#define DEBUG_TYPE "isel"
#include "ScheduleDAGSDNodes.h"
#include "SelectionDAGBuilder.h"
-#include "llvm/CodeGen/FunctionLoweringInfo.h"
-#include "llvm/CodeGen/SelectionDAGISel.h"
-#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/DebugInfo.h"
#include "llvm/Constants.h"
+#include "llvm/DebugInfo.h"
#include "llvm/Function.h"
#include "llvm/InlineAsm.h"
#include "llvm/Instructions.h"
#include "llvm/IntrinsicInst.h"
#include "llvm/LLVMContext.h"
#include "llvm/Module.h"
+#include "llvm/Analysis/AliasAnalysis.h"
+#include "llvm/Analysis/BranchProbabilityInfo.h"
#include "llvm/CodeGen/FastISel.h"
+#include "llvm/CodeGen/FunctionLoweringInfo.h"
#include "llvm/CodeGen/GCStrategy.h"
#include "llvm/CodeGen/GCMetadata.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/ScheduleHazardRecognizer.h"
#include "llvm/CodeGen/SchedulerRegistry.h"
#include "llvm/CodeGen/SelectionDAG.h"
+#include "llvm/CodeGen/SelectionDAGISel.h"
#include "llvm/Target/TargetRegisterInfo.h"
#include "llvm/Target/TargetIntrinsicInfo.h"
#include "llvm/Target/TargetInstrInfo.h"
+#include "llvm/Target/TargetLibraryInfo.h"
#include "llvm/Target/TargetLowering.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
using namespace llvm;
STATISTIC(NumFastIselFailures, "Number of instructions fast isel failed on");
+STATISTIC(NumFastIselSuccess, "Number of instructions fast isel selected");
STATISTIC(NumFastIselBlocks, "Number of blocks selected entirely by fast isel");
STATISTIC(NumDAGBlocks, "Number of blocks selected using DAG");
STATISTIC(NumDAGIselRetries,"Number of times dag isel has to try another path");
#ifndef NDEBUG
-STATISTIC(NumBBWithOutOfOrderLineInfo,
- "Number of blocks with out of order line number info");
-STATISTIC(NumMBBWithOutOfOrderLineInfo,
- "Number of machine blocks with out of order line number info");
+static cl::opt<bool>
+EnableFastISelVerbose2("fast-isel-verbose2", cl::Hidden,
+ cl::desc("Enable extra verbose messages in the \"fast\" "
+ "instruction selector"));
+ // Terminators
+STATISTIC(NumFastIselFailRet,"Fast isel fails on Ret");
+STATISTIC(NumFastIselFailBr,"Fast isel fails on Br");
+STATISTIC(NumFastIselFailSwitch,"Fast isel fails on Switch");
+STATISTIC(NumFastIselFailIndirectBr,"Fast isel fails on IndirectBr");
+STATISTIC(NumFastIselFailInvoke,"Fast isel fails on Invoke");
+STATISTIC(NumFastIselFailResume,"Fast isel fails on Resume");
+STATISTIC(NumFastIselFailUnreachable,"Fast isel fails on Unreachable");
+
+ // Standard binary operators...
+STATISTIC(NumFastIselFailAdd,"Fast isel fails on Add");
+STATISTIC(NumFastIselFailFAdd,"Fast isel fails on FAdd");
+STATISTIC(NumFastIselFailSub,"Fast isel fails on Sub");
+STATISTIC(NumFastIselFailFSub,"Fast isel fails on FSub");
+STATISTIC(NumFastIselFailMul,"Fast isel fails on Mul");
+STATISTIC(NumFastIselFailFMul,"Fast isel fails on FMul");
+STATISTIC(NumFastIselFailUDiv,"Fast isel fails on UDiv");
+STATISTIC(NumFastIselFailSDiv,"Fast isel fails on SDiv");
+STATISTIC(NumFastIselFailFDiv,"Fast isel fails on FDiv");
+STATISTIC(NumFastIselFailURem,"Fast isel fails on URem");
+STATISTIC(NumFastIselFailSRem,"Fast isel fails on SRem");
+STATISTIC(NumFastIselFailFRem,"Fast isel fails on FRem");
+
+ // Logical operators...
+STATISTIC(NumFastIselFailAnd,"Fast isel fails on And");
+STATISTIC(NumFastIselFailOr,"Fast isel fails on Or");
+STATISTIC(NumFastIselFailXor,"Fast isel fails on Xor");
+
+ // Memory instructions...
+STATISTIC(NumFastIselFailAlloca,"Fast isel fails on Alloca");
+STATISTIC(NumFastIselFailLoad,"Fast isel fails on Load");
+STATISTIC(NumFastIselFailStore,"Fast isel fails on Store");
+STATISTIC(NumFastIselFailAtomicCmpXchg,"Fast isel fails on AtomicCmpXchg");
+STATISTIC(NumFastIselFailAtomicRMW,"Fast isel fails on AtomicRWM");
+STATISTIC(NumFastIselFailFence,"Fast isel fails on Frence");
+STATISTIC(NumFastIselFailGetElementPtr,"Fast isel fails on GetElementPtr");
+
+ // Convert instructions...
+STATISTIC(NumFastIselFailTrunc,"Fast isel fails on Trunc");
+STATISTIC(NumFastIselFailZExt,"Fast isel fails on ZExt");
+STATISTIC(NumFastIselFailSExt,"Fast isel fails on SExt");
+STATISTIC(NumFastIselFailFPTrunc,"Fast isel fails on FPTrunc");
+STATISTIC(NumFastIselFailFPExt,"Fast isel fails on FPExt");
+STATISTIC(NumFastIselFailFPToUI,"Fast isel fails on FPToUI");
+STATISTIC(NumFastIselFailFPToSI,"Fast isel fails on FPToSI");
+STATISTIC(NumFastIselFailUIToFP,"Fast isel fails on UIToFP");
+STATISTIC(NumFastIselFailSIToFP,"Fast isel fails on SIToFP");
+STATISTIC(NumFastIselFailIntToPtr,"Fast isel fails on IntToPtr");
+STATISTIC(NumFastIselFailPtrToInt,"Fast isel fails on PtrToInt");
+STATISTIC(NumFastIselFailBitCast,"Fast isel fails on BitCast");
+
+ // Other instructions...
+STATISTIC(NumFastIselFailICmp,"Fast isel fails on ICmp");
+STATISTIC(NumFastIselFailFCmp,"Fast isel fails on FCmp");
+STATISTIC(NumFastIselFailPHI,"Fast isel fails on PHI");
+STATISTIC(NumFastIselFailSelect,"Fast isel fails on Select");
+STATISTIC(NumFastIselFailCall,"Fast isel fails on Call");
+STATISTIC(NumFastIselFailShl,"Fast isel fails on Shl");
+STATISTIC(NumFastIselFailLShr,"Fast isel fails on LShr");
+STATISTIC(NumFastIselFailAShr,"Fast isel fails on AShr");
+STATISTIC(NumFastIselFailVAArg,"Fast isel fails on VAArg");
+STATISTIC(NumFastIselFailExtractElement,"Fast isel fails on ExtractElement");
+STATISTIC(NumFastIselFailInsertElement,"Fast isel fails on InsertElement");
+STATISTIC(NumFastIselFailShuffleVector,"Fast isel fails on ShuffleVector");
+STATISTIC(NumFastIselFailExtractValue,"Fast isel fails on ExtractValue");
+STATISTIC(NumFastIselFailInsertValue,"Fast isel fails on InsertValue");
+STATISTIC(NumFastIselFailLandingPad,"Fast isel fails on LandingPad");
#endif
static cl::opt<bool>
EnableFastISelAbort("fast-isel-abort", cl::Hidden,
cl::desc("Enable abort calls when \"fast\" instruction fails"));
+static cl::opt<bool>
+UseMBPI("use-mbpi",
+ cl::desc("use Machine Branch Probability Info"),
+ cl::init(true), cl::Hidden);
+
#ifndef NDEBUG
static cl::opt<bool>
ViewDAGCombine1("view-dag-combine1-dags", cl::Hidden,
CodeGenOpt::Level OptLevel) {
const TargetLowering &TLI = IS->getTargetLowering();
- if (OptLevel == CodeGenOpt::None)
+ if (OptLevel == CodeGenOpt::None ||
+ TLI.getSchedulingPreference() == Sched::Source)
return createSourceListDAGScheduler(IS, OptLevel);
- if (TLI.getSchedulingPreference() == Sched::Latency)
- return createTDListDAGScheduler(IS, OptLevel);
if (TLI.getSchedulingPreference() == Sched::RegPressure)
return createBURRListDAGScheduler(IS, OptLevel);
if (TLI.getSchedulingPreference() == Sched::Hybrid)
return createHybridListDAGScheduler(IS, OptLevel);
+ if (TLI.getSchedulingPreference() == Sched::VLIW)
+ return createVLIWDAGScheduler(IS, OptLevel);
assert(TLI.getSchedulingPreference() == Sched::ILP &&
"Unknown sched type!");
return createILPListDAGScheduler(IS, OptLevel);
"TargetLowering::EmitInstrWithCustomInserter!";
#endif
llvm_unreachable(0);
- return 0;
+}
+
+void TargetLowering::AdjustInstrPostInstrSelection(MachineInstr *MI,
+ SDNode *Node) const {
+ assert(!MI->hasPostISelHook() &&
+ "If a target marks an instruction with 'hasPostISelHook', "
+ "it must implement TargetLowering::AdjustInstrPostInstrSelection!");
}
//===----------------------------------------------------------------------===//
CodeGenOpt::Level OL) :
MachineFunctionPass(ID), TM(tm), TLI(*tm.getTargetLowering()),
FuncInfo(new FunctionLoweringInfo(TLI)),
- CurDAG(new SelectionDAG(tm)),
+ CurDAG(new SelectionDAG(tm, OL)),
SDB(new SelectionDAGBuilder(*CurDAG, *FuncInfo, OL)),
GFI(),
OptLevel(OL),
DAGSize(0) {
initializeGCModuleInfoPass(*PassRegistry::getPassRegistry());
initializeAliasAnalysisAnalysisGroup(*PassRegistry::getPassRegistry());
+ initializeBranchProbabilityInfoPass(*PassRegistry::getPassRegistry());
+ initializeTargetLibraryInfoPass(*PassRegistry::getPassRegistry());
}
SelectionDAGISel::~SelectionDAGISel() {
AU.addPreserved<AliasAnalysis>();
AU.addRequired<GCModuleInfo>();
AU.addPreserved<GCModuleInfo>();
+ AU.addRequired<TargetLibraryInfo>();
+ if (UseMBPI && OptLevel != CodeGenOpt::None)
+ AU.addRequired<BranchProbabilityInfo>();
MachineFunctionPass::getAnalysisUsage(AU);
}
-/// FunctionCallsSetJmp - Return true if the function has a call to setjmp or
-/// other function that gcc recognizes as "returning twice". This is used to
-/// limit code-gen optimizations on the machine function.
-///
-/// FIXME: Remove after <rdar://problem/8031714> is fixed.
-static bool FunctionCallsSetJmp(const Function *F) {
- const Module *M = F->getParent();
- static const char *ReturnsTwiceFns[] = {
- "_setjmp",
- "setjmp",
- "sigsetjmp",
- "setjmp_syscall",
- "savectx",
- "qsetjmp",
- "vfork",
- "getcontext"
- };
-#define NUM_RETURNS_TWICE_FNS sizeof(ReturnsTwiceFns) / sizeof(const char *)
-
- for (unsigned I = 0; I < NUM_RETURNS_TWICE_FNS; ++I)
- if (const Function *Callee = M->getFunction(ReturnsTwiceFns[I])) {
- if (!Callee->use_empty())
- for (Value::const_use_iterator
- I = Callee->use_begin(), E = Callee->use_end();
- I != E; ++I)
- if (const CallInst *CI = dyn_cast<CallInst>(*I))
- if (CI->getParent()->getParent() == F)
- return true;
- }
-
- return false;
-#undef NUM_RETURNS_TWICE_FNS
-}
-
/// SplitCriticalSideEffectEdges - Look for critical edges with a PHI value that
/// may trap on it. In this case we have to split the edge so that the path
/// through the predecessor block that doesn't go to the phi block doesn't
bool SelectionDAGISel::runOnMachineFunction(MachineFunction &mf) {
// Do some sanity-checking on the command-line options.
- assert((!EnableFastISelVerbose || EnableFastISel) &&
+ assert((!EnableFastISelVerbose || TM.Options.EnableFastISel) &&
"-fast-isel-verbose requires -fast-isel");
- assert((!EnableFastISelAbort || EnableFastISel) &&
+ assert((!EnableFastISelAbort || TM.Options.EnableFastISel) &&
"-fast-isel-abort requires -fast-isel");
const Function &Fn = *mf.getFunction();
MF = &mf;
RegInfo = &MF->getRegInfo();
AA = &getAnalysis<AliasAnalysis>();
+ LibInfo = &getAnalysis<TargetLibraryInfo>();
GFI = Fn.hasGC() ? &getAnalysis<GCModuleInfo>().getFunctionInfo(Fn) : 0;
DEBUG(dbgs() << "\n\n\n=== " << Fn.getName() << "\n");
CurDAG->init(*MF);
FuncInfo->set(Fn, *MF);
- SDB->init(GFI, *AA);
+
+ if (UseMBPI && OptLevel != CodeGenOpt::None)
+ FuncInfo->BPI = &getAnalysis<BranchProbabilityInfo>();
+ else
+ FuncInfo->BPI = 0;
+
+ SDB->init(GFI, *AA, LibInfo);
SelectAllBasicBlocks(Fn);
TII.get(TargetOpcode::DBG_VALUE))
.addReg(CopyUseMI->getOperand(0).getReg(), RegState::Debug)
.addImm(Offset).addMetadata(Variable);
- EntryMBB->insertAfter(CopyUseMI, NewMI);
+ MachineBasicBlock::iterator Pos = CopyUseMI;
+ EntryMBB->insertAfter(Pos, NewMI);
}
}
}
const MachineBasicBlock *MBB = I;
for (MachineBasicBlock::const_iterator
II = MBB->begin(), IE = MBB->end(); II != IE; ++II) {
- const TargetInstrDesc &TID = TM.getInstrInfo()->get(II->getOpcode());
+ const MCInstrDesc &MCID = TM.getInstrInfo()->get(II->getOpcode());
- if ((TID.isCall() && !TID.isReturn()) ||
+ if ((MCID.isCall() && !MCID.isReturn()) ||
II->isStackAligningInlineAsm()) {
MFI->setHasCalls(true);
goto done;
}
}
}
- done:;
}
+ done:
// Determine if there is a call to setjmp in the machine function.
- MF->setCallsSetJmp(FunctionCallsSetJmp(&Fn));
+ MF->setExposesReturnsTwice(Fn.callsFunctionThatReturnsTwice());
// Replace forward-declared registers with the registers containing
// the desired value.
// If To is also scheduled to be replaced, find what its ultimate
// replacement is.
for (;;) {
- DenseMap<unsigned, unsigned>::iterator J =
- FuncInfo->RegFixups.find(To);
+ DenseMap<unsigned, unsigned>::iterator J = FuncInfo->RegFixups.find(To);
if (J == E) break;
To = J->second;
}
MRI.replaceRegWith(From, To);
}
+ // Freeze the set of reserved registers now that MachineFrameInfo has been
+ // set up. All the information required by getReservedRegs() should be
+ // available now.
+ MRI.freezeReservedRegs(*MF);
+
// Release function-specific state. SDB and CurDAG are already cleared
// at this point.
FuncInfo->clear();
Worklist.push_back(CurDAG->getRoot().getNode());
- APInt Mask;
APInt KnownZero;
APInt KnownOne;
continue;
unsigned NumSignBits = CurDAG->ComputeNumSignBits(Src);
- Mask = APInt::getAllOnesValue(SrcVT.getSizeInBits());
- CurDAG->ComputeMaskedBits(Src, Mask, KnownZero, KnownOne);
+ CurDAG->ComputeMaskedBits(Src, KnownZero, KnownOne);
FuncInfo->AddLiveOutRegInfo(DestReg, NumSignBits, KnownZero, KnownOne);
} while (!Worklist.empty());
}
GroupName = "Instruction Selection and Scheduling";
std::string BlockName;
int BlockNumber = -1;
+ (void)BlockNumber;
#ifdef NDEBUG
if (ViewDAGCombine1 || ViewLegalizeTypesDAGs || ViewLegalizeDAGs ||
ViewDAGCombine2 || ViewDAGCombineLT || ViewISelDAGs || ViewSchedDAGs ||
#endif
{
BlockNumber = FuncInfo->MBB->getNumber();
- BlockName = MF->getFunction()->getNameStr() + ":" +
- FuncInfo->MBB->getBasicBlock()->getNameStr();
+ BlockName = MF->getName().str() + ":" +
+ FuncInfo->MBB->getBasicBlock()->getName().str();
}
DEBUG(dbgs() << "Initial selection DAG: BB#" << BlockNumber
<< " '" << BlockName << "'\n"; CurDAG->dump());
// Run the DAG combiner in pre-legalize mode.
{
NamedRegionTimer T("DAG Combining 1", GroupName, TimePassesIsEnabled);
- CurDAG->Combine(Unrestricted, *AA, OptLevel);
+ CurDAG->Combine(BeforeLegalizeTypes, *AA, OptLevel);
}
DEBUG(dbgs() << "Optimized lowered selection DAG: BB#" << BlockNumber
{
NamedRegionTimer T("DAG Combining after legalize types", GroupName,
TimePassesIsEnabled);
- CurDAG->Combine(NoIllegalTypes, *AA, OptLevel);
+ CurDAG->Combine(AfterLegalizeTypes, *AA, OptLevel);
}
DEBUG(dbgs() << "Optimized type-legalized selection DAG: BB#" << BlockNumber
{
NamedRegionTimer T("DAG Combining after legalize vectors", GroupName,
TimePassesIsEnabled);
- CurDAG->Combine(NoIllegalOperations, *AA, OptLevel);
+ CurDAG->Combine(AfterLegalizeVectorOps, *AA, OptLevel);
}
DEBUG(dbgs() << "Optimized vector-legalized selection DAG: BB#"
{
NamedRegionTimer T("DAG Legalization", GroupName, TimePassesIsEnabled);
- CurDAG->Legalize(OptLevel);
+ CurDAG->Legalize();
}
DEBUG(dbgs() << "Legalized selection DAG: BB#" << BlockNumber
// Run the DAG combiner in post-legalize mode.
{
NamedRegionTimer T("DAG Combining 2", GroupName, TimePassesIsEnabled);
- CurDAG->Combine(NoIllegalOperations, *AA, OptLevel);
+ CurDAG->Combine(AfterLegalizeDAG, *AA, OptLevel);
}
DEBUG(dbgs() << "Optimized legalized selection DAG: BB#" << BlockNumber
{
NamedRegionTimer T("Instruction Scheduling", GroupName,
TimePassesIsEnabled);
- Scheduler->Run(CurDAG, FuncInfo->MBB, FuncInfo->InsertPt);
+ Scheduler->Run(CurDAG, FuncInfo->MBB);
}
if (ViewSUnitDAGs) Scheduler->viewGraph();
{
NamedRegionTimer T("Instruction Creation", GroupName, TimePassesIsEnabled);
- LastMBB = FuncInfo->MBB = Scheduler->EmitSchedule();
- FuncInfo->InsertPt = Scheduler->InsertPos;
+ // FuncInfo->InsertPt is passed by reference and set to the end of the
+ // scheduled instructions.
+ LastMBB = FuncInfo->MBB = Scheduler->EmitSchedule(FuncInfo->InsertPt);
}
// If the block was split, make sure we update any references that are used to
CurDAG->clear();
}
+namespace {
+/// ISelUpdater - helper class to handle updates of the instruction selection
+/// graph.
+class ISelUpdater : public SelectionDAG::DAGUpdateListener {
+ SelectionDAG::allnodes_iterator &ISelPosition;
+public:
+ ISelUpdater(SelectionDAG &DAG, SelectionDAG::allnodes_iterator &isp)
+ : SelectionDAG::DAGUpdateListener(DAG), ISelPosition(isp) {}
+
+ /// NodeDeleted - Handle nodes deleted from the graph. If the node being
+ /// deleted is the current ISelPosition node, update ISelPosition.
+ ///
+ virtual void NodeDeleted(SDNode *N, SDNode *E) {
+ if (ISelPosition == SelectionDAG::allnodes_iterator(N))
+ ++ISelPosition;
+ }
+};
+} // end anonymous namespace
+
void SelectionDAGISel::DoInstructionSelection() {
DEBUG(errs() << "===== Instruction selection begins: BB#"
<< FuncInfo->MBB->getNumber()
// a reference to the root node, preventing it from being deleted,
// and tracking any changes of the root.
HandleSDNode Dummy(CurDAG->getRoot());
- ISelPosition = SelectionDAG::allnodes_iterator(CurDAG->getRoot().getNode());
+ SelectionDAG::allnodes_iterator ISelPosition (CurDAG->getRoot().getNode());
++ISelPosition;
+ // Make sure that ISelPosition gets properly updated when nodes are deleted
+ // in calls made from this function.
+ ISelUpdater ISU(*CurDAG, ISelPosition);
+
// The AllNodes list is now topological-sorted. Visit the
// nodes by starting at the end of the list (the root of the
// graph) and preceding back toward the beginning (the entry
// If after the replacement this node is not used any more,
// remove this dead node.
- if (Node->use_empty()) { // Don't delete EntryToken, etc.
- ISelUpdater ISU(ISelPosition);
- CurDAG->RemoveDeadNode(Node, &ISU);
- }
+ if (Node->use_empty()) // Don't delete EntryToken, etc.
+ CurDAG->RemoveDeadNode(Node);
}
CurDAG->setRoot(Dummy.getValue());
/// PrepareEHLandingPad - Emit an EH_LABEL, set up live-in registers, and
/// do other setup for EH landing-pad blocks.
void SelectionDAGISel::PrepareEHLandingPad() {
+ MachineBasicBlock *MBB = FuncInfo->MBB;
+
// Add a label to mark the beginning of the landing pad. Deletion of the
// landing pad can thus be detected via the MachineModuleInfo.
- MCSymbol *Label = MF->getMMI().addLandingPad(FuncInfo->MBB);
+ MCSymbol *Label = MF->getMMI().addLandingPad(MBB);
+
+ // Assign the call site to the landing pad's begin label.
+ MF->getMMI().setCallSiteLandingPad(Label, SDB->LPadToCallSiteMap[MBB]);
- const TargetInstrDesc &II = TM.getInstrInfo()->get(TargetOpcode::EH_LABEL);
- BuildMI(*FuncInfo->MBB, FuncInfo->InsertPt, SDB->getCurDebugLoc(), II)
+ const MCInstrDesc &II = TM.getInstrInfo()->get(TargetOpcode::EH_LABEL);
+ BuildMI(*MBB, FuncInfo->InsertPt, SDB->getCurDebugLoc(), II)
.addSym(Label);
// Mark exception register as live in.
- unsigned Reg = TLI.getExceptionAddressRegister();
- if (Reg) FuncInfo->MBB->addLiveIn(Reg);
+ unsigned Reg = TLI.getExceptionPointerRegister();
+ if (Reg) MBB->addLiveIn(Reg);
// Mark exception selector register as live in.
Reg = TLI.getExceptionSelectorRegister();
- if (Reg) FuncInfo->MBB->addLiveIn(Reg);
-
- // FIXME: Hack around an exception handling flaw (PR1508): the personality
- // function and list of typeids logically belong to the invoke (or, if you
- // like, the basic block containing the invoke), and need to be associated
- // with it in the dwarf exception handling tables. Currently however the
- // information is provided by an intrinsic (eh.selector) that can be moved
- // to unexpected places by the optimizers: if the unwind edge is critical,
- // then breaking it can result in the intrinsics being in the successor of
- // the landing pad, not the landing pad itself. This results
- // in exceptions not being caught because no typeids are associated with
- // the invoke. This may not be the only way things can go wrong, but it
- // is the only way we try to work around for the moment.
- const BasicBlock *LLVMBB = FuncInfo->MBB->getBasicBlock();
- const BranchInst *Br = dyn_cast<BranchInst>(LLVMBB->getTerminator());
-
- if (Br && Br->isUnconditional()) { // Critical edge?
- BasicBlock::const_iterator I, E;
- for (I = LLVMBB->begin(), E = --LLVMBB->end(); I != E; ++I)
- if (isa<EHSelectorInst>(I))
- break;
-
- if (I == E)
- // No catch info found - try to extract some from the successor.
- CopyCatchInfo(Br->getSuccessor(0), LLVMBB, &MF->getMMI(), *FuncInfo);
- }
+ if (Reg) MBB->addLiveIn(Reg);
}
-
-
/// TryToFoldFastISelLoad - We're checking to see if we can fold the specified
/// load into the specified FoldInst. Note that we could have a sequence where
/// multiple LLVM IR instructions are folded into the same machineinstr. For
// isn't one of the folded instructions, then we can't succeed here. Handle
// this by scanning the single-use users of the load until we get to FoldInst.
unsigned MaxUsers = 6; // Don't scan down huge single-use chains of instrs.
-
+
const Instruction *TheUser = LI->use_back();
while (TheUser != FoldInst && // Scan up until we find FoldInst.
// Stay in the right block.
// If there are multiple or no uses of this instruction, then bail out.
if (!TheUser->hasOneUse())
return false;
-
+
TheUser = TheUser->use_back();
}
-
+
+ // If we didn't find the fold instruction, then we failed to collapse the
+ // sequence.
+ if (TheUser != FoldInst)
+ return false;
+
// Don't try to fold volatile loads. Target has to deal with alignment
// constraints.
if (LI->isVolatile()) return false;
return FastIS->TryToFoldLoad(User, RI.getOperandNo(), LI);
}
-#ifndef NDEBUG
-/// CheckLineNumbers - Check if basic block instructions follow source order
-/// or not.
-static void CheckLineNumbers(const BasicBlock *BB) {
- unsigned Line = 0;
- unsigned Col = 0;
- for (BasicBlock::const_iterator BI = BB->begin(),
- BE = BB->end(); BI != BE; ++BI) {
- const DebugLoc DL = BI->getDebugLoc();
- if (DL.isUnknown()) continue;
- unsigned L = DL.getLine();
- unsigned C = DL.getCol();
- if (L < Line || (L == Line && C < Col)) {
- ++NumBBWithOutOfOrderLineInfo;
- return;
- }
- Line = L;
- Col = C;
- }
-}
-
-/// CheckLineNumbers - Check if machine basic block instructions follow source
-/// order or not.
-static void CheckLineNumbers(const MachineBasicBlock *MBB) {
- unsigned Line = 0;
- unsigned Col = 0;
- for (MachineBasicBlock::const_iterator MBI = MBB->begin(),
- MBE = MBB->end(); MBI != MBE; ++MBI) {
- const DebugLoc DL = MBI->getDebugLoc();
- if (DL.isUnknown()) continue;
- unsigned L = DL.getLine();
- unsigned C = DL.getCol();
- if (L < Line || (L == Line && C < Col)) {
- ++NumMBBWithOutOfOrderLineInfo;
- return;
- }
- Line = L;
- Col = C;
- }
-}
-#endif
-
/// isFoldedOrDeadInstruction - Return true if the specified instruction is
/// side-effect free and is either dead or folded into a generated instruction.
/// Return false if it needs to be emitted.
return !I->mayWriteToMemory() && // Side-effecting instructions aren't folded.
!isa<TerminatorInst>(I) && // Terminators aren't folded.
!isa<DbgInfoIntrinsic>(I) && // Debug instructions aren't folded.
+ !isa<LandingPadInst>(I) && // Landingpad instructions aren't folded.
!FuncInfo->isExportedInst(I); // Exported instrs must be computed.
}
+#ifndef NDEBUG
+// Collect per Instruction statistics for fast-isel misses. Only those
+// instructions that cause the bail are accounted for. It does not account for
+// instructions higher in the block. Thus, summing the per instructions stats
+// will not add up to what is reported by NumFastIselFailures.
+static void collectFailStats(const Instruction *I) {
+ switch (I->getOpcode()) {
+ default: assert (0 && "<Invalid operator> ");
+
+ // Terminators
+ case Instruction::Ret: NumFastIselFailRet++; return;
+ case Instruction::Br: NumFastIselFailBr++; return;
+ case Instruction::Switch: NumFastIselFailSwitch++; return;
+ case Instruction::IndirectBr: NumFastIselFailIndirectBr++; return;
+ case Instruction::Invoke: NumFastIselFailInvoke++; return;
+ case Instruction::Resume: NumFastIselFailResume++; return;
+ case Instruction::Unreachable: NumFastIselFailUnreachable++; return;
+
+ // Standard binary operators...
+ case Instruction::Add: NumFastIselFailAdd++; return;
+ case Instruction::FAdd: NumFastIselFailFAdd++; return;
+ case Instruction::Sub: NumFastIselFailSub++; return;
+ case Instruction::FSub: NumFastIselFailFSub++; return;
+ case Instruction::Mul: NumFastIselFailMul++; return;
+ case Instruction::FMul: NumFastIselFailFMul++; return;
+ case Instruction::UDiv: NumFastIselFailUDiv++; return;
+ case Instruction::SDiv: NumFastIselFailSDiv++; return;
+ case Instruction::FDiv: NumFastIselFailFDiv++; return;
+ case Instruction::URem: NumFastIselFailURem++; return;
+ case Instruction::SRem: NumFastIselFailSRem++; return;
+ case Instruction::FRem: NumFastIselFailFRem++; return;
+
+ // Logical operators...
+ case Instruction::And: NumFastIselFailAnd++; return;
+ case Instruction::Or: NumFastIselFailOr++; return;
+ case Instruction::Xor: NumFastIselFailXor++; return;
+
+ // Memory instructions...
+ case Instruction::Alloca: NumFastIselFailAlloca++; return;
+ case Instruction::Load: NumFastIselFailLoad++; return;
+ case Instruction::Store: NumFastIselFailStore++; return;
+ case Instruction::AtomicCmpXchg: NumFastIselFailAtomicCmpXchg++; return;
+ case Instruction::AtomicRMW: NumFastIselFailAtomicRMW++; return;
+ case Instruction::Fence: NumFastIselFailFence++; return;
+ case Instruction::GetElementPtr: NumFastIselFailGetElementPtr++; return;
+
+ // Convert instructions...
+ case Instruction::Trunc: NumFastIselFailTrunc++; return;
+ case Instruction::ZExt: NumFastIselFailZExt++; return;
+ case Instruction::SExt: NumFastIselFailSExt++; return;
+ case Instruction::FPTrunc: NumFastIselFailFPTrunc++; return;
+ case Instruction::FPExt: NumFastIselFailFPExt++; return;
+ case Instruction::FPToUI: NumFastIselFailFPToUI++; return;
+ case Instruction::FPToSI: NumFastIselFailFPToSI++; return;
+ case Instruction::UIToFP: NumFastIselFailUIToFP++; return;
+ case Instruction::SIToFP: NumFastIselFailSIToFP++; return;
+ case Instruction::IntToPtr: NumFastIselFailIntToPtr++; return;
+ case Instruction::PtrToInt: NumFastIselFailPtrToInt++; return;
+ case Instruction::BitCast: NumFastIselFailBitCast++; return;
+
+ // Other instructions...
+ case Instruction::ICmp: NumFastIselFailICmp++; return;
+ case Instruction::FCmp: NumFastIselFailFCmp++; return;
+ case Instruction::PHI: NumFastIselFailPHI++; return;
+ case Instruction::Select: NumFastIselFailSelect++; return;
+ case Instruction::Call: NumFastIselFailCall++; return;
+ case Instruction::Shl: NumFastIselFailShl++; return;
+ case Instruction::LShr: NumFastIselFailLShr++; return;
+ case Instruction::AShr: NumFastIselFailAShr++; return;
+ case Instruction::VAArg: NumFastIselFailVAArg++; return;
+ case Instruction::ExtractElement: NumFastIselFailExtractElement++; return;
+ case Instruction::InsertElement: NumFastIselFailInsertElement++; return;
+ case Instruction::ShuffleVector: NumFastIselFailShuffleVector++; return;
+ case Instruction::ExtractValue: NumFastIselFailExtractValue++; return;
+ case Instruction::InsertValue: NumFastIselFailInsertValue++; return;
+ case Instruction::LandingPad: NumFastIselFailLandingPad++; return;
+ }
+}
+#endif
+
void SelectionDAGISel::SelectAllBasicBlocks(const Function &Fn) {
// Initialize the Fast-ISel state, if needed.
FastISel *FastIS = 0;
- if (EnableFastISel)
- FastIS = TLI.createFastISel(*FuncInfo);
+ if (TM.Options.EnableFastISel)
+ FastIS = TLI.createFastISel(*FuncInfo, LibInfo);
// Iterate over all basic blocks in the function.
ReversePostOrderTraversal<const Function*> RPOT(&Fn);
for (ReversePostOrderTraversal<const Function*>::rpo_iterator
I = RPOT.begin(), E = RPOT.end(); I != E; ++I) {
const BasicBlock *LLVMBB = *I;
-#ifndef NDEBUG
- CheckLineNumbers(LLVMBB);
-#endif
if (OptLevel != CodeGenOpt::None) {
bool AllPredsVisited = true;
FastIS->setLastLocalValue(0);
}
+ unsigned NumFastIselRemaining = std::distance(Begin, End);
// Do FastISel on as many instructions as possible.
for (; BI != Begin; --BI) {
const Instruction *Inst = llvm::prior(BI);
// If we no longer require this instruction, skip it.
- if (isFoldedOrDeadInstruction(Inst, FuncInfo))
+ if (isFoldedOrDeadInstruction(Inst, FuncInfo)) {
+ --NumFastIselRemaining;
continue;
+ }
// Bottom-up: reset the insert pos at the top, after any local-value
// instructions.
// Try to select the instruction with FastISel.
if (FastIS->SelectInstruction(Inst)) {
+ --NumFastIselRemaining;
+ ++NumFastIselSuccess;
// If fast isel succeeded, skip over all the folded instructions, and
// then see if there is a load right before the selected instructions.
// Try to fold the load if so.
}
if (BeforeInst != Inst && isa<LoadInst>(BeforeInst) &&
BeforeInst->hasOneUse() &&
- TryToFoldFastISelLoad(cast<LoadInst>(BeforeInst), Inst, FastIS))
+ TryToFoldFastISelLoad(cast<LoadInst>(BeforeInst), Inst, FastIS)) {
// If we succeeded, don't re-select the load.
BI = llvm::next(BasicBlock::const_iterator(BeforeInst));
+ --NumFastIselRemaining;
+ ++NumFastIselSuccess;
+ }
continue;
}
+#ifndef NDEBUG
+ if (EnableFastISelVerbose2)
+ collectFailStats(Inst);
+#endif
+
// Then handle certain instructions as single-LLVM-Instruction blocks.
if (isa<CallInst>(Inst)) {
- ++NumFastIselFailures;
+
if (EnableFastISelVerbose || EnableFastISelAbort) {
dbgs() << "FastISel missed call: ";
Inst->dump();
bool HadTailCall = false;
SelectBasicBlock(Inst, BI, HadTailCall);
+ // Recompute NumFastIselRemaining as Selection DAG instruction
+ // selection may have handled the call, input args, etc.
+ unsigned RemainingNow = std::distance(Begin, BI);
+ NumFastIselFailures += NumFastIselRemaining - RemainingNow;
+
// If the call was emitted as a tail call, we're done with the block.
if (HadTailCall) {
--BI;
break;
}
+ NumFastIselRemaining = RemainingNow;
continue;
}
- // Otherwise, give up on FastISel for the rest of the block.
- // For now, be a little lenient about non-branch terminators.
- if (!isa<TerminatorInst>(Inst) || isa<BranchInst>(Inst)) {
- ++NumFastIselFailures;
+ if (isa<TerminatorInst>(Inst) && !isa<BranchInst>(Inst)) {
+ // Don't abort, and use a different message for terminator misses.
+ NumFastIselFailures += NumFastIselRemaining;
+ if (EnableFastISelVerbose || EnableFastISelAbort) {
+ dbgs() << "FastISel missed terminator: ";
+ Inst->dump();
+ }
+ } else {
+ NumFastIselFailures += NumFastIselRemaining;
if (EnableFastISelVerbose || EnableFastISelAbort) {
dbgs() << "FastISel miss: ";
Inst->dump();
}
delete FastIS;
-#ifndef NDEBUG
- for (MachineFunction::const_iterator MBI = MF->begin(), MBE = MF->end();
- MBI != MBE; ++MBI)
- CheckLineNumbers(MBI);
-#endif
+ SDB->clearDanglingDebugInfo();
}
void
CodeGenAndEmitDAG();
}
+ uint32_t UnhandledWeight = 0;
+ for (unsigned j = 0, ej = SDB->BitTestCases[i].Cases.size(); j != ej; ++j)
+ UnhandledWeight += SDB->BitTestCases[i].Cases[j].ExtraWeight;
+
for (unsigned j = 0, ej = SDB->BitTestCases[i].Cases.size(); j != ej; ++j) {
+ UnhandledWeight -= SDB->BitTestCases[i].Cases[j].ExtraWeight;
// Set the current basic block to the mbb we wish to insert the code into
FuncInfo->MBB = SDB->BitTestCases[i].Cases[j].ThisBB;
FuncInfo->InsertPt = FuncInfo->MBB->end();
if (j+1 != ej)
SDB->visitBitTestCase(SDB->BitTestCases[i],
SDB->BitTestCases[i].Cases[j+1].ThisBB,
+ UnhandledWeight,
SDB->BitTestCases[i].Reg,
SDB->BitTestCases[i].Cases[j],
FuncInfo->MBB);
else
SDB->visitBitTestCase(SDB->BitTestCases[i],
SDB->BitTestCases[i].Default,
+ UnhandledWeight,
SDB->BitTestCases[i].Reg,
SDB->BitTestCases[i].Cases[j],
FuncInfo->MBB);
APInt NeededMask = DesiredMask & ~ActualMask;
APInt KnownZero, KnownOne;
- CurDAG->ComputeMaskedBits(LHS, NeededMask, KnownZero, KnownOne);
+ CurDAG->ComputeMaskedBits(LHS, KnownZero, KnownOne);
// If all the missing bits in the or are already known to be set, match!
if ((NeededMask & KnownOne) == NeededMask)
bool isMorphNodeTo) {
SmallVector<SDNode*, 4> NowDeadNodes;
- ISelUpdater ISU(ISelPosition);
-
// Now that all the normal results are replaced, we replace the chain and
// glue results if present.
if (!ChainNodesMatched.empty()) {
if (ChainVal.getValueType() == MVT::Glue)
ChainVal = ChainVal.getValue(ChainVal->getNumValues()-2);
assert(ChainVal.getValueType() == MVT::Other && "Not a chain?");
- CurDAG->ReplaceAllUsesOfValueWith(ChainVal, InputChain, &ISU);
+ CurDAG->ReplaceAllUsesOfValueWith(ChainVal, InputChain);
// If the node became dead and we haven't already seen it, delete it.
if (ChainNode->use_empty() &&
assert(FRN->getValueType(FRN->getNumValues()-1) == MVT::Glue &&
"Doesn't have a glue result");
CurDAG->ReplaceAllUsesOfValueWith(SDValue(FRN, FRN->getNumValues()-1),
- InputGlue, &ISU);
+ InputGlue);
// If the node became dead and we haven't already seen it, delete it.
if (FRN->use_empty() &&
}
if (!NowDeadNodes.empty())
- CurDAG->RemoveDeadNodes(NowDeadNodes, &ISU);
+ CurDAG->RemoveDeadNodes(NowDeadNodes);
DEBUG(errs() << "ISEL: Match complete!\n");
}
/// The walk we do here is guaranteed to be small because we quickly get down to
/// already selected nodes "below" us.
static ChainResult
-WalkChainUsers(SDNode *ChainedNode,
+WalkChainUsers(const SDNode *ChainedNode,
SmallVectorImpl<SDNode*> &ChainedNodesInPattern,
SmallVectorImpl<SDNode*> &InteriorChainedNodes) {
ChainResult Result = CR_Simple;
User->getOpcode() == ISD::HANDLENODE) // Root of the graph.
continue;
- if (User->getOpcode() == ISD::CopyToReg ||
- User->getOpcode() == ISD::CopyFromReg ||
- User->getOpcode() == ISD::INLINEASM ||
- User->getOpcode() == ISD::EH_LABEL) {
+ unsigned UserOpcode = User->getOpcode();
+ if (UserOpcode == ISD::CopyToReg ||
+ UserOpcode == ISD::CopyFromReg ||
+ UserOpcode == ISD::INLINEASM ||
+ UserOpcode == ISD::EH_LABEL ||
+ UserOpcode == ISD::LIFETIME_START ||
+ UserOpcode == ISD::LIFETIME_END) {
// If their node ID got reset to -1 then they've already been selected.
// Treat them like a MachineOpcode.
if (User->getNodeId() == -1)
return Res;
}
-/// CheckPatternPredicate - Implements OP_CheckPatternPredicate.
+/// CheckSame - Implements OP_CheckSame.
LLVM_ATTRIBUTE_ALWAYS_INLINE static bool
CheckSame(const unsigned char *MatcherTable, unsigned &MatcherIndex,
SDValue N,
/// CheckPatternPredicate - Implements OP_CheckPatternPredicate.
LLVM_ATTRIBUTE_ALWAYS_INLINE static bool
CheckPatternPredicate(const unsigned char *MatcherTable, unsigned &MatcherIndex,
- SelectionDAGISel &SDISel) {
+ const SelectionDAGISel &SDISel) {
return SDISel.CheckPatternPredicate(MatcherTable[MatcherIndex++]);
}
/// CheckNodePredicate - Implements OP_CheckNodePredicate.
LLVM_ATTRIBUTE_ALWAYS_INLINE static bool
CheckNodePredicate(const unsigned char *MatcherTable, unsigned &MatcherIndex,
- SelectionDAGISel &SDISel, SDNode *N) {
+ const SelectionDAGISel &SDISel, SDNode *N) {
return SDISel.CheckNodePredicate(N, MatcherTable[MatcherIndex++]);
}
LLVM_ATTRIBUTE_ALWAYS_INLINE static bool
CheckAndImm(const unsigned char *MatcherTable, unsigned &MatcherIndex,
- SDValue N, SelectionDAGISel &SDISel) {
+ SDValue N, const SelectionDAGISel &SDISel) {
int64_t Val = MatcherTable[MatcherIndex++];
if (Val & 128)
Val = GetVBR(Val, MatcherTable, MatcherIndex);
LLVM_ATTRIBUTE_ALWAYS_INLINE static bool
CheckOrImm(const unsigned char *MatcherTable, unsigned &MatcherIndex,
- SDValue N, SelectionDAGISel &SDISel) {
+ SDValue N, const SelectionDAGISel &SDISel) {
int64_t Val = MatcherTable[MatcherIndex++];
if (Val & 128)
Val = GetVBR(Val, MatcherTable, MatcherIndex);
/// MatcherIndex to continue with.
static unsigned IsPredicateKnownToFail(const unsigned char *Table,
unsigned Index, SDValue N,
- bool &Result, SelectionDAGISel &SDISel,
+ bool &Result,
+ const SelectionDAGISel &SDISel,
SmallVectorImpl<std::pair<SDValue, SDNode*> > &RecordedNodes) {
switch (Table[Index++]) {
default:
case ISD::EntryToken: // These nodes remain the same.
case ISD::BasicBlock:
case ISD::Register:
+ case ISD::RegisterMask:
//case ISD::VALUETYPE:
//case ISD::CONDCODE:
case ISD::HANDLENODE:
case ISD::CopyFromReg:
case ISD::CopyToReg:
case ISD::EH_LABEL:
+ case ISD::LIFETIME_START:
+ case ISD::LIFETIME_END:
NodeToMatch->setNodeId(-1); // Mark selected.
return 0;
case ISD::AssertSext:
(SDNode*) 0));
}
- } else {
+ } else if (NodeToMatch->getOpcode() != ISD::DELETED_NODE) {
Res = MorphNode(NodeToMatch, TargetOpc, VTList, Ops.data(), Ops.size(),
EmitNodeInfo);
+ } else {
+ // NodeToMatch was eliminated by CSE when the target changed the DAG.
+ // We will visit the equivalent node later.
+ DEBUG(dbgs() << "Node was eliminated by CSE\n");
+ return 0;
}
// If the node had chain/glue results, update our notion of the current
// instructions that access memory and for ComplexPatterns that match
// loads.
if (EmitNodeInfo & OPFL_MemRefs) {
+ // Only attach load or store memory operands if the generated
+ // instruction may load or store.
+ const MCInstrDesc &MCID = TM.getInstrInfo()->get(TargetOpc);
+ bool mayLoad = MCID.mayLoad();
+ bool mayStore = MCID.mayStore();
+
+ unsigned NumMemRefs = 0;
+ for (SmallVector<MachineMemOperand*, 2>::const_iterator I =
+ MatchedMemRefs.begin(), E = MatchedMemRefs.end(); I != E; ++I) {
+ if ((*I)->isLoad()) {
+ if (mayLoad)
+ ++NumMemRefs;
+ } else if ((*I)->isStore()) {
+ if (mayStore)
+ ++NumMemRefs;
+ } else {
+ ++NumMemRefs;
+ }
+ }
+
MachineSDNode::mmo_iterator MemRefs =
- MF->allocateMemRefsArray(MatchedMemRefs.size());
- std::copy(MatchedMemRefs.begin(), MatchedMemRefs.end(), MemRefs);
+ MF->allocateMemRefsArray(NumMemRefs);
+
+ MachineSDNode::mmo_iterator MemRefsPos = MemRefs;
+ for (SmallVector<MachineMemOperand*, 2>::const_iterator I =
+ MatchedMemRefs.begin(), E = MatchedMemRefs.end(); I != E; ++I) {
+ if ((*I)->isLoad()) {
+ if (mayLoad)
+ *MemRefsPos++ = *I;
+ } else if ((*I)->isStore()) {
+ if (mayStore)
+ *MemRefsPos++ = *I;
+ } else {
+ *MemRefsPos++ = *I;
+ }
+ }
+
cast<MachineSDNode>(Res)
- ->setMemRefs(MemRefs, MemRefs + MatchedMemRefs.size());
+ ->setMemRefs(MemRefs, MemRefs + NumMemRefs);
}
DEBUG(errs() << " "
N->getOpcode() != ISD::INTRINSIC_WO_CHAIN &&
N->getOpcode() != ISD::INTRINSIC_VOID) {
N->printrFull(Msg, CurDAG);
+ Msg << "\nIn function: " << MF->getName();
} else {
bool HasInputChain = N->getOperand(0).getValueType() == MVT::Other;
unsigned iid =
projects / oota-llvm.git / blobdiff