[VPlan] Add VPInstruction to VPRecipe transformation.

This patch introduces a VPInstructionToVPRecipe transformation, which
allows us to generate code for a VPInstruction based VPlan re-using the
existing infrastructure.

Reviewers: dcaballe, hsaito, mssimpso, hfinkel, rengolin, mkuper, javed.absar, sguggill

Reviewed By: dcaballe

Differential Revision: https://reviews.llvm.org/D46827

llvm-svn: 334969

6 files changed, 292 insertions, 0 deletions

diff --git a/llvm/lib/Transforms/Vectorize/CMakeLists.txt b/llvm/lib/Transforms/Vectorize/CMakeLists.txt
index 7c0a39d7ff5..27a4d241b32 100644
--- a/llvm/lib/Transforms/Vectorize/CMakeLists.txt
+++ b/llvm/lib/Transforms/Vectorize/CMakeLists.txt
@@ -6,6 +6,7 @@ add_llvm_library(LLVMVectorize
   Vectorize.cpp
   VPlan.cpp
   VPlanHCFGBuilder.cpp
+  VPlanHCFGTransforms.cpp
   VPlanVerifier.cpp
 
   ADDITIONAL_HEADER_DIRS
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 0eefa0cea42..86368a8a67f 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -568,6 +568,8 @@ public:
 /// executed, these instructions would always form a single-def expression as
 /// the VPInstruction is also a single def-use vertex.
 class VPInstruction : public VPUser, public VPRecipeBase {
+  friend class VPlanHCFGTransforms;
+
 public:
   /// VPlan opcodes, extending LLVM IR with idiomatics instructions.
   enum { Not = Instruction::OtherOpsEnd + 1 };
diff --git a/llvm/lib/Transforms/Vectorize/VPlanHCFGTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanHCFGTransforms.cpp
new file mode 100644
index 00000000000..e3cbab077e6
--- /dev/null
+++ b/llvm/lib/Transforms/Vectorize/VPlanHCFGTransforms.cpp
@@ -0,0 +1,73 @@
+//===-- VPlanHCFGTransforms.cpp - Utility VPlan to VPlan transforms -------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// This file implements a set of utility VPlan to VPlan transformations.
+///
+//===----------------------------------------------------------------------===//
+
+#include "VPlanHCFGTransforms.h"
+#include "llvm/ADT/PostOrderIterator.h"
+
+using namespace llvm;
+
+void VPlanHCFGTransforms::VPInstructionsToVPRecipes(
+    VPlanPtr &Plan,
+    LoopVectorizationLegality::InductionList *Inductions,
+    SmallPtrSetImpl<Instruction *> &DeadInstructions) {
+
+  VPRegionBlock *TopRegion = dyn_cast<VPRegionBlock>(Plan->getEntry());
+  ReversePostOrderTraversal<VPBlockBase *> RPOT(TopRegion->getEntry());
+  for (VPBlockBase *Base : RPOT) {
+    // Do not widen instructions in pre-header and exit blocks.
+    if (Base->getNumPredecessors() == 0 || Base->getNumSuccessors() == 0)
+      continue;
+
+    VPBasicBlock *VPBB = Base->getEntryBasicBlock();
+    VPRecipeBase *LastRecipe = nullptr;
+    // Introduce each ingredient into VPlan.
+    for (auto I = VPBB->begin(), E = VPBB->end(); I != E;) {
+      VPRecipeBase *Ingredient = &*I++;
+      // Can only handle VPInstructions.
+      VPInstruction *VPInst = cast<VPInstruction>(Ingredient);
+      Instruction *Inst = cast<Instruction>(VPInst->getUnderlyingValue());
+      if (DeadInstructions.count(Inst)) {
+        Ingredient->eraseFromParent();
+        continue;
+      }
+
+      VPRecipeBase *NewRecipe = nullptr;
+      // Create VPWidenMemoryInstructionRecipe for loads and stores.
+      if (isa<LoadInst>(Inst) || isa<StoreInst>(Inst))
+        NewRecipe = new VPWidenMemoryInstructionRecipe(*Inst, nullptr /*Mask*/);
+      else if (PHINode *Phi = dyn_cast<PHINode>(Inst)) {
+        InductionDescriptor II = Inductions->lookup(Phi);
+        if (II.getKind() == InductionDescriptor::IK_IntInduction ||
+            II.getKind() == InductionDescriptor::IK_FpInduction) {
+          NewRecipe = new VPWidenIntOrFpInductionRecipe(Phi);
+        } else
+          NewRecipe = new VPWidenPHIRecipe(Phi);
+      } else {
+        // If the last recipe is a VPWidenRecipe, add Inst to it instead of
+        // creating a new recipe.
+        if (VPWidenRecipe *WidenRecipe =
+                dyn_cast_or_null<VPWidenRecipe>(LastRecipe)) {
+          WidenRecipe->appendInstruction(Inst);
+          Ingredient->eraseFromParent();
+          continue;
+        }
+        NewRecipe = new VPWidenRecipe(Inst);
+      }
+
+      NewRecipe->insertBefore(Ingredient);
+      LastRecipe = NewRecipe;
+      Ingredient->eraseFromParent();
+    }
+  }
+}
diff --git a/llvm/lib/Transforms/Vectorize/VPlanHCFGTransforms.h b/llvm/lib/Transforms/Vectorize/VPlanHCFGTransforms.h
new file mode 100644
index 00000000000..ae549c6871b
--- /dev/null
+++ b/llvm/lib/Transforms/Vectorize/VPlanHCFGTransforms.h
@@ -0,0 +1,36 @@
+//===- VPlanHCFGTransforms.h - Utility VPlan to VPlan transforms ----------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// This file provides utility VPlan to VPlan transformations.
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_TRANSFORMS_VECTORIZE_VPLANHCFGTRANSFORMS_H
+#define LLVM_TRANSFORMS_VECTORIZE_VPLANHCFGTRANSFORMS_H
+
+#include "VPlan.h"
+#include "llvm/IR/Instruction.h"
+#include "llvm/Transforms/Vectorize/LoopVectorizationLegality.h"
+
+namespace llvm {
+
+class VPlanHCFGTransforms {
+
+public:
+  /// Replaces the VPInstructions in \p Plan with corresponding
+  /// widen recipes.
+  static void VPInstructionsToVPRecipes(
+      VPlanPtr &Plan,
+      LoopVectorizationLegality::InductionList *Inductions,
+      SmallPtrSetImpl<Instruction *> &DeadInstructions);
+};
+
+} // namespace llvm
+
+#endif // LLVM_TRANSFORMS_VECTORIZE_VPLANHCFGTRANSFORMS_H
diff --git a/llvm/unittests/Transforms/Vectorize/CMakeLists.txt b/llvm/unittests/Transforms/Vectorize/CMakeLists.txt
index 0bb63562f42..0957a17025f 100644
--- a/llvm/unittests/Transforms/Vectorize/CMakeLists.txt
+++ b/llvm/unittests/Transforms/Vectorize/CMakeLists.txt
@@ -1,7 +1,9 @@
 set(LLVM_LINK_COMPONENTS
   Vectorize
+  AsmParser
   )
 
 add_llvm_unittest(VectorizeTests
   VPlanTest.cpp
+  VPlanHCFGTest.cpp
   )
diff --git a/llvm/unittests/Transforms/Vectorize/VPlanHCFGTest.cpp b/llvm/unittests/Transforms/Vectorize/VPlanHCFGTest.cpp
new file mode 100644
index 00000000000..282a995da45
--- /dev/null
+++ b/llvm/unittests/Transforms/Vectorize/VPlanHCFGTest.cpp
@@ -0,0 +1,178 @@
+//===- llvm/unittest/Transforms/Vectorize/VPlanHCFGTest.cpp ---------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../lib/Transforms/Vectorize/VPlan.h"
+#include "../lib/Transforms/Vectorize/VPlanHCFGBuilder.h"
+#include "../lib/Transforms/Vectorize/VPlanHCFGTransforms.h"
+#include "llvm/AsmParser/Parser.h"
+#include "llvm/IR/Dominators.h"
+#include "gtest/gtest.h"
+
+namespace llvm {
+namespace {
+
+class VPlanHCFGTest : public testing::Test {
+protected:
+  std::unique_ptr<DominatorTree> DT;
+  std::unique_ptr<LoopInfo> LI;
+
+  VPlanHCFGTest() {}
+
+  VPlanPtr doBuildPlan(BasicBlock *LoopHeader) {
+    DT.reset(new DominatorTree(*LoopHeader->getParent()));
+    LI.reset(new LoopInfo(*DT));
+
+    auto Plan = llvm::make_unique<VPlan>();
+    VPlanHCFGBuilder HCFGBuilder(LI->getLoopFor(LoopHeader), LI.get());
+    HCFGBuilder.buildHierarchicalCFG(*Plan.get());
+    return Plan;
+  }
+};
+
+TEST_F(VPlanHCFGTest, testBuildHCFGInnerLoop) {
+  LLVMContext Ctx;
+  const char *ModuleString =
+      "define void @f(i32* %A, i64 %N) {\n"
+      "entry:\n"
+      "  br label %for.body\n"
+      "for.body:\n"
+      "  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]\n"
+      "  %arr.idx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv\n"
+      "  %l1 = load i32, i32* %arr.idx, align 4\n"
+      "  %res = add i32 %l1, 10\n"
+      "  store i32 %res, i32* %arr.idx, align 4\n"
+      "  %indvars.iv.next = add i64 %indvars.iv, 1\n"
+      "  %exitcond = icmp ne i64 %indvars.iv.next, %N\n"
+      "  br i1 %exitcond, label %for.body, label %for.end\n"
+      "for.end:\n"
+      "  ret void\n"
+      "}\n";
+
+  SMDiagnostic Err;
+  std::unique_ptr<Module> M = parseAssemblyString(ModuleString, Err, Ctx);
+
+  Function *F = M->getFunction("f");
+  BasicBlock *LoopHeader = F->getEntryBlock().getSingleSuccessor();
+  auto Plan = doBuildPlan(LoopHeader);
+
+  VPBlockBase *Entry = Plan->getEntry()->getEntryBasicBlock();
+  EXPECT_NE(nullptr, Entry->getSingleSuccessor());
+  EXPECT_EQ(0u, Entry->getNumPredecessors());
+  EXPECT_EQ(1u, Entry->getNumSuccessors());
+
+  VPBasicBlock *VecBB = Entry->getSingleSuccessor()->getEntryBasicBlock();
+  EXPECT_EQ(7u, VecBB->size());
+  EXPECT_EQ(2u, VecBB->getNumPredecessors());
+  EXPECT_EQ(2u, VecBB->getNumSuccessors());
+
+  auto Iter = VecBB->begin();
+  VPInstruction *Phi = dyn_cast<VPInstruction>(&*Iter++);
+  EXPECT_EQ(Instruction::PHI, Phi->getOpcode());
+
+  VPInstruction *Idx = dyn_cast<VPInstruction>(&*Iter++);
+  EXPECT_EQ(Instruction::GetElementPtr, Idx->getOpcode());
+  EXPECT_EQ(2u, Idx->getNumOperands());
+  EXPECT_EQ(Phi, Idx->getOperand(1));
+
+  VPInstruction *Load = dyn_cast<VPInstruction>(&*Iter++);
+  EXPECT_EQ(Instruction::Load, Load->getOpcode());
+  EXPECT_EQ(1u, Load->getNumOperands());
+  EXPECT_EQ(Idx, Load->getOperand(0));
+
+  VPInstruction *Add = dyn_cast<VPInstruction>(&*Iter++);
+  EXPECT_EQ(Instruction::Add, Add->getOpcode());
+  EXPECT_EQ(2u, Add->getNumOperands());
+  EXPECT_EQ(Load, Add->getOperand(0));
+
+  VPInstruction *Store = dyn_cast<VPInstruction>(&*Iter++);
+  EXPECT_EQ(Instruction::Store, Store->getOpcode());
+  EXPECT_EQ(2u, Store->getNumOperands());
+  EXPECT_EQ(Add, Store->getOperand(0));
+  EXPECT_EQ(Idx, Store->getOperand(1));
+
+  VPInstruction *IndvarAdd = dyn_cast<VPInstruction>(&*Iter++);
+  EXPECT_EQ(Instruction::Add, IndvarAdd->getOpcode());
+  EXPECT_EQ(2u, IndvarAdd->getNumOperands());
+  EXPECT_EQ(Phi, IndvarAdd->getOperand(0));
+
+  VPInstruction *ICmp = dyn_cast<VPInstruction>(&*Iter++);
+  EXPECT_EQ(Instruction::ICmp, ICmp->getOpcode());
+  EXPECT_EQ(2u, ICmp->getNumOperands());
+  EXPECT_EQ(IndvarAdd, ICmp->getOperand(0));
+
+  LoopVectorizationLegality::InductionList Inductions;
+  SmallPtrSet<Instruction *, 1> DeadInstructions;
+  VPlanHCFGTransforms::VPInstructionsToVPRecipes(Plan, &Inductions,
+                                                 DeadInstructions);
+}
+
+TEST_F(VPlanHCFGTest, testVPInstructionToVPRecipesInner) {
+  LLVMContext Ctx;
+  const char *ModuleString =
+      "define void @f(i32* %A, i64 %N) {\n"
+      "entry:\n"
+      "  br label %for.body\n"
+      "for.body:\n"
+      "  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]\n"
+      "  %arr.idx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv\n"
+      "  %l1 = load i32, i32* %arr.idx, align 4\n"
+      "  %res = add i32 %l1, 10\n"
+      "  store i32 %res, i32* %arr.idx, align 4\n"
+      "  %indvars.iv.next = add i64 %indvars.iv, 1\n"
+      "  %exitcond = icmp ne i64 %indvars.iv.next, %N\n"
+      "  br i1 %exitcond, label %for.body, label %for.end\n"
+      "for.end:\n"
+      "  ret void\n"
+      "}\n";
+
+  SMDiagnostic Err;
+  std::unique_ptr<Module> M = parseAssemblyString(ModuleString, Err, Ctx);
+
+  Function *F = M->getFunction("f");
+  BasicBlock *LoopHeader = F->getEntryBlock().getSingleSuccessor();
+  auto Plan = doBuildPlan(LoopHeader);
+
+  LoopVectorizationLegality::InductionList Inductions;
+  SmallPtrSet<Instruction *, 1> DeadInstructions;
+  VPlanHCFGTransforms::VPInstructionsToVPRecipes(Plan, &Inductions,
+                                                 DeadInstructions);
+
+  VPBlockBase *Entry = Plan->getEntry()->getEntryBasicBlock();
+  EXPECT_NE(nullptr, Entry->getSingleSuccessor());
+  EXPECT_EQ(0u, Entry->getNumPredecessors());
+  EXPECT_EQ(1u, Entry->getNumSuccessors());
+
+  VPBasicBlock *VecBB = Entry->getSingleSuccessor()->getEntryBasicBlock();
+  EXPECT_EQ(6u, VecBB->size());
+  EXPECT_EQ(2u, VecBB->getNumPredecessors());
+  EXPECT_EQ(2u, VecBB->getNumSuccessors());
+
+  auto Iter = VecBB->begin();
+  auto *Phi = dyn_cast<VPWidenPHIRecipe>(&*Iter++);
+  EXPECT_NE(nullptr, Phi);
+
+  auto *Idx = dyn_cast<VPWidenRecipe>(&*Iter++);
+  EXPECT_NE(nullptr, Idx);
+
+  auto *Load = dyn_cast<VPWidenMemoryInstructionRecipe>(&*Iter++);
+  EXPECT_NE(nullptr, Load);
+
+  auto *Add = dyn_cast<VPWidenRecipe>(&*Iter++);
+  EXPECT_NE(nullptr, Add);
+
+  auto *Store = dyn_cast<VPWidenMemoryInstructionRecipe>(&*Iter++);
+  EXPECT_NE(nullptr, Store);
+
+  auto *LastWiden = dyn_cast<VPWidenRecipe>(&*Iter++);
+  EXPECT_NE(nullptr, LastWiden);
+  EXPECT_EQ(VecBB->end(), Iter);
+}
+
+} // namespace
+} // namespace llvm