+/// Add common passes that perform LLVM IR to IR transforms in preparation for
+/// instruction selection.
+void TargetPassConfig::addISelPrepare() {
+ if (getOptLevel() != CodeGenOpt::None && !DisableCGP)
+ PM.add(createCodeGenPreparePass(getTargetLowering()));
+
+ PM.add(createStackProtectorPass(getTargetLowering()));
+
+ addPreISel();
+
+ if (PrintISelInput)
+ PM.add(createPrintFunctionPass("\n\n"
+ "*** Final LLVM Code input to ISel ***\n",
+ &dbgs()));
+
+ // All passes which modify the LLVM IR are now complete; run the verifier
+ // to ensure that the IR is valid.
+ if (!DisableVerify)
+ PM.add(createVerifierPass());
+}
+
+/// Add the complete set of target-independent postISel code generator passes.
+///
+/// This can be read as the standard order of major LLVM CodeGen stages. Stages
+/// with nontrivial configuration or multiple passes are broken out below in
+/// add%Stage routines.
+///
+/// Any TargetPassConfig::addXX routine may be overriden by the Target. The
+/// addPre/Post methods with empty header implementations allow injecting
+/// target-specific fixups just before or after major stages. Additionally,
+/// targets have the flexibility to change pass order within a stage by
+/// overriding default implementation of add%Stage routines below. Each
+/// technique has maintainability tradeoffs because alternate pass orders are
+/// not well supported. addPre/Post works better if the target pass is easily
+/// tied to a common pass. But if it has subtle dependencies on multiple passes,
+/// the target should override the stage instead.
+///
+/// TODO: We could use a single addPre/Post(ID) hook to allow pass injection
+/// before/after any target-independent pass. But it's currently overkill.
+void TargetPassConfig::addMachinePasses() {
+ // Print the instruction selected machine code...
+ printAndVerify("After Instruction Selection");
+
+ // Expand pseudo-instructions emitted by ISel.
+ addPass(ExpandISelPseudosID);
+
+ // Add passes that optimize machine instructions in SSA form.
+ if (getOptLevel() != CodeGenOpt::None) {
+ addMachineSSAOptimization();
+ }
+ else {
+ // If the target requests it, assign local variables to stack slots relative
+ // to one another and simplify frame index references where possible.
+ addPass(LocalStackSlotAllocationID);
+ }
+
+ // Run pre-ra passes.
+ if (addPreRegAlloc())
+ printAndVerify("After PreRegAlloc passes");
+
+ // Run register allocation and passes that are tightly coupled with it,
+ // including phi elimination and scheduling.
+ if (getOptimizeRegAlloc())
+ addOptimizedRegAlloc(createRegAllocPass(true));
+ else
+ addFastRegAlloc(createRegAllocPass(false));
+
+ // Run post-ra passes.
+ if (addPostRegAlloc())
+ printAndVerify("After PostRegAlloc passes");
+
+ // Insert prolog/epilog code. Eliminate abstract frame index references...
+ addPass(PrologEpilogCodeInserterID);
+ printAndVerify("After PrologEpilogCodeInserter");
+
+ /// Add passes that optimize machine instructions after register allocation.
+ if (getOptLevel() != CodeGenOpt::None)
+ addMachineLateOptimization();
+
+ // Expand pseudo instructions before second scheduling pass.
+ addPass(ExpandPostRAPseudosID);
+ printNoVerify("After ExpandPostRAPseudos");
+
+ // Run pre-sched2 passes.
+ if (addPreSched2())
+ printNoVerify("After PreSched2 passes");
+
+ // Second pass scheduler.
+ if (getOptLevel() != CodeGenOpt::None) {
+ addPass(PostRASchedulerID);
+ printNoVerify("After PostRAScheduler");
+ }
+
+ // GC
+ addPass(GCMachineCodeAnalysisID);
+ if (PrintGCInfo)
+ PM.add(createGCInfoPrinter(dbgs()));
+
+ // Basic block placement.
+ if (getOptLevel() != CodeGenOpt::None)
+ addBlockPlacement();
+
+ if (addPreEmitPass())
+ printNoVerify("After PreEmit passes");
+}
+
+/// Add passes that optimize machine instructions in SSA form.
+void TargetPassConfig::addMachineSSAOptimization() {
+ // Pre-ra tail duplication.
+ if (addPass(EarlyTailDuplicateID) != &NoPassID)
+ printAndVerify("After Pre-RegAlloc TailDuplicate");
+
+ // Optimize PHIs before DCE: removing dead PHI cycles may make more
+ // instructions dead.
+ addPass(OptimizePHIsID);
+
+ // If the target requests it, assign local variables to stack slots relative
+ // to one another and simplify frame index references where possible.
+ addPass(LocalStackSlotAllocationID);
+
+ // With optimization, dead code should already be eliminated. However
+ // there is one known exception: lowered code for arguments that are only
+ // used by tail calls, where the tail calls reuse the incoming stack
+ // arguments directly (see t11 in test/CodeGen/X86/sibcall.ll).
+ addPass(DeadMachineInstructionElimID);
+ printAndVerify("After codegen DCE pass");
+
+ addPass(MachineLICMID);
+ addPass(MachineCSEID);
+ addPass(MachineSinkingID);
+ printAndVerify("After Machine LICM, CSE and Sinking passes");
+
+ addPass(PeepholeOptimizerID);
+ printAndVerify("After codegen peephole optimization pass");
+}
+
+//===---------------------------------------------------------------------===//
+/// Register Allocation Pass Configuration
+//===---------------------------------------------------------------------===//
+
+bool TargetPassConfig::getOptimizeRegAlloc() const {
+ switch (OptimizeRegAlloc) {
+ case cl::BOU_UNSET: return getOptLevel() != CodeGenOpt::None;
+ case cl::BOU_TRUE: return true;
+ case cl::BOU_FALSE: return false;
+ }
+ llvm_unreachable("Invalid optimize-regalloc state");
+}
+
+/// RegisterRegAlloc's global Registry tracks allocator registration.
+MachinePassRegistry RegisterRegAlloc::Registry;
+
+/// A dummy default pass factory indicates whether the register allocator is
+/// overridden on the command line.
+static FunctionPass *useDefaultRegisterAllocator() { return 0; }
+static RegisterRegAlloc
+defaultRegAlloc("default",
+ "pick register allocator based on -O option",
+ useDefaultRegisterAllocator);
+
+/// -regalloc=... command line option.
+static cl::opt<RegisterRegAlloc::FunctionPassCtor, false,
+ RegisterPassParser<RegisterRegAlloc> >
+RegAlloc("regalloc",
+ cl::init(&useDefaultRegisterAllocator),
+ cl::desc("Register allocator to use"));
+
+
+/// Instantiate the default register allocator pass for this target for either
+/// the optimized or unoptimized allocation path. This will be added to the pass
+/// manager by addFastRegAlloc in the unoptimized case or addOptimizedRegAlloc
+/// in the optimized case.
+///
+/// A target that uses the standard regalloc pass order for fast or optimized
+/// allocation may still override this for per-target regalloc
+/// selection. But -regalloc=... always takes precedence.
+FunctionPass *TargetPassConfig::createTargetRegisterAllocator(bool Optimized) {
+ if (Optimized)
+ return createGreedyRegisterAllocator();
+ else
+ return createFastRegisterAllocator();
+}
+
+/// Find and instantiate the register allocation pass requested by this target
+/// at the current optimization level. Different register allocators are
+/// defined as separate passes because they may require different analysis.
+///
+/// This helper ensures that the regalloc= option is always available,
+/// even for targets that override the default allocator.
+///
+/// FIXME: When MachinePassRegistry register pass IDs instead of function ptrs,
+/// this can be folded into addPass.
+FunctionPass *TargetPassConfig::createRegAllocPass(bool Optimized) {
+ RegisterRegAlloc::FunctionPassCtor Ctor = RegisterRegAlloc::getDefault();
+
+ // Initialize the global default.
+ if (!Ctor) {
+ Ctor = RegAlloc;
+ RegisterRegAlloc::setDefault(RegAlloc);
+ }
+ if (Ctor != useDefaultRegisterAllocator)
+ return Ctor();
+
+ // With no -regalloc= override, ask the target for a regalloc pass.
+ return createTargetRegisterAllocator(Optimized);
+}
+
+/// Add the minimum set of target-independent passes that are required for
+/// register allocation. No coalescing or scheduling.
+void TargetPassConfig::addFastRegAlloc(FunctionPass *RegAllocPass) {
+ addPass(PHIEliminationID);
+ addPass(TwoAddressInstructionPassID);
+
+ PM.add(RegAllocPass);
+ printAndVerify("After Register Allocation");
+}
+
+/// Add standard target-independent passes that are tightly coupled with
+/// optimized register allocation, including coalescing, machine instruction
+/// scheduling, and register allocation itself.
+void TargetPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
+ // LiveVariables currently requires pure SSA form.
+ //
+ // FIXME: Once TwoAddressInstruction pass no longer uses kill flags,
+ // LiveVariables can be removed completely, and LiveIntervals can be directly
+ // computed. (We still either need to regenerate kill flags after regalloc, or
+ // preferably fix the scavenger to not depend on them).
+ addPass(LiveVariablesID);
+
+ // Add passes that move from transformed SSA into conventional SSA. This is a
+ // "copy coalescing" problem.
+ //
+ if (!EnableStrongPHIElim) {
+ // Edge splitting is smarter with machine loop info.
+ addPass(MachineLoopInfoID);
+ addPass(PHIEliminationID);
+ }
+ addPass(TwoAddressInstructionPassID);
+
+ // FIXME: Either remove this pass completely, or fix it so that it works on
+ // SSA form. We could modify LiveIntervals to be independent of this pass, But
+ // it would be even better to simply eliminate *all* IMPLICIT_DEFs before
+ // leaving SSA.
+ addPass(ProcessImplicitDefsID);
+
+ if (EnableStrongPHIElim)
+ addPass(StrongPHIEliminationID);
+
+ addPass(RegisterCoalescerID);
+
+ // PreRA instruction scheduling.
+ if (addPass(MachineSchedulerID) != &NoPassID)
+ printAndVerify("After Machine Scheduling");
+
+ // Add the selected register allocation pass.
+ PM.add(RegAllocPass);
+ printAndVerify("After Register Allocation");
+
+ // FinalizeRegAlloc is convenient until MachineInstrBundles is more mature,
+ // but eventually, all users of it should probably be moved to addPostRA and
+ // it can go away. Currently, it's the intended place for targets to run
+ // FinalizeMachineBundles, because passes other than MachineScheduling an
+ // RegAlloc itself may not be aware of bundles.
+ if (addFinalizeRegAlloc())
+ printAndVerify("After RegAlloc finalization");
+
+ // Perform stack slot coloring and post-ra machine LICM.
+ //
+ // FIXME: Re-enable coloring with register when it's capable of adding
+ // kill markers.
+ addPass(StackSlotColoringID);
+
+ // Run post-ra machine LICM to hoist reloads / remats.
+ //
+ // FIXME: can this move into MachineLateOptimization?
+ addPass(PostRAMachineLICMID);
+
+ printAndVerify("After StackSlotColoring and postra Machine LICM");
+}
+
+//===---------------------------------------------------------------------===//
+/// Post RegAlloc Pass Configuration
+//===---------------------------------------------------------------------===//
+
+/// Add passes that optimize machine instructions after register allocation.
+void TargetPassConfig::addMachineLateOptimization() {
+ // Branch folding must be run after regalloc and prolog/epilog insertion.
+ if (addPass(BranchFolderPassID) != &NoPassID)
+ printNoVerify("After BranchFolding");
+
+ // Tail duplication.
+ if (addPass(TailDuplicateID) != &NoPassID)
+ printNoVerify("After TailDuplicate");
+
+ // Copy propagation.
+ if (addPass(MachineCopyPropagationID) != &NoPassID)
+ printNoVerify("After copy propagation pass");
+}
+
+/// Add standard basic block placement passes.
+void TargetPassConfig::addBlockPlacement() {
+ AnalysisID ID = &NoPassID;
+ if (EnableBlockPlacement) {
+ // MachineBlockPlacement is an experimental pass which is disabled by
+ // default currently. Eventually it should subsume CodePlacementOpt, so
+ // when enabled, the other is disabled.
+ ID = addPass(MachineBlockPlacementID);
+ } else {
+ ID = addPass(CodePlacementOptID);
+ }
+ if (ID != &NoPassID) {
+ // Run a separate pass to collect block placement statistics.
+ if (EnableBlockPlacementStats)
+ addPass(MachineBlockPlacementStatsID);
+
+ printNoVerify("After machine block placement.");
+ }
+}