Initial checkin of a simple loop unroller. This pass is extremely basic and

limited.  Even in it's extremely simple state (it can only *fully* unroll single
basic block loops that execute a constant number of times), it already helps improve
performance a LOT on some benchmarks, particularly with the native code generators.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13028 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Scalar/LoopUnroll.cpp b/lib/Transforms/Scalar/LoopUnroll.cpp
new file mode 100644 (file)
index 0000000..d337a43
--- /dev/null
+++ b/lib/Transforms/Scalar/LoopUnroll.cpp
@@ -0,0 +1,247 @@
+//===-- LoopUnroll.cpp - Loop unroller pass -------------------------------===//
+// 
+//                     The LLVM Compiler Infrastructure
+//
+// This file was developed by the LLVM research group and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+// 
+//===----------------------------------------------------------------------===//
+//
+// This pass implements a simple loop unroller.  It works best when loops have
+// been canonicalized by the -indvars pass, allowing it to determine the trip
+// counts of loops easily.
+//
+// This pass is currently extremely limited.  It only currently only unrolls
+// single basic block loops that execute a constant number of times.
+//
+//===----------------------------------------------------------------------===//
+
+#define DEBUG_TYPE "loop-unroll"
+#include "llvm/Transforms/Scalar.h"
+#include "llvm/Constants.h"
+#include "llvm/Function.h"
+#include "llvm/Instructions.h"
+#include "llvm/Analysis/LoopInfo.h"
+#include "llvm/Transforms/Utils/Cloning.h"
+#include "llvm/Transforms/Utils/Local.h"
+#include "Support/CommandLine.h"
+#include "Support/Debug.h"
+#include "Support/Statistic.h"
+#include <cstdio>
+using namespace llvm;
+
+namespace {
+  Statistic<> NumUnrolled("loop-unroll", "Number of loops completely unrolled");
+
+  cl::opt<unsigned>
+  UnrollThreshold("unroll-threshold", cl::init(100), cl::Hidden,
+                  cl::desc("The cut-off point for loop unrolling"));
+
+  class LoopUnroll : public FunctionPass {
+    LoopInfo *LI;  // The current loop information
+  public:
+    virtual bool runOnFunction(Function &F);
+    bool visitLoop(Loop *L);
+
+    /// This transformation requires natural loop information & requires that
+    /// loop preheaders be inserted into the CFG...
+    ///
+    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+      AU.setPreservesCFG();
+      AU.addRequiredID(LoopSimplifyID);
+      AU.addRequired<LoopInfo>();
+    }
+  };
+  RegisterOpt<LoopUnroll> X("loop-unroll", "Unroll loops");
+}
+
+FunctionPass *llvm::createLoopUnrollPass() { return new LoopUnroll(); }
+
+bool LoopUnroll::runOnFunction(Function &F) {
+  bool Changed = false;
+  LI = &getAnalysis<LoopInfo>();
+
+  for (LoopInfo::iterator I = LI->begin(), E = LI->end(); I != E; ++I)
+    Changed |= visitLoop(*I);
+
+  return Changed;
+}
+
+/// ApproximateLoopSize - Approximate the size of the loop after it has been
+/// unrolled.
+static unsigned ApproximateLoopSize(const Loop *L) {
+  unsigned Size = 0;
+  for (unsigned i = 0, e = L->getBlocks().size(); i != e; ++i) {
+    BasicBlock *BB = L->getBlocks()[i];
+    Instruction *Term = BB->getTerminator();
+    for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
+      if (isa<PHINode>(I) && BB == L->getHeader()) {
+        // Ignore PHI nodes in the header.
+      } else if (I->hasOneUse() && I->use_back() == Term) {
+        // Ignore instructions only used by the loop terminator.
+      } else {
+        ++Size;
+      }
+
+      // TODO: Ignore expressions derived from PHI and constants if inval of phi
+      // is a constant, or if operation is associative.  This will get induction
+      // variables.
+    }
+  }
+
+  return Size;
+}
+
+// RemapInstruction - Convert the instruction operands from referencing the 
+// current values into those specified by ValueMap.
+//
+static inline void RemapInstruction(Instruction *I, 
+                                    std::map<const Value *, Value*> &ValueMap) {
+  for (unsigned op = 0, E = I->getNumOperands(); op != E; ++op) {
+    Value *Op = I->getOperand(op);
+    std::map<const Value *, Value*>::iterator It = ValueMap.find(Op);
+    if (It != ValueMap.end()) Op = It->second;
+    I->setOperand(op, Op);
+  }
+}
+
+
+bool LoopUnroll::visitLoop(Loop *L) {
+  bool Changed = false;
+
+  // Recurse through all subloops before we process this loop.  Copy the loop
+  // list so that the child can update the loop tree if it needs to delete the
+  // loop.
+  std::vector<Loop*> SubLoops(L->begin(), L->end());
+  for (unsigned i = 0, e = SubLoops.size(); i != e; ++i)
+    Changed |= visitLoop(SubLoops[i]);
+
+  // We only handle single basic block loops right now.
+  if (L->getBlocks().size() != 1)
+    return Changed;
+
+  BasicBlock *BB = L->getHeader();
+  BranchInst *BI = dyn_cast<BranchInst>(BB->getTerminator());
+  if (BI == 0) return Changed;  // Must end in a conditional branch
+
+  ConstantInt *TripCountC = dyn_cast_or_null<ConstantInt>(L->getTripCount());
+  if (!TripCountC) return Changed;  // Must have constant trip count!
+
+  unsigned TripCount = TripCountC->getRawValue();
+  if (TripCount != TripCountC->getRawValue())
+    return Changed; // More than 2^32 iterations???
+
+  unsigned LoopSize = ApproximateLoopSize(L);
+  DEBUG(std::cerr << "Loop Unroll: F[" << BB->getParent()->getName()
+        << "] Loop %" << BB->getName() << " Loop Size = " << LoopSize
+        << " Trip Count = " << TripCount << " - ");
+  if (LoopSize*TripCount > UnrollThreshold) {
+    DEBUG(std::cerr << "TOO LARGE: " << LoopSize*TripCount << ">"
+                    << UnrollThreshold << "\n");
+    return Changed;
+  }
+  DEBUG(std::cerr << "UNROLLING!\n");
+  
+  assert(L->getExitBlocks().size() == 1 && "Must have exactly one exit block!");
+  BasicBlock *LoopExit = L->getExitBlocks()[0];
+
+  // Create a new basic block to temporarily hold all of the cloned code.
+  BasicBlock *NewBlock = new BasicBlock();
+
+  // For the first iteration of the loop, we should use the precloned values for
+  // PHI nodes.  Insert associations now.
+  std::map<const Value*, Value*> LastValueMap;
+  std::vector<PHINode*> OrigPHINode;
+  for (BasicBlock::iterator I = BB->begin();
+       PHINode *PN = dyn_cast<PHINode>(I); ++I) {
+    OrigPHINode.push_back(PN);
+    if (Instruction *I =dyn_cast<Instruction>(PN->getIncomingValueForBlock(BB)))
+      if (I->getParent() == BB)
+        LastValueMap[I] = I;
+  }
+
+  // Remove the exit branch from the loop
+  BB->getInstList().erase(BI);
+
+  assert(TripCount != 0 && "Trip count of 0 is impossible!");
+  for (unsigned It = 1; It != TripCount; ++It) {
+    char SuffixBuffer[100];
+    sprintf(SuffixBuffer, ".%d", It);
+    std::map<const Value*, Value*> ValueMap;
+    BasicBlock *New = CloneBasicBlock(BB, ValueMap, SuffixBuffer);
+
+    // Loop over all of the PHI nodes in the block, changing them to use the
+    // incoming values from the previous block.
+    for (unsigned i = 0, e = OrigPHINode.size(); i != e; ++i) {
+      PHINode *NewPHI = cast<PHINode>(ValueMap[OrigPHINode[i]]);
+      Value *InVal = NewPHI->getIncomingValueForBlock(BB);
+      if (Instruction *InValI = dyn_cast<Instruction>(InVal))
+        if (InValI->getParent() == BB)
+          InVal = LastValueMap[InValI];
+      ValueMap[OrigPHINode[i]] = InVal;
+      New->getInstList().erase(NewPHI);
+    }
+
+    for (BasicBlock::iterator I = New->begin(), E = New->end(); I != E; ++I)
+      RemapInstruction(I, ValueMap);
+
+    // Now that all of the instructions are remapped, splice them into the end
+    // of the NewBlock.
+    NewBlock->getInstList().splice(NewBlock->end(), New->getInstList());
+    delete New;
+
+    // LastValue map now contains values from this iteration.
+    std::swap(LastValueMap, ValueMap);
+  }
+
+  // If there was more than one iteration, replace any uses of values computed
+  // in the loop with values computed during last iteration of the loop.
+  if (TripCount != 1)
+    for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
+      std::vector<User*> Users(I->use_begin(), I->use_end());
+      for (unsigned i = 0, e = Users.size(); i != e; ++i) {
+        Instruction *UI = cast<Instruction>(Users[i]);
+        if (UI->getParent() != BB && UI->getParent() != NewBlock)
+          UI->replaceUsesOfWith(I, LastValueMap[I]);
+      }
+    }
+
+  // Now that we cloned the block as many times as we needed, stitch the new
+  // code into the original block and delete the temporary block.
+  BB->getInstList().splice(BB->end(), NewBlock->getInstList());
+  delete NewBlock;
+
+  // Now loop over the PHI nodes in the original block, setting them to their
+  // incoming values.
+  BasicBlock *Preheader = L->getLoopPreheader();
+  for (unsigned i = 0, e = OrigPHINode.size(); i != e; ++i) {
+    PHINode *PN = OrigPHINode[i];
+    PN->replaceAllUsesWith(PN->getIncomingValueForBlock(Preheader));
+    BB->getInstList().erase(PN);
+  }
+ 
+  // Finally, add an unconditional branch to the block to continue into the exit
+  // block.
+  new BranchInst(LoopExit, BB);
+
+  // At this point, the code is well formed.  We now do a quick sweep over the
+  // inserted code, doing constant propagation and dead code elimination as we
+  // go.
+  for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ) {
+    Instruction *Inst = I++;
+    
+    if (isInstructionTriviallyDead(Inst))
+      BB->getInstList().erase(Inst);
+    else if (Constant *C = ConstantFoldInstruction(Inst)) {
+      Inst->replaceAllUsesWith(C);
+      BB->getInstList().erase(Inst);
+    }
+  }
+
+  // FIXME: Should update analyses
+
+  // FIXME: Should fold into preheader and exit block
+
+  ++NumUnrolled;
+  return true;
+}