth(bool Bottom) const { return std::max(Instrs, PRMax); } - -unsigned MachineTraceMetrics::Trace:: -getResourceLength(ArrayRef Extrablocks, - ArrayRef ExtraInstrs) const { +unsigned MachineTraceMetrics::Trace::getResourceLength( + ArrayRef Extrablocks, + ArrayRef ExtraInstrs, + ArrayRef RemoveInstrs) const { // Add up resources above and below the center block. ArrayRef PRDepths = TE.getProcResourceDepths(getBlockNum()); ArrayRef PRHeights = TE.getProcResourceHeights(getBlockNum()); unsigned PRMax = 0; - for (unsigned K = 0; K != PRDepths.size(); ++K) { - unsigned PRCycles = PRDepths[K] + PRHeights[K]; - for (unsigned I = 0; I != Extrablocks.size(); ++I) - PRCycles += TE.MTM.getProcResourceCycles(Extrablocks[I]->getNumber())[K]; - for (unsigned I = 0; I != ExtraInstrs.size(); ++I) { - const MCSchedClassDesc* SC = ExtraInstrs[I]; + + // Capture computing cycles from extra instructions + auto extraCycles = [this](ArrayRef Instrs, + unsigned ResourceIdx) + ->unsigned { + unsigned Cycles = 0; + for (unsigned I = 0; I != Instrs.size(); ++I) { + const MCSchedClassDesc *SC = Instrs[I]; if (!SC->isValid()) continue; for (TargetSchedModel::ProcResIter - PI = TE.MTM.SchedModel.getWriteProcResBegin(SC), - PE = TE.MTM.SchedModel.getWriteProcResEnd(SC); PI != PE; ++PI) { - if (PI->ProcResourceIdx != K) + PI = TE.MTM.SchedModel.getWriteProcResBegin(SC), + PE = TE.MTM.SchedModel.getWriteProcResEnd(SC); + PI != PE; ++PI) { + if (PI->ProcResourceIdx != ResourceIdx) continue; - PRCycles += (PI->Cycles * TE.MTM.SchedModel.getResourceFactor(K)); + Cycles += + (PI->Cycles * TE.MTM.SchedModel.getResourceFactor(ResourceIdx)); } } + return Cycles; + }; + + for (unsigned K = 0; K != PRDepths.size(); ++K) { + unsigned PRCycles = PRDepths[K] + PRHeights[K]; + for (unsigned I = 0; I != Extrablocks.size(); ++I) + PRCycles += TE.MTM.getProcResourceCycles(Extrablocks[I]->getNumber())[K]; + PRCycles += extraCycles(ExtraInstrs, K); + PRCycles -= extraCycles(RemoveInstrs, K); PRMax = std::max(PRMax, PRCycles); } // Convert to cycle count. PRMax = TE.MTM.getCycles(PRMax); + // Instrs: #instructions in current trace outside current block. unsigned Instrs = TBI.InstrDepth + TBI.InstrHeight; + // Add instruction count from the extra blocks. for (unsigned i = 0, e = Extrablocks.size(); i != e; ++i) Instrs += TE.MTM.getResources(Extrablocks[i])->InstrCount; + Instrs += ExtraInstrs.size(); + Instrs -= RemoveInstrs.size(); if (unsigned IW = TE.MTM.SchedModel.getIssueWidth()) Instrs /= IW; // Assume issue width 1 without a schedule model. return std::max(Instrs, PRMax); } +bool MachineTraceMetrics::Trace::isDepInTrace(const MachineInstr *DefMI, + const MachineInstr *UseMI) const { + if (DefMI->getParent() == UseMI->getParent()) + return true; + + const TraceBlockInfo &DepTBI = TE.BlockInfo[DefMI->getParent()->getNumber()]; + const TraceBlockInfo &TBI = TE.BlockInfo[UseMI->getParent()->getNumber()]; + + return DepTBI.isUsefulDominator(TBI); +} + void MachineTraceMetrics::Ensemble::print(raw_ostream &OS) const { OS << getName() << " ensemble:\n"; for (unsigned i = 0, e = BlockInfo.size(); i != e; ++i) {