| 1 | //===-- SPIRVStructurizer.cpp ----------------------*- C++ -*-===// |
| 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 | // |
| 9 | //===----------------------------------------------------------------------===// |
| 10 | |
| 11 | #include "Analysis/SPIRVConvergenceRegionAnalysis.h" |
| 12 | #include "SPIRV.h" |
| 13 | #include "SPIRVStructurizerWrapper.h" |
| 14 | #include "SPIRVSubtarget.h" |
| 15 | #include "SPIRVUtils.h" |
| 16 | #include "llvm/ADT/DenseMap.h" |
| 17 | #include "llvm/ADT/SmallPtrSet.h" |
| 18 | #include "llvm/Analysis/LoopInfo.h" |
| 19 | #include "llvm/CodeGen/IntrinsicLowering.h" |
| 20 | #include "llvm/IR/CFG.h" |
| 21 | #include "llvm/IR/Dominators.h" |
| 22 | #include "llvm/IR/IRBuilder.h" |
| 23 | #include "llvm/IR/IntrinsicInst.h" |
| 24 | #include "llvm/IR/Intrinsics.h" |
| 25 | #include "llvm/IR/IntrinsicsSPIRV.h" |
| 26 | #include "llvm/IR/LegacyPassManager.h" |
| 27 | #include "llvm/InitializePasses.h" |
| 28 | #include "llvm/Transforms/Utils.h" |
| 29 | #include "llvm/Transforms/Utils/Cloning.h" |
| 30 | #include "llvm/Transforms/Utils/LoopSimplify.h" |
| 31 | #include "llvm/Transforms/Utils/LowerMemIntrinsics.h" |
| 32 | #include <stack> |
| 33 | #include <unordered_set> |
| 34 | |
| 35 | using namespace llvm; |
| 36 | using namespace SPIRV; |
| 37 | |
| 38 | using BlockSet = std::unordered_set<BasicBlock *>; |
| 39 | using Edge = std::pair<BasicBlock *, BasicBlock *>; |
| 40 | |
| 41 | // Helper function to do a partial order visit from the block |Start|, calling |
| 42 | // |Op| on each visited node. |
| 43 | static void partialOrderVisit(BasicBlock &Start, |
| 44 | std::function<bool(BasicBlock *)> Op) { |
| 45 | PartialOrderingVisitor V(*Start.getParent()); |
| 46 | V.partialOrderVisit(Start, Op); |
| 47 | } |
| 48 | |
| 49 | // Returns the exact convergence region in the tree defined by `Node` for which |
| 50 | // `BB` is the header, nullptr otherwise. |
| 51 | static const ConvergenceRegion * |
| 52 | (const ConvergenceRegion *Node, BasicBlock *BB) { |
| 53 | if (Node->Entry == BB) |
| 54 | return Node; |
| 55 | |
| 56 | for (auto *Child : Node->Children) { |
| 57 | const auto *CR = getRegionForHeader(Node: Child, BB); |
| 58 | if (CR != nullptr) |
| 59 | return CR; |
| 60 | } |
| 61 | return nullptr; |
| 62 | } |
| 63 | |
| 64 | // Returns the single BasicBlock exiting the convergence region `CR`, |
| 65 | // nullptr if no such exit exists. |
| 66 | static BasicBlock *getExitFor(const ConvergenceRegion *CR) { |
| 67 | std::unordered_set<BasicBlock *> ExitTargets; |
| 68 | for (BasicBlock *Exit : CR->Exits) { |
| 69 | for (BasicBlock *Successor : successors(BB: Exit)) { |
| 70 | if (CR->Blocks.count(Ptr: Successor) == 0) |
| 71 | ExitTargets.insert(x: Successor); |
| 72 | } |
| 73 | } |
| 74 | |
| 75 | assert(ExitTargets.size() <= 1); |
| 76 | if (ExitTargets.size() == 0) |
| 77 | return nullptr; |
| 78 | |
| 79 | return *ExitTargets.begin(); |
| 80 | } |
| 81 | |
| 82 | // Returns the merge block designated by I if I is a merge instruction, nullptr |
| 83 | // otherwise. |
| 84 | static BasicBlock *getDesignatedMergeBlock(Instruction *I) { |
| 85 | IntrinsicInst *II = dyn_cast_or_null<IntrinsicInst>(Val: I); |
| 86 | if (II == nullptr) |
| 87 | return nullptr; |
| 88 | |
| 89 | if (II->getIntrinsicID() != Intrinsic::spv_loop_merge && |
| 90 | II->getIntrinsicID() != Intrinsic::spv_selection_merge) |
| 91 | return nullptr; |
| 92 | |
| 93 | BlockAddress *BA = cast<BlockAddress>(Val: II->getOperand(i_nocapture: 0)); |
| 94 | return BA->getBasicBlock(); |
| 95 | } |
| 96 | |
| 97 | // Returns the continue block designated by I if I is an OpLoopMerge, nullptr |
| 98 | // otherwise. |
| 99 | static BasicBlock *getDesignatedContinueBlock(Instruction *I) { |
| 100 | IntrinsicInst *II = dyn_cast_or_null<IntrinsicInst>(Val: I); |
| 101 | if (II == nullptr) |
| 102 | return nullptr; |
| 103 | |
| 104 | if (II->getIntrinsicID() != Intrinsic::spv_loop_merge) |
| 105 | return nullptr; |
| 106 | |
| 107 | BlockAddress *BA = cast<BlockAddress>(Val: II->getOperand(i_nocapture: 1)); |
| 108 | return BA->getBasicBlock(); |
| 109 | } |
| 110 | |
| 111 | // Returns true if Header has one merge instruction which designated Merge as |
| 112 | // merge block. |
| 113 | static bool isDefinedAsSelectionMergeBy(BasicBlock &, BasicBlock &Merge) { |
| 114 | for (auto &I : Header) { |
| 115 | BasicBlock *MB = getDesignatedMergeBlock(I: &I); |
| 116 | if (MB == &Merge) |
| 117 | return true; |
| 118 | } |
| 119 | return false; |
| 120 | } |
| 121 | |
| 122 | // Returns true if the BB has one OpLoopMerge instruction. |
| 123 | static bool hasLoopMergeInstruction(BasicBlock &BB) { |
| 124 | for (auto &I : BB) |
| 125 | if (getDesignatedContinueBlock(I: &I)) |
| 126 | return true; |
| 127 | return false; |
| 128 | } |
| 129 | |
| 130 | // Returns true is I is an OpSelectionMerge or OpLoopMerge instruction, false |
| 131 | // otherwise. |
| 132 | static bool isMergeInstruction(Instruction *I) { |
| 133 | return getDesignatedMergeBlock(I) != nullptr; |
| 134 | } |
| 135 | |
| 136 | // Returns all blocks in F having at least one OpLoopMerge or OpSelectionMerge |
| 137 | // instruction. |
| 138 | static SmallPtrSet<BasicBlock *, 2> (Function &F) { |
| 139 | SmallPtrSet<BasicBlock *, 2> Output; |
| 140 | for (BasicBlock &BB : F) { |
| 141 | for (Instruction &I : BB) { |
| 142 | if (getDesignatedMergeBlock(I: &I) != nullptr) |
| 143 | Output.insert(Ptr: &BB); |
| 144 | } |
| 145 | } |
| 146 | return Output; |
| 147 | } |
| 148 | |
| 149 | // Returns all basic blocks in |F| referenced by at least 1 |
| 150 | // OpSelectionMerge/OpLoopMerge instruction. |
| 151 | static SmallPtrSet<BasicBlock *, 2> getMergeBlocks(Function &F) { |
| 152 | SmallPtrSet<BasicBlock *, 2> Output; |
| 153 | for (BasicBlock &BB : F) { |
| 154 | for (Instruction &I : BB) { |
| 155 | BasicBlock *MB = getDesignatedMergeBlock(I: &I); |
| 156 | if (MB != nullptr) |
| 157 | Output.insert(Ptr: MB); |
| 158 | } |
| 159 | } |
| 160 | return Output; |
| 161 | } |
| 162 | |
| 163 | // Return all the merge instructions contained in BB. |
| 164 | // Note: the SPIR-V spec doesn't allow a single BB to contain more than 1 merge |
| 165 | // instruction, but this can happen while we structurize the CFG. |
| 166 | static std::vector<Instruction *> getMergeInstructions(BasicBlock &BB) { |
| 167 | std::vector<Instruction *> Output; |
| 168 | for (Instruction &I : BB) |
| 169 | if (isMergeInstruction(I: &I)) |
| 170 | Output.push_back(x: &I); |
| 171 | return Output; |
| 172 | } |
| 173 | |
| 174 | // Returns all basic blocks in |F| referenced as continue target by at least 1 |
| 175 | // OpLoopMerge instruction. |
| 176 | static SmallPtrSet<BasicBlock *, 2> getContinueBlocks(Function &F) { |
| 177 | SmallPtrSet<BasicBlock *, 2> Output; |
| 178 | for (BasicBlock &BB : F) { |
| 179 | for (Instruction &I : BB) { |
| 180 | BasicBlock *MB = getDesignatedContinueBlock(I: &I); |
| 181 | if (MB != nullptr) |
| 182 | Output.insert(Ptr: MB); |
| 183 | } |
| 184 | } |
| 185 | return Output; |
| 186 | } |
| 187 | |
| 188 | // Do a preorder traversal of the CFG starting from the BB |Start|. |
| 189 | // point. Calls |op| on each basic block encountered during the traversal. |
| 190 | static void visit(BasicBlock &Start, std::function<bool(BasicBlock *)> op) { |
| 191 | std::stack<BasicBlock *> ToVisit; |
| 192 | SmallPtrSet<BasicBlock *, 8> Seen; |
| 193 | |
| 194 | ToVisit.push(x: &Start); |
| 195 | Seen.insert(Ptr: ToVisit.top()); |
| 196 | while (ToVisit.size() != 0) { |
| 197 | BasicBlock *BB = ToVisit.top(); |
| 198 | ToVisit.pop(); |
| 199 | |
| 200 | if (!op(BB)) |
| 201 | continue; |
| 202 | |
| 203 | for (auto Succ : successors(BB)) { |
| 204 | if (Seen.contains(Ptr: Succ)) |
| 205 | continue; |
| 206 | ToVisit.push(x: Succ); |
| 207 | Seen.insert(Ptr: Succ); |
| 208 | } |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | // Replaces the conditional and unconditional branch targets of |BB| by |
| 213 | // |NewTarget| if the target was |OldTarget|. This function also makes sure the |
| 214 | // associated merge instruction gets updated accordingly. |
| 215 | static void replaceIfBranchTargets(BasicBlock *BB, BasicBlock *OldTarget, |
| 216 | BasicBlock *NewTarget) { |
| 217 | auto *BI = cast<BranchInst>(Val: BB->getTerminator()); |
| 218 | |
| 219 | // 1. Replace all matching successors. |
| 220 | for (size_t i = 0; i < BI->getNumSuccessors(); i++) { |
| 221 | if (BI->getSuccessor(i) == OldTarget) |
| 222 | BI->setSuccessor(idx: i, NewSucc: NewTarget); |
| 223 | } |
| 224 | |
| 225 | // Branch was unconditional, no fixup required. |
| 226 | if (BI->isUnconditional()) |
| 227 | return; |
| 228 | |
| 229 | // Branch had 2 successors, maybe now both are the same? |
| 230 | if (BI->getSuccessor(i: 0) != BI->getSuccessor(i: 1)) |
| 231 | return; |
| 232 | |
| 233 | // Note: we may end up here because the original IR had such branches. |
| 234 | // This means Target is not necessarily equal to NewTarget. |
| 235 | IRBuilder<> Builder(BB); |
| 236 | Builder.SetInsertPoint(BI); |
| 237 | Builder.CreateBr(Dest: BI->getSuccessor(i: 0)); |
| 238 | BI->eraseFromParent(); |
| 239 | |
| 240 | // The branch was the only instruction, nothing else to do. |
| 241 | if (BB->size() == 1) |
| 242 | return; |
| 243 | |
| 244 | // Otherwise, we need to check: was there an OpSelectionMerge before this |
| 245 | // branch? If we removed the OpBranchConditional, we must also remove the |
| 246 | // OpSelectionMerge. This is not valid for OpLoopMerge: |
| 247 | IntrinsicInst *II = |
| 248 | dyn_cast<IntrinsicInst>(Val: BB->getTerminator()->getPrevNode()); |
| 249 | if (!II || II->getIntrinsicID() != Intrinsic::spv_selection_merge) |
| 250 | return; |
| 251 | |
| 252 | Constant *C = cast<Constant>(Val: II->getOperand(i_nocapture: 0)); |
| 253 | II->eraseFromParent(); |
| 254 | if (!C->isConstantUsed()) |
| 255 | C->destroyConstant(); |
| 256 | } |
| 257 | |
| 258 | // Replaces the target of branch instruction in |BB| with |NewTarget| if it |
| 259 | // was |OldTarget|. This function also fixes the associated merge instruction. |
| 260 | // Note: this function does not simplify branching instructions, it only updates |
| 261 | // targets. See also: simplifyBranches. |
| 262 | static void replaceBranchTargets(BasicBlock *BB, BasicBlock *OldTarget, |
| 263 | BasicBlock *NewTarget) { |
| 264 | auto *T = BB->getTerminator(); |
| 265 | if (isa<ReturnInst>(Val: T)) |
| 266 | return; |
| 267 | |
| 268 | if (isa<BranchInst>(Val: T)) |
| 269 | return replaceIfBranchTargets(BB, OldTarget, NewTarget); |
| 270 | |
| 271 | if (auto *SI = dyn_cast<SwitchInst>(Val: T)) { |
| 272 | for (size_t i = 0; i < SI->getNumSuccessors(); i++) { |
| 273 | if (SI->getSuccessor(idx: i) == OldTarget) |
| 274 | SI->setSuccessor(idx: i, NewSucc: NewTarget); |
| 275 | } |
| 276 | return; |
| 277 | } |
| 278 | |
| 279 | assert(false && "Unhandled terminator type." ); |
| 280 | } |
| 281 | |
| 282 | namespace { |
| 283 | // Given a reducible CFG, produces a structurized CFG in the SPIR-V sense, |
| 284 | // adding merge instructions when required. |
| 285 | class SPIRVStructurizer : public FunctionPass { |
| 286 | struct DivergentConstruct; |
| 287 | // Represents a list of condition/loops/switch constructs. |
| 288 | // See SPIR-V 2.11.2. Structured Control-flow Constructs for the list of |
| 289 | // constructs. |
| 290 | using ConstructList = std::vector<std::unique_ptr<DivergentConstruct>>; |
| 291 | |
| 292 | // Represents a divergent construct in the SPIR-V sense. |
| 293 | // Such constructs are represented by a header (entry), a merge block (exit), |
| 294 | // and possibly a continue block (back-edge). A construct can contain other |
| 295 | // constructs, but their boundaries do not cross. |
| 296 | struct DivergentConstruct { |
| 297 | BasicBlock * = nullptr; |
| 298 | BasicBlock *Merge = nullptr; |
| 299 | BasicBlock *Continue = nullptr; |
| 300 | |
| 301 | DivergentConstruct *Parent = nullptr; |
| 302 | ConstructList Children; |
| 303 | }; |
| 304 | |
| 305 | // An helper class to clean the construct boundaries. |
| 306 | // It is used to gather the list of blocks that should belong to each |
| 307 | // divergent construct, and possibly modify CFG edges when exits would cross |
| 308 | // the boundary of multiple constructs. |
| 309 | struct Splitter { |
| 310 | Function &F; |
| 311 | LoopInfo &LI; |
| 312 | DomTreeBuilder::BBDomTree DT; |
| 313 | DomTreeBuilder::BBPostDomTree PDT; |
| 314 | |
| 315 | Splitter(Function &F, LoopInfo &LI) : F(F), LI(LI) { invalidate(); } |
| 316 | |
| 317 | void invalidate() { |
| 318 | PDT.recalculate(Func&: F); |
| 319 | DT.recalculate(Func&: F); |
| 320 | } |
| 321 | |
| 322 | // Returns the list of blocks that belong to a SPIR-V loop construct, |
| 323 | // including the continue construct. |
| 324 | std::vector<BasicBlock *> getLoopConstructBlocks(BasicBlock *, |
| 325 | BasicBlock *Merge) { |
| 326 | assert(DT.dominates(Header, Merge)); |
| 327 | std::vector<BasicBlock *> Output; |
| 328 | partialOrderVisit(Start&: *Header, Op: [&](BasicBlock *BB) { |
| 329 | if (BB == Merge) |
| 330 | return false; |
| 331 | if (DT.dominates(A: Merge, B: BB) || !DT.dominates(A: Header, B: BB)) |
| 332 | return false; |
| 333 | Output.push_back(x: BB); |
| 334 | return true; |
| 335 | }); |
| 336 | return Output; |
| 337 | } |
| 338 | |
| 339 | // Returns the list of blocks that belong to a SPIR-V selection construct. |
| 340 | std::vector<BasicBlock *> |
| 341 | getSelectionConstructBlocks(DivergentConstruct *Node) { |
| 342 | assert(DT.dominates(Node->Header, Node->Merge)); |
| 343 | BlockSet OutsideBlocks; |
| 344 | OutsideBlocks.insert(x: Node->Merge); |
| 345 | |
| 346 | for (DivergentConstruct *It = Node->Parent; It != nullptr; |
| 347 | It = It->Parent) { |
| 348 | OutsideBlocks.insert(x: It->Merge); |
| 349 | if (It->Continue) |
| 350 | OutsideBlocks.insert(x: It->Continue); |
| 351 | } |
| 352 | |
| 353 | std::vector<BasicBlock *> Output; |
| 354 | partialOrderVisit(Start&: *Node->Header, Op: [&](BasicBlock *BB) { |
| 355 | if (OutsideBlocks.count(x: BB) != 0) |
| 356 | return false; |
| 357 | if (DT.dominates(A: Node->Merge, B: BB) || !DT.dominates(A: Node->Header, B: BB)) |
| 358 | return false; |
| 359 | Output.push_back(x: BB); |
| 360 | return true; |
| 361 | }); |
| 362 | return Output; |
| 363 | } |
| 364 | |
| 365 | // Returns the list of blocks that belong to a SPIR-V switch construct. |
| 366 | std::vector<BasicBlock *> getSwitchConstructBlocks(BasicBlock *, |
| 367 | BasicBlock *Merge) { |
| 368 | assert(DT.dominates(Header, Merge)); |
| 369 | |
| 370 | std::vector<BasicBlock *> Output; |
| 371 | partialOrderVisit(Start&: *Header, Op: [&](BasicBlock *BB) { |
| 372 | // the blocks structurally dominated by a switch header, |
| 373 | if (!DT.dominates(A: Header, B: BB)) |
| 374 | return false; |
| 375 | // excluding blocks structurally dominated by the switch header’s merge |
| 376 | // block. |
| 377 | if (DT.dominates(A: Merge, B: BB) || BB == Merge) |
| 378 | return false; |
| 379 | Output.push_back(x: BB); |
| 380 | return true; |
| 381 | }); |
| 382 | return Output; |
| 383 | } |
| 384 | |
| 385 | // Returns the list of blocks that belong to a SPIR-V case construct. |
| 386 | std::vector<BasicBlock *> getCaseConstructBlocks(BasicBlock *Target, |
| 387 | BasicBlock *Merge) { |
| 388 | assert(DT.dominates(Target, Merge)); |
| 389 | |
| 390 | std::vector<BasicBlock *> Output; |
| 391 | partialOrderVisit(Start&: *Target, Op: [&](BasicBlock *BB) { |
| 392 | // the blocks structurally dominated by an OpSwitch Target or Default |
| 393 | // block |
| 394 | if (!DT.dominates(A: Target, B: BB)) |
| 395 | return false; |
| 396 | // excluding the blocks structurally dominated by the OpSwitch |
| 397 | // construct’s corresponding merge block. |
| 398 | if (DT.dominates(A: Merge, B: BB) || BB == Merge) |
| 399 | return false; |
| 400 | Output.push_back(x: BB); |
| 401 | return true; |
| 402 | }); |
| 403 | return Output; |
| 404 | } |
| 405 | |
| 406 | // Splits the given edges by recreating proxy nodes so that the destination |
| 407 | // has unique incoming edges from this region. |
| 408 | // |
| 409 | // clang-format off |
| 410 | // |
| 411 | // In SPIR-V, constructs must have a single exit/merge. |
| 412 | // Given nodes A and B in the construct, a node C outside, and the following edges. |
| 413 | // A -> C |
| 414 | // B -> C |
| 415 | // |
| 416 | // In such cases, we must create a new exit node D, that belong to the construct to make is viable: |
| 417 | // A -> D -> C |
| 418 | // B -> D -> C |
| 419 | // |
| 420 | // This is fine (assuming C has no PHI nodes), but requires handling the merge instruction here. |
| 421 | // By adding a proxy node, we create a regular divergent shape which can easily be regularized later on. |
| 422 | // A -> D -> D1 -> C |
| 423 | // B -> D -> D2 -> C |
| 424 | // |
| 425 | // A, B, D belongs to the construct. D is the exit. D1 and D2 are empty. |
| 426 | // |
| 427 | // clang-format on |
| 428 | std::vector<Edge> |
| 429 | createAliasBlocksForComplexEdges(std::vector<Edge> Edges) { |
| 430 | std::unordered_set<BasicBlock *> Seen; |
| 431 | std::vector<Edge> Output; |
| 432 | Output.reserve(n: Edges.size()); |
| 433 | |
| 434 | for (auto &[Src, Dst] : Edges) { |
| 435 | auto [Iterator, Inserted] = Seen.insert(x: Src); |
| 436 | if (!Inserted) { |
| 437 | // Src already a source node. Cannot have 2 edges from A to B. |
| 438 | // Creating alias source block. |
| 439 | BasicBlock *NewSrc = BasicBlock::Create( |
| 440 | Context&: F.getContext(), Name: Src->getName() + ".new.src" , Parent: &F); |
| 441 | replaceBranchTargets(BB: Src, OldTarget: Dst, NewTarget: NewSrc); |
| 442 | IRBuilder<> Builder(NewSrc); |
| 443 | Builder.CreateBr(Dest: Dst); |
| 444 | Src = NewSrc; |
| 445 | } |
| 446 | |
| 447 | Output.emplace_back(args&: Src, args&: Dst); |
| 448 | } |
| 449 | |
| 450 | return Output; |
| 451 | } |
| 452 | |
| 453 | AllocaInst *CreateVariable(Function &F, Type *Type, |
| 454 | BasicBlock::iterator Position) { |
| 455 | const DataLayout &DL = F.getDataLayout(); |
| 456 | return new AllocaInst(Type, DL.getAllocaAddrSpace(), nullptr, "reg" , |
| 457 | Position); |
| 458 | } |
| 459 | |
| 460 | // Given a construct defined by |Header|, and a list of exiting edges |
| 461 | // |Edges|, creates a new single exit node, fixing up those edges. |
| 462 | BasicBlock *createSingleExitNode(BasicBlock *, |
| 463 | std::vector<Edge> &Edges) { |
| 464 | |
| 465 | std::vector<Edge> FixedEdges = createAliasBlocksForComplexEdges(Edges); |
| 466 | |
| 467 | std::vector<BasicBlock *> Dsts; |
| 468 | std::unordered_map<BasicBlock *, ConstantInt *> DstToIndex; |
| 469 | auto NewExit = BasicBlock::Create(Context&: F.getContext(), |
| 470 | Name: Header->getName() + ".new.exit" , Parent: &F); |
| 471 | IRBuilder<> ExitBuilder(NewExit); |
| 472 | for (auto &[Src, Dst] : FixedEdges) { |
| 473 | if (DstToIndex.count(x: Dst) != 0) |
| 474 | continue; |
| 475 | DstToIndex.emplace(args&: Dst, args: ExitBuilder.getInt32(C: DstToIndex.size())); |
| 476 | Dsts.push_back(x: Dst); |
| 477 | } |
| 478 | |
| 479 | if (Dsts.size() == 1) { |
| 480 | for (auto &[Src, Dst] : FixedEdges) { |
| 481 | replaceBranchTargets(BB: Src, OldTarget: Dst, NewTarget: NewExit); |
| 482 | } |
| 483 | ExitBuilder.CreateBr(Dest: Dsts[0]); |
| 484 | return NewExit; |
| 485 | } |
| 486 | |
| 487 | AllocaInst *Variable = CreateVariable(F, Type: ExitBuilder.getInt32Ty(), |
| 488 | Position: F.begin()->getFirstInsertionPt()); |
| 489 | for (auto &[Src, Dst] : FixedEdges) { |
| 490 | IRBuilder<> B2(Src); |
| 491 | B2.SetInsertPoint(Src->getFirstInsertionPt()); |
| 492 | B2.CreateStore(Val: DstToIndex[Dst], Ptr: Variable); |
| 493 | replaceBranchTargets(BB: Src, OldTarget: Dst, NewTarget: NewExit); |
| 494 | } |
| 495 | |
| 496 | Value *Load = ExitBuilder.CreateLoad(Ty: ExitBuilder.getInt32Ty(), Ptr: Variable); |
| 497 | |
| 498 | // If we can avoid an OpSwitch, generate an OpBranch. Reason is some |
| 499 | // OpBranch are allowed to exist without a new OpSelectionMerge if one of |
| 500 | // the branch is the parent's merge node, while OpSwitches are not. |
| 501 | if (Dsts.size() == 2) { |
| 502 | Value *Condition = |
| 503 | ExitBuilder.CreateCmp(Pred: CmpInst::ICMP_EQ, LHS: DstToIndex[Dsts[0]], RHS: Load); |
| 504 | ExitBuilder.CreateCondBr(Cond: Condition, True: Dsts[0], False: Dsts[1]); |
| 505 | return NewExit; |
| 506 | } |
| 507 | |
| 508 | SwitchInst *Sw = ExitBuilder.CreateSwitch(V: Load, Dest: Dsts[0], NumCases: Dsts.size() - 1); |
| 509 | for (BasicBlock *BB : drop_begin(RangeOrContainer&: Dsts)) |
| 510 | Sw->addCase(OnVal: DstToIndex[BB], Dest: BB); |
| 511 | return NewExit; |
| 512 | } |
| 513 | }; |
| 514 | |
| 515 | /// Create a value in BB set to the value associated with the branch the block |
| 516 | /// terminator will take. |
| 517 | Value *createExitVariable( |
| 518 | BasicBlock *BB, |
| 519 | const DenseMap<BasicBlock *, ConstantInt *> &TargetToValue) { |
| 520 | auto *T = BB->getTerminator(); |
| 521 | if (isa<ReturnInst>(Val: T)) |
| 522 | return nullptr; |
| 523 | |
| 524 | IRBuilder<> Builder(BB); |
| 525 | Builder.SetInsertPoint(T); |
| 526 | |
| 527 | if (auto *BI = dyn_cast<BranchInst>(Val: T)) { |
| 528 | |
| 529 | BasicBlock *LHSTarget = BI->getSuccessor(i: 0); |
| 530 | BasicBlock *RHSTarget = |
| 531 | BI->isConditional() ? BI->getSuccessor(i: 1) : nullptr; |
| 532 | |
| 533 | Value *LHS = TargetToValue.lookup(Val: LHSTarget); |
| 534 | Value *RHS = TargetToValue.lookup(Val: RHSTarget); |
| 535 | |
| 536 | if (LHS == nullptr || RHS == nullptr) |
| 537 | return LHS == nullptr ? RHS : LHS; |
| 538 | return Builder.CreateSelect(C: BI->getCondition(), True: LHS, False: RHS); |
| 539 | } |
| 540 | |
| 541 | // TODO: add support for switch cases. |
| 542 | llvm_unreachable("Unhandled terminator type." ); |
| 543 | } |
| 544 | |
| 545 | // Creates a new basic block in F with a single OpUnreachable instruction. |
| 546 | BasicBlock *CreateUnreachable(Function &F) { |
| 547 | BasicBlock *BB = BasicBlock::Create(Context&: F.getContext(), Name: "unreachable" , Parent: &F); |
| 548 | IRBuilder<> Builder(BB); |
| 549 | Builder.CreateUnreachable(); |
| 550 | return BB; |
| 551 | } |
| 552 | |
| 553 | // Add OpLoopMerge instruction on cycles. |
| 554 | bool addMergeForLoops(Function &F) { |
| 555 | LoopInfo &LI = getAnalysis<LoopInfoWrapperPass>().getLoopInfo(); |
| 556 | auto *TopLevelRegion = |
| 557 | getAnalysis<SPIRVConvergenceRegionAnalysisWrapperPass>() |
| 558 | .getRegionInfo() |
| 559 | .getTopLevelRegion(); |
| 560 | |
| 561 | bool Modified = false; |
| 562 | for (auto &BB : F) { |
| 563 | // Not a loop header. Ignoring for now. |
| 564 | if (!LI.isLoopHeader(BB: &BB)) |
| 565 | continue; |
| 566 | auto *L = LI.getLoopFor(BB: &BB); |
| 567 | |
| 568 | // This loop header is not the entrance of a convergence region. Ignoring |
| 569 | // this block. |
| 570 | auto *CR = getRegionForHeader(Node: TopLevelRegion, BB: &BB); |
| 571 | if (CR == nullptr) |
| 572 | continue; |
| 573 | |
| 574 | IRBuilder<> Builder(&BB); |
| 575 | |
| 576 | auto *Merge = getExitFor(CR); |
| 577 | // We are indeed in a loop, but there are no exits (infinite loop). |
| 578 | // This could be caused by a bad shader, but also could be an artifact |
| 579 | // from an earlier optimization. It is not always clear if structurally |
| 580 | // reachable means runtime reachable, so we cannot error-out. What we must |
| 581 | // do however is to make is legal on the SPIR-V point of view, hence |
| 582 | // adding an unreachable merge block. |
| 583 | if (Merge == nullptr) { |
| 584 | BranchInst *Br = cast<BranchInst>(Val: BB.getTerminator()); |
| 585 | assert(Br->isUnconditional()); |
| 586 | |
| 587 | Merge = CreateUnreachable(F); |
| 588 | Builder.SetInsertPoint(Br); |
| 589 | Builder.CreateCondBr(Cond: Builder.getFalse(), True: Merge, False: Br->getSuccessor(i: 0)); |
| 590 | Br->eraseFromParent(); |
| 591 | } |
| 592 | |
| 593 | auto *Continue = L->getLoopLatch(); |
| 594 | |
| 595 | Builder.SetInsertPoint(BB.getTerminator()); |
| 596 | auto MergeAddress = BlockAddress::get(F: Merge->getParent(), BB: Merge); |
| 597 | auto ContinueAddress = BlockAddress::get(F: Continue->getParent(), BB: Continue); |
| 598 | SmallVector<Value *, 2> Args = {MergeAddress, ContinueAddress}; |
| 599 | SmallVector<unsigned, 1> LoopControlImms = |
| 600 | getSpirvLoopControlOperandsFromLoopMetadata(L); |
| 601 | for (unsigned Imm : LoopControlImms) |
| 602 | Args.emplace_back(Args: ConstantInt::get(Ty: Builder.getInt32Ty(), V: Imm)); |
| 603 | Builder.CreateIntrinsic(ID: Intrinsic::spv_loop_merge, Args: {Args}); |
| 604 | Modified = true; |
| 605 | } |
| 606 | |
| 607 | return Modified; |
| 608 | } |
| 609 | |
| 610 | // Adds an OpSelectionMerge to the immediate dominator or each node with an |
| 611 | // in-degree of 2 or more which is not already the merge target of an |
| 612 | // OpLoopMerge/OpSelectionMerge. |
| 613 | bool addMergeForNodesWithMultiplePredecessors(Function &F) { |
| 614 | DomTreeBuilder::BBDomTree DT; |
| 615 | DT.recalculate(Func&: F); |
| 616 | |
| 617 | bool Modified = false; |
| 618 | for (auto &BB : F) { |
| 619 | if (pred_size(BB: &BB) <= 1) |
| 620 | continue; |
| 621 | |
| 622 | if (hasLoopMergeInstruction(BB) && pred_size(BB: &BB) <= 2) |
| 623 | continue; |
| 624 | |
| 625 | assert(DT.getNode(&BB)->getIDom()); |
| 626 | BasicBlock * = DT.getNode(BB: &BB)->getIDom()->getBlock(); |
| 627 | |
| 628 | if (isDefinedAsSelectionMergeBy(Header&: *Header, Merge&: BB)) |
| 629 | continue; |
| 630 | |
| 631 | IRBuilder<> Builder(Header); |
| 632 | Builder.SetInsertPoint(Header->getTerminator()); |
| 633 | |
| 634 | auto MergeAddress = BlockAddress::get(F: BB.getParent(), BB: &BB); |
| 635 | createOpSelectMerge(Builder: &Builder, MergeAddress); |
| 636 | |
| 637 | Modified = true; |
| 638 | } |
| 639 | |
| 640 | return Modified; |
| 641 | } |
| 642 | |
| 643 | // When a block has multiple OpSelectionMerge/OpLoopMerge instructions, sorts |
| 644 | // them to put the "largest" first. A merge instruction is defined as larger |
| 645 | // than another when its target merge block post-dominates the other target's |
| 646 | // merge block. (This ordering should match the nesting ordering of the source |
| 647 | // HLSL). |
| 648 | bool sortSelectionMerge(Function &F, BasicBlock &Block) { |
| 649 | std::vector<Instruction *> MergeInstructions; |
| 650 | for (Instruction &I : Block) |
| 651 | if (isMergeInstruction(I: &I)) |
| 652 | MergeInstructions.push_back(x: &I); |
| 653 | |
| 654 | if (MergeInstructions.size() <= 1) |
| 655 | return false; |
| 656 | |
| 657 | Instruction *InsertionPoint = *MergeInstructions.begin(); |
| 658 | |
| 659 | PartialOrderingVisitor Visitor(F); |
| 660 | std::sort(first: MergeInstructions.begin(), last: MergeInstructions.end(), |
| 661 | comp: [&Visitor](Instruction *Left, Instruction *Right) { |
| 662 | if (Left == Right) |
| 663 | return false; |
| 664 | BasicBlock *RightMerge = getDesignatedMergeBlock(I: Right); |
| 665 | BasicBlock *LeftMerge = getDesignatedMergeBlock(I: Left); |
| 666 | return !Visitor.compare(LHS: RightMerge, RHS: LeftMerge); |
| 667 | }); |
| 668 | |
| 669 | for (Instruction *I : MergeInstructions) { |
| 670 | I->moveBefore(InsertPos: InsertionPoint->getIterator()); |
| 671 | InsertionPoint = I; |
| 672 | } |
| 673 | |
| 674 | return true; |
| 675 | } |
| 676 | |
| 677 | // Sorts selection merge headers in |F|. |
| 678 | // A is sorted before B if the merge block designated by B is an ancestor of |
| 679 | // the one designated by A. |
| 680 | bool (Function &F) { |
| 681 | bool Modified = false; |
| 682 | for (BasicBlock &BB : F) { |
| 683 | Modified |= sortSelectionMerge(F, Block&: BB); |
| 684 | } |
| 685 | return Modified; |
| 686 | } |
| 687 | |
| 688 | // Split basic blocks containing multiple OpLoopMerge/OpSelectionMerge |
| 689 | // instructions so each basic block contains only a single merge instruction. |
| 690 | bool (Function &F) { |
| 691 | std::stack<BasicBlock *> Work; |
| 692 | for (auto &BB : F) { |
| 693 | std::vector<Instruction *> MergeInstructions = getMergeInstructions(BB); |
| 694 | if (MergeInstructions.size() <= 1) |
| 695 | continue; |
| 696 | Work.push(x: &BB); |
| 697 | } |
| 698 | |
| 699 | const bool Modified = Work.size() > 0; |
| 700 | while (Work.size() > 0) { |
| 701 | BasicBlock * = Work.top(); |
| 702 | Work.pop(); |
| 703 | |
| 704 | std::vector<Instruction *> MergeInstructions = |
| 705 | getMergeInstructions(BB&: *Header); |
| 706 | for (unsigned i = 1; i < MergeInstructions.size(); i++) { |
| 707 | BasicBlock *NewBlock = |
| 708 | Header->splitBasicBlock(I: MergeInstructions[i], BBName: "new.header" ); |
| 709 | |
| 710 | if (getDesignatedContinueBlock(I: MergeInstructions[0]) == nullptr) { |
| 711 | BasicBlock *Unreachable = CreateUnreachable(F); |
| 712 | |
| 713 | BranchInst *BI = cast<BranchInst>(Val: Header->getTerminator()); |
| 714 | IRBuilder<> Builder(Header); |
| 715 | Builder.SetInsertPoint(BI); |
| 716 | Builder.CreateCondBr(Cond: Builder.getTrue(), True: NewBlock, False: Unreachable); |
| 717 | BI->eraseFromParent(); |
| 718 | } |
| 719 | |
| 720 | Header = NewBlock; |
| 721 | } |
| 722 | } |
| 723 | |
| 724 | return Modified; |
| 725 | } |
| 726 | |
| 727 | // Adds an OpSelectionMerge to each block with an out-degree >= 2 which |
| 728 | // doesn't already have an OpSelectionMerge. |
| 729 | bool addMergeForDivergentBlocks(Function &F) { |
| 730 | DomTreeBuilder::BBPostDomTree PDT; |
| 731 | PDT.recalculate(Func&: F); |
| 732 | bool Modified = false; |
| 733 | |
| 734 | auto MergeBlocks = getMergeBlocks(F); |
| 735 | auto ContinueBlocks = getContinueBlocks(F); |
| 736 | |
| 737 | for (auto &BB : F) { |
| 738 | if (getMergeInstructions(BB).size() != 0) |
| 739 | continue; |
| 740 | |
| 741 | std::vector<BasicBlock *> Candidates; |
| 742 | for (BasicBlock *Successor : successors(BB: &BB)) { |
| 743 | if (MergeBlocks.contains(Ptr: Successor)) |
| 744 | continue; |
| 745 | if (ContinueBlocks.contains(Ptr: Successor)) |
| 746 | continue; |
| 747 | Candidates.push_back(x: Successor); |
| 748 | } |
| 749 | |
| 750 | if (Candidates.size() <= 1) |
| 751 | continue; |
| 752 | |
| 753 | Modified = true; |
| 754 | BasicBlock *Merge = Candidates[0]; |
| 755 | |
| 756 | auto MergeAddress = BlockAddress::get(F: Merge->getParent(), BB: Merge); |
| 757 | IRBuilder<> Builder(&BB); |
| 758 | Builder.SetInsertPoint(BB.getTerminator()); |
| 759 | createOpSelectMerge(Builder: &Builder, MergeAddress); |
| 760 | } |
| 761 | |
| 762 | return Modified; |
| 763 | } |
| 764 | |
| 765 | // Gather all the exit nodes for the construct header by |Header| and |
| 766 | // containing the blocks |Construct|. |
| 767 | std::vector<Edge> getExitsFrom(const BlockSet &Construct, |
| 768 | BasicBlock &) { |
| 769 | std::vector<Edge> Output; |
| 770 | visit(Start&: Header, op: [&](BasicBlock *Item) { |
| 771 | if (Construct.count(x: Item) == 0) |
| 772 | return false; |
| 773 | |
| 774 | for (BasicBlock *Successor : successors(BB: Item)) { |
| 775 | if (Construct.count(x: Successor) == 0) |
| 776 | Output.emplace_back(args&: Item, args&: Successor); |
| 777 | } |
| 778 | return true; |
| 779 | }); |
| 780 | |
| 781 | return Output; |
| 782 | } |
| 783 | |
| 784 | // Build a divergent construct tree searching from |BB|. |
| 785 | // If |Parent| is not null, this tree is attached to the parent's tree. |
| 786 | void constructDivergentConstruct(BlockSet &Visited, Splitter &S, |
| 787 | BasicBlock *BB, DivergentConstruct *Parent) { |
| 788 | if (Visited.count(x: BB) != 0) |
| 789 | return; |
| 790 | Visited.insert(x: BB); |
| 791 | |
| 792 | auto MIS = getMergeInstructions(BB&: *BB); |
| 793 | if (MIS.size() == 0) { |
| 794 | for (BasicBlock *Successor : successors(BB)) |
| 795 | constructDivergentConstruct(Visited, S, BB: Successor, Parent); |
| 796 | return; |
| 797 | } |
| 798 | |
| 799 | assert(MIS.size() == 1); |
| 800 | Instruction *MI = MIS[0]; |
| 801 | |
| 802 | BasicBlock *Merge = getDesignatedMergeBlock(I: MI); |
| 803 | BasicBlock *Continue = getDesignatedContinueBlock(I: MI); |
| 804 | |
| 805 | auto Output = std::make_unique<DivergentConstruct>(); |
| 806 | Output->Header = BB; |
| 807 | Output->Merge = Merge; |
| 808 | Output->Continue = Continue; |
| 809 | Output->Parent = Parent; |
| 810 | |
| 811 | constructDivergentConstruct(Visited, S, BB: Merge, Parent); |
| 812 | if (Continue) |
| 813 | constructDivergentConstruct(Visited, S, BB: Continue, Parent: Output.get()); |
| 814 | |
| 815 | for (BasicBlock *Successor : successors(BB)) |
| 816 | constructDivergentConstruct(Visited, S, BB: Successor, Parent: Output.get()); |
| 817 | |
| 818 | if (Parent) |
| 819 | Parent->Children.emplace_back(args: std::move(Output)); |
| 820 | } |
| 821 | |
| 822 | // Returns the blocks belonging to the divergent construct |Node|. |
| 823 | BlockSet getConstructBlocks(Splitter &S, DivergentConstruct *Node) { |
| 824 | assert(Node->Header && Node->Merge); |
| 825 | |
| 826 | if (Node->Continue) { |
| 827 | auto LoopBlocks = S.getLoopConstructBlocks(Header: Node->Header, Merge: Node->Merge); |
| 828 | return BlockSet(LoopBlocks.begin(), LoopBlocks.end()); |
| 829 | } |
| 830 | |
| 831 | auto SelectionBlocks = S.getSelectionConstructBlocks(Node); |
| 832 | return BlockSet(SelectionBlocks.begin(), SelectionBlocks.end()); |
| 833 | } |
| 834 | |
| 835 | // Fixup the construct |Node| to respect a set of rules defined by the SPIR-V |
| 836 | // spec. |
| 837 | bool fixupConstruct(Splitter &S, DivergentConstruct *Node) { |
| 838 | bool Modified = false; |
| 839 | for (auto &Child : Node->Children) |
| 840 | Modified |= fixupConstruct(S, Node: Child.get()); |
| 841 | |
| 842 | // This construct is the root construct. Does not represent any real |
| 843 | // construct, just a way to access the first level of the forest. |
| 844 | if (Node->Parent == nullptr) |
| 845 | return Modified; |
| 846 | |
| 847 | // This node's parent is the root. Meaning this is a top-level construct. |
| 848 | // There can be multiple exists, but all are guaranteed to exit at most 1 |
| 849 | // construct since we are at first level. |
| 850 | if (Node->Parent->Header == nullptr) |
| 851 | return Modified; |
| 852 | |
| 853 | // Health check for the structure. |
| 854 | assert(Node->Header && Node->Merge); |
| 855 | assert(Node->Parent->Header && Node->Parent->Merge); |
| 856 | |
| 857 | BlockSet ConstructBlocks = getConstructBlocks(S, Node); |
| 858 | auto Edges = getExitsFrom(Construct: ConstructBlocks, Header&: *Node->Header); |
| 859 | |
| 860 | // No edges exiting the construct. |
| 861 | if (Edges.size() < 1) |
| 862 | return Modified; |
| 863 | |
| 864 | bool HasBadEdge = Node->Merge == Node->Parent->Merge || |
| 865 | Node->Merge == Node->Parent->Continue; |
| 866 | // BasicBlock *Target = Edges[0].second; |
| 867 | for (auto &[Src, Dst] : Edges) { |
| 868 | // - Breaking from a selection construct: S is a selection construct, S is |
| 869 | // the innermost structured |
| 870 | // control-flow construct containing A, and B is the merge block for S |
| 871 | // - Breaking from the innermost loop: S is the innermost loop construct |
| 872 | // containing A, |
| 873 | // and B is the merge block for S |
| 874 | if (Node->Merge == Dst) |
| 875 | continue; |
| 876 | |
| 877 | // Entering the innermost loop’s continue construct: S is the innermost |
| 878 | // loop construct containing A, and B is the continue target for S |
| 879 | if (Node->Continue == Dst) |
| 880 | continue; |
| 881 | |
| 882 | // TODO: what about cases branching to another case in the switch? Seems |
| 883 | // to work, but need to double check. |
| 884 | HasBadEdge = true; |
| 885 | } |
| 886 | |
| 887 | if (!HasBadEdge) |
| 888 | return Modified; |
| 889 | |
| 890 | // Create a single exit node gathering all exit edges. |
| 891 | BasicBlock *NewExit = S.createSingleExitNode(Header: Node->Header, Edges); |
| 892 | |
| 893 | // Fixup this construct's merge node to point to the new exit. |
| 894 | // Note: this algorithm fixes inner-most divergence construct first. So |
| 895 | // recursive structures sharing a single merge node are fixed from the |
| 896 | // inside toward the outside. |
| 897 | auto MergeInstructions = getMergeInstructions(BB&: *Node->Header); |
| 898 | assert(MergeInstructions.size() == 1); |
| 899 | Instruction *I = MergeInstructions[0]; |
| 900 | BlockAddress *BA = cast<BlockAddress>(Val: I->getOperand(i: 0)); |
| 901 | if (BA->getBasicBlock() == Node->Merge) { |
| 902 | auto MergeAddress = BlockAddress::get(F: NewExit->getParent(), BB: NewExit); |
| 903 | I->setOperand(i: 0, Val: MergeAddress); |
| 904 | } |
| 905 | |
| 906 | // Clean up of the possible dangling BockAddr operands to prevent MIR |
| 907 | // comments about "address of removed block taken". |
| 908 | if (!BA->isConstantUsed()) |
| 909 | BA->destroyConstant(); |
| 910 | |
| 911 | Node->Merge = NewExit; |
| 912 | // Regenerate the dom trees. |
| 913 | S.invalidate(); |
| 914 | return true; |
| 915 | } |
| 916 | |
| 917 | bool splitCriticalEdges(Function &F) { |
| 918 | LoopInfo &LI = getAnalysis<LoopInfoWrapperPass>().getLoopInfo(); |
| 919 | Splitter S(F, LI); |
| 920 | |
| 921 | DivergentConstruct Root; |
| 922 | BlockSet Visited; |
| 923 | constructDivergentConstruct(Visited, S, BB: &*F.begin(), Parent: &Root); |
| 924 | return fixupConstruct(S, Node: &Root); |
| 925 | } |
| 926 | |
| 927 | // Simplify branches when possible: |
| 928 | // - if the 2 sides of a conditional branch are the same, transforms it to an |
| 929 | // unconditional branch. |
| 930 | // - if a switch has only 2 distinct successors, converts it to a conditional |
| 931 | // branch. |
| 932 | bool simplifyBranches(Function &F) { |
| 933 | bool Modified = false; |
| 934 | |
| 935 | for (BasicBlock &BB : F) { |
| 936 | SwitchInst *SI = dyn_cast<SwitchInst>(Val: BB.getTerminator()); |
| 937 | if (!SI) |
| 938 | continue; |
| 939 | if (SI->getNumCases() > 1) |
| 940 | continue; |
| 941 | |
| 942 | Modified = true; |
| 943 | IRBuilder<> Builder(&BB); |
| 944 | Builder.SetInsertPoint(SI); |
| 945 | |
| 946 | if (SI->getNumCases() == 0) { |
| 947 | Builder.CreateBr(Dest: SI->getDefaultDest()); |
| 948 | } else { |
| 949 | Value *Condition = |
| 950 | Builder.CreateCmp(Pred: CmpInst::ICMP_EQ, LHS: SI->getCondition(), |
| 951 | RHS: SI->case_begin()->getCaseValue()); |
| 952 | Builder.CreateCondBr(Cond: Condition, True: SI->case_begin()->getCaseSuccessor(), |
| 953 | False: SI->getDefaultDest()); |
| 954 | } |
| 955 | SI->eraseFromParent(); |
| 956 | } |
| 957 | |
| 958 | return Modified; |
| 959 | } |
| 960 | |
| 961 | // Makes sure every case target in |F| is unique. If 2 cases branch to the |
| 962 | // same basic block, one of the targets is updated so it jumps to a new basic |
| 963 | // block ending with a single unconditional branch to the original target. |
| 964 | bool splitSwitchCases(Function &F) { |
| 965 | bool Modified = false; |
| 966 | |
| 967 | for (BasicBlock &BB : F) { |
| 968 | SwitchInst *SI = dyn_cast<SwitchInst>(Val: BB.getTerminator()); |
| 969 | if (!SI) |
| 970 | continue; |
| 971 | |
| 972 | BlockSet Seen; |
| 973 | Seen.insert(x: SI->getDefaultDest()); |
| 974 | |
| 975 | auto It = SI->case_begin(); |
| 976 | while (It != SI->case_end()) { |
| 977 | BasicBlock *Target = It->getCaseSuccessor(); |
| 978 | if (Seen.count(x: Target) == 0) { |
| 979 | Seen.insert(x: Target); |
| 980 | ++It; |
| 981 | continue; |
| 982 | } |
| 983 | |
| 984 | Modified = true; |
| 985 | BasicBlock *NewTarget = |
| 986 | BasicBlock::Create(Context&: F.getContext(), Name: "new.sw.case" , Parent: &F); |
| 987 | IRBuilder<> Builder(NewTarget); |
| 988 | Builder.CreateBr(Dest: Target); |
| 989 | SI->addCase(OnVal: It->getCaseValue(), Dest: NewTarget); |
| 990 | It = SI->removeCase(I: It); |
| 991 | } |
| 992 | } |
| 993 | |
| 994 | return Modified; |
| 995 | } |
| 996 | |
| 997 | // Removes blocks not contributing to any structured CFG. This assumes there |
| 998 | // is no PHI nodes. |
| 999 | bool removeUselessBlocks(Function &F) { |
| 1000 | std::vector<BasicBlock *> ToRemove; |
| 1001 | |
| 1002 | auto MergeBlocks = getMergeBlocks(F); |
| 1003 | auto ContinueBlocks = getContinueBlocks(F); |
| 1004 | |
| 1005 | for (BasicBlock &BB : F) { |
| 1006 | if (BB.size() != 1) |
| 1007 | continue; |
| 1008 | |
| 1009 | if (isa<ReturnInst>(Val: BB.getTerminator())) |
| 1010 | continue; |
| 1011 | |
| 1012 | if (MergeBlocks.count(Ptr: &BB) != 0 || ContinueBlocks.count(Ptr: &BB) != 0) |
| 1013 | continue; |
| 1014 | |
| 1015 | if (BB.getUniqueSuccessor() == nullptr) |
| 1016 | continue; |
| 1017 | |
| 1018 | BasicBlock *Successor = BB.getUniqueSuccessor(); |
| 1019 | std::vector<BasicBlock *> Predecessors(predecessors(BB: &BB).begin(), |
| 1020 | predecessors(BB: &BB).end()); |
| 1021 | for (BasicBlock *Predecessor : Predecessors) |
| 1022 | replaceBranchTargets(BB: Predecessor, OldTarget: &BB, NewTarget: Successor); |
| 1023 | ToRemove.push_back(x: &BB); |
| 1024 | } |
| 1025 | |
| 1026 | for (BasicBlock *BB : ToRemove) |
| 1027 | BB->eraseFromParent(); |
| 1028 | |
| 1029 | return ToRemove.size() != 0; |
| 1030 | } |
| 1031 | |
| 1032 | bool addHeaderToRemainingDivergentDAG(Function &F) { |
| 1033 | bool Modified = false; |
| 1034 | |
| 1035 | auto MergeBlocks = getMergeBlocks(F); |
| 1036 | auto ContinueBlocks = getContinueBlocks(F); |
| 1037 | auto = getHeaderBlocks(F); |
| 1038 | |
| 1039 | DomTreeBuilder::BBDomTree DT; |
| 1040 | DomTreeBuilder::BBPostDomTree PDT; |
| 1041 | PDT.recalculate(Func&: F); |
| 1042 | DT.recalculate(Func&: F); |
| 1043 | |
| 1044 | for (BasicBlock &BB : F) { |
| 1045 | if (HeaderBlocks.count(Ptr: &BB) != 0) |
| 1046 | continue; |
| 1047 | if (succ_size(BB: &BB) < 2) |
| 1048 | continue; |
| 1049 | |
| 1050 | size_t CandidateEdges = 0; |
| 1051 | for (BasicBlock *Successor : successors(BB: &BB)) { |
| 1052 | if (MergeBlocks.count(Ptr: Successor) != 0 || |
| 1053 | ContinueBlocks.count(Ptr: Successor) != 0) |
| 1054 | continue; |
| 1055 | if (HeaderBlocks.count(Ptr: Successor) != 0) |
| 1056 | continue; |
| 1057 | CandidateEdges += 1; |
| 1058 | } |
| 1059 | |
| 1060 | if (CandidateEdges <= 1) |
| 1061 | continue; |
| 1062 | |
| 1063 | BasicBlock * = &BB; |
| 1064 | BasicBlock *Merge = PDT.getNode(BB: &BB)->getIDom()->getBlock(); |
| 1065 | |
| 1066 | bool HasBadBlock = false; |
| 1067 | visit(Start&: *Header, op: [&](const BasicBlock *Node) { |
| 1068 | if (DT.dominates(A: Header, B: Node)) |
| 1069 | return false; |
| 1070 | if (PDT.dominates(A: Merge, B: Node)) |
| 1071 | return false; |
| 1072 | if (Node == Header || Node == Merge) |
| 1073 | return true; |
| 1074 | |
| 1075 | HasBadBlock |= MergeBlocks.count(Ptr: Node) != 0 || |
| 1076 | ContinueBlocks.count(Ptr: Node) != 0 || |
| 1077 | HeaderBlocks.count(Ptr: Node) != 0; |
| 1078 | return !HasBadBlock; |
| 1079 | }); |
| 1080 | |
| 1081 | if (HasBadBlock) |
| 1082 | continue; |
| 1083 | |
| 1084 | Modified = true; |
| 1085 | |
| 1086 | if (Merge == nullptr) { |
| 1087 | Merge = *successors(BB: Header).begin(); |
| 1088 | IRBuilder<> Builder(Header); |
| 1089 | Builder.SetInsertPoint(Header->getTerminator()); |
| 1090 | |
| 1091 | auto MergeAddress = BlockAddress::get(F: Merge->getParent(), BB: Merge); |
| 1092 | createOpSelectMerge(Builder: &Builder, MergeAddress); |
| 1093 | continue; |
| 1094 | } |
| 1095 | |
| 1096 | Instruction *SplitInstruction = Merge->getTerminator(); |
| 1097 | if (isMergeInstruction(I: SplitInstruction->getPrevNode())) |
| 1098 | SplitInstruction = SplitInstruction->getPrevNode(); |
| 1099 | BasicBlock *NewMerge = |
| 1100 | Merge->splitBasicBlockBefore(I: SplitInstruction, BBName: "new.merge" ); |
| 1101 | |
| 1102 | IRBuilder<> Builder(Header); |
| 1103 | Builder.SetInsertPoint(Header->getTerminator()); |
| 1104 | |
| 1105 | auto MergeAddress = BlockAddress::get(F: NewMerge->getParent(), BB: NewMerge); |
| 1106 | createOpSelectMerge(Builder: &Builder, MergeAddress); |
| 1107 | } |
| 1108 | |
| 1109 | return Modified; |
| 1110 | } |
| 1111 | |
| 1112 | public: |
| 1113 | static char ID; |
| 1114 | |
| 1115 | SPIRVStructurizer() : FunctionPass(ID) {} |
| 1116 | |
| 1117 | virtual bool runOnFunction(Function &F) override { |
| 1118 | bool Modified = false; |
| 1119 | |
| 1120 | // In LLVM, Switches are allowed to have several cases branching to the same |
| 1121 | // basic block. This is allowed in SPIR-V, but can make structurizing SPIR-V |
| 1122 | // harder, so first remove edge cases. |
| 1123 | Modified |= splitSwitchCases(F); |
| 1124 | |
| 1125 | // LLVM allows conditional branches to have both side jumping to the same |
| 1126 | // block. It also allows switched to have a single default, or just one |
| 1127 | // case. Cleaning this up now. |
| 1128 | Modified |= simplifyBranches(F); |
| 1129 | |
| 1130 | // At this state, we should have a reducible CFG with cycles. |
| 1131 | // STEP 1: Adding OpLoopMerge instructions to loop headers. |
| 1132 | Modified |= addMergeForLoops(F); |
| 1133 | |
| 1134 | // STEP 2: adding OpSelectionMerge to each node with an in-degree >= 2. |
| 1135 | Modified |= addMergeForNodesWithMultiplePredecessors(F); |
| 1136 | |
| 1137 | // STEP 3: |
| 1138 | // Sort selection merge, the largest construct goes first. |
| 1139 | // This simplifies the next step. |
| 1140 | Modified |= sortSelectionMergeHeaders(F); |
| 1141 | |
| 1142 | // STEP 4: As this stage, we can have a single basic block with multiple |
| 1143 | // OpLoopMerge/OpSelectionMerge instructions. Splitting this block so each |
| 1144 | // BB has a single merge instruction. |
| 1145 | Modified |= splitBlocksWithMultipleHeaders(F); |
| 1146 | |
| 1147 | // STEP 5: In the previous steps, we added merge blocks the loops and |
| 1148 | // natural merge blocks (in-degree >= 2). What remains are conditions with |
| 1149 | // an exiting branch (return, unreachable). In such case, we must start from |
| 1150 | // the header, and add headers to divergent construct with no headers. |
| 1151 | Modified |= addMergeForDivergentBlocks(F); |
| 1152 | |
| 1153 | // STEP 6: At this stage, we have several divergent construct defines by a |
| 1154 | // header and a merge block. But their boundaries have no constraints: a |
| 1155 | // construct exit could be outside of the parents' construct exit. Such |
| 1156 | // edges are called critical edges. What we need is to split those edges |
| 1157 | // into several parts. Each part exiting the parent's construct by its merge |
| 1158 | // block. |
| 1159 | Modified |= splitCriticalEdges(F); |
| 1160 | |
| 1161 | // STEP 7: The previous steps possibly created a lot of "proxy" blocks. |
| 1162 | // Blocks with a single unconditional branch, used to create a valid |
| 1163 | // divergent construct tree. Some nodes are still requires (e.g: nodes |
| 1164 | // allowing a valid exit through the parent's merge block). But some are |
| 1165 | // left-overs of past transformations, and could cause actual validation |
| 1166 | // issues. E.g: the SPIR-V spec allows a construct to break to the parents |
| 1167 | // loop construct without an OpSelectionMerge, but this requires a straight |
| 1168 | // jump. If a proxy block lies between the conditional branch and the |
| 1169 | // parent's merge, the CFG is not valid. |
| 1170 | Modified |= removeUselessBlocks(F); |
| 1171 | |
| 1172 | // STEP 8: Final fix-up steps: our tree boundaries are correct, but some |
| 1173 | // blocks are branching with no header. Those are often simple conditional |
| 1174 | // branches with 1 or 2 returning edges. Adding a header for those. |
| 1175 | Modified |= addHeaderToRemainingDivergentDAG(F); |
| 1176 | |
| 1177 | // STEP 9: sort basic blocks to match both the LLVM & SPIR-V requirements. |
| 1178 | Modified |= sortBlocks(F); |
| 1179 | |
| 1180 | return Modified; |
| 1181 | } |
| 1182 | |
| 1183 | void getAnalysisUsage(AnalysisUsage &AU) const override { |
| 1184 | AU.addRequired<DominatorTreeWrapperPass>(); |
| 1185 | AU.addRequired<LoopInfoWrapperPass>(); |
| 1186 | AU.addRequired<SPIRVConvergenceRegionAnalysisWrapperPass>(); |
| 1187 | |
| 1188 | AU.addPreserved<SPIRVConvergenceRegionAnalysisWrapperPass>(); |
| 1189 | FunctionPass::getAnalysisUsage(AU); |
| 1190 | } |
| 1191 | |
| 1192 | void createOpSelectMerge(IRBuilder<> *Builder, BlockAddress *MergeAddress) { |
| 1193 | Instruction *BBTerminatorInst = Builder->GetInsertBlock()->getTerminator(); |
| 1194 | |
| 1195 | MDNode *MDNode = BBTerminatorInst->getMetadata(Kind: "hlsl.controlflow.hint" ); |
| 1196 | |
| 1197 | ConstantInt *BranchHint = ConstantInt::get(Ty: Builder->getInt32Ty(), V: 0); |
| 1198 | |
| 1199 | if (MDNode) { |
| 1200 | assert(MDNode->getNumOperands() == 2 && |
| 1201 | "invalid metadata hlsl.controlflow.hint" ); |
| 1202 | BranchHint = mdconst::extract<ConstantInt>(MD: MDNode->getOperand(I: 1)); |
| 1203 | } |
| 1204 | |
| 1205 | SmallVector<Value *, 2> Args = {MergeAddress, BranchHint}; |
| 1206 | |
| 1207 | Builder->CreateIntrinsic(ID: Intrinsic::spv_selection_merge, |
| 1208 | Types: {MergeAddress->getType()}, Args); |
| 1209 | } |
| 1210 | }; |
| 1211 | } // anonymous namespace |
| 1212 | |
| 1213 | char SPIRVStructurizer::ID = 0; |
| 1214 | |
| 1215 | INITIALIZE_PASS_BEGIN(SPIRVStructurizer, "spirv-structurizer" , |
| 1216 | "structurize SPIRV" , false, false) |
| 1217 | INITIALIZE_PASS_DEPENDENCY(LoopSimplify) |
| 1218 | INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass) |
| 1219 | INITIALIZE_PASS_DEPENDENCY(LoopInfoWrapperPass) |
| 1220 | INITIALIZE_PASS_DEPENDENCY(SPIRVConvergenceRegionAnalysisWrapperPass) |
| 1221 | |
| 1222 | INITIALIZE_PASS_END(SPIRVStructurizer, "spirv-structurizer" , |
| 1223 | "structurize SPIRV" , false, false) |
| 1224 | |
| 1225 | FunctionPass *llvm::createSPIRVStructurizerPass() { |
| 1226 | return new SPIRVStructurizer(); |
| 1227 | } |
| 1228 | |
| 1229 | PreservedAnalyses SPIRVStructurizerWrapper::run(Function &F, |
| 1230 | FunctionAnalysisManager &AF) { |
| 1231 | |
| 1232 | auto FPM = legacy::FunctionPassManager(F.getParent()); |
| 1233 | FPM.add(P: createSPIRVStructurizerPass()); |
| 1234 | |
| 1235 | if (!FPM.run(F)) |
| 1236 | return PreservedAnalyses::all(); |
| 1237 | PreservedAnalyses PA; |
| 1238 | PA.preserveSet<CFGAnalyses>(); |
| 1239 | return PA; |
| 1240 | } |
| 1241 | |