[LV][VPlan] Build plain CFG with simple VPInstructions for outer loops.

Patch brson#3 from VPlan Outer Loop Vectorization Patch Series brson#1
(RFC: http://lists.llvm.org/pipermail/llvm-dev/2017-December/119523.html).

Expected to be NFC for the current inner loop vectorization path. It
introduces the basic algorithm to build the VPlan plain CFG (single-level
CFG, no hierarchical CFG (H-CFG), yet) in the VPlan-native vectorization
path using VPInstructions. It includes:
  - VPlanHCFGBuilder: Main class to build the VPlan H-CFG (plain CFG without nested regions, for now).
  - VPlanVerifier: Main class with utilities to check the consistency of a H-CFG.
  - VPlanBlockUtils: Main class with utilities to manipulate VPBlockBases in VPlan.

Reviewers: rengolin, fhahn, mkuper, mssimpso, a.elovikov, hfinkel, aprantl.

Differential Revision: https://reviews.llvm.org/D44338



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@332654 91177308-0d34-0410-b5e6-96231b3b80d8

Author: dcaballe

lib/Transforms/Vectorize/CMakeLists.txt

@@ -5,6 +5,8 @@ add_llvm_library(LLVMVectorize
   SLPVectorizer.cpp
   Vectorize.cpp
   VPlan.cpp
+  VPlanHCFGBuilder.cpp
+  VPlanVerifier.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${LLVM_MAIN_INCLUDE_DIR}/llvm/Transforms

lib/Transforms/Vectorize/LoopVectorizationPlanner.h

@@ -39,23 +39,94 @@ class VPBuilder {
   VPBasicBlock::iterator InsertPt = VPBasicBlock::iterator();
 
   VPInstruction *createInstruction(unsigned Opcode,
-                                   std::initializer_list<VPValue *> Operands) {
+                                   ArrayRef<VPValue *> Operands) {
     VPInstruction *Instr = new VPInstruction(Opcode, Operands);
-    BB->insert(Instr, InsertPt);
+    if (BB)
+      BB->insert(Instr, InsertPt);
     return Instr;
   }
 
+  VPInstruction *createInstruction(unsigned Opcode,
+                                   std::initializer_list<VPValue *> Operands) {
+    return createInstruction(Opcode, ArrayRef<VPValue *>(Operands));
+  }
+
 public:
   VPBuilder() {}
 
-  /// This specifies that created VPInstructions should be appended to
-  /// the end of the specified block.
+  /// Clear the insertion point: created instructions will not be inserted into
+  /// a block.
+  void clearInsertionPoint() {
+    BB = nullptr;
+    InsertPt = VPBasicBlock::iterator();
+  }
+
+  VPBasicBlock *getInsertBlock() const { return BB; }
+  VPBasicBlock::iterator getInsertPoint() const { return InsertPt; }
+
+  /// InsertPoint - A saved insertion point.
+  class VPInsertPoint {
+    VPBasicBlock *Block = nullptr;
+    VPBasicBlock::iterator Point;
+
+  public:
+    /// Creates a new insertion point which doesn't point to anything.
+    VPInsertPoint() = default;
+
+    /// Creates a new insertion point at the given location.
+    VPInsertPoint(VPBasicBlock *InsertBlock, VPBasicBlock::iterator InsertPoint)
+        : Block(InsertBlock), Point(InsertPoint) {}
+
+    /// Returns true if this insert point is set.
+    bool isSet() const { return Block != nullptr; }
+
+    VPBasicBlock *getBlock() const { return Block; }
+    VPBasicBlock::iterator getPoint() const { return Point; }
+  };
+
+  /// Sets the current insert point to a previously-saved location.
+  void restoreIP(VPInsertPoint IP) {
+    if (IP.isSet())
+      setInsertPoint(IP.getBlock(), IP.getPoint());
+    else
+      clearInsertionPoint();
+  }
+
+  /// This specifies that created VPInstructions should be appended to the end
+  /// of the specified block.
   void setInsertPoint(VPBasicBlock *TheBB) {
     assert(TheBB && "Attempting to set a null insert point");
     BB = TheBB;
     InsertPt = BB->end();
   }
 
+  /// This specifies that created instructions should be inserted at the
+  /// specified point.
+  void setInsertPoint(VPBasicBlock *TheBB, VPBasicBlock::iterator IP) {
+    BB = TheBB;
+    InsertPt = IP;
+  }
+
+  /// Insert and return the specified instruction.
+  VPInstruction *insert(VPInstruction *I) const {
+    BB->insert(I, InsertPt);
+    return I;
+  }
+
+  /// Create an N-ary operation with \p Opcode, \p Operands and set \p Inst as
+  /// its underlying Instruction.
+  VPValue *createNaryOp(unsigned Opcode, ArrayRef<VPValue *> Operands,
+                        Instruction *Inst = nullptr) {
+    VPInstruction *NewVPInst = createInstruction(Opcode, Operands);
+    NewVPInst->setUnderlyingValue(Inst);
+    return NewVPInst;
+  }
+  VPValue *createNaryOp(unsigned Opcode,
+                        std::initializer_list<VPValue *> Operands,
+                        Instruction *Inst = nullptr) {
+    return createNaryOp(Opcode, ArrayRef<VPValue *>(Operands), Inst);
+  }
+
   VPValue *createNot(VPValue *Operand) {
     return createInstruction(VPInstruction::Not, {Operand});
   }
@@ -67,8 +138,28 @@ class VPBuilder {
   VPValue *createOr(VPValue *LHS, VPValue *RHS) {
     return createInstruction(Instruction::BinaryOps::Or, {LHS, RHS});
   }
-};
 
+  //===--------------------------------------------------------------------===//
+  // RAII helpers.
+  //===--------------------------------------------------------------------===//
+
+  /// RAII object that stores the current insertion point and restores it when
+  /// the object is destroyed.
+  class InsertPointGuard {
+    VPBuilder &Builder;
+    VPBasicBlock *Block;
+    VPBasicBlock::iterator Point;
+
+  public:
+    InsertPointGuard(VPBuilder &B)
+        : Builder(B), Block(B.getInsertBlock()), Point(B.getInsertPoint()) {}
+
+    InsertPointGuard(const InsertPointGuard &) = delete;
+    InsertPointGuard &operator=(const InsertPointGuard &) = delete;
+
+    ~InsertPointGuard() { Builder.restoreIP(VPInsertPoint(Block, Point)); }
+  };
+};
 
 /// TODO: The following VectorizationFactor was pulled out of
 /// LoopVectorizationCostModel class. LV also deals with

lib/Transforms/Vectorize/LoopVectorize.cpp

@@ -56,6 +56,7 @@
 
 #include "llvm/Transforms/Vectorize/LoopVectorize.h"
 #include "LoopVectorizationPlanner.h"
+#include "VPlanHCFGBuilder.h"
 #include "llvm/ADT/APInt.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
@@ -244,6 +245,17 @@ static cl::opt<bool> EnableVPlanNativePath(
     cl::desc("Enable VPlan-native vectorization path with "
              "support for outer loop vectorization."));
 
+// This flag enables the stress testing of the VPlan H-CFG construction in the
+// VPlan-native vectorization path. It must be used in conjuction with
+// -enable-vplan-native-path. -vplan-verify-hcfg can also be used to enable the
+// verification of the H-CFGs built.
+static cl::opt<bool> VPlanBuildStressTest(
+    "vplan-build-stress-test", cl::init(false), cl::Hidden,
+    cl::desc(
+        "Build VPlan for every supported loop nest in the function and bail "
+        "out right after the build (stress test the VPlan H-CFG construction "
+        "in the VPlan-native vectorization path)."));
+
 /// A helper function for converting Scalar types to vector types.
 /// If the incoming type is void, we return void. If the VF is 1, we return
 /// the scalar type.
@@ -1653,8 +1665,11 @@ static void collectSupportedLoops(Loop &L, LoopInfo *LI,
                                   OptimizationRemarkEmitter *ORE,
                                   SmallVectorImpl<Loop *> &V) {
   // Collect inner loops and outer loops without irreducible control flow. For
-  // now, only collect outer loops that have explicit vectorization hints.
-  if (L.empty() || (EnableVPlanNativePath && isExplicitVecOuterLoop(&L, ORE))) {
+  // now, only collect outer loops that have explicit vectorization hints. If we
+  // are stress testing the VPlan H-CFG construction, we collect the outermost
+  // loop of every loop nest.
+  if (L.empty() || VPlanBuildStressTest ||
+      (EnableVPlanNativePath && isExplicitVecOuterLoop(&L, ORE))) {
     LoopBlocksRPO RPOT(&L);
     RPOT.perform(LI);
     if (!containsIrreducibleCFG<const BasicBlock *>(RPOT, *LI)) {
@@ -6254,20 +6269,30 @@ void LoopVectorizationCostModel::collectValuesToIgnore() {
 VectorizationFactor
 LoopVectorizationPlanner::planInVPlanNativePath(bool OptForSize,
                                                 unsigned UserVF) {
-  // Width 1 means no vectorize, cost 0 means uncomputed cost.
+  // Width 1 means no vectorization, cost 0 means uncomputed cost.
   const VectorizationFactor NoVectorization = {1U, 0U};
 
   // Outer loop handling: They may require CFG and instruction level
   // transformations before even evaluating whether vectorization is profitable.
   // Since we cannot modify the incoming IR, we need to build VPlan upfront in
   // the vectorization pipeline.
   if (!OrigLoop->empty()) {
+    // TODO: If UserVF is not provided, we set UserVF to 4 for stress testing.
+    // This won't be necessary when UserVF is not required in the VPlan-native
+    // path.
+    if (VPlanBuildStressTest && !UserVF)
+      UserVF = 4;
+
     assert(EnableVPlanNativePath && "VPlan-native path is not enabled.");
     assert(UserVF && "Expected UserVF for outer loop vectorization.");
     assert(isPowerOf2_32(UserVF) && "VF needs to be a power of two");
     LLVM_DEBUG(dbgs() << "LV: Using user VF " << UserVF << ".\n");
     buildVPlans(UserVF, UserVF);
 
+    // For VPlan build stress testing, we bail out after VPlan construction.
+    if (VPlanBuildStressTest)
+      return NoVectorization;
+
     return {UserVF, 0};
   }
 
@@ -6280,7 +6305,7 @@ LoopVectorizationPlanner::planInVPlanNativePath(bool OptForSize,
 VectorizationFactor
 LoopVectorizationPlanner::plan(bool OptForSize, unsigned UserVF) {
   assert(OrigLoop->empty() && "Inner loop expected.");
-  // Width 1 means no vectorize, cost 0 means uncomputed cost.
+  // Width 1 means no vectorization, cost 0 means uncomputed cost.
   const VectorizationFactor NoVectorization = {1U, 0U};
   Optional<unsigned> MaybeMaxVF = CM.computeMaxVF(OptForSize);
   if (!MaybeMaxVF.hasValue()) // Cases considered too costly to vectorize.
@@ -6806,9 +6831,11 @@ VPBasicBlock *LoopVectorizationPlanner::handleReplication(
          "VPBB has successors when handling predicated replication.");
   // Record predicated instructions for above packing optimizations.
   PredInst2Recipe[I] = Recipe;
-  VPBlockBase *Region =
-    VPBB->setOneSuccessor(createReplicateRegion(I, Recipe, Plan));
-  return cast<VPBasicBlock>(Region->setOneSuccessor(new VPBasicBlock()));
+  VPBlockBase *Region = createReplicateRegion(I, Recipe, Plan);
+  VPBlockUtils::insertBlockAfter(Region, VPBB);
+  auto *RegSucc = new VPBasicBlock();
+  VPBlockUtils::insertBlockAfter(RegSucc, Region);
+  return RegSucc;
 }
 
 VPRegionBlock *
@@ -6834,8 +6861,8 @@ LoopVectorizationPlanner::createReplicateRegion(Instruction *Instr,
 
   // Note: first set Entry as region entry and then connect successors starting
   // from it in order, to propagate the "parent" of each VPBasicBlock.
-  Entry->setTwoSuccessors(Pred, Exit);
-  Pred->setOneSuccessor(Exit);
+  VPBlockUtils::insertTwoBlocksAfter(Pred, Exit, Entry);
+  VPBlockUtils::connectBlocks(Pred, Exit);
 
   return Region;
 }
@@ -6852,6 +6879,11 @@ LoopVectorizationPlanner::buildVPlan(VFRange &Range,
 
     // Create new empty VPlan
     auto Plan = llvm::make_unique<VPlan>();
+
+    // Build hierarchical CFG
+    VPlanHCFGBuilder HCFGBuilder(OrigLoop, LI);
+    HCFGBuilder.buildHierarchicalCFG(*Plan.get());
+
     return Plan;
   }
 
@@ -6893,7 +6925,7 @@ LoopVectorizationPlanner::buildVPlan(VFRange &Range,
     // ingredients and fill a new VPBasicBlock.
     unsigned VPBBsForBB = 0;
     auto *FirstVPBBForBB = new VPBasicBlock(BB->getName());
-    VPBB->setOneSuccessor(FirstVPBBForBB);
+    VPBlockUtils::insertBlockAfter(FirstVPBBForBB, VPBB);
     VPBB = FirstVPBBForBB;
     Builder.setInsertPoint(VPBB);
 
@@ -6997,7 +7029,7 @@ LoopVectorizationPlanner::buildVPlan(VFRange &Range,
   VPBasicBlock *PreEntry = cast<VPBasicBlock>(Plan->getEntry());
   assert(PreEntry->empty() && "Expecting empty pre-entry block.");
   VPBlockBase *Entry = Plan->setEntry(PreEntry->getSingleSuccessor());
-  PreEntry->disconnectSuccessor(Entry);
+  VPBlockUtils::disconnectBlocks(PreEntry, Entry);
   delete PreEntry;
 
   std::string PlanName;