| 1 | //===- IndirectBrExpandPass.cpp - Expand indirectbr to switch -------------===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | /// \file |
| 9 | /// |
| 10 | /// Implements an expansion pass to turn `indirectbr` instructions in the IR |
| 11 | /// into `switch` instructions. This works by enumerating the basic blocks in |
| 12 | /// a dense range of integers, replacing each `blockaddr` constant with the |
| 13 | /// corresponding integer constant, and then building a switch that maps from |
| 14 | /// the integers to the actual blocks. All of the indirectbr instructions in the |
| 15 | /// function are redirected to this common switch. |
| 16 | /// |
| 17 | /// While this is generically useful if a target is unable to codegen |
| 18 | /// `indirectbr` natively, it is primarily useful when there is some desire to |
| 19 | /// get the builtin non-jump-table lowering of a switch even when the input |
| 20 | /// source contained an explicit indirect branch construct. |
| 21 | /// |
| 22 | /// Note that it doesn't make any sense to enable this pass unless a target also |
| 23 | /// disables jump-table lowering of switches. Doing that is likely to pessimize |
| 24 | /// the code. |
| 25 | /// |
| 26 | //===----------------------------------------------------------------------===// |
| 27 | |
| 28 | #include "llvm/ADT/Sequence.h" |
| 29 | #include "llvm/ADT/SmallVector.h" |
| 30 | #include "llvm/Analysis/DomTreeUpdater.h" |
| 31 | #include "llvm/CodeGen/IndirectBrExpand.h" |
| 32 | #include "llvm/CodeGen/TargetPassConfig.h" |
| 33 | #include "llvm/CodeGen/TargetSubtargetInfo.h" |
| 34 | #include "llvm/IR/BasicBlock.h" |
| 35 | #include "llvm/IR/Constants.h" |
| 36 | #include "llvm/IR/Dominators.h" |
| 37 | #include "llvm/IR/Function.h" |
| 38 | #include "llvm/IR/Instructions.h" |
| 39 | #include "llvm/InitializePasses.h" |
| 40 | #include "llvm/Pass.h" |
| 41 | #include "llvm/Support/ErrorHandling.h" |
| 42 | #include "llvm/Target/TargetMachine.h" |
| 43 | #include <optional> |
| 44 | |
| 45 | using namespace llvm; |
| 46 | |
| 47 | #define DEBUG_TYPE "indirectbr-expand" |
| 48 | |
| 49 | namespace { |
| 50 | |
| 51 | class IndirectBrExpandLegacyPass : public FunctionPass { |
| 52 | public: |
| 53 | static char ID; // Pass identification, replacement for typeid |
| 54 | |
| 55 | IndirectBrExpandLegacyPass() : FunctionPass(ID) {} |
| 56 | |
| 57 | void getAnalysisUsage(AnalysisUsage &AU) const override { |
| 58 | AU.addPreserved<DominatorTreeWrapperPass>(); |
| 59 | } |
| 60 | |
| 61 | bool runOnFunction(Function &F) override; |
| 62 | }; |
| 63 | |
| 64 | } // end anonymous namespace |
| 65 | |
| 66 | static bool runImpl(Function &F, const TargetLowering *TLI, |
| 67 | DomTreeUpdater *DTU); |
| 68 | |
| 69 | PreservedAnalyses IndirectBrExpandPass::run(Function &F, |
| 70 | FunctionAnalysisManager &FAM) { |
| 71 | auto *STI = TM->getSubtargetImpl(F); |
| 72 | if (!STI->enableIndirectBrExpand()) |
| 73 | return PreservedAnalyses::all(); |
| 74 | |
| 75 | auto *TLI = STI->getTargetLowering(); |
| 76 | auto *DT = FAM.getCachedResult<DominatorTreeAnalysis>(IR&: F); |
| 77 | DomTreeUpdater DTU(DT, DomTreeUpdater::UpdateStrategy::Lazy); |
| 78 | |
| 79 | bool Changed = runImpl(F, TLI, DTU: DT ? &DTU : nullptr); |
| 80 | if (!Changed) |
| 81 | return PreservedAnalyses::all(); |
| 82 | PreservedAnalyses PA; |
| 83 | PA.preserve<DominatorTreeAnalysis>(); |
| 84 | return PA; |
| 85 | } |
| 86 | |
| 87 | char IndirectBrExpandLegacyPass::ID = 0; |
| 88 | |
| 89 | INITIALIZE_PASS_BEGIN(IndirectBrExpandLegacyPass, DEBUG_TYPE, |
| 90 | "Expand indirectbr instructions" , false, false) |
| 91 | INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass) |
| 92 | INITIALIZE_PASS_END(IndirectBrExpandLegacyPass, DEBUG_TYPE, |
| 93 | "Expand indirectbr instructions" , false, false) |
| 94 | |
| 95 | FunctionPass *llvm::createIndirectBrExpandPass() { |
| 96 | return new IndirectBrExpandLegacyPass(); |
| 97 | } |
| 98 | |
| 99 | bool runImpl(Function &F, const TargetLowering *TLI, DomTreeUpdater *DTU) { |
| 100 | auto &DL = F.getDataLayout(); |
| 101 | |
| 102 | SmallVector<IndirectBrInst *, 1> IndirectBrs; |
| 103 | |
| 104 | // Set of all potential successors for indirectbr instructions. |
| 105 | SmallPtrSet<BasicBlock *, 4> IndirectBrSuccs; |
| 106 | |
| 107 | // Build a list of indirectbrs that we want to rewrite. |
| 108 | for (BasicBlock &BB : F) |
| 109 | if (auto *IBr = dyn_cast<IndirectBrInst>(Val: BB.getTerminator())) { |
| 110 | // Handle the degenerate case of no successors by replacing the indirectbr |
| 111 | // with unreachable as there is no successor available. |
| 112 | if (IBr->getNumSuccessors() == 0) { |
| 113 | (void)new UnreachableInst(F.getContext(), IBr->getIterator()); |
| 114 | IBr->eraseFromParent(); |
| 115 | continue; |
| 116 | } |
| 117 | |
| 118 | IndirectBrs.push_back(Elt: IBr); |
| 119 | IndirectBrSuccs.insert_range(R: IBr->successors()); |
| 120 | } |
| 121 | |
| 122 | if (IndirectBrs.empty()) |
| 123 | return false; |
| 124 | |
| 125 | // If we need to replace any indirectbrs we need to establish integer |
| 126 | // constants that will correspond to each of the basic blocks in the function |
| 127 | // whose address escapes. We do that here and rewrite all the blockaddress |
| 128 | // constants to just be those integer constants cast to a pointer type. |
| 129 | SmallVector<BasicBlock *, 4> BBs; |
| 130 | |
| 131 | for (BasicBlock &BB : F) { |
| 132 | // Skip blocks that aren't successors to an indirectbr we're going to |
| 133 | // rewrite. |
| 134 | if (!IndirectBrSuccs.count(Ptr: &BB)) |
| 135 | continue; |
| 136 | |
| 137 | auto *BA = BlockAddress::lookup(BB: &BB); |
| 138 | |
| 139 | // Skip if the constant was formed but ended up not being used (due to DCE |
| 140 | // or whatever). |
| 141 | if (!BA || !BA->isConstantUsed()) |
| 142 | continue; |
| 143 | |
| 144 | // Compute the index we want to use for this basic block. We can't use zero |
| 145 | // because null can be compared with block addresses. |
| 146 | int BBIndex = BBs.size() + 1; |
| 147 | BBs.push_back(Elt: &BB); |
| 148 | |
| 149 | auto *ITy = cast<IntegerType>(Val: DL.getIntPtrType(BA->getType())); |
| 150 | ConstantInt *BBIndexC = ConstantInt::get(Ty: ITy, V: BBIndex); |
| 151 | |
| 152 | // Now rewrite the blockaddress to an integer constant based on the index. |
| 153 | // FIXME: This part doesn't properly recognize other uses of blockaddress |
| 154 | // expressions, for instance, where they are used to pass labels to |
| 155 | // asm-goto. This part of the pass needs a rework. |
| 156 | BA->replaceAllUsesWith(V: ConstantExpr::getIntToPtr(C: BBIndexC, Ty: BA->getType())); |
| 157 | } |
| 158 | |
| 159 | if (BBs.empty()) { |
| 160 | // There are no blocks whose address is taken, so any indirectbr instruction |
| 161 | // cannot get a valid input and we can replace all of them with unreachable. |
| 162 | SmallVector<DominatorTree::UpdateType, 8> Updates; |
| 163 | if (DTU) |
| 164 | Updates.reserve(N: IndirectBrSuccs.size()); |
| 165 | for (auto *IBr : IndirectBrs) { |
| 166 | if (DTU) { |
| 167 | for (BasicBlock *SuccBB : IBr->successors()) |
| 168 | Updates.push_back(Elt: {DominatorTree::Delete, IBr->getParent(), SuccBB}); |
| 169 | } |
| 170 | (void)new UnreachableInst(F.getContext(), IBr->getIterator()); |
| 171 | IBr->eraseFromParent(); |
| 172 | } |
| 173 | if (DTU) { |
| 174 | assert(Updates.size() == IndirectBrSuccs.size() && |
| 175 | "Got unexpected update count." ); |
| 176 | DTU->applyUpdates(Updates); |
| 177 | } |
| 178 | return true; |
| 179 | } |
| 180 | |
| 181 | BasicBlock *SwitchBB; |
| 182 | Value *SwitchValue; |
| 183 | |
| 184 | // Compute a common integer type across all the indirectbr instructions. |
| 185 | IntegerType *CommonITy = nullptr; |
| 186 | for (auto *IBr : IndirectBrs) { |
| 187 | auto *ITy = |
| 188 | cast<IntegerType>(Val: DL.getIntPtrType(IBr->getAddress()->getType())); |
| 189 | if (!CommonITy || ITy->getBitWidth() > CommonITy->getBitWidth()) |
| 190 | CommonITy = ITy; |
| 191 | } |
| 192 | |
| 193 | auto GetSwitchValue = [CommonITy](IndirectBrInst *IBr) { |
| 194 | return CastInst::CreatePointerCast(S: IBr->getAddress(), Ty: CommonITy, |
| 195 | Name: Twine(IBr->getAddress()->getName()) + |
| 196 | ".switch_cast" , |
| 197 | InsertBefore: IBr->getIterator()); |
| 198 | }; |
| 199 | |
| 200 | SmallVector<DominatorTree::UpdateType, 8> Updates; |
| 201 | |
| 202 | if (IndirectBrs.size() == 1) { |
| 203 | // If we only have one indirectbr, we can just directly replace it within |
| 204 | // its block. |
| 205 | IndirectBrInst *IBr = IndirectBrs[0]; |
| 206 | SwitchBB = IBr->getParent(); |
| 207 | SwitchValue = GetSwitchValue(IBr); |
| 208 | if (DTU) { |
| 209 | Updates.reserve(N: IndirectBrSuccs.size()); |
| 210 | for (BasicBlock *SuccBB : IBr->successors()) |
| 211 | Updates.push_back(Elt: {DominatorTree::Delete, IBr->getParent(), SuccBB}); |
| 212 | assert(Updates.size() == IndirectBrSuccs.size() && |
| 213 | "Got unexpected update count." ); |
| 214 | } |
| 215 | IBr->eraseFromParent(); |
| 216 | } else { |
| 217 | // Otherwise we need to create a new block to hold the switch across BBs, |
| 218 | // jump to that block instead of each indirectbr, and phi together the |
| 219 | // values for the switch. |
| 220 | SwitchBB = BasicBlock::Create(Context&: F.getContext(), Name: "switch_bb" , Parent: &F); |
| 221 | auto *SwitchPN = PHINode::Create(Ty: CommonITy, NumReservedValues: IndirectBrs.size(), |
| 222 | NameStr: "switch_value_phi" , InsertBefore: SwitchBB); |
| 223 | SwitchValue = SwitchPN; |
| 224 | |
| 225 | // Now replace the indirectbr instructions with direct branches to the |
| 226 | // switch block and fill out the PHI operands. |
| 227 | if (DTU) |
| 228 | Updates.reserve(N: IndirectBrs.size() + 2 * IndirectBrSuccs.size()); |
| 229 | for (auto *IBr : IndirectBrs) { |
| 230 | SwitchPN->addIncoming(V: GetSwitchValue(IBr), BB: IBr->getParent()); |
| 231 | UncondBrInst::Create(Target: SwitchBB, InsertBefore: IBr->getIterator()); |
| 232 | if (DTU) { |
| 233 | Updates.push_back(Elt: {DominatorTree::Insert, IBr->getParent(), SwitchBB}); |
| 234 | for (BasicBlock *SuccBB : IBr->successors()) |
| 235 | Updates.push_back(Elt: {DominatorTree::Delete, IBr->getParent(), SuccBB}); |
| 236 | } |
| 237 | IBr->eraseFromParent(); |
| 238 | } |
| 239 | } |
| 240 | |
| 241 | // Now build the switch in the block. The block will have no terminator |
| 242 | // already. |
| 243 | auto *SI = SwitchInst::Create(Value: SwitchValue, Default: BBs[0], NumCases: BBs.size(), InsertBefore: SwitchBB); |
| 244 | |
| 245 | // Add a case for each block. |
| 246 | for (int i : llvm::seq<int>(Begin: 1, End: BBs.size())) |
| 247 | SI->addCase(OnVal: ConstantInt::get(Ty: CommonITy, V: i + 1), Dest: BBs[i]); |
| 248 | |
| 249 | if (DTU) { |
| 250 | // If there were multiple indirectbr's, they may have common successors, |
| 251 | // but in the dominator tree, we only track unique edges. |
| 252 | SmallPtrSet<BasicBlock *, 8> UniqueSuccessors; |
| 253 | Updates.reserve(N: Updates.size() + BBs.size()); |
| 254 | for (BasicBlock *BB : BBs) { |
| 255 | if (UniqueSuccessors.insert(Ptr: BB).second) |
| 256 | Updates.push_back(Elt: {DominatorTree::Insert, SwitchBB, BB}); |
| 257 | } |
| 258 | DTU->applyUpdates(Updates); |
| 259 | } |
| 260 | |
| 261 | return true; |
| 262 | } |
| 263 | |
| 264 | bool IndirectBrExpandLegacyPass::runOnFunction(Function &F) { |
| 265 | auto *TPC = getAnalysisIfAvailable<TargetPassConfig>(); |
| 266 | if (!TPC) |
| 267 | return false; |
| 268 | |
| 269 | auto &TM = TPC->getTM<TargetMachine>(); |
| 270 | auto &STI = *TM.getSubtargetImpl(F); |
| 271 | if (!STI.enableIndirectBrExpand()) |
| 272 | return false; |
| 273 | auto *TLI = STI.getTargetLowering(); |
| 274 | |
| 275 | std::optional<DomTreeUpdater> DTU; |
| 276 | if (auto *DTWP = getAnalysisIfAvailable<DominatorTreeWrapperPass>()) |
| 277 | DTU.emplace(args&: DTWP->getDomTree(), args: DomTreeUpdater::UpdateStrategy::Lazy); |
| 278 | |
| 279 | return runImpl(F, TLI, DTU: DTU ? &*DTU : nullptr); |
| 280 | } |
| 281 | |