//===-- WebAssemblyCFGSort.cpp - CFG Sorting ------------------------------===// // // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. // See https://llvm.org/LICENSE.txt for license information. // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// /// /// \file /// This file implements a CFG sorting pass. /// /// This pass reorders the blocks in a function to put them into topological /// order, ignoring loop backedges, and without any loop or exception being /// interrupted by a block not dominated by the its header, with special care /// to keep the order as similar as possible to the original order. /// ////===----------------------------------------------------------------------===// #include "MCTargetDesc/WebAssemblyMCTargetDesc.h" #include "WebAssembly.h" #include "WebAssemblyExceptionInfo.h" #include "WebAssemblySubtarget.h" #include "WebAssemblyUtilities.h" #include "llvm/ADT/PriorityQueue.h" #include "llvm/ADT/SetVector.h" #include "llvm/CodeGen/MachineDominators.h" #include "llvm/CodeGen/MachineFunction.h" #include "llvm/CodeGen/MachineLoopInfo.h" #include "llvm/CodeGen/MachineRegisterInfo.h" #include "llvm/CodeGen/Passes.h" #include "llvm/Support/Debug.h" #include "llvm/Support/raw_ostream.h" using namespace llvm; #define DEBUG_TYPE "wasm-cfg-sort" // Option to disable EH pad first sorting. Only for testing unwind destination // mismatches in CFGStackify. static cl::opt WasmDisableEHPadSort( "wasm-disable-ehpad-sort", cl::ReallyHidden, cl::desc( "WebAssembly: Disable EH pad-first sort order. Testing purpose only."), cl::init(false)); namespace { // Wrapper for loops and exceptions class Region { public: virtual ~Region() = default; virtual MachineBasicBlock *getHeader() const = 0; virtual bool contains(const MachineBasicBlock *MBB) const = 0; virtual unsigned getNumBlocks() const = 0; using block_iterator = typename ArrayRef::const_iterator; virtual iterator_range blocks() const = 0; virtual bool isLoop() const = 0; }; template class ConcreteRegion : public Region { const T *Region; public: ConcreteRegion(const T *Region) : Region(Region) {} MachineBasicBlock *getHeader() const override { return Region->getHeader(); } bool contains(const MachineBasicBlock *MBB) const override { return Region->contains(MBB); } unsigned getNumBlocks() const override { return Region->getNumBlocks(); } iterator_range blocks() const override { return Region->blocks(); } bool isLoop() const override { return false; } }; template <> bool ConcreteRegion::isLoop() const { return true; } // This class has information of nested Regions; this is analogous to what // LoopInfo is for loops. class RegionInfo { const MachineLoopInfo &MLI; const WebAssemblyExceptionInfo &WEI; DenseMap> LoopMap; DenseMap> ExceptionMap; public: RegionInfo(const MachineLoopInfo &MLI, const WebAssemblyExceptionInfo &WEI) : MLI(MLI), WEI(WEI) {} // Returns a smallest loop or exception that contains MBB const Region *getRegionFor(const MachineBasicBlock *MBB) { const auto *ML = MLI.getLoopFor(MBB); const auto *WE = WEI.getExceptionFor(MBB); if (!ML && !WE) return nullptr; // We determine subregion relationship by domination of their headers, i.e., // if region A's header dominates region B's header, B is a subregion of A. // WebAssemblyException contains BBs in all its subregions (loops or // exceptions), but MachineLoop may not, because MachineLoop does not contain // BBs that don't have a path to its header even if they are dominated by // its header. So here we should use WE->contains(ML->getHeader()), but not // ML->contains(WE->getHeader()). if ((ML && !WE) || (ML && WE && WE->contains(ML->getHeader()))) { // If the smallest region containing MBB is a loop if (LoopMap.count(ML)) return LoopMap[ML].get(); LoopMap[ML] = std::make_unique>(ML); return LoopMap[ML].get(); } else { // If the smallest region containing MBB is an exception if (ExceptionMap.count(WE)) return ExceptionMap[WE].get(); ExceptionMap[WE] = std::make_unique>(WE); return ExceptionMap[WE].get(); } } }; class WebAssemblyCFGSort final : public MachineFunctionPass { StringRef getPassName() const override { return "WebAssembly CFG Sort"; } void getAnalysisUsage(AnalysisUsage &AU) const override { AU.setPreservesCFG(); AU.addRequired(); AU.addPreserved(); AU.addRequired(); AU.addPreserved(); AU.addRequired(); AU.addPreserved(); MachineFunctionPass::getAnalysisUsage(AU); } bool runOnMachineFunction(MachineFunction &MF) override; public: static char ID; // Pass identification, replacement for typeid WebAssemblyCFGSort() : MachineFunctionPass(ID) {} }; } // end anonymous namespace char WebAssemblyCFGSort::ID = 0; INITIALIZE_PASS(WebAssemblyCFGSort, DEBUG_TYPE, "Reorders blocks in topological order", false, false) FunctionPass *llvm::createWebAssemblyCFGSort() { return new WebAssemblyCFGSort(); } static void maybeUpdateTerminator(MachineBasicBlock *MBB) { #ifndef NDEBUG bool AnyBarrier = false; #endif bool AllAnalyzable = true; for (const MachineInstr &Term : MBB->terminators()) { #ifndef NDEBUG AnyBarrier |= Term.isBarrier(); #endif AllAnalyzable &= Term.isBranch() && !Term.isIndirectBranch(); } assert((AnyBarrier || AllAnalyzable) && "analyzeBranch needs to analyze any block with a fallthrough"); // Find the layout successor from the original block order. MachineFunction *MF = MBB->getParent(); MachineBasicBlock *OriginalSuccessor = unsigned(MBB->getNumber() + 1) < MF->getNumBlockIDs() ? MF->getBlockNumbered(MBB->getNumber() + 1) : nullptr; if (AllAnalyzable) MBB->updateTerminator(OriginalSuccessor); } namespace { // EH pads are selected first regardless of the block comparison order. // When only one of the BBs is an EH pad, we give a higher priority to it, to // prevent common mismatches between possibly throwing calls and ehpads they // unwind to, as in the example below: // // bb0: // call @foo // If this throws, unwind to bb2 // bb1: // call @bar // If this throws, unwind to bb3 // bb2 (ehpad): // handler_bb2 // bb3 (ehpad): // handler_bb3 // continuing code // // Because this pass tries to preserve the original BB order, this order will // not change. But this will result in this try-catch structure in CFGStackify, // resulting in a mismatch: // try // try // call @foo // call @bar // This should unwind to bb3, not bb2! // catch // handler_bb2 // end // catch // handler_bb3 // end // continuing code // // If we give a higher priority to an EH pad whenever it is ready in this // example, when both bb1 and bb2 are ready, we would pick up bb2 first. /// Sort blocks by their number. struct CompareBlockNumbers { bool operator()(const MachineBasicBlock *A, const MachineBasicBlock *B) const { if (!WasmDisableEHPadSort) { if (A->isEHPad() && !B->isEHPad()) return false; if (!A->isEHPad() && B->isEHPad()) return true; } return A->getNumber() > B->getNumber(); } }; /// Sort blocks by their number in the opposite order.. struct CompareBlockNumbersBackwards { bool operator()(const MachineBasicBlock *A, const MachineBasicBlock *B) const { if (!WasmDisableEHPadSort) { if (A->isEHPad() && !B->isEHPad()) return false; if (!A->isEHPad() && B->isEHPad()) return true; } return A->getNumber() < B->getNumber(); } }; /// Bookkeeping for a region to help ensure that we don't mix blocks not /// dominated by the its header among its blocks. struct Entry { const Region *TheRegion; unsigned NumBlocksLeft; /// List of blocks not dominated by Loop's header that are deferred until /// after all of Loop's blocks have been seen. std::vector Deferred; explicit Entry(const class Region *R) : TheRegion(R), NumBlocksLeft(R->getNumBlocks()) {} }; } // end anonymous namespace /// Sort the blocks, taking special care to make sure that regions are not /// interrupted by blocks not dominated by their header. /// TODO: There are many opportunities for improving the heuristics here. /// Explore them. static void sortBlocks(MachineFunction &MF, const MachineLoopInfo &MLI, const WebAssemblyExceptionInfo &WEI, const MachineDominatorTree &MDT) { // Remember original layout ordering, so we can update terminators after // reordering to point to the original layout successor. MF.RenumberBlocks(); // Prepare for a topological sort: Record the number of predecessors each // block has, ignoring loop backedges. SmallVector NumPredsLeft(MF.getNumBlockIDs(), 0); for (MachineBasicBlock &MBB : MF) { unsigned N = MBB.pred_size(); if (MachineLoop *L = MLI.getLoopFor(&MBB)) if (L->getHeader() == &MBB) for (const MachineBasicBlock *Pred : MBB.predecessors()) if (L->contains(Pred)) --N; NumPredsLeft[MBB.getNumber()] = N; } // Topological sort the CFG, with additional constraints: // - Between a region header and the last block in the region, there can be // no blocks not dominated by its header. // - It's desirable to preserve the original block order when possible. // We use two ready lists; Preferred and Ready. Preferred has recently // processed successors, to help preserve block sequences from the original // order. Ready has the remaining ready blocks. EH blocks are picked first // from both queues. PriorityQueue, CompareBlockNumbers> Preferred; PriorityQueue, CompareBlockNumbersBackwards> Ready; RegionInfo RI(MLI, WEI); SmallVector Entries; for (MachineBasicBlock *MBB = &MF.front();;) { const Region *R = RI.getRegionFor(MBB); if (R) { // If MBB is a region header, add it to the active region list. We can't // put any blocks that it doesn't dominate until we see the end of the // region. if (R->getHeader() == MBB) Entries.push_back(Entry(R)); // For each active region the block is in, decrement the count. If MBB is // the last block in an active region, take it off the list and pick up // any blocks deferred because the header didn't dominate them. for (Entry &E : Entries) if (E.TheRegion->contains(MBB) && --E.NumBlocksLeft == 0) for (auto DeferredBlock : E.Deferred) Ready.push(DeferredBlock); while (!Entries.empty() && Entries.back().NumBlocksLeft == 0) Entries.pop_back(); } // The main topological sort logic. for (MachineBasicBlock *Succ : MBB->successors()) { // Ignore backedges. if (MachineLoop *SuccL = MLI.getLoopFor(Succ)) if (SuccL->getHeader() == Succ && SuccL->contains(MBB)) continue; // Decrement the predecessor count. If it's now zero, it's ready. if (--NumPredsLeft[Succ->getNumber()] == 0) Preferred.push(Succ); } // Determine the block to follow MBB. First try to find a preferred block, // to preserve the original block order when possible. MachineBasicBlock *Next = nullptr; while (!Preferred.empty()) { Next = Preferred.top(); Preferred.pop(); // If X isn't dominated by the top active region header, defer it until // that region is done. if (!Entries.empty() && !MDT.dominates(Entries.back().TheRegion->getHeader(), Next)) { Entries.back().Deferred.push_back(Next); Next = nullptr; continue; } // If Next was originally ordered before MBB, and it isn't because it was // loop-rotated above the header, it's not preferred. if (Next->getNumber() < MBB->getNumber() && (WasmDisableEHPadSort || !Next->isEHPad()) && (!R || !R->contains(Next) || R->getHeader()->getNumber() < Next->getNumber())) { Ready.push(Next); Next = nullptr; continue; } break; } // If we didn't find a suitable block in the Preferred list, check the // general Ready list. if (!Next) { // If there are no more blocks to process, we're done. if (Ready.empty()) { maybeUpdateTerminator(MBB); break; } for (;;) { Next = Ready.top(); Ready.pop(); // If Next isn't dominated by the top active region header, defer it // until that region is done. if (!Entries.empty() && !MDT.dominates(Entries.back().TheRegion->getHeader(), Next)) { Entries.back().Deferred.push_back(Next); continue; } break; } } // Move the next block into place and iterate. Next->moveAfter(MBB); maybeUpdateTerminator(MBB); MBB = Next; } assert(Entries.empty() && "Active sort region list not finished"); MF.RenumberBlocks(); #ifndef NDEBUG SmallSetVector OnStack; // Insert a sentinel representing the degenerate loop that starts at the // function entry block and includes the entire function as a "loop" that // executes once. OnStack.insert(nullptr); for (auto &MBB : MF) { assert(MBB.getNumber() >= 0 && "Renumbered blocks should be non-negative."); const Region *Region = RI.getRegionFor(&MBB); if (Region && &MBB == Region->getHeader()) { // Region header. if (Region->isLoop()) { // Loop header. The loop predecessor should be sorted above, and the // other predecessors should be backedges below. for (auto Pred : MBB.predecessors()) assert( (Pred->getNumber() < MBB.getNumber() || Region->contains(Pred)) && "Loop header predecessors must be loop predecessors or " "backedges"); } else { // Exception header. All predecessors should be sorted above. for (auto Pred : MBB.predecessors()) assert(Pred->getNumber() < MBB.getNumber() && "Non-loop-header predecessors should be topologically sorted"); } assert(OnStack.insert(Region) && "Regions should be declared at most once."); } else { // Not a region header. All predecessors should be sorted above. for (auto Pred : MBB.predecessors()) assert(Pred->getNumber() < MBB.getNumber() && "Non-loop-header predecessors should be topologically sorted"); assert(OnStack.count(RI.getRegionFor(&MBB)) && "Blocks must be nested in their regions"); } while (OnStack.size() > 1 && &MBB == WebAssembly::getBottom(OnStack.back())) OnStack.pop_back(); } assert(OnStack.pop_back_val() == nullptr && "The function entry block shouldn't actually be a region header"); assert(OnStack.empty() && "Control flow stack pushes and pops should be balanced."); #endif } bool WebAssemblyCFGSort::runOnMachineFunction(MachineFunction &MF) { LLVM_DEBUG(dbgs() << "********** CFG Sorting **********\n" "********** Function: " << MF.getName() << '\n'); const auto &MLI = getAnalysis(); const auto &WEI = getAnalysis(); auto &MDT = getAnalysis(); // Liveness is not tracked for VALUE_STACK physreg. MF.getRegInfo().invalidateLiveness(); // Sort the blocks, with contiguous sort regions. sortBlocks(MF, MLI, WEI, MDT); return true; }