This target is no longer built. The ,v files now live in the reoptimizer.

llvm-svn: 27885

15 files changed, 0 insertions, 10154 deletions

diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/DependenceAnalyzer.cpp b/llvm/lib/Target/SparcV9/ModuloScheduling/DependenceAnalyzer.cpp
deleted file mode 100644
index 9c3422d54a6..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/DependenceAnalyzer.cpp
+++ /dev/null
@@ -1,305 +0,0 @@
-//===-- DependenceAnalyzer.cpp - DependenceAnalyzer  ------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-//
-//
-//
-//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "ModuloSched"
-
-#include "DependenceAnalyzer.h"
-#include "llvm/Type.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/ADT/Statistic.h"
-#include "llvm/Constants.h"
-#include <iostream>
-using namespace llvm;
-
-namespace llvm {
-
-  /// Create ModuloSchedulingPass
-  FunctionPass *createDependenceAnalyzer() {
-    return new DependenceAnalyzer();
-  }
-}
-
-Statistic<> NoDeps("depanalyzer-nodeps", "Number of dependences eliminated");
-Statistic<> NumDeps("depanalyzer-deps",
-                    "Number of dependences could not eliminate");
-Statistic<> AdvDeps("depanalyzer-advdeps",
-                    "Number of dependences using advanced techniques");
-
-bool DependenceAnalyzer::runOnFunction(Function &F) {
-  AA = &getAnalysis<AliasAnalysis>();
-  TD = &getAnalysis<TargetData>();
-  SE = &getAnalysis<ScalarEvolution>();
-
-  return  false;
-}
-
-static RegisterAnalysis<DependenceAnalyzer>X("depanalyzer",
-                                             "Dependence Analyzer");
-
-//  - Get inter and intra dependences between loads and stores
-//
-// Overview of Method:
-// Step 1: Use alias analysis to determine dependencies if values are loop
-//       invariant
-// Step 2: If pointers are not GEP, then there is a dependence.
-// Step 3: Compare GEP base pointers with AA. If no alias, no dependence.
-//         If may alias, then add a dependence. If must alias, then analyze
-//         further (Step 4)
-// Step 4: do advanced analysis
-void DependenceAnalyzer::AnalyzeDeps(Value *val, Value *val2, bool valLoad,
-                                     bool val2Load,
-                                     std::vector<Dependence> &deps,
-                                     BasicBlock *BB,
-                                     bool srcBeforeDest) {
-
-  bool loopInvariant = true;
-
-  //Check if both are instructions and prove not loop invariant if possible
-  if(Instruction *valInst = dyn_cast<Instruction>(val))
-    if(valInst->getParent() == BB)
-      loopInvariant = false;
-  if(Instruction *val2Inst = dyn_cast<Instruction>(val2))
-    if(val2Inst->getParent() == BB)
-      loopInvariant = false;
-
-
-  //If Loop invariant, let AA decide
-  if(loopInvariant) {
-    if(AA->alias(val, (unsigned)TD->getTypeSize(val->getType()),
-                 val2,(unsigned)TD->getTypeSize(val2->getType()))
-       != AliasAnalysis::NoAlias) {
-      createDep(deps, valLoad, val2Load, srcBeforeDest);
-    }
-    else
-      ++NoDeps;
-    return;
-  }
-
-  //Otherwise, continue with step 2
-
-  GetElementPtrInst *GP = dyn_cast<GetElementPtrInst>(val);
-  GetElementPtrInst *GP2 = dyn_cast<GetElementPtrInst>(val2);
-
-  //If both are not GP instructions, we can not do further analysis
-  if(!GP || !GP2) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-  }
-
-
-  //Otherwise, compare GEP bases (op #0) with Alias Analysis
-
-  Value *GPop = GP->getOperand(0);
-  Value *GP2op = GP2->getOperand(0);
-  int alias = AA->alias(GPop, (unsigned)TD->getTypeSize(GPop->getType()),
-                        GP2op,(unsigned)TD->getTypeSize(GP2op->getType()));
-
-
-  if(alias == AliasAnalysis::MustAlias) {
-    //Further dep analysis to do
-    advancedDepAnalysis(GP, GP2, valLoad, val2Load, deps, srcBeforeDest);
-    ++AdvDeps;
-  }
-  else if(alias == AliasAnalysis::MayAlias) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-  }
-  //Otherwise no dependence since there is no alias
-  else
-    ++NoDeps;
-}
-
-
-// advancedDepAnalysis - Do advanced data dependence tests
-void DependenceAnalyzer::advancedDepAnalysis(GetElementPtrInst *gp1,
-                                             GetElementPtrInst *gp2,
-                                             bool valLoad,
-                                             bool val2Load,
-                                             std::vector<Dependence> &deps,
-                                             bool srcBeforeDest) {
-
-  //Check if both GEPs are in a simple form: 3 ops, constant 0 as second arg
-  if(gp1->getNumOperands() != 3 || gp2->getNumOperands() != 3) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-  }
-
-  //Check second arg is constant 0
-  bool GPok = false;
-  if(Constant *c1 = dyn_cast<Constant>(gp1->getOperand(1)))
-    if(Constant *c2 = dyn_cast<Constant>(gp2->getOperand(1)))
-      if(c1->isNullValue() && c2->isNullValue())
-        GPok = true;
-
-  if(!GPok) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-
-  }
-
-  Value *Gep1Idx = gp1->getOperand(2);
-  Value *Gep2Idx = gp2->getOperand(2);
-
-  if(CastInst *c1 = dyn_cast<CastInst>(Gep1Idx))
-    Gep1Idx = c1->getOperand(0);
-  if(CastInst *c2 = dyn_cast<CastInst>(Gep2Idx))
-    Gep2Idx = c2->getOperand(0);
-
-  //Get SCEV for each index into the area
-  SCEVHandle SV1 = SE->getSCEV(Gep1Idx);
-  SCEVHandle SV2 = SE->getSCEV(Gep2Idx);
-
-  //Now handle special cases of dependence analysis
-  //SV1->print(std::cerr);
-  //std::cerr << "\n";
-  //SV2->print(std::cerr);
-  //std::cerr << "\n";
-
-  //Check if we have an SCEVAddExpr, cause we can only handle those
-  SCEVAddRecExpr *SVAdd1 = dyn_cast<SCEVAddRecExpr>(SV1);
-  SCEVAddRecExpr *SVAdd2 = dyn_cast<SCEVAddRecExpr>(SV2);
-
-  //Default to having a dependence since we can't analyze further
-  if(!SVAdd1 || !SVAdd2) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-  }
-
-  //Check if not Affine, we can't handle those
-  if(!SVAdd1->isAffine( ) || !SVAdd2->isAffine()) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-  }
-
-  //We know the SCEV is in the form A + B*x, check that B is the same for both
-  SCEVConstant *B1 = dyn_cast<SCEVConstant>(SVAdd1->getOperand(1));
-  SCEVConstant *B2 = dyn_cast<SCEVConstant>(SVAdd2->getOperand(1));
-
-  if(B1->getValue() != B2->getValue()) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-  }
-
-  if(B1->getValue()->getRawValue() != 1 || B2->getValue()->getRawValue() != 1) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-  }
-
-
-  SCEVConstant *A1 = dyn_cast<SCEVConstant>(SVAdd1->getOperand(0));
-  SCEVConstant *A2 = dyn_cast<SCEVConstant>(SVAdd2->getOperand(0));
-
-  //Come back and deal with nested SCEV!
-  if(!A1 || !A2) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-  }
-
-  //If equal, create dep as normal
-  if(A1->getValue() == A2->getValue()) {
-    createDep(deps, valLoad, val2Load, srcBeforeDest);
-    return;
-  }
-  //Eliminate a dep if this is a intra dep
-  else if(srcBeforeDest) {
-    ++NoDeps;
-    return;
-  }
-
-  //Find constant index difference
-  int diff = A1->getValue()->getRawValue() - A2->getValue()->getRawValue();
-  //std::cerr << diff << "\n";
-  if(diff > 5)
-    diff = 2;
-
-  if(diff > 0)
-    createDep(deps, valLoad, val2Load, srcBeforeDest, diff);
-
-  //assert(diff > 0 && "Expected diff to be greater then 0");
-}
-
-// Create dependences once its determined these two instructions
-// references the same memory
-void DependenceAnalyzer::createDep(std::vector<Dependence> &deps,
-                                   bool valLoad, bool val2Load,
-                                   bool srcBeforeDest, int diff) {
-
-  //If the source instruction occurs after the destination instruction
-  //(execution order), then this dependence is across iterations
-  if(!srcBeforeDest && (diff==0))
-    diff = 1;
-
-  //If load/store pair
-  if(valLoad && !val2Load) {
-    if(srcBeforeDest)
-      //Anti Dep
-      deps.push_back(Dependence(diff, Dependence::AntiDep));
-    else
-      deps.push_back(Dependence(diff, Dependence::TrueDep));
-
-    ++NumDeps;
-  }
-  //If store/load pair
-  else if(!valLoad && val2Load) {
-    if(srcBeforeDest)
-      //True Dep
-      deps.push_back(Dependence(diff, Dependence::TrueDep));
-    else
-      deps.push_back(Dependence(diff, Dependence::AntiDep));
-    ++NumDeps;
-  }
-  //If store/store pair
-  else if(!valLoad && !val2Load) {
-    //True Dep
-    deps.push_back(Dependence(diff, Dependence::OutputDep));
-    ++NumDeps;
-  }
-}
-
-
-
-//Get Dependence Info for a pair of Instructions
-DependenceResult DependenceAnalyzer::getDependenceInfo(Instruction *inst1,
-                                                       Instruction *inst2,
-                                                       bool srcBeforeDest) {
-  std::vector<Dependence> deps;
-
-  DEBUG(std::cerr << "Inst1: " << *inst1 << "\n");
-  DEBUG(std::cerr << "Inst2: " << *inst2 << "\n");
-
-  //No self deps
-  if(inst1 == inst2)
-    return DependenceResult(deps);
-
-  if(LoadInst *ldInst = dyn_cast<LoadInst>(inst1)) {
-
-    if(StoreInst *stInst = dyn_cast<StoreInst>(inst2))
-      AnalyzeDeps(ldInst->getOperand(0), stInst->getOperand(1),
-                  true, false, deps, ldInst->getParent(), srcBeforeDest);
-  }
-  else if(StoreInst *stInst = dyn_cast<StoreInst>(inst1)) {
-
-    if(LoadInst *ldInst = dyn_cast<LoadInst>(inst2))
-      AnalyzeDeps(stInst->getOperand(1), ldInst->getOperand(0), false, true,
-                  deps, ldInst->getParent(), srcBeforeDest);
-
-    else if(StoreInst *stInst2 = dyn_cast<StoreInst>(inst2))
-      AnalyzeDeps(stInst->getOperand(1), stInst2->getOperand(1), false, false,
-                  deps, stInst->getParent(), srcBeforeDest);
-  }
-  else
-    assert(0 && "Expected a load or a store\n");
-
-  DependenceResult dr = DependenceResult(deps);
-  return dr;
-}
-
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/DependenceAnalyzer.h b/llvm/lib/Target/SparcV9/ModuloScheduling/DependenceAnalyzer.h
deleted file mode 100644
index f9aac5c0101..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/DependenceAnalyzer.h
+++ /dev/null
@@ -1,92 +0,0 @@
-//===-- DependenceAnalyzer.h - Dependence Analyzer--------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_DEPENDENCEANALYZER_H
-#define LLVM_DEPENDENCEANALYZER_H
-
-#include "llvm/Instructions.h"
-#include "llvm/Function.h"
-#include "llvm/Pass.h"
-#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/ScalarEvolutionExpressions.h"
-#include "llvm/Target/TargetData.h"
-#include <vector>
-
-namespace llvm {
-
-
-  //class to represent a dependence
-  struct Dependence {
-
-    enum DataDepType {
-      TrueDep, AntiDep, OutputDep, NonDateDep,
-    };
-
-    Dependence(int diff, DataDepType dep) : iteDiff(diff), depType(dep) {}
-    unsigned getIteDiff() { return iteDiff; }
-    unsigned getDepType() { return depType; }
-
-    private:
-
-    unsigned iteDiff;
-    unsigned depType;
-  };
-
-
-  struct DependenceResult {
-    std::vector<Dependence> dependences;
-    DependenceResult(const std::vector<Dependence> &d) : dependences(d) {}
-  };
-
-
-  class DependenceAnalyzer : public FunctionPass {
-
-
-    AliasAnalysis *AA;
-    TargetData *TD;
-    ScalarEvolution *SE;
-
-    void advancedDepAnalysis(GetElementPtrInst *gp1, GetElementPtrInst *gp2,
-                             bool valLoad, bool val2Load,
-                             std::vector<Dependence> &deps, bool srcBeforeDest);
-
-    void AnalyzeDeps(Value *val, Value *val2, bool val1Load, bool val2Load,
-                     std::vector<Dependence> &deps, BasicBlock *BB,
-                     bool srcBeforeDest);
-
-    void createDep(std::vector<Dependence> &deps, bool valLoad, bool val2Load,
-                   bool srcBeforeDest, int diff = 0);
-
-  public:
-    DependenceAnalyzer() { AA = 0; TD = 0; SE = 0; }
-    virtual bool runOnFunction(Function &F);
-    virtual const char* getPassName() const { return "DependenceAnalyzer"; }
-
-    // getAnalysisUsage
-    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
-      AU.addRequired<AliasAnalysis>();
-      AU.addRequired<TargetData>();
-      AU.addRequired<ScalarEvolution>();
-      AU.setPreservesAll();
-    }
-
-    //get dependence info
-    DependenceResult getDependenceInfo(Instruction *inst1, Instruction *inst2,
-                                       bool srcBeforeDest);
-
-  };
-
-}
-
-
-
-#endif
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/MSSchedule.cpp b/llvm/lib/Target/SparcV9/ModuloScheduling/MSSchedule.cpp
deleted file mode 100644
index 52d53243f9e..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/MSSchedule.cpp
+++ /dev/null
@@ -1,309 +0,0 @@
-//===-- MSSchedule.cpp Schedule ---------------------------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-//
-//
-//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "ModuloSched"
-
-#include "MSSchedule.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/Target/TargetSchedInfo.h"
-#include "../SparcV9Internals.h"
-#include "llvm/CodeGen/MachineInstr.h"
-#include <iostream>
-using namespace llvm;
-
-//Check if all resources are free
-bool resourcesFree(MSchedGraphNode*, int,
-std::map<int, std::map<int, int> > &resourceNumPerCycle);
-
-//Returns a boolean indicating if the start cycle needs to be increased/decreased
-bool MSSchedule::insert(MSchedGraphNode *node, int cycle, int II) {
-
-  //First, check if the cycle has a spot free to start
-  if(schedule.find(cycle) != schedule.end()) {
-    //Check if we have a free issue slot at this cycle
-    if (schedule[cycle].size() < numIssue) {
-      //Now check if all the resources in their respective cycles are available
-      if(resourcesFree(node, cycle, II)) {
-        //Insert to preserve dependencies
-        addToSchedule(cycle,node);
-        DEBUG(std::cerr << "Found spot in map, and there is an issue slot\n");
-        return false;
-      }
-    }
-  }
-  //Not in the map yet so put it in
-  else {
-    if(resourcesFree(node,cycle,II)) {
-      std::vector<MSchedGraphNode*> nodes;
-      nodes.push_back(node);
-      schedule[cycle] = nodes;
-      DEBUG(std::cerr << "Nothing in map yet so taking an issue slot\n");
-      return false;
-    }
-  }
-
-  DEBUG(std::cerr << "All issue slots taken\n");
-  return true;
-
-}
-
-void MSSchedule::addToSchedule(int cycle, MSchedGraphNode *node) {
-  std::vector<MSchedGraphNode*> nodesAtCycle = schedule[cycle];
-
-  std::map<unsigned, MSchedGraphNode*> indexMap;
-  for(unsigned i=0; i < nodesAtCycle.size(); ++i) {
-    indexMap[nodesAtCycle[i]->getIndex()] = nodesAtCycle[i];
-  }
-
-  indexMap[node->getIndex()] = node;
-
-  std::vector<MSchedGraphNode*> nodes;
-  for(std::map<unsigned, MSchedGraphNode*>::iterator I = indexMap.begin(), E = indexMap.end(); I != E; ++I)
-    nodes.push_back(I->second);
-
-  schedule[cycle] =  nodes;
-}
-
-bool MSSchedule::resourceAvailable(int resourceNum, int cycle) {
-  bool isFree = true;
-
-  //Get Map for this cycle
-  if(resourceNumPerCycle.count(cycle)) {
-    if(resourceNumPerCycle[cycle].count(resourceNum)) {
-      int maxRes = CPUResource::getCPUResource(resourceNum)->maxNumUsers;
-      if(resourceNumPerCycle[cycle][resourceNum] >= maxRes)
-        isFree = false;
-    }
-  }
-
-  return isFree;
-}
-
-void MSSchedule::useResource(int resourceNum, int cycle) {
-
-  //Get Map for this cycle
-  if(resourceNumPerCycle.count(cycle)) {
-    if(resourceNumPerCycle[cycle].count(resourceNum)) {
-      resourceNumPerCycle[cycle][resourceNum]++;
-    }
-    else {
-      resourceNumPerCycle[cycle][resourceNum] = 1;
-    }
-  }
-  //If no map, create one!
-  else {
-    std::map<int, int> resourceUse;
-    resourceUse[resourceNum] = 1;
-    resourceNumPerCycle[cycle] = resourceUse;
-  }
-
-}
-
-bool MSSchedule::resourcesFree(MSchedGraphNode *node, int cycle, int II) {
-
-  //Get Resource usage for this instruction
-  const TargetSchedInfo *msi = node->getParent()->getTarget()->getSchedInfo();
-  int currentCycle = cycle;
-  bool success = true;
-
-  //Create vector of starting cycles
-  std::vector<int> cyclesMayConflict;
-  cyclesMayConflict.push_back(cycle);
-
-  if(resourceNumPerCycle.size() > 0) {
-    for(int i = cycle-II; i >= (resourceNumPerCycle.begin()->first); i-=II)
-      cyclesMayConflict.push_back(i);
-    for(int i = cycle+II; i <= resourceNumPerCycle.end()->first; i+=II)
-      cyclesMayConflict.push_back(i);
-  }
-
-  //Now check all cycles for conflicts
-  for(int index = 0; index < (int) cyclesMayConflict.size(); ++index) {
-    currentCycle = cyclesMayConflict[index];
-
-    //Get resource usage for this instruction
-    InstrRUsage rUsage = msi->getInstrRUsage(node->getInst()->getOpcode());
-    std::vector<std::vector<resourceId_t> > resources = rUsage.resourcesByCycle;
-
-    //Loop over resources in each cycle and increments their usage count
-    for(unsigned i=0; i < resources.size(); ++i) {
-      for(unsigned j=0; j < resources[i].size(); ++j) {
-
-        //Get Resource to check its availability
-        int resourceNum = resources[i][j];
-
-        DEBUG(std::cerr << "Attempting to schedule Resource Num: " << resourceNum << " in cycle: " << currentCycle << "\n");
-
-        success = resourceAvailable(resourceNum, currentCycle);
-
-        if(!success)
-          break;
-
-      }
-
-      if(!success)
-        break;
-
-      //Increase cycle
-      currentCycle++;
-    }
-
-    if(!success)
-      return false;
-  }
-
-  //Actually put resources into the map
-  if(success) {
-
-    int currentCycle = cycle;
-    //Get resource usage for this instruction
-    InstrRUsage rUsage = msi->getInstrRUsage(node->getInst()->getOpcode());
-    std::vector<std::vector<resourceId_t> > resources = rUsage.resourcesByCycle;
-
-    //Loop over resources in each cycle and increments their usage count
-    for(unsigned i=0; i < resources.size(); ++i) {
-      for(unsigned j=0; j < resources[i].size(); ++j) {
-        int resourceNum = resources[i][j];
-        useResource(resourceNum, currentCycle);
-      }
-      currentCycle++;
-    }
-  }
-
-
-  return true;
-
-}
-
-bool MSSchedule::constructKernel(int II, std::vector<MSchedGraphNode*> &branches, std::map<const MachineInstr*, unsigned> &indVar) {
-
-  //Our schedule is allowed to have negative numbers, so lets calculate this offset
-  int offset = schedule.begin()->first;
-  if(offset > 0)
-    offset = 0;
-
-  DEBUG(std::cerr << "Offset: " << offset << "\n");
-
-  //Using the schedule, fold up into kernel and check resource conflicts as we go
-  std::vector<std::pair<MSchedGraphNode*, int> > tempKernel;
-
-  int stageNum = ((schedule.rbegin()->first-offset)+1)/ II;
-  int maxSN = 0;
-
-  DEBUG(std::cerr << "Number of Stages: " << stageNum << "\n");
-
-  for(int index = offset; index < (II+offset); ++index) {
-    int count = 0;
-    for(int i = index; i <= (schedule.rbegin()->first); i+=II) {
-      if(schedule.count(i)) {
-        for(std::vector<MSchedGraphNode*>::iterator I = schedule[i].begin(),
-              E = schedule[i].end(); I != E; ++I) {
-          //Check if its a branch
-          assert(!(*I)->isBranch() && "Branch should not be schedule!");
-
-          tempKernel.push_back(std::make_pair(*I, count));
-          maxSN = std::max(maxSN, count);
-
-        }
-      }
-      ++count;
-    }
-  }
-
-
-  //Add in induction var code
-  for(std::vector<std::pair<MSchedGraphNode*, int> >::iterator I = tempKernel.begin(), IE = tempKernel.end();
-      I != IE; ++I) {
-    //Add indVar instructions before this one for the current iteration
-    if(I->second == 0) {
-      std::map<unsigned, MachineInstr*> tmpMap;
-
-      //Loop over induction variable instructions in the map that come before this instr
-      for(std::map<const MachineInstr*, unsigned>::iterator N = indVar.begin(), NE = indVar.end(); N != NE; ++N) {
-
-
-        if(N->second < I->first->getIndex())
-          tmpMap[N->second] = (MachineInstr*) N->first;
-      }
-
-      //Add to kernel, and delete from indVar
-      for(std::map<unsigned, MachineInstr*>::iterator N = tmpMap.begin(), NE = tmpMap.end(); N != NE; ++N) {
-        kernel.push_back(std::make_pair(N->second, 0));
-        indVar.erase(N->second);
-      }
-    }
-
-    kernel.push_back(std::make_pair((MachineInstr*) I->first->getInst(), I->second));
-
-  }
-
-  std::map<unsigned, MachineInstr*> tmpMap;
-
-  //Add remaining invar instructions
-  for(std::map<const MachineInstr*, unsigned>::iterator N = indVar.begin(), NE = indVar.end(); N != NE; ++N) {
-    tmpMap[N->second] = (MachineInstr*) N->first;
-  }
-
-  //Add to kernel, and delete from indVar
-  for(std::map<unsigned, MachineInstr*>::iterator N = tmpMap.begin(), NE = tmpMap.end(); N != NE; ++N) {
-    kernel.push_back(std::make_pair(N->second, 0));
-    indVar.erase(N->second);
-  }
-
-
-  maxStage = maxSN;
-
-
-  return true;
-}
-
-bool MSSchedule::defPreviousStage(Value *def, int stage) {
-
-  //Loop over kernel and determine if value is being defined in previous stage
-  for(std::vector<std::pair<MachineInstr*, int> >::iterator P = kernel.begin(), PE = kernel.end(); P != PE; ++P) {
-    MachineInstr* inst = P->first;
-
-    //Loop over Machine Operands
-    for(unsigned i=0; i < inst->getNumOperands(); ++i) {
-      //get machine operand
-     const MachineOperand &mOp = inst->getOperand(i);
-     if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef()) {
-       if(def == mOp.getVRegValue()) {
-         if(P->second >= stage)
-           return false;
-         else
-           return true;
-       }
-     }
-    }
-  }
-
-  assert(0 && "We should always have found the def in our kernel\n");
-  abort();
-}
-
-
-void MSSchedule::print(std::ostream &os) const {
-  os << "Schedule:\n";
-
-  for(schedule_const_iterator I =  schedule.begin(), E = schedule.end(); I != E; ++I) {
-    os << "Cycle: " << I->first << "\n";
-    for(std::vector<MSchedGraphNode*>::const_iterator node = I->second.begin(), nodeEnd = I->second.end(); node != nodeEnd; ++node)
-    os << **node << "\n";
-  }
-
-  os << "Kernel:\n";
-  for(std::vector<std::pair<MachineInstr*, int> >::const_iterator I = kernel.begin(),
-        E = kernel.end(); I != E; ++I)
-    os << "Node: " << *(I->first) << " Stage: " << I->second << "\n";
-}
-
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/MSSchedule.h b/llvm/lib/Target/SparcV9/ModuloScheduling/MSSchedule.h
deleted file mode 100644
index 34a37db8b65..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/MSSchedule.h
+++ /dev/null
@@ -1,72 +0,0 @@
-//===-- MSSchedule.h - Schedule ------- -------------------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-// The schedule generated by a scheduling algorithm
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_MSSCHEDULE_H
-#define LLVM_MSSCHEDULE_H
-
-#include "MSchedGraph.h"
-#include <vector>
-#include <set>
-
-namespace llvm {
-
-  class MSSchedule {
-    std::map<int, std::vector<MSchedGraphNode*> > schedule;
-    unsigned numIssue;
-
-    //Internal map to keep track of explicit resources
-    std::map<int, std::map<int, int> > resourceNumPerCycle;
-
-    //Check if all resources are free
-    bool resourcesFree(MSchedGraphNode*, int, int II);
-    bool resourceAvailable(int resourceNum, int cycle);
-    void useResource(int resourceNum, int cycle);
-
-    //Resulting kernel
-    std::vector<std::pair<MachineInstr*, int> > kernel;
-
-    //Max stage count
-    int maxStage;
-
-    //add at the right spot in the schedule
-    void addToSchedule(int, MSchedGraphNode*);
-
-  public:
-    MSSchedule(int num) : numIssue(num) {}
-    MSSchedule() : numIssue(4) {}
-    bool insert(MSchedGraphNode *node, int cycle, int II);
-    int getStartCycle(MSchedGraphNode *node);
-    void clear() { schedule.clear(); resourceNumPerCycle.clear(); kernel.clear(); }
-    std::vector<std::pair<MachineInstr*, int> >* getKernel() { return &kernel; }
-    bool constructKernel(int II, std::vector<MSchedGraphNode*> &branches, std::map<const MachineInstr*, unsigned> &indVar);
-    int getMaxStage() { return maxStage; }
-    bool defPreviousStage(Value *def, int stage);
-
-    //iterators
-    typedef std::map<int, std::vector<MSchedGraphNode*> >::iterator schedule_iterator;
-    typedef std::map<int, std::vector<MSchedGraphNode*> >::const_iterator schedule_const_iterator;
-    schedule_iterator begin() { return schedule.begin(); };
-    schedule_iterator end() { return schedule.end(); };
-    void print(std::ostream &os) const;
-
-    typedef std::vector<std::pair<MachineInstr*, int> >::iterator kernel_iterator;
-    typedef std::vector<std::pair<MachineInstr*, int> >::const_iterator kernel_const_iterator;
-    kernel_iterator kernel_begin() { return kernel.begin(); }
-    kernel_iterator kernel_end() { return kernel.end(); }
-
-  };
-
-}
-
-
-#endif
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/MSScheduleSB.cpp b/llvm/lib/Target/SparcV9/ModuloScheduling/MSScheduleSB.cpp
deleted file mode 100644
index 487fb336e06..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/MSScheduleSB.cpp
+++ /dev/null
@@ -1,325 +0,0 @@
-//===-- MSScheduleSB.cpp Schedule ---------------------------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-//
-//
-//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "ModuloSchedSB"
-
-#include "MSScheduleSB.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/Target/TargetSchedInfo.h"
-#include "../SparcV9Internals.h"
-#include "llvm/CodeGen/MachineInstr.h"
-#include <iostream>
-using namespace llvm;
-
-//Check if all resources are free
-bool resourcesFree(MSchedGraphSBNode*, int,
-std::map<int, std::map<int, int> > &resourceNumPerCycle);
-
-//Returns a boolean indicating if the start cycle needs to be increased/decreased
-bool MSScheduleSB::insert(MSchedGraphSBNode *node, int cycle, int II) {
-
-  //First, check if the cycle has a spot free to start
-  if(schedule.find(cycle) != schedule.end()) {
-    //Check if we have a free issue slot at this cycle
-    if (schedule[cycle].size() < numIssue) {
-      //Now check if all the resources in their respective cycles are available
-      if(resourcesFree(node, cycle, II)) {
-        //Insert to preserve dependencies
-        addToSchedule(cycle,node);
-        DEBUG(std::cerr << "Found spot in map, and there is an issue slot\n");
-        return false;
-      }
-    }
-  }
-  //Not in the map yet so put it in
-  else {
-    if(resourcesFree(node,cycle,II)) {
-      std::vector<MSchedGraphSBNode*> nodes;
-      nodes.push_back(node);
-      schedule[cycle] = nodes;
-      DEBUG(std::cerr << "Nothing in map yet so taking an issue slot\n");
-      return false;
-    }
-  }
-
-  DEBUG(std::cerr << "All issue slots taken\n");
-  return true;
-
-}
-
-void MSScheduleSB::addToSchedule(int cycle, MSchedGraphSBNode *node) {
-  std::vector<MSchedGraphSBNode*> nodesAtCycle = schedule[cycle];
-
-  std::map<unsigned, MSchedGraphSBNode*> indexMap;
-  for(unsigned i=0; i < nodesAtCycle.size(); ++i) {
-    indexMap[nodesAtCycle[i]->getIndex()] = nodesAtCycle[i];
-  }
-
-  indexMap[node->getIndex()] = node;
-
-  std::vector<MSchedGraphSBNode*> nodes;
-  for(std::map<unsigned, MSchedGraphSBNode*>::iterator I = indexMap.begin(), E = indexMap.end(); I != E; ++I)
-    nodes.push_back(I->second);
-
-  schedule[cycle] =  nodes;
-}
-
-bool MSScheduleSB::resourceAvailable(int resourceNum, int cycle) {
-  bool isFree = true;
-
-  //Get Map for this cycle
-  if(resourceNumPerCycle.count(cycle)) {
-    if(resourceNumPerCycle[cycle].count(resourceNum)) {
-      int maxRes = CPUResource::getCPUResource(resourceNum)->maxNumUsers;
-      if(resourceNumPerCycle[cycle][resourceNum] >= maxRes)
-        isFree = false;
-    }
-  }
-
-  return isFree;
-}
-
-void MSScheduleSB::useResource(int resourceNum, int cycle) {
-
-  //Get Map for this cycle
-  if(resourceNumPerCycle.count(cycle)) {
-    if(resourceNumPerCycle[cycle].count(resourceNum)) {
-      resourceNumPerCycle[cycle][resourceNum]++;
-    }
-    else {
-      resourceNumPerCycle[cycle][resourceNum] = 1;
-    }
-  }
-  //If no map, create one!
-  else {
-    std::map<int, int> resourceUse;
-    resourceUse[resourceNum] = 1;
-    resourceNumPerCycle[cycle] = resourceUse;
-  }
-
-}
-
-bool MSScheduleSB::resourcesFree(MSchedGraphSBNode *node, int cycle, int II) {
-
-  //Get Resource usage for this instruction
-  const TargetSchedInfo *msi = node->getParent()->getTarget()->getSchedInfo();
-  int currentCycle = cycle;
-  bool success = true;
-
-  //Create vector of starting cycles
-  std::vector<int> cyclesMayConflict;
-  cyclesMayConflict.push_back(cycle);
-
-  if(resourceNumPerCycle.size() > 0) {
-    for(int i = cycle-II; i >= (resourceNumPerCycle.begin()->first); i-=II)
-      cyclesMayConflict.push_back(i);
-    for(int i = cycle+II; i <= resourceNumPerCycle.end()->first; i+=II)
-      cyclesMayConflict.push_back(i);
-  }
-
-  //Now check all cycles for conflicts
-  for(int index = 0; index < (int) cyclesMayConflict.size(); ++index) {
-    currentCycle = cyclesMayConflict[index];
-
-    //Get resource usage for this instruction
-    InstrRUsage rUsage = msi->getInstrRUsage(node->getInst()->getOpcode());
-    std::vector<std::vector<resourceId_t> > resources = rUsage.resourcesByCycle;
-
-    //Loop over resources in each cycle and increments their usage count
-    for(unsigned i=0; i < resources.size(); ++i) {
-      for(unsigned j=0; j < resources[i].size(); ++j) {
-
-        //Get Resource to check its availability
-        int resourceNum = resources[i][j];
-
-        DEBUG(std::cerr << "Attempting to schedule Resource Num: " << resourceNum << " in cycle: " << currentCycle << "\n");
-
-        success = resourceAvailable(resourceNum, currentCycle);
-
-        if(!success)
-          break;
-
-      }
-
-      if(!success)
-        break;
-
-      //Increase cycle
-      currentCycle++;
-    }
-
-    if(!success)
-      return false;
-  }
-
-  //Actually put resources into the map
-  if(success) {
-
-    int currentCycle = cycle;
-    //Get resource usage for this instruction
-    InstrRUsage rUsage = msi->getInstrRUsage(node->getInst()->getOpcode());
-    std::vector<std::vector<resourceId_t> > resources = rUsage.resourcesByCycle;
-
-    //Loop over resources in each cycle and increments their usage count
-    for(unsigned i=0; i < resources.size(); ++i) {
-      for(unsigned j=0; j < resources[i].size(); ++j) {
-        int resourceNum = resources[i][j];
-        useResource(resourceNum, currentCycle);
-      }
-      currentCycle++;
-    }
-  }
-
-
-  return true;
-
-}
-
-bool MSScheduleSB::constructKernel(int II, std::vector<MSchedGraphSBNode*> &branches, std::map<const MachineInstr*, unsigned> &indVar) {
-
-  //Our schedule is allowed to have negative numbers, so lets calculate this offset
-  int offset = schedule.begin()->first;
-  if(offset > 0)
-    offset = 0;
-
-  DEBUG(std::cerr << "Offset: " << offset << "\n");
-
-  //Using the schedule, fold up into kernel and check resource conflicts as we go
-  std::vector<std::pair<MSchedGraphSBNode*, int> > tempKernel;
-
-  int stageNum = ((schedule.rbegin()->first-offset)+1)/ II;
-  int maxSN = 0;
-
-  DEBUG(std::cerr << "Number of Stages: " << stageNum << "\n");
-
-  for(int index = offset; index < (II+offset); ++index) {
-    int count = 0;
-    for(int i = index; i <= (schedule.rbegin()->first); i+=II) {
-      if(schedule.count(i)) {
-        for(std::vector<MSchedGraphSBNode*>::iterator I = schedule[i].begin(),
-              E = schedule[i].end(); I != E; ++I) {
-          //Check if its a branch
-          assert(!(*I)->isBranch() && "Branch should not be schedule!");
-
-          tempKernel.push_back(std::make_pair(*I, count));
-          maxSN = std::max(maxSN, count);
-
-        }
-      }
-      ++count;
-    }
-  }
-
-
-  //Add in induction var code
-  for(std::vector<std::pair<MSchedGraphSBNode*, int> >::iterator I = tempKernel.begin(), IE = tempKernel.end();
-      I != IE; ++I) {
-    //Add indVar instructions before this one for the current iteration
-    if(I->second == 0) {
-      std::map<unsigned, MachineInstr*> tmpMap;
-
-      //Loop over induction variable instructions in the map that come before this instr
-      for(std::map<const MachineInstr*, unsigned>::iterator N = indVar.begin(), NE = indVar.end(); N != NE; ++N) {
-
-
-        if(N->second < I->first->getIndex())
-          tmpMap[N->second] = (MachineInstr*) N->first;
-      }
-
-      //Add to kernel, and delete from indVar
-      for(std::map<unsigned, MachineInstr*>::iterator N = tmpMap.begin(), NE = tmpMap.end(); N != NE; ++N) {
-        kernel.push_back(std::make_pair(N->second, 0));
-        indVar.erase(N->second);
-      }
-    }
-
-    kernel.push_back(std::make_pair((MachineInstr*) I->first->getInst(), I->second));
-    if(I->first->isPredicate()) {
-      //assert(I->second == 0 && "Predicate node must be from current iteration\n");
-      std::vector<const MachineInstr*> otherInstrs = I->first->getOtherInstrs();
-      for(std::vector<const MachineInstr*>::iterator O = otherInstrs.begin(), OE = otherInstrs.end(); O != OE; ++O) {
-        kernel.push_back(std::make_pair((MachineInstr*) *O, I->second));
-      }
-    }
-
-  }
-
-  std::map<unsigned, MachineInstr*> tmpMap;
-
-  //Add remaining invar instructions
-  for(std::map<const MachineInstr*, unsigned>::iterator N = indVar.begin(), NE = indVar.end(); N != NE; ++N) {
-    tmpMap[N->second] = (MachineInstr*) N->first;
-  }
-
-  //Add to kernel, and delete from indVar
-  for(std::map<unsigned, MachineInstr*>::iterator N = tmpMap.begin(), NE = tmpMap.end(); N != NE; ++N) {
-    kernel.push_back(std::make_pair(N->second, 0));
-    indVar.erase(N->second);
-  }
-
-
-  maxStage = maxSN;
-
-
-  return true;
-}
-
-bool MSScheduleSB::defPreviousStage(Value *def, int stage) {
-
-  //Loop over kernel and determine if value is being defined in previous stage
-  for(std::vector<std::pair<MachineInstr*, int> >::iterator P = kernel.begin(), PE = kernel.end(); P != PE; ++P) {
-    MachineInstr* inst = P->first;
-
-    //Loop over Machine Operands
-    for(unsigned i=0; i < inst->getNumOperands(); ++i) {
-      //get machine operand
-     const MachineOperand &mOp = inst->getOperand(i);
-     if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef()) {
-       if(def == mOp.getVRegValue()) {
-         if(P->second >= stage)
-           return false;
-         else
-           return true;
-       }
-     }
-    }
-  }
-
-  assert(0 && "We should always have found the def in our kernel\n");
-  abort();
-}
-
-
-void MSScheduleSB::print(std::ostream &os) const {
-  os << "Schedule:\n";
-
-  for(schedule_const_iterator I =  schedule.begin(), E = schedule.end(); I != E; ++I) {
-    os << "Cycle: " << I->first << "\n";
-    for(std::vector<MSchedGraphSBNode*>::const_iterator node = I->second.begin(), nodeEnd = I->second.end(); node != nodeEnd; ++node)
-    os << **node << "\n";
-  }
-
-  os << "Kernel:\n";
-  for(std::vector<std::pair<MachineInstr*, int> >::const_iterator I = kernel.begin(),
-        E = kernel.end(); I != E; ++I)
-    os << "Node: " << *(I->first) << " Stage: " << I->second << "\n";
-}
-
-void MSScheduleSB::printSchedule(std::ostream &os) const {
-  os << "Schedule:\n";
-
-  for(schedule_const_iterator I =  schedule.begin(), E = schedule.end(); I != E; ++I) {
-    os << "Cycle: " << I->first << "\n";
-    for(std::vector<MSchedGraphSBNode*>::const_iterator node = I->second.begin(), nodeEnd = I->second.end(); node != nodeEnd; ++node)
-    os << **node << "\n";
-  }
-}
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/MSScheduleSB.h b/llvm/lib/Target/SparcV9/ModuloScheduling/MSScheduleSB.h
deleted file mode 100644
index 40bcb873e22..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/MSScheduleSB.h
+++ /dev/null
@@ -1,73 +0,0 @@
-//===-- MSScheduleSB.h - Schedule ------- -------------------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-// The schedule generated by a scheduling algorithm
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_MSSCHEDULESB_H
-#define LLVM_MSSCHEDULESB_H
-
-#include "MSchedGraphSB.h"
-#include <vector>
-#include <set>
-
-namespace llvm {
-
-  class MSScheduleSB {
-    std::map<int, std::vector<MSchedGraphSBNode*> > schedule;
-    unsigned numIssue;
-
-    //Internal map to keep track of explicit resources
-    std::map<int, std::map<int, int> > resourceNumPerCycle;
-
-    //Check if all resources are free
-    bool resourcesFree(MSchedGraphSBNode*, int, int II);
-    bool resourceAvailable(int resourceNum, int cycle);
-    void useResource(int resourceNum, int cycle);
-
-    //Resulting kernel
-    std::vector<std::pair<MachineInstr*, int> > kernel;
-
-    //Max stage count
-    int maxStage;
-
-    //add at the right spot in the schedule
-    void addToSchedule(int, MSchedGraphSBNode*);
-
-  public:
-    MSScheduleSB(int num) : numIssue(num) {}
-    MSScheduleSB() : numIssue(4) {}
-    bool insert(MSchedGraphSBNode *node, int cycle, int II);
-    int getStartCycle(MSchedGraphSBNode *node);
-    void clear() { schedule.clear(); resourceNumPerCycle.clear(); kernel.clear(); }
-    std::vector<std::pair<MachineInstr*, int> >* getKernel() { return &kernel; }
-    bool constructKernel(int II, std::vector<MSchedGraphSBNode*> &branches, std::map<const MachineInstr*, unsigned> &indVar);
-    int getMaxStage() { return maxStage; }
-    bool defPreviousStage(Value *def, int stage);
-
-    //iterators
-    typedef std::map<int, std::vector<MSchedGraphSBNode*> >::iterator schedule_iterator;
-    typedef std::map<int, std::vector<MSchedGraphSBNode*> >::const_iterator schedule_const_iterator;
-    schedule_iterator begin() { return schedule.begin(); };
-    schedule_iterator end() { return schedule.end(); };
-    void print(std::ostream &os) const;
-    void printSchedule(std::ostream &os) const;
-
-    typedef std::vector<std::pair<MachineInstr*, int> >::iterator kernel_iterator;
-    typedef std::vector<std::pair<MachineInstr*, int> >::const_iterator kernel_const_iterator;
-    kernel_iterator kernel_begin() { return kernel.begin(); }
-    kernel_iterator kernel_end() { return kernel.end(); }
-
-  };
-
-}
-
-
-#endif
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraph.cpp b/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraph.cpp
deleted file mode 100644
index d9b1f9b1202..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraph.cpp
+++ /dev/null
@@ -1,804 +0,0 @@
-//===-- MSchedGraph.cpp - Scheduling Graph ----------------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-// A graph class for dependencies. This graph only contains true, anti, and
-// output data dependencies for a given MachineBasicBlock. Dependencies
-// across iterations are also computed. Unless data dependence analysis
-// is provided, a conservative approach of adding dependencies between all
-// loads and stores is taken.
-//===----------------------------------------------------------------------===//
-
-#define DEBUG_TYPE "ModuloSched"
-#include "MSchedGraph.h"
-#include "../SparcV9RegisterInfo.h"
-#include "../MachineCodeForInstruction.h"
-#include "llvm/BasicBlock.h"
-#include "llvm/Constants.h"
-#include "llvm/Instructions.h"
-#include "llvm/Type.h"
-#include "llvm/CodeGen/MachineBasicBlock.h"
-#include "llvm/Target/TargetInstrInfo.h"
-#include "llvm/Support/Debug.h"
-#include <cstdlib>
-#include <algorithm>
-#include <set>
-#include <iostream>
-using namespace llvm;
-
-//MSchedGraphNode constructor
-MSchedGraphNode::MSchedGraphNode(const MachineInstr* inst,
-                                 MSchedGraph *graph, unsigned idx,
-                                 unsigned late, bool isBranch)
-  : Inst(inst), Parent(graph), index(idx), latency(late),
-    isBranchInstr(isBranch) {
-
-  //Add to the graph
-  graph->addNode(inst, this);
-}
-
-//MSchedGraphNode copy constructor
-MSchedGraphNode::MSchedGraphNode(const MSchedGraphNode &N)
-  : Predecessors(N.Predecessors), Successors(N.Successors) {
-
-  Inst = N.Inst;
-  Parent = N.Parent;
-  index = N.index;
-  latency = N.latency;
-  isBranchInstr = N.isBranchInstr;
-
-}
-
-//Print the node (instruction and latency)
-void MSchedGraphNode::print(std::ostream &os) const {
-  os << "MSchedGraphNode: Inst=" << *Inst << ", latency= " << latency << "\n";
-}
-
-
-//Get the edge from a predecessor to this node
-MSchedGraphEdge MSchedGraphNode::getInEdge(MSchedGraphNode *pred) {
-  //Loop over all the successors of our predecessor
-  //return the edge the corresponds to this in edge
-  for (MSchedGraphNode::succ_iterator I = pred->succ_begin(),
-         E = pred->succ_end(); I != E; ++I) {
-    if (*I == this)
-      return I.getEdge();
-  }
-  assert(0 && "Should have found edge between this node and its predecessor!");
-  abort();
-}
-
-//Get the iteration difference for the edge from this node to its successor
-unsigned MSchedGraphNode::getIteDiff(MSchedGraphNode *succ) {
-  for(std::vector<MSchedGraphEdge>::iterator I = Successors.begin(),
-        E = Successors.end();
-      I != E; ++I) {
-    if(I->getDest() == succ)
-      return I->getIteDiff();
-  }
-  return 0;
-}
-
-//Get the index into the vector of edges for the edge from pred to this node
-unsigned MSchedGraphNode::getInEdgeNum(MSchedGraphNode *pred) {
-  //Loop over all the successors of our predecessor
-  //return the edge the corresponds to this in edge
-  int count = 0;
-  for(MSchedGraphNode::succ_iterator I = pred->succ_begin(),
-        E = pred->succ_end();
-      I != E; ++I) {
-    if(*I == this)
-      return count;
-    count++;
-  }
-  assert(0 && "Should have found edge between this node and its predecessor!");
-  abort();
-}
-
-//Determine if succ is a successor of this node
-bool MSchedGraphNode::isSuccessor(MSchedGraphNode *succ) {
-  for(succ_iterator I = succ_begin(), E = succ_end(); I != E; ++I)
-    if(*I == succ)
-      return true;
-  return false;
-}
-
-//Dtermine if pred is a predecessor of this node
-bool MSchedGraphNode::isPredecessor(MSchedGraphNode *pred) {
-  if(std::find( Predecessors.begin(),  Predecessors.end(),
-                pred) !=   Predecessors.end())
-    return true;
-  else
-    return false;
-}
-
-//Add a node to the graph
-void MSchedGraph::addNode(const MachineInstr *MI,
-                          MSchedGraphNode *node) {
-
-  //Make sure node does not already exist
-  assert(GraphMap.find(MI) == GraphMap.end()
-         && "New MSchedGraphNode already exists for this instruction");
-
-  GraphMap[MI] = node;
-}
-
-//Delete a node to the graph
-void MSchedGraph::deleteNode(MSchedGraphNode *node) {
-
-  //Delete the edge to this node from all predecessors
-  while(node->pred_size() > 0) {
-    //DEBUG(std::cerr << "Delete edge from: " << **P << " to " << *node << "\n");
-    MSchedGraphNode *pred = *(node->pred_begin());
-    pred->deleteSuccessor(node);
-  }
-
-  //Remove this node from the graph
-  GraphMap.erase(node->getInst());
-
-}
-
-
-//Create a graph for a machine block. The ignoreInstrs map is so that
-//we ignore instructions associated to the index variable since this
-//is a special case in Modulo Scheduling.  We only want to deal with
-//the body of the loop.
-MSchedGraph::MSchedGraph(const MachineBasicBlock *bb,
-                         const TargetMachine &targ,
-                         std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                         DependenceAnalyzer &DA,
-                         std::map<MachineInstr*, Instruction*> &machineTollvm)
-  : Target(targ) {
-
-  //Make sure BB is not null,
-  assert(bb != NULL && "Basic Block is null");
-
-  BBs.push_back(bb);
-
-  //Create nodes and edges for this BB
-  buildNodesAndEdges(ignoreInstrs, DA, machineTollvm);
-
-  //Experimental!
-  //addBranchEdges();
-}
-
-//Create a graph for a machine block. The ignoreInstrs map is so that
-//we ignore instructions associated to the index variable since this
-//is a special case in Modulo Scheduling.  We only want to deal with
-//the body of the loop.
-MSchedGraph::MSchedGraph(std::vector<const MachineBasicBlock*> &bbs,
-                         const TargetMachine &targ,
-                         std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                         DependenceAnalyzer &DA,
-                         std::map<MachineInstr*, Instruction*> &machineTollvm)
-  : BBs(bbs), Target(targ) {
-
-  //Make sure there is at least one BB and it is not null,
-  assert(((bbs.size() >= 1) &&  bbs[1] != NULL) && "Basic Block is null");
-
-  //Create nodes and edges for this BB
-  buildNodesAndEdges(ignoreInstrs, DA, machineTollvm);
-
-  //Experimental!
-  //addBranchEdges();
-}
-
-
-//Copies the graph and keeps a map from old to new nodes
-MSchedGraph::MSchedGraph(const MSchedGraph &G,
-                         std::map<MSchedGraphNode*, MSchedGraphNode*> &newNodes)
-  : Target(G.Target) {
-
-  BBs = G.BBs;
-
-  std::map<MSchedGraphNode*, MSchedGraphNode*> oldToNew;
-  //Copy all nodes
-  for(MSchedGraph::const_iterator N = G.GraphMap.begin(),
-        NE = G.GraphMap.end(); N != NE; ++N) {
-
-    MSchedGraphNode *newNode = new MSchedGraphNode(*(N->second));
-    oldToNew[&*(N->second)] = newNode;
-    newNodes[newNode] = &*(N->second);
-    GraphMap[&*(N->first)] = newNode;
-  }
-
-  //Loop over nodes and update edges to point to new nodes
-  for(MSchedGraph::iterator N = GraphMap.begin(), NE = GraphMap.end();
-      N != NE; ++N) {
-
-    //Get the node we are dealing with
-    MSchedGraphNode *node = &*(N->second);
-
-    node->setParent(this);
-
-    //Loop over nodes successors and predecessors and update to the new nodes
-    for(unsigned i = 0; i < node->pred_size(); ++i) {
-      node->setPredecessor(i, oldToNew[node->getPredecessor(i)]);
-    }
-
-    for(unsigned i = 0; i < node->succ_size(); ++i) {
-      MSchedGraphEdge *edge = node->getSuccessor(i);
-      MSchedGraphNode *oldDest = edge->getDest();
-      edge->setDest(oldToNew[oldDest]);
-    }
-  }
-}
-
-//Deconstructor, deletes all nodes in the graph
-MSchedGraph::~MSchedGraph () {
-  for(MSchedGraph::iterator I = GraphMap.begin(), E = GraphMap.end();
-      I != E; ++I)
-    delete I->second;
-}
-
-//Print out graph
-void MSchedGraph::print(std::ostream &os) const {
-  for(MSchedGraph::const_iterator N = GraphMap.begin(), NE = GraphMap.end();
-      N != NE; ++N) {
-
-    //Get the node we are dealing with
-    MSchedGraphNode *node = &*(N->second);
-
-    os << "Node Start\n";
-    node->print(os);
-    os << "Successors:\n";
-    //print successors
-    for(unsigned i = 0; i < node->succ_size(); ++i) {
-      MSchedGraphEdge *edge = node->getSuccessor(i);
-      MSchedGraphNode *oldDest = edge->getDest();
-      oldDest->print(os);
-    }
-    os << "Node End\n";
-  }
-}
-
-//Calculate total delay
-int MSchedGraph::totalDelay() {
-  int sum = 0;
-
-  for(MSchedGraph::const_iterator N = GraphMap.begin(), NE = GraphMap.end();
-      N != NE; ++N) {
-
-    //Get the node we are dealing with
-    MSchedGraphNode *node = &*(N->second);
-    sum += node->getLatency();
-  }
-  return sum;
-}
-//Experimental code to add edges from the branch to all nodes dependent upon it.
-void hasPath(MSchedGraphNode *node, std::set<MSchedGraphNode*> &visited,
-           std::set<MSchedGraphNode*> &branches, MSchedGraphNode *startNode,
-           std::set<std::pair<MSchedGraphNode*,MSchedGraphNode*> > &newEdges ) {
-
-  visited.insert(node);
-  DEBUG(std::cerr << "Visiting: " << *node << "\n");
-  //Loop over successors
-  for(unsigned i = 0; i < node->succ_size(); ++i) {
-    MSchedGraphEdge *edge = node->getSuccessor(i);
-    MSchedGraphNode *dest = edge->getDest();
-    if(branches.count(dest))
-      newEdges.insert(std::make_pair(dest, startNode));
-
-    //only visit if we have not already
-    else if(!visited.count(dest)) {
-      if(edge->getIteDiff() == 0)
-        hasPath(dest, visited, branches, startNode, newEdges);}
-
-  }
-
-}
-
-//Experimental code to add edges from the branch to all nodes dependent upon it.
-void MSchedGraph::addBranchEdges() {
-  std::set<MSchedGraphNode*> branches;
-  std::set<MSchedGraphNode*> nodes;
-
-  for(MSchedGraph::iterator I = GraphMap.begin(), E = GraphMap.end();
-      I != E; ++I) {
-    if(I->second->isBranch())
-      if(I->second->hasPredecessors())
-        branches.insert(I->second);
-  }
-
-  //See if there is a path first instruction to the branches, if so, add an
-  //iteration dependence between that node and the branch
-  std::set<std::pair<MSchedGraphNode*, MSchedGraphNode*> > newEdges;
-  for(MSchedGraph::iterator I = GraphMap.begin(), E = GraphMap.end();
-      I != E; ++I) {
-    std::set<MSchedGraphNode*> visited;
-    hasPath((I->second), visited, branches, (I->second), newEdges);
-  }
-
-  //Spit out all edges we are going to add
-  unsigned min = GraphMap.size();
-  if(newEdges.size() == 1) {
-    ((newEdges.begin())->first)->addOutEdge(((newEdges.begin())->second),
-                           MSchedGraphEdge::BranchDep,
-                           MSchedGraphEdge::NonDataDep, 1);
-  }
-  else {
-
-    unsigned count = 0;
-    MSchedGraphNode *start;
-    MSchedGraphNode *end;
-    for(std::set<std::pair<MSchedGraphNode*, MSchedGraphNode*> >::iterator I = newEdges.begin(), E = newEdges.end(); I != E; ++I) {
-
-      DEBUG(std::cerr << "Branch Edge from: " << *(I->first) << " to " << *(I->second) << "\n");
-
-      //      if(I->second->getIndex() <= min) {
-        start = I->first;
-        end = I->second;
-        //min = I->second->getIndex();
-        //}
-        start->addOutEdge(end,
-                          MSchedGraphEdge::BranchDep,
-                          MSchedGraphEdge::NonDataDep, 1);
-    }
-  }
-}
-
-
-//Add edges between the nodes
-void MSchedGraph::buildNodesAndEdges(std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                                     DependenceAnalyzer &DA,
-                       std::map<MachineInstr*, Instruction*> &machineTollvm) {
-
-
-  //Get Machine target information for calculating latency
-  const TargetInstrInfo *MTI = Target.getInstrInfo();
-
-  std::vector<MSchedGraphNode*> memInstructions;
-  std::map<int, std::vector<OpIndexNodePair> > regNumtoNodeMap;
-  std::map<const Value*, std::vector<OpIndexNodePair> > valuetoNodeMap;
-
-  //Save PHI instructions to deal with later
-  std::vector<const MachineInstr*> phiInstrs;
-  unsigned index = 0;
-
-  for(std::vector<const MachineBasicBlock*>::iterator B = BBs.begin(),
-        BE = BBs.end(); B != BE; ++B) {
-
-    const MachineBasicBlock *BB = *B;
-
-    //Loop over instructions in MBB and add nodes and edges
-    for (MachineBasicBlock::const_iterator MI = BB->begin(), e = BB->end();
-         MI != e; ++MI) {
-
-      //Ignore indvar instructions
-      if(ignoreInstrs.count(MI)) {
-        ++index;
-        continue;
-      }
-
-      //Get each instruction of machine basic block, get the delay
-      //using the op code, create a new node for it, and add to the
-      //graph.
-
-      MachineOpCode opCode = MI->getOpcode();
-      int delay;
-
-#if 0  // FIXME: LOOK INTO THIS
-      //Check if subsequent instructions can be issued before
-      //the result is ready, if so use min delay.
-      if(MTI->hasResultInterlock(MIopCode))
-        delay = MTI->minLatency(MIopCode);
-      else
-#endif
-        //Get delay
-        delay = MTI->maxLatency(opCode);
-
-      //Create new node for this machine instruction and add to the graph.
-      //Create only if not a nop
-      if(MTI->isNop(opCode))
-        continue;
-
-      //Sparc BE does not use PHI opcode, so assert on this case
-      assert(opCode != TargetInstrInfo::PHI && "Did not expect PHI opcode");
-
-      bool isBranch = false;
-
-      //We want to flag the branch node to treat it special
-      if(MTI->isBranch(opCode))
-        isBranch = true;
-
-      //Node is created and added to the graph automatically
-      MSchedGraphNode *node =  new MSchedGraphNode(MI, this, index, delay,
-                                                   isBranch);
-
-      DEBUG(std::cerr << "Created Node: " << *node << "\n");
-
-      //Check OpCode to keep track of memory operations to add memory
-      //dependencies later.
-      if(MTI->isLoad(opCode) || MTI->isStore(opCode))
-        memInstructions.push_back(node);
-
-      //Loop over all operands, and put them into the register number to
-      //graph node map for determining dependencies
-      //If an operands is a use/def, we have an anti dependence to itself
-      for(unsigned i=0; i < MI->getNumOperands(); ++i) {
-        //Get Operand
-        const MachineOperand &mOp = MI->getOperand(i);
-
-        //Check if it has an allocated register
-        if(mOp.hasAllocatedReg()) {
-          int regNum = mOp.getReg();
-
-          if(regNum != SparcV9::g0) {
-            //Put into our map
-            regNumtoNodeMap[regNum].push_back(std::make_pair(i, node));
-          }
-          continue;
-        }
-
-
-        //Add virtual registers dependencies
-        //Check if any exist in the value map already and create dependencies
-        //between them.
-        if(mOp.getType() == MachineOperand::MO_VirtualRegister
-           ||  mOp.getType() == MachineOperand::MO_CCRegister) {
-
-          //Make sure virtual register value is not null
-          assert((mOp.getVRegValue() != NULL) && "Null value is defined");
-
-          //Check if this is a read operation in a phi node, if so DO NOT PROCESS
-          if(mOp.isUse() && (opCode == TargetInstrInfo::PHI)) {
-            DEBUG(std::cerr << "Read Operation in a PHI node\n");
-            continue;
-          }
-
-          if (const Value* srcI = mOp.getVRegValue()) {
-
-            //Find value in the map
-            std::map<const Value*, std::vector<OpIndexNodePair> >::iterator V
-              = valuetoNodeMap.find(srcI);
-
-            //If there is something in the map already, add edges from
-            //those instructions
-            //to this one we are processing
-            if(V != valuetoNodeMap.end()) {
-              addValueEdges(V->second, node, mOp.isUse(), mOp.isDef(), phiInstrs);
-
-              //Add to value map
-              V->second.push_back(std::make_pair(i,node));
-            }
-            //Otherwise put it in the map
-            else
-              //Put into value map
-              valuetoNodeMap[mOp.getVRegValue()].push_back(std::make_pair(i, node));
-          }
-        }
-      }
-      ++index;
-    }
-
-    //Loop over LLVM BB, examine phi instructions, and add them to our
-    //phiInstr list to process
-    const BasicBlock *llvm_bb = BB->getBasicBlock();
-    for(BasicBlock::const_iterator I = llvm_bb->begin(), E = llvm_bb->end();
-        I != E; ++I) {
-      if(const PHINode *PN = dyn_cast<PHINode>(I)) {
-        MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(PN);
-        for (unsigned j = 0; j < tempMvec.size(); j++) {
-          if(!ignoreInstrs.count(tempMvec[j])) {
-            DEBUG(std::cerr << "Inserting phi instr into map: " << *tempMvec[j] << "\n");
-            phiInstrs.push_back((MachineInstr*) tempMvec[j]);
-          }
-        }
-      }
-
-    }
-
-    addMemEdges(memInstructions, DA, machineTollvm);
-    addMachRegEdges(regNumtoNodeMap);
-
-    //Finally deal with PHI Nodes and Value*
-    for(std::vector<const MachineInstr*>::iterator I = phiInstrs.begin(),
-          E = phiInstrs.end(); I != E;  ++I) {
-
-      //Get Node for this instruction
-      std::map<const MachineInstr*, MSchedGraphNode*>::iterator X;
-      X = find(*I);
-
-      if(X == GraphMap.end())
-        continue;
-
-      MSchedGraphNode *node = X->second;
-
-      DEBUG(std::cerr << "Adding ite diff edges for node: " << *node << "\n");
-
-      //Loop over operands for this instruction and add value edges
-      for(unsigned i=0; i < (*I)->getNumOperands(); ++i) {
-        //Get Operand
-        const MachineOperand &mOp = (*I)->getOperand(i);
-        if((mOp.getType() == MachineOperand::MO_VirtualRegister
-            ||  mOp.getType() == MachineOperand::MO_CCRegister) && mOp.isUse()) {
-
-          //find the value in the map
-          if (const Value* srcI = mOp.getVRegValue()) {
-
-            //Find value in the map
-            std::map<const Value*, std::vector<OpIndexNodePair> >::iterator V
-              = valuetoNodeMap.find(srcI);
-
-            //If there is something in the map already, add edges from
-            //those instructions
-            //to this one we are processing
-            if(V != valuetoNodeMap.end()) {
-              addValueEdges(V->second, node, mOp.isUse(), mOp.isDef(),
-                            phiInstrs, 1);
-            }
-          }
-        }
-      }
-    }
-  }
-}
-//Add dependencies for Value*s
-void MSchedGraph::addValueEdges(std::vector<OpIndexNodePair> &NodesInMap,
-                                MSchedGraphNode *destNode, bool nodeIsUse,
-                                bool nodeIsDef, std::vector<const MachineInstr*> &phiInstrs, int diff) {
-
-  for(std::vector<OpIndexNodePair>::iterator I = NodesInMap.begin(),
-        E = NodesInMap.end(); I != E; ++I) {
-
-    //Get node in vectors machine operand that is the same value as node
-    MSchedGraphNode *srcNode = I->second;
-    MachineOperand mOp = srcNode->getInst()->getOperand(I->first);
-
-    if(diff > 0)
-      if(std::find(phiInstrs.begin(), phiInstrs.end(), srcNode->getInst()) == phiInstrs.end())
-        continue;
-
-    //Node is a Def, so add output dep.
-    if(nodeIsDef) {
-      if(mOp.isUse()) {
-        DEBUG(std::cerr << "Edge from " << *srcNode << " to " << *destNode << " (itediff=" << diff << ", type=anti)\n");
-        srcNode->addOutEdge(destNode, MSchedGraphEdge::ValueDep,
-                            MSchedGraphEdge::AntiDep, diff);
-      }
-      if(mOp.isDef()) {
-        DEBUG(std::cerr << "Edge from " << *srcNode << " to " << *destNode << " (itediff=" << diff << ", type=output)\n");
-        srcNode->addOutEdge(destNode, MSchedGraphEdge::ValueDep,
-                            MSchedGraphEdge::OutputDep, diff);
-      }
-    }
-    if(nodeIsUse) {
-      if(mOp.isDef()) {
-        DEBUG(std::cerr << "Edge from " << *srcNode << " to " << *destNode << " (itediff=" << diff << ", type=true)\n");
-        srcNode->addOutEdge(destNode, MSchedGraphEdge::ValueDep,
-                            MSchedGraphEdge::TrueDep, diff);
-      }
-    }
-  }
-}
-
-//Add dependencies for machine registers across iterations
-void MSchedGraph::addMachRegEdges(std::map<int, std::vector<OpIndexNodePair> >& regNumtoNodeMap) {
-  //Loop over all machine registers in the map, and add dependencies
-  //between the instructions that use it
-  typedef std::map<int, std::vector<OpIndexNodePair> > regNodeMap;
-  for(regNodeMap::iterator I = regNumtoNodeMap.begin();
-      I != regNumtoNodeMap.end(); ++I) {
-    //Get the register number
-    int regNum = (*I).first;
-
-    //Get Vector of nodes that use this register
-    std::vector<OpIndexNodePair> Nodes = (*I).second;
-
-    //Loop over nodes and determine the dependence between the other
-    //nodes in the vector
-    for(unsigned i =0; i < Nodes.size(); ++i) {
-
-      //Get src node operator index that uses this machine register
-      int srcOpIndex = Nodes[i].first;
-
-      //Get the actual src Node
-      MSchedGraphNode *srcNode = Nodes[i].second;
-
-      //Get Operand
-      const MachineOperand &srcMOp = srcNode->getInst()->getOperand(srcOpIndex);
-
-      bool srcIsUseandDef = srcMOp.isDef() && srcMOp.isUse();
-      bool srcIsUse = srcMOp.isUse() && !srcMOp.isDef();
-
-
-      //Look at all instructions after this in execution order
-      for(unsigned j=i+1; j < Nodes.size(); ++j) {
-
-        //Sink node is a write
-        if(Nodes[j].second->getInst()->getOperand(Nodes[j].first).isDef()) {
-                      //Src only uses the register (read)
-            if(srcIsUse)
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphEdge::MachineRegister,
-                                  MSchedGraphEdge::AntiDep);
-
-            else if(srcIsUseandDef) {
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphEdge::MachineRegister,
-                                  MSchedGraphEdge::AntiDep);
-
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphEdge::MachineRegister,
-                                  MSchedGraphEdge::OutputDep);
-            }
-            else
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphEdge::MachineRegister,
-                                  MSchedGraphEdge::OutputDep);
-        }
-        //Dest node is a read
-        else {
-          if(!srcIsUse || srcIsUseandDef)
-            srcNode->addOutEdge(Nodes[j].second,
-                                MSchedGraphEdge::MachineRegister,
-                                MSchedGraphEdge::TrueDep);
-        }
-
-      }
-
-      //Look at all the instructions before this one since machine registers
-      //could live across iterations.
-      for(unsigned j = 0; j < i; ++j) {
-                //Sink node is a write
-        if(Nodes[j].second->getInst()->getOperand(Nodes[j].first).isDef()) {
-                      //Src only uses the register (read)
-            if(srcIsUse)
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphEdge::MachineRegister,
-                                  MSchedGraphEdge::AntiDep, 1);
-            else if(srcIsUseandDef) {
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphEdge::MachineRegister,
-                                  MSchedGraphEdge::AntiDep, 1);
-
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphEdge::MachineRegister,
-                                  MSchedGraphEdge::OutputDep, 1);
-            }
-            else
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphEdge::MachineRegister,
-                                  MSchedGraphEdge::OutputDep, 1);
-        }
-        //Dest node is a read
-        else {
-          if(!srcIsUse || srcIsUseandDef)
-            srcNode->addOutEdge(Nodes[j].second,
-                                MSchedGraphEdge::MachineRegister,
-                                MSchedGraphEdge::TrueDep,1 );
-        }
-
-
-      }
-
-    }
-
-  }
-
-}
-
-//Add edges between all loads and stores
-//Can be less strict with alias analysis and data dependence analysis.
-void MSchedGraph::addMemEdges(const std::vector<MSchedGraphNode*>& memInst,
-                      DependenceAnalyzer &DA,
-                      std::map<MachineInstr*, Instruction*> &machineTollvm) {
-
-  //Get Target machine instruction info
-  const TargetInstrInfo *TMI = Target.getInstrInfo();
-
-  //Loop over all memory instructions in the vector
-  //Knowing that they are in execution, add true, anti, and output dependencies
-  for (unsigned srcIndex = 0; srcIndex < memInst.size(); ++srcIndex) {
-
-    MachineInstr *srcInst = (MachineInstr*) memInst[srcIndex]->getInst();
-
-    //Get the machine opCode to determine type of memory instruction
-    MachineOpCode srcNodeOpCode = srcInst->getOpcode();
-
-    //All instructions after this one in execution order have an
-    //iteration delay of 0
-    for(unsigned destIndex = 0; destIndex < memInst.size(); ++destIndex) {
-
-      //No self loops
-      if(destIndex == srcIndex)
-        continue;
-
-      MachineInstr *destInst = (MachineInstr*) memInst[destIndex]->getInst();
-
-      DEBUG(std::cerr << "MInst1: " << *srcInst << "\n");
-      DEBUG(std::cerr << "MInst2: " << *destInst << "\n");
-
-      //Assuming instructions without corresponding llvm instructions
-      //are from constant pools.
-      if (!machineTollvm.count(srcInst) || !machineTollvm.count(destInst))
-        continue;
-
-      bool useDepAnalyzer = true;
-
-      //Some machine loads and stores are generated by casts, so be
-      //conservative and always add deps
-      Instruction *srcLLVM = machineTollvm[srcInst];
-      Instruction *destLLVM = machineTollvm[destInst];
-      if(!isa<LoadInst>(srcLLVM)
-         && !isa<StoreInst>(srcLLVM)) {
-        if(isa<BinaryOperator>(srcLLVM)) {
-          if(isa<ConstantFP>(srcLLVM->getOperand(0)) || isa<ConstantFP>(srcLLVM->getOperand(1)))
-            continue;
-        }
-        useDepAnalyzer = false;
-      }
-      if(!isa<LoadInst>(destLLVM)
-         && !isa<StoreInst>(destLLVM)) {
-        if(isa<BinaryOperator>(destLLVM)) {
-          if(isa<ConstantFP>(destLLVM->getOperand(0)) || isa<ConstantFP>(destLLVM->getOperand(1)))
-            continue;
-        }
-        useDepAnalyzer = false;
-      }
-
-      //Use dep analysis when we have corresponding llvm loads/stores
-      if(useDepAnalyzer) {
-        bool srcBeforeDest = true;
-        if(destIndex < srcIndex)
-          srcBeforeDest = false;
-
-        DependenceResult dr = DA.getDependenceInfo(machineTollvm[srcInst],
-                                                   machineTollvm[destInst],
-                                                   srcBeforeDest);
-
-        for(std::vector<Dependence>::iterator d = dr.dependences.begin(),
-              de = dr.dependences.end(); d != de; ++d) {
-          //Add edge from load to store
-          memInst[srcIndex]->addOutEdge(memInst[destIndex],
-                                        MSchedGraphEdge::MemoryDep,
-                                        d->getDepType(), d->getIteDiff());
-
-        }
-      }
-      //Otherwise, we can not do any further analysis and must make a dependence
-      else {
-
-        //Get the machine opCode to determine type of memory instruction
-        MachineOpCode destNodeOpCode = destInst->getOpcode();
-
-        //Get the Value* that we are reading from the load, always the first op
-        const MachineOperand &mOp = srcInst->getOperand(0);
-        const MachineOperand &mOp2 = destInst->getOperand(0);
-
-        if(mOp.hasAllocatedReg())
-          if(mOp.getReg() == SparcV9::g0)
-            continue;
-        if(mOp2.hasAllocatedReg())
-          if(mOp2.getReg() == SparcV9::g0)
-            continue;
-
-        DEBUG(std::cerr << "Adding dependence for machine instructions\n");
-        //Load-Store deps
-        if(TMI->isLoad(srcNodeOpCode)) {
-
-          if(TMI->isStore(destNodeOpCode))
-            memInst[srcIndex]->addOutEdge(memInst[destIndex],
-                                          MSchedGraphEdge::MemoryDep,
-                                          MSchedGraphEdge::AntiDep, 0);
-        }
-        else if(TMI->isStore(srcNodeOpCode)) {
-          if(TMI->isStore(destNodeOpCode))
-            memInst[srcIndex]->addOutEdge(memInst[destIndex],
-                                          MSchedGraphEdge::MemoryDep,
-                                          MSchedGraphEdge::OutputDep, 0);
-
-          else
-            memInst[srcIndex]->addOutEdge(memInst[destIndex],
-                                          MSchedGraphEdge::MemoryDep,
-                                          MSchedGraphEdge::TrueDep, 0);
-        }
-      }
-    }
-  }
-}
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraph.h b/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraph.h
deleted file mode 100644
index 201b3083a19..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraph.h
+++ /dev/null
@@ -1,398 +0,0 @@
-//===-- MSchedGraph.h - Scheduling Graph ------------------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-// A graph class for dependencies. This graph only contains true, anti, and
-// output data dependencies for a given MachineBasicBlock. Dependencies
-// across iterations are also computed. Unless data dependence analysis
-// is provided, a conservative approach of adding dependencies between all
-// loads and stores is taken.
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_MSCHEDGRAPH_H
-#define LLVM_MSCHEDGRAPH_H
-#include "DependenceAnalyzer.h"
-#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/CodeGen/MachineInstr.h"
-#include "llvm/Target/TargetMachine.h"
-#include "llvm/Target/TargetData.h"
-#include "llvm/ADT/GraphTraits.h"
-#include "llvm/ADT/STLExtras.h"
-#include "llvm/ADT/iterator"
-#include <vector>
-
-namespace llvm {
-
-  class MSchedGraph;
-  class MSchedGraphNode;
-  template<class IteratorType, class NodeType>
-  class MSchedGraphNodeIterator;
-
-  //MSchedGraphEdge encapsulates the data dependence between nodes. It
-  //identifies the dependence type, on what, and the iteration
-  //difference
-  struct MSchedGraphEdge {
-    enum DataDepOrderType {
-      TrueDep, AntiDep, OutputDep, NonDataDep
-    };
-
-    enum MSchedGraphEdgeType {
-      MemoryDep, ValueDep, MachineRegister, BranchDep
-    };
-
-    //Get or set edge data
-    MSchedGraphNode *getDest() const { return dest; }
-    unsigned getIteDiff() { return iteDiff; }
-    unsigned getDepOrderType() { return depOrderType; }
-    void setDest(MSchedGraphNode *newDest) { dest = newDest; }
-
-  private:
-    friend class MSchedGraphNode;
-    MSchedGraphEdge(MSchedGraphNode *destination, MSchedGraphEdgeType type,
-                    unsigned deptype, unsigned diff)
-      : dest(destination), depType(type), depOrderType(deptype), iteDiff(diff) {}
-
-    MSchedGraphNode *dest;
-    MSchedGraphEdgeType depType;
-    unsigned depOrderType;
-    unsigned iteDiff;
-  };
-
-  //MSchedGraphNode represents a machine instruction and its
-  //corresponding latency. Each node also contains a list of its
-  //predecessors and sucessors.
-  class MSchedGraphNode {
-
-    const MachineInstr* Inst; //Machine Instruction
-    MSchedGraph* Parent; //Graph this node belongs to
-    unsigned index; //Index in BB
-    unsigned latency; //Latency of Instruction
-    bool isBranchInstr; //Is this node the branch instr or not
-
-    std::vector<MSchedGraphNode*> Predecessors; //Predecessor Nodes
-    std::vector<MSchedGraphEdge> Successors; //Successor edges
-
-  public:
-    MSchedGraphNode(const MachineInstr *inst, MSchedGraph *graph,
-                    unsigned index, unsigned late=0, bool isBranch=false);
-
-    MSchedGraphNode(const MSchedGraphNode &N);
-
-    //Iterators - Predecessor and Succussor
-    typedef std::vector<MSchedGraphNode*>::iterator pred_iterator;
-    pred_iterator pred_begin() { return Predecessors.begin(); }
-    pred_iterator pred_end() { return Predecessors.end(); }
-    unsigned pred_size() { return Predecessors.size(); }
-
-    typedef std::vector<MSchedGraphNode*>::const_iterator pred_const_iterator;
-    pred_const_iterator pred_begin() const { return Predecessors.begin(); }
-    pred_const_iterator pred_end() const { return Predecessors.end(); }
-
-    typedef MSchedGraphNodeIterator<std::vector<MSchedGraphEdge>::const_iterator,
-                                    const MSchedGraphNode> succ_const_iterator;
-    succ_const_iterator succ_begin() const;
-    succ_const_iterator succ_end() const;
-
-    typedef MSchedGraphNodeIterator<std::vector<MSchedGraphEdge>::iterator,
-                                    MSchedGraphNode> succ_iterator;
-    succ_iterator succ_begin();
-    succ_iterator succ_end();
-    unsigned succ_size() { return Successors.size(); }
-
-    //Get or set predecessor nodes, or successor edges
-    void setPredecessor(unsigned index, MSchedGraphNode *dest) {
-      Predecessors[index] = dest;
-    }
-
-    MSchedGraphNode* getPredecessor(unsigned index) {
-      return Predecessors[index];
-    }
-
-    MSchedGraphEdge* getSuccessor(unsigned index) {
-      return &Successors[index];
-    }
-
-    void deleteSuccessor(MSchedGraphNode *node) {
-      for (unsigned i = 0; i != Successors.size(); ++i)
-        if (Successors[i].getDest() == node) {
-          Successors.erase(Successors.begin()+i);
-          node->Predecessors.erase(std::find(node->Predecessors.begin(),
-                                             node->Predecessors.end(), this));
-          --i; //Decrease index var since we deleted a node
-        }
-    }
-
-    void addOutEdge(MSchedGraphNode *destination,
-                    MSchedGraphEdge::MSchedGraphEdgeType type,
-                    unsigned deptype, unsigned diff=0) {
-      Successors.push_back(MSchedGraphEdge(destination, type, deptype,diff));
-      destination->Predecessors.push_back(this);
-    }
-
-    //General methods to get and set data for the node
-    const MachineInstr* getInst() { return Inst; }
-    MSchedGraph* getParent() { return Parent; }
-    bool hasPredecessors() { return (Predecessors.size() > 0); }
-    bool hasSuccessors() { return (Successors.size() > 0); }
-    unsigned getLatency() { return latency; }
-    unsigned getLatency() const { return latency; }
-    unsigned getIndex() { return index; }
-    unsigned getIteDiff(MSchedGraphNode *succ);
-    MSchedGraphEdge getInEdge(MSchedGraphNode *pred);
-    unsigned getInEdgeNum(MSchedGraphNode *pred);
-    bool isSuccessor(MSchedGraphNode *);
-    bool isPredecessor(MSchedGraphNode *);
-    bool isBranch() { return isBranchInstr; }
-
-    //Debug support
-    void print(std::ostream &os) const;
-    void setParent(MSchedGraph *p) { Parent = p; }
-  };
-
-  //Node iterator for graph generation
-  template<class IteratorType, class NodeType>
-  class MSchedGraphNodeIterator : public forward_iterator<NodeType*, ptrdiff_t> {
-    IteratorType I;   // std::vector<MSchedGraphEdge>::iterator or const_iterator
-  public:
-    MSchedGraphNodeIterator(IteratorType i) : I(i) {}
-
-    bool operator==(const MSchedGraphNodeIterator RHS) const { return I == RHS.I; }
-    bool operator!=(const MSchedGraphNodeIterator RHS) const { return I != RHS.I; }
-
-    const MSchedGraphNodeIterator &operator=(const MSchedGraphNodeIterator &RHS) {
-      I = RHS.I;
-      return *this;
-    }
-
-    NodeType* operator*() const {
-      return I->getDest();
-    }
-    NodeType* operator->() const { return operator*(); }
-
-    MSchedGraphNodeIterator& operator++() {                // Preincrement
-      ++I;
-      return *this;
-    }
-    MSchedGraphNodeIterator operator++(int) { // Postincrement
-      MSchedGraphNodeIterator tmp = *this; ++*this; return tmp;
-    }
-
-    MSchedGraphEdge &getEdge() {
-      return *I;
-    }
-    const MSchedGraphEdge &getEdge() const {
-      return *I;
-    }
-  };
-
-  inline MSchedGraphNode::succ_const_iterator MSchedGraphNode::succ_begin() const {
-    return succ_const_iterator(Successors.begin());
-  }
-  inline MSchedGraphNode::succ_const_iterator MSchedGraphNode::succ_end() const {
-    return succ_const_iterator(Successors.end());
-  }
-  inline MSchedGraphNode::succ_iterator MSchedGraphNode::succ_begin() {
-    return succ_iterator(Successors.begin());
-  }
-  inline MSchedGraphNode::succ_iterator MSchedGraphNode::succ_end() {
-    return succ_iterator(Successors.end());
-  }
-
-  // ostream << operator for MSGraphNode class
-  inline std::ostream &operator<<(std::ostream &os,
-                                  const MSchedGraphNode &node) {
-    node.print(os);
-    return os;
-  }
-
-
-  // Provide specializations of GraphTraits to be able to use graph
-  // iterators on the scheduling graph!
-  //
-  template <> struct GraphTraits<MSchedGraphNode*> {
-    typedef MSchedGraphNode NodeType;
-    typedef MSchedGraphNode::succ_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->succ_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->succ_end();
-    }
-
-    static NodeType *getEntryNode(NodeType* N) { return N; }
-  };
-
-
-
-  //Graph class to represent dependence graph
-  class MSchedGraph {
-
-    std::vector<const MachineBasicBlock *> BBs; //Machine basic block
-    const TargetMachine &Target; //Target Machine
-
-    //Nodes
-    std::map<const MachineInstr*, MSchedGraphNode*> GraphMap;
-
-    //Add Nodes and Edges to this graph for our BB
-    typedef std::pair<int, MSchedGraphNode*> OpIndexNodePair;
-    void buildNodesAndEdges(std::map<const MachineInstr*, unsigned> &ignoreInstrs, DependenceAnalyzer &DA, std::map<MachineInstr*, Instruction*> &machineTollvm);
-    void addValueEdges(std::vector<OpIndexNodePair> &NodesInMap,
-                       MSchedGraphNode *node,
-                       bool nodeIsUse, bool nodeIsDef, std::vector<const MachineInstr*> &phiInstrs, int diff=0);
-    void addMachRegEdges(std::map<int,
-                         std::vector<OpIndexNodePair> >& regNumtoNodeMap);
-    void addMemEdges(const std::vector<MSchedGraphNode*>& memInst,
-                     DependenceAnalyzer &DA, std::map<MachineInstr*, Instruction*> &machineTollvm);
-    void addBranchEdges();
-
-  public:
-    MSchedGraph(const MachineBasicBlock *bb, const TargetMachine &targ,
-                std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                DependenceAnalyzer &DA, std::map<MachineInstr*, Instruction*> &machineTollvm);
-
-    //Copy constructor with maps to link old nodes to new nodes
-    MSchedGraph(const MSchedGraph &G, std::map<MSchedGraphNode*, MSchedGraphNode*> &newNodes);
-
-    MSchedGraph(std::vector<const MachineBasicBlock*> &bbs,
-                const TargetMachine &targ,
-                std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                DependenceAnalyzer &DA,
-                std::map<MachineInstr*, Instruction*> &machineTollvm);
-
-    //Print graph
-    void print(std::ostream &os) const;
-
-    //Deconstructor!
-    ~MSchedGraph();
-
-    //Add or delete nodes from the Graph
-    void addNode(const MachineInstr* MI, MSchedGraphNode *node);
-    void deleteNode(MSchedGraphNode *node);
-    int totalDelay();
-
-    //iterators
-    typedef std::map<const MachineInstr*, MSchedGraphNode*>::iterator iterator;
-    typedef std::map<const MachineInstr*, MSchedGraphNode*>::const_iterator const_iterator;
-    typedef std::map<const MachineInstr*, MSchedGraphNode*>::reverse_iterator reverse_iterator;
-    iterator find(const MachineInstr* I) { return GraphMap.find(I); }
-    iterator end() { return GraphMap.end(); }
-    iterator begin() { return GraphMap.begin(); }
-    unsigned size() { return GraphMap.size(); }
-    reverse_iterator rbegin() { return GraphMap.rbegin(); }
-    reverse_iterator rend() { return GraphMap.rend(); }
-
-    //Get Target or original machine basic block
-    const TargetMachine* getTarget() { return &Target; }
-    std::vector<const MachineBasicBlock*> getBBs() { return BBs; }
-  };
-
-
-
-
-
-  // Provide specializations of GraphTraits to be able to use graph
-  // iterators on the scheduling graph
-  static MSchedGraphNode& getSecond(std::pair<const MachineInstr* const,
-                                    MSchedGraphNode*> &Pair) {
-    return *Pair.second;
-  }
-
-  template <> struct GraphTraits<MSchedGraph*> {
-    typedef MSchedGraphNode NodeType;
-    typedef MSchedGraphNode::succ_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->succ_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->succ_end();
-    }
-
-    typedef std::pointer_to_unary_function<std::pair<const MachineInstr* const,
-           MSchedGraphNode*>&, MSchedGraphNode&> DerefFun;
-
-    typedef mapped_iterator<MSchedGraph::iterator, DerefFun> nodes_iterator;
-    static nodes_iterator nodes_begin(MSchedGraph *G) {
-      return map_iterator(((MSchedGraph*)G)->begin(), DerefFun(getSecond));
-    }
-    static nodes_iterator nodes_end(MSchedGraph *G) {
-      return map_iterator(((MSchedGraph*)G)->end(), DerefFun(getSecond));
-    }
-
-  };
-
-  template <> struct GraphTraits<const MSchedGraph*> {
-    typedef const MSchedGraphNode NodeType;
-    typedef MSchedGraphNode::succ_const_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->succ_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->succ_end();
-    }
-    typedef std::pointer_to_unary_function<std::pair<const MachineInstr* const,
-                                                     MSchedGraphNode*>&, MSchedGraphNode&> DerefFun;
-
-    typedef mapped_iterator<MSchedGraph::iterator, DerefFun> nodes_iterator;
-    static nodes_iterator nodes_begin(MSchedGraph *G) {
-      return map_iterator(((MSchedGraph*)G)->begin(), DerefFun(getSecond));
-    }
-    static nodes_iterator nodes_end(MSchedGraph *G) {
-      return map_iterator(((MSchedGraph*)G)->end(), DerefFun(getSecond));
-    }
-  };
-
-  template <> struct GraphTraits<Inverse<MSchedGraph*> > {
-    typedef MSchedGraphNode NodeType;
-    typedef MSchedGraphNode::pred_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->pred_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->pred_end();
-    }
-    typedef std::pointer_to_unary_function<std::pair<const MachineInstr* const,
-           MSchedGraphNode*>&, MSchedGraphNode&> DerefFun;
-
-    typedef mapped_iterator<MSchedGraph::iterator, DerefFun> nodes_iterator;
-    static nodes_iterator nodes_begin(MSchedGraph *G) {
-      return map_iterator(((MSchedGraph*)G)->begin(), DerefFun(getSecond));
-    }
-    static nodes_iterator nodes_end(MSchedGraph *G) {
-      return map_iterator(((MSchedGraph*)G)->end(), DerefFun(getSecond));
-    }
-  };
-
-  template <> struct GraphTraits<Inverse<const MSchedGraph*> > {
-    typedef const MSchedGraphNode NodeType;
-    typedef MSchedGraphNode::pred_const_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->pred_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->pred_end();
-    }
-
-    typedef std::pointer_to_unary_function<std::pair<const MachineInstr* const,
-                                                     MSchedGraphNode*>&, MSchedGraphNode&> DerefFun;
-
-    typedef mapped_iterator<MSchedGraph::iterator, DerefFun> nodes_iterator;
-    static nodes_iterator nodes_begin(MSchedGraph *G) {
-      return map_iterator(((MSchedGraph*)G)->begin(), DerefFun(getSecond));
-    }
-    static nodes_iterator nodes_end(MSchedGraph *G) {
-      return map_iterator(((MSchedGraph*)G)->end(), DerefFun(getSecond));
-    }
-  };
-}
-
-#endif
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraphSB.cpp b/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraphSB.cpp
deleted file mode 100644
index bd879f87786..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraphSB.cpp
+++ /dev/null
@@ -1,870 +0,0 @@
-//===-- MSchedGraphSB.cpp - Scheduling Graph ----------------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-// A graph class for dependencies. This graph only contains true, anti, and
-// output data dependencies for a given MachineBasicBlock. Dependencies
-// across iterations are also computed. Unless data dependence analysis
-// is provided, a conservative approach of adding dependencies between all
-// loads and stores is taken.
-//===----------------------------------------------------------------------===//
-
-#define DEBUG_TYPE "ModuloSchedSB"
-#include "MSchedGraphSB.h"
-#include "../SparcV9RegisterInfo.h"
-#include "../MachineCodeForInstruction.h"
-#include "llvm/BasicBlock.h"
-#include "llvm/Constants.h"
-#include "llvm/Instructions.h"
-#include "llvm/Type.h"
-#include "llvm/CodeGen/MachineBasicBlock.h"
-#include "llvm/Target/TargetInstrInfo.h"
-#include "llvm/Support/Debug.h"
-#include <cstdlib>
-#include <algorithm>
-#include <set>
-#include "llvm/Target/TargetSchedInfo.h"
-#include "../SparcV9Internals.h"
-#include <iostream>
-using namespace llvm;
-
-//MSchedGraphSBNode constructor
-MSchedGraphSBNode::MSchedGraphSBNode(const MachineInstr* inst,
-                                 MSchedGraphSB *graph, unsigned idx,
-                                 unsigned late, bool isBranch)
-  : Inst(inst), Parent(graph), index(idx), latency(late),
-    isBranchInstr(isBranch) {
-
-  //Add to the graph
-  graph->addNode(inst, this);
-}
-
-//MSchedGraphSBNode constructor
-MSchedGraphSBNode::MSchedGraphSBNode(const MachineInstr* inst,
-                                     std::vector<const MachineInstr*> &other,
-                                     MSchedGraphSB *graph, unsigned idx,
-                                     unsigned late, bool isPNode)
-  : Inst(inst), otherInstrs(other), Parent(graph), index(idx), latency(late), isPredicateNode(isPNode) {
-
-
-  isBranchInstr = false;
-
-  //Add to the graph
-  graph->addNode(inst, this);
-}
-
-//MSchedGraphSBNode copy constructor
-MSchedGraphSBNode::MSchedGraphSBNode(const MSchedGraphSBNode &N)
-  : Predecessors(N.Predecessors), Successors(N.Successors) {
-
-  Inst = N.Inst;
-  Parent = N.Parent;
-  index = N.index;
-  latency = N.latency;
-  isBranchInstr = N.isBranchInstr;
-  otherInstrs = N.otherInstrs;
-}
-
-//Print the node (instruction and latency)
-void MSchedGraphSBNode::print(std::ostream &os) const {
-  if(!isPredicate())
-    os << "MSchedGraphSBNode: Inst=" << *Inst << ", latency= " << latency << "\n";
-  else
-    os << "Pred Node\n";
-}
-
-
-//Get the edge from a predecessor to this node
-MSchedGraphSBEdge MSchedGraphSBNode::getInEdge(MSchedGraphSBNode *pred) {
-  //Loop over all the successors of our predecessor
-  //return the edge the corresponds to this in edge
-  for (MSchedGraphSBNode::succ_iterator I = pred->succ_begin(),
-         E = pred->succ_end(); I != E; ++I) {
-    if (*I == this)
-      return I.getEdge();
-  }
-  assert(0 && "Should have found edge between this node and its predecessor!");
-  abort();
-}
-
-//Get the iteration difference for the edge from this node to its successor
-unsigned MSchedGraphSBNode::getIteDiff(MSchedGraphSBNode *succ) {
-  for(std::vector<MSchedGraphSBEdge>::iterator I = Successors.begin(),
-        E = Successors.end();
-      I != E; ++I) {
-    if(I->getDest() == succ)
-      return I->getIteDiff();
-  }
-  return 0;
-}
-
-//Get the index into the vector of edges for the edge from pred to this node
-unsigned MSchedGraphSBNode::getInEdgeNum(MSchedGraphSBNode *pred) {
-  //Loop over all the successors of our predecessor
-  //return the edge the corresponds to this in edge
-  int count = 0;
-  for(MSchedGraphSBNode::succ_iterator I = pred->succ_begin(),
-        E = pred->succ_end();
-      I != E; ++I) {
-    if(*I == this)
-      return count;
-    count++;
-  }
-  assert(0 && "Should have found edge between this node and its predecessor!");
-  abort();
-}
-
-//Determine if succ is a successor of this node
-bool MSchedGraphSBNode::isSuccessor(MSchedGraphSBNode *succ) {
-  for(succ_iterator I = succ_begin(), E = succ_end(); I != E; ++I)
-    if(*I == succ)
-      return true;
-  return false;
-}
-
-//Dtermine if pred is a predecessor of this node
-bool MSchedGraphSBNode::isPredecessor(MSchedGraphSBNode *pred) {
-  if(std::find( Predecessors.begin(),  Predecessors.end(),
-                pred) !=   Predecessors.end())
-    return true;
-  else
-    return false;
-}
-
-//Add a node to the graph
-void MSchedGraphSB::addNode(const MachineInstr* MI,
-                          MSchedGraphSBNode *node) {
-
-  //Make sure node does not already exist
-  assert(GraphMap.find(MI) == GraphMap.end()
-         && "New MSchedGraphSBNode already exists for this instruction");
-
-  GraphMap[MI] = node;
-}
-
-//Delete a node to the graph
-void MSchedGraphSB::deleteNode(MSchedGraphSBNode *node) {
-
-  //Delete the edge to this node from all predecessors
-  while(node->pred_size() > 0) {
-    //DEBUG(std::cerr << "Delete edge from: " << **P << " to " << *node << "\n");
-    MSchedGraphSBNode *pred = *(node->pred_begin());
-    pred->deleteSuccessor(node);
-  }
-
-  //Remove this node from the graph
-  GraphMap.erase(node->getInst());
-
-}
-
-
-//Create a graph for a machine block. The ignoreInstrs map is so that
-//we ignore instructions associated to the index variable since this
-//is a special case in Modulo Scheduling.  We only want to deal with
-//the body of the loop.
-MSchedGraphSB::MSchedGraphSB(std::vector<const MachineBasicBlock*> &bbs,
-                         const TargetMachine &targ,
-                         std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                         DependenceAnalyzer &DA,
-                         std::map<MachineInstr*, Instruction*> &machineTollvm)
-  : BBs(bbs), Target(targ) {
-
-  //Make sure there is at least one BB and it is not null,
-  assert(((bbs.size() >= 1) &&  bbs[1] != NULL) && "Basic Block is null");
-
-  std::map<MSchedGraphSBNode*, std::set<MachineInstr*> > liveOutsideTrace;
-  std::set<const BasicBlock*> llvmBBs;
-
-  for(std::vector<const MachineBasicBlock*>::iterator MBB = bbs.begin(), ME = bbs.end()-1;
-      MBB != ME; ++MBB)
-    llvmBBs.insert((*MBB)->getBasicBlock());
-
-  //create predicate nodes
-  DEBUG(std::cerr << "Create predicate nodes\n");
-  for(std::vector<const MachineBasicBlock*>::iterator MBB = bbs.begin(), ME = bbs.end()-1;
-       MBB != ME; ++MBB) {
-    //Get LLVM basic block
-    BasicBlock *BB = (BasicBlock*) (*MBB)->getBasicBlock();
-
-    //Get Terminator
-    BranchInst *b = dyn_cast<BranchInst>(BB->getTerminator());
-
-    std::vector<const MachineInstr*> otherInstrs;
-    MachineInstr *instr = 0;
-
-    //Get the condition for the branch (we already checked if it was conditional)
-    if(b->isConditional()) {
-
-      Value *cond = b->getCondition();
-
-      DEBUG(std::cerr << "Condition: " << *cond << "\n");
-
-      assert(cond && "Condition must not be null!");
-
-      if(Instruction *I = dyn_cast<Instruction>(cond)) {
-        MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(I);
-        if(tempMvec.size() > 0) {
-          DEBUG(std::cerr << *(tempMvec[tempMvec.size()-1]) << "\n");;
-          instr = (MachineInstr*) tempMvec[tempMvec.size()-1];
-        }
-      }
-    }
-
-    //Get Machine target information for calculating latency
-    const TargetInstrInfo *MTI = Target.getInstrInfo();
-
-    MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(b);
-    int offset = tempMvec.size();
-    for (unsigned j = 0; j < tempMvec.size(); j++) {
-      MachineInstr *mi = tempMvec[j];
-      if(MTI->isNop(mi->getOpcode()))
-        continue;
-
-      if(!instr) {
-        instr = mi;
-        DEBUG(std::cerr << "No Cond MI: " << *mi << "\n");
-      }
-      else {
-        DEBUG(std::cerr << *mi << "\n");;
-        otherInstrs.push_back(mi);
-      }
-    }
-
-    //Node is created and added to the graph automatically
-    MSchedGraphSBNode *node =  new MSchedGraphSBNode(instr, otherInstrs, this, (*MBB)->size()-offset-1, 3, true);
-
-    DEBUG(std::cerr << "Created Node: " << *node << "\n");
-
-    //Now loop over all instructions and see if their def is live outside the trace
-    MachineBasicBlock *mb = (MachineBasicBlock*) *MBB;
-    for(MachineBasicBlock::iterator I = mb->begin(), E = mb->end(); I != E; ++I) {
-      MachineInstr *instr = I;
-      if(MTI->isNop(instr->getOpcode()) || MTI->isBranch(instr->getOpcode()))
-        continue;
-      if(node->getInst() == instr)
-        continue;
-
-      for(unsigned i=0; i < instr->getNumOperands(); ++i) {
-        MachineOperand &mOp = instr->getOperand(i);
-        if(mOp.isDef() && mOp.getType() == MachineOperand::MO_VirtualRegister) {
-          Value *val = mOp.getVRegValue();
-          //Check if there is a use not in the trace
-          for(Value::use_iterator V = val->use_begin(), VE = val->use_end(); V != VE; ++V) {
-            if (Instruction *Inst = dyn_cast<Instruction>(*V)) {
-              if(llvmBBs.count(Inst->getParent()))
-                 liveOutsideTrace[node].insert(instr);
-            }
-          }
-        }
-      }
-    }
-
-
-  }
-
-  //Create nodes and edges for this BB
-  buildNodesAndEdges(ignoreInstrs, DA, machineTollvm, liveOutsideTrace);
-
-}
-
-
-//Copies the graph and keeps a map from old to new nodes
-MSchedGraphSB::MSchedGraphSB(const MSchedGraphSB &G,
-                         std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> &newNodes)
-  : Target(G.Target) {
-
-  BBs = G.BBs;
-
-  std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> oldToNew;
-  //Copy all nodes
-  for(MSchedGraphSB::const_iterator N = G.GraphMap.begin(),
-        NE = G.GraphMap.end(); N != NE; ++N) {
-
-    MSchedGraphSBNode *newNode = new MSchedGraphSBNode(*(N->second));
-    oldToNew[&*(N->second)] = newNode;
-    newNodes[newNode] = &*(N->second);
-    GraphMap[&*(N->first)] = newNode;
-  }
-
-  //Loop over nodes and update edges to point to new nodes
-  for(MSchedGraphSB::iterator N = GraphMap.begin(), NE = GraphMap.end();
-      N != NE; ++N) {
-
-    //Get the node we are dealing with
-    MSchedGraphSBNode *node = &*(N->second);
-
-    node->setParent(this);
-
-    //Loop over nodes successors and predecessors and update to the new nodes
-    for(unsigned i = 0; i < node->pred_size(); ++i) {
-      node->setPredecessor(i, oldToNew[node->getPredecessor(i)]);
-    }
-
-    for(unsigned i = 0; i < node->succ_size(); ++i) {
-      MSchedGraphSBEdge *edge = node->getSuccessor(i);
-      MSchedGraphSBNode *oldDest = edge->getDest();
-      edge->setDest(oldToNew[oldDest]);
-    }
-  }
-}
-
-//Deconstructor, deletes all nodes in the graph
-MSchedGraphSB::~MSchedGraphSB () {
-  for(MSchedGraphSB::iterator I = GraphMap.begin(), E = GraphMap.end();
-      I != E; ++I)
-    delete I->second;
-}
-
-//Print out graph
-void MSchedGraphSB::print(std::ostream &os) const {
-  for(MSchedGraphSB::const_iterator N = GraphMap.begin(), NE = GraphMap.end();
-      N != NE; ++N) {
-
-    //Get the node we are dealing with
-    MSchedGraphSBNode *node = &*(N->second);
-
-    os << "Node Start\n";
-    node->print(os);
-    os << "Successors:\n";
-    //print successors
-    for(unsigned i = 0; i < node->succ_size(); ++i) {
-      MSchedGraphSBEdge *edge = node->getSuccessor(i);
-      MSchedGraphSBNode *oldDest = edge->getDest();
-      oldDest->print(os);
-    }
-    os << "Node End\n";
-  }
-}
-
-//Calculate total delay
-int MSchedGraphSB::totalDelay() {
-  int sum = 0;
-
-  for(MSchedGraphSB::const_iterator N = GraphMap.begin(), NE = GraphMap.end();
-      N != NE; ++N) {
-
-    //Get the node we are dealing with
-    MSchedGraphSBNode *node = &*(N->second);
-    sum += node->getLatency();
-  }
-  return sum;
-}
-
-bool MSchedGraphSB::instrCauseException(MachineOpCode opCode) {
-  //Check for integer divide
-  if(opCode == V9::SDIVXr || opCode == V9::SDIVXi
-     || opCode == V9::UDIVXr || opCode == V9::UDIVXi)
-    return true;
-
-  //Check for loads or stores
-  const TargetInstrInfo *MTI = Target.getInstrInfo();
-  //if( MTI->isLoad(opCode) ||
-  if(MTI->isStore(opCode))
-    return true;
-
-  //Check for any floating point operation
-  const TargetSchedInfo *msi = Target.getSchedInfo();
-  InstrSchedClass sc = msi->getSchedClass(opCode);
-
-  //FIXME: Should check for floating point instructions!
-  //if(sc == SPARC_FGA || sc == SPARC_FGM)
-  //return true;
-
-  return false;
-}
-
-
-//Add edges between the nodes
-void MSchedGraphSB::buildNodesAndEdges(std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                                     DependenceAnalyzer &DA,
-                                       std::map<MachineInstr*, Instruction*> &machineTollvm,
-                                       std::map<MSchedGraphSBNode*, std::set<MachineInstr*> > &liveOutsideTrace) {
-
-
-  //Get Machine target information for calculating latency
-  const TargetInstrInfo *MTI = Target.getInstrInfo();
-
-  std::vector<MSchedGraphSBNode*> memInstructions;
-  std::map<int, std::vector<OpIndexNodePair> > regNumtoNodeMap;
-  std::map<const Value*, std::vector<OpIndexNodePair> > valuetoNodeMap;
-
-  //Save PHI instructions to deal with later
-  std::vector<const MachineInstr*> phiInstrs;
-  unsigned index = 0;
-
-  MSchedGraphSBNode *lastPred = 0;
-
-
-  for(std::vector<const MachineBasicBlock*>::iterator B = BBs.begin(),
-        BE = BBs.end(); B != BE; ++B) {
-
-    const MachineBasicBlock *BB = *B;
-
-
-    //Loop over instructions in MBB and add nodes and edges
-    for (MachineBasicBlock::const_iterator MI = BB->begin(), e = BB->end();
-         MI != e; ++MI) {
-
-      //Ignore indvar instructions
-      if(ignoreInstrs.count(MI)) {
-        ++index;
-        continue;
-      }
-
-      //Get each instruction of machine basic block, get the delay
-      //using the op code, create a new node for it, and add to the
-      //graph.
-
-      MachineOpCode opCode = MI->getOpcode();
-      int delay;
-
-      //Get delay
-      delay = MTI->maxLatency(opCode);
-
-      //Create new node for this machine instruction and add to the graph.
-      //Create only if not a nop
-      if(MTI->isNop(opCode))
-        continue;
-
-      //Sparc BE does not use PHI opcode, so assert on this case
-      assert(opCode != TargetInstrInfo::PHI && "Did not expect PHI opcode");
-
-      bool isBranch = false;
-
-      //Skip branches
-      if(MTI->isBranch(opCode))
-        continue;
-
-      //Node is created and added to the graph automatically
-      MSchedGraphSBNode *node = 0;
-      if(!GraphMap.count(MI)){
-        node =  new MSchedGraphSBNode(MI, this, index, delay);
-        DEBUG(std::cerr << "Created Node: " << *node << "\n");
-      }
-      else {
-        node = GraphMap[MI];
-        if(node->isPredicate()) {
-          //Create edge between this node and last pred, then switch to new pred
-          if(lastPred) {
-            lastPred->addOutEdge(node, MSchedGraphSBEdge::PredDep,
-            MSchedGraphSBEdge::NonDataDep, 0);
-
-            if(liveOutsideTrace.count(lastPred)) {
-              for(std::set<MachineInstr*>::iterator L = liveOutsideTrace[lastPred].begin(), LE = liveOutsideTrace[lastPred].end(); L != LE; ++L)
-                lastPred->addOutEdge(GraphMap[*L], MSchedGraphSBEdge::PredDep,
-                                     MSchedGraphSBEdge::NonDataDep, 1);
-            }
-
-          }
-
-          lastPred = node;
-        }
-      }
-
-      //Add dependencies to instructions that cause exceptions
-      if(lastPred)
-        lastPred->print(std::cerr);
-
-      if(!node->isPredicate() && instrCauseException(opCode)) {
-        if(lastPred) {
-          lastPred->addOutEdge(node, MSchedGraphSBEdge::PredDep,
-                               MSchedGraphSBEdge::NonDataDep, 0);
-        }
-      }
-
-
-      //Check OpCode to keep track of memory operations to add memory
-      //dependencies later.
-      if(MTI->isLoad(opCode) || MTI->isStore(opCode))
-        memInstructions.push_back(node);
-
-      //Loop over all operands, and put them into the register number to
-      //graph node map for determining dependencies
-      //If an operands is a use/def, we have an anti dependence to itself
-      for(unsigned i=0; i < MI->getNumOperands(); ++i) {
-        //Get Operand
-        const MachineOperand &mOp = MI->getOperand(i);
-
-        //Check if it has an allocated register
-        if(mOp.hasAllocatedReg()) {
-          int regNum = mOp.getReg();
-
-          if(regNum != SparcV9::g0) {
-            //Put into our map
-            regNumtoNodeMap[regNum].push_back(std::make_pair(i, node));
-          }
-          continue;
-        }
-
-
-        //Add virtual registers dependencies
-        //Check if any exist in the value map already and create dependencies
-        //between them.
-        if(mOp.getType() == MachineOperand::MO_VirtualRegister
-           ||  mOp.getType() == MachineOperand::MO_CCRegister) {
-
-          //Make sure virtual register value is not null
-          assert((mOp.getVRegValue() != NULL) && "Null value is defined");
-
-          //Check if this is a read operation in a phi node, if so DO NOT PROCESS
-          if(mOp.isUse() && (opCode == TargetInstrInfo::PHI)) {
-            DEBUG(std::cerr << "Read Operation in a PHI node\n");
-            continue;
-          }
-
-          if (const Value* srcI = mOp.getVRegValue()) {
-
-            //Find value in the map
-            std::map<const Value*, std::vector<OpIndexNodePair> >::iterator V
-              = valuetoNodeMap.find(srcI);
-
-            //If there is something in the map already, add edges from
-            //those instructions
-            //to this one we are processing
-            if(V != valuetoNodeMap.end()) {
-              addValueEdges(V->second, node, mOp.isUse(), mOp.isDef(), phiInstrs);
-
-              //Add to value map
-              V->second.push_back(std::make_pair(i,node));
-            }
-            //Otherwise put it in the map
-            else
-              //Put into value map
-              valuetoNodeMap[mOp.getVRegValue()].push_back(std::make_pair(i, node));
-          }
-        }
-      }
-      ++index;
-    }
-
-    //Loop over LLVM BB, examine phi instructions, and add them to our
-    //phiInstr list to process
-    const BasicBlock *llvm_bb = BB->getBasicBlock();
-    for(BasicBlock::const_iterator I = llvm_bb->begin(), E = llvm_bb->end();
-        I != E; ++I) {
-      if(const PHINode *PN = dyn_cast<PHINode>(I)) {
-        MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(PN);
-        for (unsigned j = 0; j < tempMvec.size(); j++) {
-          if(!ignoreInstrs.count(tempMvec[j])) {
-            DEBUG(std::cerr << "Inserting phi instr into map: " << *tempMvec[j] << "\n");
-            phiInstrs.push_back((MachineInstr*) tempMvec[j]);
-          }
-        }
-      }
-
-    }
-
-    addMemEdges(memInstructions, DA, machineTollvm);
-    addMachRegEdges(regNumtoNodeMap);
-
-    //Finally deal with PHI Nodes and Value*
-    for(std::vector<const MachineInstr*>::iterator I = phiInstrs.begin(),
-          E = phiInstrs.end(); I != E;  ++I) {
-
-      //Get Node for this instruction
-      std::map<const MachineInstr*, MSchedGraphSBNode*>::iterator X;
-      X = find(*I);
-
-      if(X == GraphMap.end())
-        continue;
-
-      MSchedGraphSBNode *node = X->second;
-
-      DEBUG(std::cerr << "Adding ite diff edges for node: " << *node << "\n");
-
-      //Loop over operands for this instruction and add value edges
-      for(unsigned i=0; i < (*I)->getNumOperands(); ++i) {
-        //Get Operand
-        const MachineOperand &mOp = (*I)->getOperand(i);
-        if((mOp.getType() == MachineOperand::MO_VirtualRegister
-            ||  mOp.getType() == MachineOperand::MO_CCRegister) && mOp.isUse()) {
-
-          //find the value in the map
-          if (const Value* srcI = mOp.getVRegValue()) {
-
-            //Find value in the map
-            std::map<const Value*, std::vector<OpIndexNodePair> >::iterator V
-              = valuetoNodeMap.find(srcI);
-
-            //If there is something in the map already, add edges from
-            //those instructions
-            //to this one we are processing
-            if(V != valuetoNodeMap.end()) {
-              addValueEdges(V->second, node, mOp.isUse(), mOp.isDef(),
-                            phiInstrs, 1);
-            }
-          }
-        }
-      }
-    }
-  }
-}
-//Add dependencies for Value*s
-void MSchedGraphSB::addValueEdges(std::vector<OpIndexNodePair> &NodesInMap,
-                                MSchedGraphSBNode *destNode, bool nodeIsUse,
-                                bool nodeIsDef, std::vector<const MachineInstr*> &phiInstrs, int diff) {
-
-  for(std::vector<OpIndexNodePair>::iterator I = NodesInMap.begin(),
-        E = NodesInMap.end(); I != E; ++I) {
-
-    //Get node in vectors machine operand that is the same value as node
-    MSchedGraphSBNode *srcNode = I->second;
-    MachineOperand mOp = srcNode->getInst()->getOperand(I->first);
-
-    if(diff > 0)
-      if(std::find(phiInstrs.begin(), phiInstrs.end(), srcNode->getInst()) == phiInstrs.end())
-        continue;
-
-    //Node is a Def, so add output dep.
-    if(nodeIsDef) {
-      if(mOp.isUse()) {
-        DEBUG(std::cerr << "Edge from " << *srcNode << " to " << *destNode << " (itediff=" << diff << ", type=anti)\n");
-        srcNode->addOutEdge(destNode, MSchedGraphSBEdge::ValueDep,
-                            MSchedGraphSBEdge::AntiDep, diff);
-      }
-      if(mOp.isDef()) {
-        DEBUG(std::cerr << "Edge from " << *srcNode << " to " << *destNode << " (itediff=" << diff << ", type=output)\n");
-        srcNode->addOutEdge(destNode, MSchedGraphSBEdge::ValueDep,
-                            MSchedGraphSBEdge::OutputDep, diff);
-      }
-    }
-    if(nodeIsUse) {
-      if(mOp.isDef()) {
-        DEBUG(std::cerr << "Edge from " << *srcNode << " to " << *destNode << " (itediff=" << diff << ", type=true)\n");
-        srcNode->addOutEdge(destNode, MSchedGraphSBEdge::ValueDep,
-                            MSchedGraphSBEdge::TrueDep, diff);
-      }
-    }
-  }
-}
-
-//Add dependencies for machine registers across iterations
-void MSchedGraphSB::addMachRegEdges(std::map<int, std::vector<OpIndexNodePair> >& regNumtoNodeMap) {
-  //Loop over all machine registers in the map, and add dependencies
-  //between the instructions that use it
-  typedef std::map<int, std::vector<OpIndexNodePair> > regNodeMap;
-  for(regNodeMap::iterator I = regNumtoNodeMap.begin();
-      I != regNumtoNodeMap.end(); ++I) {
-    //Get the register number
-    int regNum = (*I).first;
-
-    //Get Vector of nodes that use this register
-    std::vector<OpIndexNodePair> Nodes = (*I).second;
-
-    //Loop over nodes and determine the dependence between the other
-    //nodes in the vector
-    for(unsigned i =0; i < Nodes.size(); ++i) {
-
-      //Get src node operator index that uses this machine register
-      int srcOpIndex = Nodes[i].first;
-
-      //Get the actual src Node
-      MSchedGraphSBNode *srcNode = Nodes[i].second;
-
-      //Get Operand
-      const MachineOperand &srcMOp = srcNode->getInst()->getOperand(srcOpIndex);
-
-      bool srcIsUseandDef = srcMOp.isDef() && srcMOp.isUse();
-      bool srcIsUse = srcMOp.isUse() && !srcMOp.isDef();
-
-
-      //Look at all instructions after this in execution order
-      for(unsigned j=i+1; j < Nodes.size(); ++j) {
-
-        //Sink node is a write
-        if(Nodes[j].second->getInst()->getOperand(Nodes[j].first).isDef()) {
-                      //Src only uses the register (read)
-            if(srcIsUse)
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphSBEdge::MachineRegister,
-                                  MSchedGraphSBEdge::AntiDep);
-
-            else if(srcIsUseandDef) {
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphSBEdge::MachineRegister,
-                                  MSchedGraphSBEdge::AntiDep);
-
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphSBEdge::MachineRegister,
-                                  MSchedGraphSBEdge::OutputDep);
-            }
-            else
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphSBEdge::MachineRegister,
-                                  MSchedGraphSBEdge::OutputDep);
-        }
-        //Dest node is a read
-        else {
-          if(!srcIsUse || srcIsUseandDef)
-            srcNode->addOutEdge(Nodes[j].second,
-                                MSchedGraphSBEdge::MachineRegister,
-                                MSchedGraphSBEdge::TrueDep);
-        }
-
-      }
-
-      //Look at all the instructions before this one since machine registers
-      //could live across iterations.
-      for(unsigned j = 0; j < i; ++j) {
-                //Sink node is a write
-        if(Nodes[j].second->getInst()->getOperand(Nodes[j].first).isDef()) {
-                      //Src only uses the register (read)
-            if(srcIsUse)
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphSBEdge::MachineRegister,
-                                  MSchedGraphSBEdge::AntiDep, 1);
-            else if(srcIsUseandDef) {
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphSBEdge::MachineRegister,
-                                  MSchedGraphSBEdge::AntiDep, 1);
-
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphSBEdge::MachineRegister,
-                                  MSchedGraphSBEdge::OutputDep, 1);
-            }
-            else
-              srcNode->addOutEdge(Nodes[j].second,
-                                  MSchedGraphSBEdge::MachineRegister,
-                                  MSchedGraphSBEdge::OutputDep, 1);
-        }
-        //Dest node is a read
-        else {
-          if(!srcIsUse || srcIsUseandDef)
-            srcNode->addOutEdge(Nodes[j].second,
-                                MSchedGraphSBEdge::MachineRegister,
-                                MSchedGraphSBEdge::TrueDep,1 );
-        }
-
-
-      }
-
-    }
-
-  }
-
-}
-
-//Add edges between all loads and stores
-//Can be less strict with alias analysis and data dependence analysis.
-void MSchedGraphSB::addMemEdges(const std::vector<MSchedGraphSBNode*>& memInst,
-                      DependenceAnalyzer &DA,
-                      std::map<MachineInstr*, Instruction*> &machineTollvm) {
-
-  //Get Target machine instruction info
-  const TargetInstrInfo *TMI = Target.getInstrInfo();
-
-  //Loop over all memory instructions in the vector
-  //Knowing that they are in execution, add true, anti, and output dependencies
-  for (unsigned srcIndex = 0; srcIndex < memInst.size(); ++srcIndex) {
-
-    MachineInstr *srcInst = (MachineInstr*) memInst[srcIndex]->getInst();
-
-    //Get the machine opCode to determine type of memory instruction
-    MachineOpCode srcNodeOpCode = srcInst->getOpcode();
-
-    //All instructions after this one in execution order have an
-    //iteration delay of 0
-    for(unsigned destIndex = 0; destIndex < memInst.size(); ++destIndex) {
-
-      //No self loops
-      if(destIndex == srcIndex)
-        continue;
-
-      MachineInstr *destInst = (MachineInstr*) memInst[destIndex]->getInst();
-
-      DEBUG(std::cerr << "MInst1: " << *srcInst << "\n");
-      DEBUG(std::cerr << "MInst2: " << *destInst << "\n");
-
-      //Assuming instructions without corresponding llvm instructions
-      //are from constant pools.
-      if (!machineTollvm.count(srcInst) || !machineTollvm.count(destInst))
-        continue;
-
-      bool useDepAnalyzer = true;
-
-      //Some machine loads and stores are generated by casts, so be
-      //conservative and always add deps
-      Instruction *srcLLVM = machineTollvm[srcInst];
-      Instruction *destLLVM = machineTollvm[destInst];
-      if(!isa<LoadInst>(srcLLVM)
-         && !isa<StoreInst>(srcLLVM)) {
-        if(isa<BinaryOperator>(srcLLVM)) {
-          if(isa<ConstantFP>(srcLLVM->getOperand(0)) || isa<ConstantFP>(srcLLVM->getOperand(1)))
-            continue;
-        }
-        useDepAnalyzer = false;
-      }
-      if(!isa<LoadInst>(destLLVM)
-         && !isa<StoreInst>(destLLVM)) {
-        if(isa<BinaryOperator>(destLLVM)) {
-          if(isa<ConstantFP>(destLLVM->getOperand(0)) || isa<ConstantFP>(destLLVM->getOperand(1)))
-            continue;
-        }
-        useDepAnalyzer = false;
-      }
-
-      //Use dep analysis when we have corresponding llvm loads/stores
-      if(useDepAnalyzer) {
-        bool srcBeforeDest = true;
-        if(destIndex < srcIndex)
-          srcBeforeDest = false;
-
-        DependenceResult dr = DA.getDependenceInfo(machineTollvm[srcInst],
-                                                   machineTollvm[destInst],
-                                                   srcBeforeDest);
-
-        for(std::vector<Dependence>::iterator d = dr.dependences.begin(),
-              de = dr.dependences.end(); d != de; ++d) {
-          //Add edge from load to store
-          memInst[srcIndex]->addOutEdge(memInst[destIndex],
-                                        MSchedGraphSBEdge::MemoryDep,
-                                        d->getDepType(), d->getIteDiff());
-
-        }
-      }
-      //Otherwise, we can not do any further analysis and must make a dependence
-      else {
-
-        //Get the machine opCode to determine type of memory instruction
-        MachineOpCode destNodeOpCode = destInst->getOpcode();
-
-        //Get the Value* that we are reading from the load, always the first op
-        const MachineOperand &mOp = srcInst->getOperand(0);
-        const MachineOperand &mOp2 = destInst->getOperand(0);
-
-        if(mOp.hasAllocatedReg())
-          if(mOp.getReg() == SparcV9::g0)
-            continue;
-        if(mOp2.hasAllocatedReg())
-          if(mOp2.getReg() == SparcV9::g0)
-            continue;
-
-        DEBUG(std::cerr << "Adding dependence for machine instructions\n");
-        //Load-Store deps
-        if(TMI->isLoad(srcNodeOpCode)) {
-
-          if(TMI->isStore(destNodeOpCode))
-            memInst[srcIndex]->addOutEdge(memInst[destIndex],
-                                          MSchedGraphSBEdge::MemoryDep,
-                                          MSchedGraphSBEdge::AntiDep, 0);
-        }
-        else if(TMI->isStore(srcNodeOpCode)) {
-          if(TMI->isStore(destNodeOpCode))
-            memInst[srcIndex]->addOutEdge(memInst[destIndex],
-                                          MSchedGraphSBEdge::MemoryDep,
-                                          MSchedGraphSBEdge::OutputDep, 0);
-
-          else
-            memInst[srcIndex]->addOutEdge(memInst[destIndex],
-                                          MSchedGraphSBEdge::MemoryDep,
-                                          MSchedGraphSBEdge::TrueDep, 0);
-        }
-      }
-    }
-  }
-}
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraphSB.h b/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraphSB.h
deleted file mode 100644
index 2e43f7ed6f6..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/MSchedGraphSB.h
+++ /dev/null
@@ -1,410 +0,0 @@
-//===-- MSchedGraphSB.h - Scheduling Graph ------------------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-// A graph class for dependencies. This graph only contains true, anti, and
-// output data dependencies for a vector of MachineBasicBlock. Dependencies
-// across iterations are also computed. Unless data dependence analysis
-// is provided, a conservative approach of adding dependencies between all
-// loads and stores is taken. It also includes pseudo predicate nodes for
-// modulo scheduling superblocks.
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_MSCHEDGRAPHSB_H
-#define LLVM_MSCHEDGRAPHSB_H
-#include "DependenceAnalyzer.h"
-#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/CodeGen/MachineInstr.h"
-#include "llvm/Target/TargetMachine.h"
-#include "llvm/Target/TargetData.h"
-#include "llvm/ADT/GraphTraits.h"
-#include "llvm/ADT/STLExtras.h"
-#include "llvm/ADT/iterator"
-#include <vector>
-
-
-namespace llvm {
-
-  class MSchedGraphSB;
-  class MSchedGraphSBNode;
-  template<class IteratorType, class NodeType>
-  class MSchedGraphSBNodeIterator;
-
-  //MSchedGraphSBEdge encapsulates the data dependence between nodes. It
-  //identifies the dependence type, on what, and the iteration
-  //difference
-  struct MSchedGraphSBEdge {
-    enum DataDepOrderType {
-      TrueDep, AntiDep, OutputDep, NonDataDep
-    };
-
-    enum MSchedGraphSBEdgeType {
-      MemoryDep, ValueDep, MachineRegister, PredDep
-    };
-
-    //Get or set edge data
-    MSchedGraphSBNode *getDest() const { return dest; }
-    unsigned getIteDiff() { return iteDiff; }
-    unsigned getDepOrderType() { return depOrderType; }
-    void setDest(MSchedGraphSBNode *newDest) { dest = newDest; }
-
-  private:
-    friend class MSchedGraphSBNode;
-    MSchedGraphSBEdge(MSchedGraphSBNode *destination, MSchedGraphSBEdgeType type,
-                    unsigned deptype, unsigned diff)
-      : dest(destination), depType(type), depOrderType(deptype), iteDiff(diff) {}
-
-    MSchedGraphSBNode *dest;
-    MSchedGraphSBEdgeType depType;
-    unsigned depOrderType;
-    unsigned iteDiff;
-  };
-
-  //MSchedGraphSBNode represents a machine instruction and its
-  //corresponding latency. Each node also contains a list of its
-  //predecessors and sucessors.
-  class MSchedGraphSBNode {
-
-    const MachineInstr* Inst; //Machine Instruction
-    std::vector<const MachineInstr*> otherInstrs;
-
-    MSchedGraphSB* Parent; //Graph this node belongs to
-    unsigned index; //Index in BB
-    unsigned latency; //Latency of Instruction
-    bool isBranchInstr; //Is this node the branch instr or not
-    bool isPredicateNode; //Indicate if this node should be treated like a predicate
-
-    std::vector<MSchedGraphSBNode*> Predecessors; //Predecessor Nodes
-    std::vector<MSchedGraphSBEdge> Successors; //Successor edges
-
-  public:
-    MSchedGraphSBNode(const MachineInstr* inst, MSchedGraphSB *graph,
-                    unsigned index, unsigned late=0, bool isBranch=false);
-    MSchedGraphSBNode(const MachineInstr* inst, std::vector<const MachineInstr*> &other,
-                      MSchedGraphSB *graph,
-                      unsigned index, unsigned late=0, bool isPNode=true);
-    MSchedGraphSBNode(const MSchedGraphSBNode &N);
-
-    //Iterators - Predecessor and Succussor
-    typedef std::vector<MSchedGraphSBNode*>::iterator pred_iterator;
-    pred_iterator pred_begin() { return Predecessors.begin(); }
-    pred_iterator pred_end() { return Predecessors.end(); }
-    unsigned pred_size() { return Predecessors.size(); }
-
-    typedef std::vector<MSchedGraphSBNode*>::const_iterator pred_const_iterator;
-    pred_const_iterator pred_begin() const { return Predecessors.begin(); }
-    pred_const_iterator pred_end() const { return Predecessors.end(); }
-
-    typedef MSchedGraphSBNodeIterator<std::vector<MSchedGraphSBEdge>::const_iterator,
-                                    const MSchedGraphSBNode> succ_const_iterator;
-    succ_const_iterator succ_begin() const;
-    succ_const_iterator succ_end() const;
-
-    typedef MSchedGraphSBNodeIterator<std::vector<MSchedGraphSBEdge>::iterator,
-                                    MSchedGraphSBNode> succ_iterator;
-    succ_iterator succ_begin();
-    succ_iterator succ_end();
-    unsigned succ_size() { return Successors.size(); }
-
-    //Get or set predecessor nodes, or successor edges
-    void setPredecessor(unsigned index, MSchedGraphSBNode *dest) {
-      Predecessors[index] = dest;
-    }
-
-    MSchedGraphSBNode* getPredecessor(unsigned index) {
-      return Predecessors[index];
-    }
-
-    MSchedGraphSBEdge* getSuccessor(unsigned index) {
-      return &Successors[index];
-    }
-
-    void deleteSuccessor(MSchedGraphSBNode *node) {
-      for (unsigned i = 0; i != Successors.size(); ++i)
-        if (Successors[i].getDest() == node) {
-          Successors.erase(Successors.begin()+i);
-          node->Predecessors.erase(std::find(node->Predecessors.begin(),
-                                             node->Predecessors.end(), this));
-          --i; //Decrease index var since we deleted a node
-        }
-    }
-
-    void addOutEdge(MSchedGraphSBNode *destination,
-                    MSchedGraphSBEdge::MSchedGraphSBEdgeType type,
-                    unsigned deptype, unsigned diff=0) {
-      Successors.push_back(MSchedGraphSBEdge(destination, type, deptype,diff));
-      destination->Predecessors.push_back(this);
-    }
-
-    //General methods to get and set data for the node
-    const MachineInstr* getInst() { return Inst; }
-    MSchedGraphSB* getParent() { return Parent; }
-    bool hasPredecessors() { return (Predecessors.size() > 0); }
-    bool hasSuccessors() { return (Successors.size() > 0); }
-    unsigned getLatency() { return latency; }
-    unsigned getLatency() const { return latency; }
-    unsigned getIndex() { return index; }
-    unsigned getIteDiff(MSchedGraphSBNode *succ);
-    MSchedGraphSBEdge getInEdge(MSchedGraphSBNode *pred);
-    unsigned getInEdgeNum(MSchedGraphSBNode *pred);
-    bool isSuccessor(MSchedGraphSBNode *);
-    bool isPredecessor(MSchedGraphSBNode *);
-    bool isBranch() { return isBranchInstr; }
-    bool isPredicate() { return isPredicateNode; }
-    bool isPredicate() const { return isPredicateNode; }
-    std::vector<const MachineInstr*> getOtherInstrs() { return otherInstrs; }
-
-    //Debug support
-    void print(std::ostream &os) const;
-    void setParent(MSchedGraphSB *p) { Parent = p; }
-  };
-
-  //Node iterator for graph generation
-  template<class IteratorType, class NodeType>
-  class MSchedGraphSBNodeIterator : public forward_iterator<NodeType*, ptrdiff_t> {
-    IteratorType I;   // std::vector<MSchedGraphSBEdge>::iterator or const_iterator
-  public:
-    MSchedGraphSBNodeIterator(IteratorType i) : I(i) {}
-
-    bool operator==(const MSchedGraphSBNodeIterator RHS) const { return I == RHS.I; }
-    bool operator!=(const MSchedGraphSBNodeIterator RHS) const { return I != RHS.I; }
-
-    const MSchedGraphSBNodeIterator &operator=(const MSchedGraphSBNodeIterator &RHS) {
-      I = RHS.I;
-      return *this;
-    }
-
-    NodeType* operator*() const {
-      return I->getDest();
-    }
-    NodeType* operator->() const { return operator*(); }
-
-    MSchedGraphSBNodeIterator& operator++() {                // Preincrement
-      ++I;
-      return *this;
-    }
-    MSchedGraphSBNodeIterator operator++(int) { // Postincrement
-      MSchedGraphSBNodeIterator tmp = *this; ++*this; return tmp;
-    }
-
-    MSchedGraphSBEdge &getEdge() {
-      return *I;
-    }
-    const MSchedGraphSBEdge &getEdge() const {
-      return *I;
-    }
-  };
-
-  inline MSchedGraphSBNode::succ_const_iterator MSchedGraphSBNode::succ_begin() const {
-    return succ_const_iterator(Successors.begin());
-  }
-  inline MSchedGraphSBNode::succ_const_iterator MSchedGraphSBNode::succ_end() const {
-    return succ_const_iterator(Successors.end());
-  }
-  inline MSchedGraphSBNode::succ_iterator MSchedGraphSBNode::succ_begin() {
-    return succ_iterator(Successors.begin());
-  }
-  inline MSchedGraphSBNode::succ_iterator MSchedGraphSBNode::succ_end() {
-    return succ_iterator(Successors.end());
-  }
-
-  // ostream << operator for MSGraphNode class
-  inline std::ostream &operator<<(std::ostream &os,
-                                  const MSchedGraphSBNode &node) {
-    node.print(os);
-    return os;
-  }
-
-
-  // Provide specializations of GraphTraits to be able to use graph
-  // iterators on the scheduling graph!
-  //
-  template <> struct GraphTraits<MSchedGraphSBNode*> {
-    typedef MSchedGraphSBNode NodeType;
-    typedef MSchedGraphSBNode::succ_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->succ_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->succ_end();
-    }
-
-    static NodeType *getEntryNode(NodeType* N) { return N; }
-  };
-
-
-
-  //Graph class to represent dependence graph
-  class MSchedGraphSB {
-
-    std::vector<const MachineBasicBlock *> BBs; //Machine basic block
-    const TargetMachine &Target; //Target Machine
-
-    //Nodes
-    std::map<const MachineInstr*, MSchedGraphSBNode*> GraphMap;
-
-    //Add Nodes and Edges to this graph for our BB
-    typedef std::pair<int, MSchedGraphSBNode*> OpIndexNodePair;
-    void buildNodesAndEdges(std::map<const MachineInstr*, unsigned> &ignoreInstrs, DependenceAnalyzer &DA, std::map<MachineInstr*, Instruction*> &machineTollvm, std::map<MSchedGraphSBNode*, std::set<MachineInstr*> > &liveOutsideTrace);
-    void addValueEdges(std::vector<OpIndexNodePair> &NodesInMap,
-                       MSchedGraphSBNode *node,
-                       bool nodeIsUse, bool nodeIsDef, std::vector<const MachineInstr*> &phiInstrs, int diff=0);
-    void addMachRegEdges(std::map<int,
-                         std::vector<OpIndexNodePair> >& regNumtoNodeMap);
-    void addMemEdges(const std::vector<MSchedGraphSBNode*>& memInst,
-                     DependenceAnalyzer &DA, std::map<MachineInstr*, Instruction*> &machineTollvm);
-
-
-    bool instrCauseException(MachineOpCode opCode);
-
-  public:
-    MSchedGraphSB(const MachineBasicBlock *bb, const TargetMachine &targ,
-                std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                DependenceAnalyzer &DA, std::map<MachineInstr*, Instruction*> &machineTollvm);
-
-    //Copy constructor with maps to link old nodes to new nodes
-    MSchedGraphSB(const MSchedGraphSB &G, std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> &newNodes);
-
-    MSchedGraphSB(std::vector<const MachineBasicBlock*> &bbs,
-                const TargetMachine &targ,
-                std::map<const MachineInstr*, unsigned> &ignoreInstrs,
-                DependenceAnalyzer &DA,
-                std::map<MachineInstr*, Instruction*> &machineTollvm);
-
-    //Print graph
-    void print(std::ostream &os) const;
-
-    //Deconstructor!
-    ~MSchedGraphSB();
-
-    //Add or delete nodes from the Graph
-    void addNode(const MachineInstr* MI, MSchedGraphSBNode *node);
-    void deleteNode(MSchedGraphSBNode *node);
-    int totalDelay();
-
-    //iterators
-    typedef std::map<const MachineInstr*, MSchedGraphSBNode*>::iterator iterator;
-    typedef std::map<const MachineInstr*, MSchedGraphSBNode*>::const_iterator const_iterator;
-    typedef std::map<const MachineInstr*, MSchedGraphSBNode*>::reverse_iterator reverse_iterator;
-    iterator find(const MachineInstr* I) { return GraphMap.find(I); }
-    iterator end() { return GraphMap.end(); }
-    iterator begin() { return GraphMap.begin(); }
-    unsigned size() { return GraphMap.size(); }
-    reverse_iterator rbegin() { return GraphMap.rbegin(); }
-    reverse_iterator rend() { return GraphMap.rend(); }
-
-    //Get Target or original machine basic block
-    const TargetMachine* getTarget() { return &Target; }
-    std::vector<const MachineBasicBlock*> getBBs() { return BBs; }
-  };
-
-
-
-
-
-  // Provide specializations of GraphTraits to be able to use graph
-  // iterators on the scheduling graph
-  static MSchedGraphSBNode& getSecond(std::pair<const MachineInstr* const,
-                                    MSchedGraphSBNode*> &Pair) {
-    return *Pair.second;
-  }
-
-  template <> struct GraphTraits<MSchedGraphSB*> {
-    typedef MSchedGraphSBNode NodeType;
-    typedef MSchedGraphSBNode::succ_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->succ_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->succ_end();
-    }
-
-    typedef std::pointer_to_unary_function<std::pair<const MachineInstr* const,
-           MSchedGraphSBNode*>&, MSchedGraphSBNode&> DerefFun;
-
-    typedef mapped_iterator<MSchedGraphSB::iterator, DerefFun> nodes_iterator;
-    static nodes_iterator nodes_begin(MSchedGraphSB *G) {
-      return map_iterator(((MSchedGraphSB*)G)->begin(), DerefFun(getSecond));
-    }
-    static nodes_iterator nodes_end(MSchedGraphSB *G) {
-      return map_iterator(((MSchedGraphSB*)G)->end(), DerefFun(getSecond));
-    }
-
-  };
-
-  template <> struct GraphTraits<const MSchedGraphSB*> {
-    typedef const MSchedGraphSBNode NodeType;
-    typedef MSchedGraphSBNode::succ_const_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->succ_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->succ_end();
-    }
-    typedef std::pointer_to_unary_function<std::pair<const MachineInstr* const,
-                                                     MSchedGraphSBNode*>&, MSchedGraphSBNode&> DerefFun;
-
-    typedef mapped_iterator<MSchedGraphSB::iterator, DerefFun> nodes_iterator;
-    static nodes_iterator nodes_begin(MSchedGraphSB *G) {
-      return map_iterator(((MSchedGraphSB*)G)->begin(), DerefFun(getSecond));
-    }
-    static nodes_iterator nodes_end(MSchedGraphSB *G) {
-      return map_iterator(((MSchedGraphSB*)G)->end(), DerefFun(getSecond));
-    }
-  };
-
-  template <> struct GraphTraits<Inverse<MSchedGraphSB*> > {
-    typedef MSchedGraphSBNode NodeType;
-    typedef MSchedGraphSBNode::pred_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->pred_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->pred_end();
-    }
-    typedef std::pointer_to_unary_function<std::pair<const MachineInstr* const,
-           MSchedGraphSBNode*>&, MSchedGraphSBNode&> DerefFun;
-
-    typedef mapped_iterator<MSchedGraphSB::iterator, DerefFun> nodes_iterator;
-    static nodes_iterator nodes_begin(MSchedGraphSB *G) {
-      return map_iterator(((MSchedGraphSB*)G)->begin(), DerefFun(getSecond));
-    }
-    static nodes_iterator nodes_end(MSchedGraphSB *G) {
-      return map_iterator(((MSchedGraphSB*)G)->end(), DerefFun(getSecond));
-    }
-  };
-
-  template <> struct GraphTraits<Inverse<const MSchedGraphSB*> > {
-    typedef const MSchedGraphSBNode NodeType;
-    typedef MSchedGraphSBNode::pred_const_iterator ChildIteratorType;
-
-    static inline ChildIteratorType child_begin(NodeType *N) {
-      return N->pred_begin();
-    }
-    static inline ChildIteratorType child_end(NodeType *N) {
-      return N->pred_end();
-    }
-
-    typedef std::pointer_to_unary_function<std::pair<const MachineInstr* const,
-                                                     MSchedGraphSBNode*>&, MSchedGraphSBNode&> DerefFun;
-
-    typedef mapped_iterator<MSchedGraphSB::iterator, DerefFun> nodes_iterator;
-    static nodes_iterator nodes_begin(MSchedGraphSB *G) {
-      return map_iterator(((MSchedGraphSB*)G)->begin(), DerefFun(getSecond));
-    }
-    static nodes_iterator nodes_end(MSchedGraphSB *G) {
-      return map_iterator(((MSchedGraphSB*)G)->end(), DerefFun(getSecond));
-    }
-  };
-}
-
-#endif
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/Makefile b/llvm/lib/Target/SparcV9/ModuloScheduling/Makefile
deleted file mode 100644
index 2ec0503a173..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/Makefile
+++ /dev/null
@@ -1,14 +0,0 @@
-##===- lib/Target/SparcV9/ModuloScheduling/Makefile --------*- Makefile -*-===##
-# 
-#                     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.
-# 
-##===----------------------------------------------------------------------===##
-
-LEVEL = ../../../..
-DIRS  = 
-LIBRARYNAME = LLVMSparcV9ModuloSched
-
-include $(LEVEL)/Makefile.common
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloScheduling.cpp b/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloScheduling.cpp
deleted file mode 100644
index a5e9661f1cd..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloScheduling.cpp
+++ /dev/null
@@ -1,2964 +0,0 @@
-//===-- ModuloScheduling.cpp - ModuloScheduling  ----------------*- C++ -*-===//
-//
-//                     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 ModuloScheduling pass is based on the Swing Modulo Scheduling
-//  algorithm.
-//
-//===----------------------------------------------------------------------===//
-
-#define DEBUG_TYPE "ModuloSched"
-
-#include "ModuloScheduling.h"
-#include "llvm/Constants.h"
-#include "llvm/Instructions.h"
-#include "llvm/Function.h"
-#include "llvm/CodeGen/MachineFunction.h"
-#include "llvm/CodeGen/Passes.h"
-#include "llvm/Support/CFG.h"
-#include "llvm/Target/TargetSchedInfo.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/Support/GraphWriter.h"
-#include "llvm/ADT/SCCIterator.h"
-#include "llvm/ADT/StringExtras.h"
-#include "llvm/ADT/Statistic.h"
-#include "llvm/Support/Timer.h"
-#include <cmath>
-#include <algorithm>
-#include <fstream>
-#include <sstream>
-#include <utility>
-#include <vector>
-#include "../MachineCodeForInstruction.h"
-#include "../SparcV9TmpInstr.h"
-#include "../SparcV9Internals.h"
-#include "../SparcV9RegisterInfo.h"
-using namespace llvm;
-
-/// Create ModuloSchedulingPass
-///
-FunctionPass *llvm::createModuloSchedulingPass(TargetMachine & targ) {
-  DEBUG(std::cerr << "Created ModuloSchedulingPass\n");
-  return new ModuloSchedulingPass(targ);
-}
-
-
-//Graph Traits for printing out the dependence graph
-template<typename GraphType>
-static void WriteGraphToFile(std::ostream &O, const std::string &GraphName,
-                             const GraphType &GT) {
-  std::string Filename = GraphName + ".dot";
-  O << "Writing '" << Filename << "'...";
-  std::ofstream F(Filename.c_str());
-
-  if (F.good())
-    WriteGraph(F, GT);
-  else
-    O << "  error opening file for writing!";
-  O << "\n";
-};
-
-
-#if 1
-#define TIME_REGION(VARNAME, DESC) \
-   NamedRegionTimer VARNAME(DESC)
-#else
-#define TIME_REGION(VARNAME, DESC)
-#endif
-
-
-//Graph Traits for printing out the dependence graph
-namespace llvm {
-
-  //Loop statistics
-  Statistic<> ValidLoops("modulosched-validLoops", "Number of candidate loops modulo-scheduled");
-  Statistic<> JumboBB("modulosched-jumboBB", "Basic Blocks with more then 100 instructions");
-  Statistic<> LoopsWithCalls("modulosched-loopCalls", "Loops with calls");
-  Statistic<> LoopsWithCondMov("modulosched-loopCondMov", "Loops with conditional moves");
-  Statistic<> InvalidLoops("modulosched-invalidLoops", "Loops with unknown trip counts or loop invariant trip counts");
-  Statistic<> SingleBBLoops("modulosched-singeBBLoops", "Number of single basic block loops");
-
-  //Scheduling Statistics
-  Statistic<> MSLoops("modulosched-schedLoops", "Number of loops successfully modulo-scheduled");
-  Statistic<> NoSched("modulosched-noSched", "No schedule");
-  Statistic<> SameStage("modulosched-sameStage", "Max stage is 0");
-  Statistic<> ResourceConstraint("modulosched-resourceConstraint", "Loops constrained by resources");
-  Statistic<> RecurrenceConstraint("modulosched-recurrenceConstraint", "Loops constrained by recurrences");
-   Statistic<> FinalIISum("modulosched-finalIISum", "Sum of all final II");
-  Statistic<> IISum("modulosched-IISum", "Sum of all theoretical II");
-
-  template<>
-  struct DOTGraphTraits<MSchedGraph*> : public DefaultDOTGraphTraits {
-    static std::string getGraphName(MSchedGraph *F) {
-      return "Dependence Graph";
-    }
-
-    static std::string getNodeLabel(MSchedGraphNode *Node, MSchedGraph *Graph) {
-      if (Node->getInst()) {
-        std::stringstream ss;
-        ss << *(Node->getInst());
-        return ss.str(); //((MachineInstr*)Node->getInst());
-      }
-      else
-        return "No Inst";
-    }
-    static std::string getEdgeSourceLabel(MSchedGraphNode *Node,
-                                          MSchedGraphNode::succ_iterator I) {
-      //Label each edge with the type of dependence
-      std::string edgelabel = "";
-      switch (I.getEdge().getDepOrderType()) {
-
-      case MSchedGraphEdge::TrueDep:
-        edgelabel = "True";
-        break;
-
-      case MSchedGraphEdge::AntiDep:
-        edgelabel =  "Anti";
-        break;
-
-      case MSchedGraphEdge::OutputDep:
-        edgelabel = "Output";
-        break;
-
-      default:
-        edgelabel = "Unknown";
-        break;
-      }
-
-      //FIXME
-      int iteDiff = I.getEdge().getIteDiff();
-      std::string intStr = "(IteDiff: ";
-      intStr += itostr(iteDiff);
-
-      intStr += ")";
-      edgelabel += intStr;
-
-      return edgelabel;
-    }
-  };
-}
-
-
-#include <unistd.h>
-
-/// ModuloScheduling::runOnFunction - main transformation entry point
-/// The Swing Modulo Schedule algorithm has three basic steps:
-/// 1) Computation and Analysis of the dependence graph
-/// 2) Ordering of the nodes
-/// 3) Scheduling
-///
-bool ModuloSchedulingPass::runOnFunction(Function &F) {
-  alarm(100);
-
-  bool Changed = false;
-  int numMS = 0;
-
-  DEBUG(std::cerr << "Creating ModuloSchedGraph for each valid BasicBlock in " + F.getName() + "\n");
-
-  //Get MachineFunction
-  MachineFunction &MF = MachineFunction::get(&F);
-
-  DependenceAnalyzer &DA = getAnalysis<DependenceAnalyzer>();
-
-
-  //Worklist
-  std::vector<MachineBasicBlock*> Worklist;
-
-  //Iterate over BasicBlocks and put them into our worklist if they are valid
-  for (MachineFunction::iterator BI = MF.begin(); BI != MF.end(); ++BI)
-    if(MachineBBisValid(BI)) {
-      if(BI->size() < 100) {
-        Worklist.push_back(&*BI);
-        ++ValidLoops;
-      }
-      else
-        ++JumboBB;
-
-    }
-
-  defaultInst = 0;
-
-  DEBUG(if(Worklist.size() == 0) std::cerr << "No single basic block loops in function to ModuloSchedule\n");
-
-  //Iterate over the worklist and perform scheduling
-  for(std::vector<MachineBasicBlock*>::iterator BI = Worklist.begin(),
-        BE = Worklist.end(); BI != BE; ++BI) {
-
-    //Print out BB for debugging
-    DEBUG(std::cerr << "BB Size: " << (*BI)->size() << "\n");
-    DEBUG(std::cerr << "ModuloScheduling BB: \n"; (*BI)->print(std::cerr));
-
-    //Print out LLVM BB
-    DEBUG(std::cerr << "ModuloScheduling LLVMBB: \n"; (*BI)->getBasicBlock()->print(std::cerr));
-
-    //Catch the odd case where we only have TmpInstructions and no real Value*s
-    if(!CreateDefMap(*BI)) {
-      //Clear out our maps for the next basic block that is processed
-      nodeToAttributesMap.clear();
-      partialOrder.clear();
-      recurrenceList.clear();
-      FinalNodeOrder.clear();
-      schedule.clear();
-      defMap.clear();
-      continue;
-    }
-
-    MSchedGraph *MSG = new MSchedGraph(*BI, target, indVarInstrs[*BI], DA, machineTollvm[*BI]);
-
-    //Write Graph out to file
-    DEBUG(WriteGraphToFile(std::cerr, F.getName(), MSG));
-    DEBUG(MSG->print(std::cerr));
-
-    //Calculate Resource II
-    int ResMII = calculateResMII(*BI);
-
-    //Calculate Recurrence II
-    int RecMII = calculateRecMII(MSG, ResMII);
-
-    DEBUG(std::cerr << "Number of reccurrences found: " << recurrenceList.size() << "\n");
-
-    //Our starting initiation interval is the maximum of RecMII and ResMII
-    if(RecMII < ResMII)
-      ++RecurrenceConstraint;
-    else
-      ++ResourceConstraint;
-
-    II = std::max(RecMII, ResMII);
-    int mII = II;
-
-    //Print out II, RecMII, and ResMII
-    DEBUG(std::cerr << "II starts out as " << II << " ( RecMII=" << RecMII << " and ResMII=" << ResMII << ")\n");
-
-    //Dump node properties if in debug mode
-    DEBUG(for(std::map<MSchedGraphNode*, MSNodeAttributes>::iterator I =  nodeToAttributesMap.begin(),
-                E = nodeToAttributesMap.end(); I !=E; ++I) {
-            std::cerr << "Node: " << *(I->first) << " ASAP: " << I->second.ASAP << " ALAP: "
-                      << I->second.ALAP << " MOB: " << I->second.MOB << " Depth: " << I->second.depth
-                      << " Height: " << I->second.height << "\n";
-          });
-
-    //Calculate Node Properties
-    calculateNodeAttributes(MSG, ResMII);
-
-    //Dump node properties if in debug mode
-    DEBUG(for(std::map<MSchedGraphNode*, MSNodeAttributes>::iterator I =  nodeToAttributesMap.begin(),
-                E = nodeToAttributesMap.end(); I !=E; ++I) {
-            std::cerr << "Node: " << *(I->first) << " ASAP: " << I->second.ASAP << " ALAP: "
-                      << I->second.ALAP << " MOB: " << I->second.MOB << " Depth: " << I->second.depth
-                      << " Height: " << I->second.height << "\n";
-          });
-
-    //Put nodes in order to schedule them
-    computePartialOrder();
-
-    //Dump out partial order
-    DEBUG(for(std::vector<std::set<MSchedGraphNode*> >::iterator I = partialOrder.begin(),
-                E = partialOrder.end(); I !=E; ++I) {
-            std::cerr << "Start set in PO\n";
-            for(std::set<MSchedGraphNode*>::iterator J = I->begin(), JE = I->end(); J != JE; ++J)
-              std::cerr << "PO:" << **J << "\n";
-          });
-
-    //Place nodes in final order
-    orderNodes();
-
-    //Dump out order of nodes
-    DEBUG(for(std::vector<MSchedGraphNode*>::iterator I = FinalNodeOrder.begin(), E = FinalNodeOrder.end(); I != E; ++I) {
-            std::cerr << "FO:" << **I << "\n";
-          });
-
-    //Finally schedule nodes
-    bool haveSched = computeSchedule(*BI, MSG);
-
-    //Print out final schedule
-    DEBUG(schedule.print(std::cerr));
-
-    //Final scheduling step is to reconstruct the loop only if we actual have
-    //stage > 0
-    if(haveSched) {
-      reconstructLoop(*BI);
-      ++MSLoops;
-      Changed = true;
-      FinalIISum += II;
-      IISum += mII;
-
-      if(schedule.getMaxStage() == 0)
-        ++SameStage;
-    }
-    else {
-      ++NoSched;
-    }
-
-    //Clear out our maps for the next basic block that is processed
-    nodeToAttributesMap.clear();
-    partialOrder.clear();
-    recurrenceList.clear();
-    FinalNodeOrder.clear();
-    schedule.clear();
-    defMap.clear();
-    //Clean up. Nuke old MachineBB and llvmBB
-    //BasicBlock *llvmBB = (BasicBlock*) (*BI)->getBasicBlock();
-    //Function *parent = (Function*) llvmBB->getParent();
-    //Should't std::find work??
-    //parent->getBasicBlockList().erase(std::find(parent->getBasicBlockList().begin(), parent->getBasicBlockList().end(), *llvmBB));
-    //parent->getBasicBlockList().erase(llvmBB);
-
-    //delete(llvmBB);
-    //delete(*BI);
-  }
-
-  alarm(0);
-  return Changed;
-}
-
-bool ModuloSchedulingPass::CreateDefMap(MachineBasicBlock *BI) {
-  defaultInst = 0;
-
-  for(MachineBasicBlock::iterator I = BI->begin(), E = BI->end(); I != E; ++I) {
-    for(unsigned opNum = 0; opNum < I->getNumOperands(); ++opNum) {
-      const MachineOperand &mOp = I->getOperand(opNum);
-      if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef()) {
-        //assert if this is the second def we have seen
-        //DEBUG(std::cerr << "Putting " << *(mOp.getVRegValue()) << " into map\n");
-        //assert(!defMap.count(mOp.getVRegValue()) && "Def already in the map");
-        if(defMap.count(mOp.getVRegValue()))
-          return false;
-
-        defMap[mOp.getVRegValue()] = &*I;
-      }
-
-      //See if we can use this Value* as our defaultInst
-      if(!defaultInst && mOp.getType() == MachineOperand::MO_VirtualRegister) {
-        Value *V = mOp.getVRegValue();
-        if(!isa<TmpInstruction>(V) && !isa<Argument>(V) && !isa<Constant>(V) && !isa<PHINode>(V))
-          defaultInst = (Instruction*) V;
-      }
-    }
-  }
-
-  if(!defaultInst)
-    return false;
-
-  return true;
-
-}
-/// This function checks if a Machine Basic Block is valid for modulo
-/// scheduling. This means that it has no control flow (if/else or
-/// calls) in the block.  Currently ModuloScheduling only works on
-/// single basic block loops.
-bool ModuloSchedulingPass::MachineBBisValid(const MachineBasicBlock *BI) {
-
-  bool isLoop = false;
-
-  //Check first if its a valid loop
-  for(succ_const_iterator I = succ_begin(BI->getBasicBlock()),
-        E = succ_end(BI->getBasicBlock()); I != E; ++I) {
-    if (*I == BI->getBasicBlock())    // has single block loop
-      isLoop = true;
-  }
-
-  if(!isLoop)
-    return false;
-
-  //Check that we have a conditional branch (avoiding MS infinite loops)
-  if(BranchInst *b = dyn_cast<BranchInst>(((BasicBlock*) BI->getBasicBlock())->getTerminator()))
-    if(b->isUnconditional())
-      return false;
-
-  //Check size of our basic block.. make sure we have more then just the terminator in it
-  if(BI->getBasicBlock()->size() == 1)
-    return false;
-
-  //Increase number of single basic block loops for stats
-  ++SingleBBLoops;
-
-  //Get Target machine instruction info
-  const TargetInstrInfo *TMI = target.getInstrInfo();
-
-  //Check each instruction and look for calls, keep map to get index later
-  std::map<const MachineInstr*, unsigned> indexMap;
-
-  unsigned count = 0;
-  for(MachineBasicBlock::const_iterator I = BI->begin(), E = BI->end(); I != E; ++I) {
-    //Get opcode to check instruction type
-    MachineOpCode OC = I->getOpcode();
-
-    //Look for calls
-    if(TMI->isCall(OC)) {
-      ++LoopsWithCalls;
-      return false;
-    }
-
-    //Look for conditional move
-    if(OC == V9::MOVRZr || OC == V9::MOVRZi || OC == V9::MOVRLEZr || OC == V9::MOVRLEZi
-       || OC == V9::MOVRLZr || OC == V9::MOVRLZi || OC == V9::MOVRNZr || OC == V9::MOVRNZi
-       || OC == V9::MOVRGZr || OC == V9::MOVRGZi || OC == V9::MOVRGEZr
-       || OC == V9::MOVRGEZi || OC == V9::MOVLEr || OC == V9::MOVLEi || OC == V9::MOVLEUr
-       || OC == V9::MOVLEUi || OC == V9::MOVFLEr || OC == V9::MOVFLEi
-       || OC == V9::MOVNEr || OC == V9::MOVNEi || OC == V9::MOVNEGr || OC == V9::MOVNEGi
-       || OC == V9::MOVFNEr || OC == V9::MOVFNEi || OC == V9::MOVGr || OC == V9::MOVGi) {
-      ++LoopsWithCondMov;
-      return false;
-    }
-
-    indexMap[I] = count;
-
-    if(TMI->isNop(OC))
-      continue;
-
-    ++count;
-  }
-
-  //Apply a simple pattern match to make sure this loop can be modulo scheduled
-  //This means only loops with a branch associated to the iteration count
-
-  //Get the branch
-  BranchInst *b = dyn_cast<BranchInst>(((BasicBlock*) BI->getBasicBlock())->getTerminator());
-
-  //Get the condition for the branch (we already checked if it was conditional)
-  Value *cond = b->getCondition();
-
-  DEBUG(std::cerr << "Condition: " << *cond << "\n");
-
-  //List of instructions associated with induction variable
-  std::set<Instruction*> indVar;
-  std::vector<Instruction*> stack;
-
-  BasicBlock *BB = (BasicBlock*) BI->getBasicBlock();
-
-  //Add branch
-  indVar.insert(b);
-
-  if(Instruction *I = dyn_cast<Instruction>(cond))
-    if(I->getParent() == BB) {
-      if (!assocIndVar(I, indVar, stack, BB)) {
-        ++InvalidLoops;
-        return false;
-      }
-    }
-    else {
-      ++InvalidLoops;
-      return false;
-    }
-  else {
-    ++InvalidLoops;
-    return false;
-  }
-  //The indVar set must be >= 3 instructions for this loop to match (FIX ME!)
-  if(indVar.size() < 3 )
-    return false;
-
-  //Dump out instructions associate with indvar for debug reasons
-  DEBUG(for(std::set<Instruction*>::iterator N = indVar.begin(), NE = indVar.end(); N != NE; ++N) {
-          std::cerr << **N << "\n";
-        });
-
-  //Create map of machine instr to llvm instr
-  std::map<MachineInstr*, Instruction*> mllvm;
-  for(BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
-    MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(I);
-    for (unsigned j = 0; j < tempMvec.size(); j++) {
-      mllvm[tempMvec[j]] = I;
-    }
-  }
-
-  //Convert list of LLVM Instructions to list of Machine instructions
-  std::map<const MachineInstr*, unsigned> mIndVar;
-  for(std::set<Instruction*>::iterator N = indVar.begin(), NE = indVar.end(); N != NE; ++N) {
-
-    //If we have a load, we can't handle this loop because there is no way to preserve dependences
-    //between loads and stores
-    if(isa<LoadInst>(*N))
-      return false;
-
-    MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(*N);
-    for (unsigned j = 0; j < tempMvec.size(); j++) {
-      MachineOpCode OC = (tempMvec[j])->getOpcode();
-      if(TMI->isNop(OC))
-        continue;
-      if(!indexMap.count(tempMvec[j]))
-        continue;
-      mIndVar[(MachineInstr*) tempMvec[j]] = indexMap[(MachineInstr*) tempMvec[j]];
-      DEBUG(std::cerr << *(tempMvec[j]) << " at index " << indexMap[(MachineInstr*) tempMvec[j]] << "\n");
-    }
-  }
-
-   //Must have some guts to the loop body (more then 1 instr, dont count nops in size)
-  if(mIndVar.size() >= (BI->size()-3))
-    return false;
-
-  //Put into a map for future access
-  indVarInstrs[BI] = mIndVar;
-  machineTollvm[BI] = mllvm;
-  return true;
-}
-
-bool ModuloSchedulingPass::assocIndVar(Instruction *I, std::set<Instruction*> &indVar,
-                                       std::vector<Instruction*> &stack, BasicBlock *BB) {
-
-  stack.push_back(I);
-
-  //If this is a phi node, check if its the canonical indvar
-  if(PHINode *PN = dyn_cast<PHINode>(I)) {
-    if (Instruction *Inc =
-        dyn_cast<Instruction>(PN->getIncomingValueForBlock(BB)))
-      if (Inc->getOpcode() == Instruction::Add && Inc->getOperand(0) == PN)
-        if (ConstantInt *CI = dyn_cast<ConstantInt>(Inc->getOperand(1)))
-          if (CI->equalsInt(1)) {
-            //We have found the indvar, so add the stack, and inc instruction to the set
-            indVar.insert(stack.begin(), stack.end());
-            indVar.insert(Inc);
-            stack.pop_back();
-            return true;
-          }
-    return false;
-  }
-  else {
-    //Loop over each of the instructions operands, check if they are an instruction and in this BB
-    for(unsigned i = 0; i < I->getNumOperands(); ++i) {
-      if(Instruction *N =  dyn_cast<Instruction>(I->getOperand(i))) {
-        if(N->getParent() == BB)
-          if(!assocIndVar(N, indVar, stack, BB))
-            return false;
-      }
-    }
-  }
-
-  stack.pop_back();
-  return true;
-}
-
-//ResMII is calculated by determining the usage count for each resource
-//and using the maximum.
-//FIXME: In future there should be a way to get alternative resources
-//for each instruction
-int ModuloSchedulingPass::calculateResMII(const MachineBasicBlock *BI) {
-
-  TIME_REGION(X, "calculateResMII");
-
-  const TargetInstrInfo *mii = target.getInstrInfo();
-  const TargetSchedInfo *msi = target.getSchedInfo();
-
-  int ResMII = 0;
-
-  //Map to keep track of usage count of each resource
-  std::map<unsigned, unsigned> resourceUsageCount;
-
-  for(MachineBasicBlock::const_iterator I = BI->begin(), E = BI->end(); I != E; ++I) {
-
-    //Get resource usage for this instruction
-    InstrRUsage rUsage = msi->getInstrRUsage(I->getOpcode());
-    std::vector<std::vector<resourceId_t> > resources = rUsage.resourcesByCycle;
-
-    //Loop over resources in each cycle and increments their usage count
-    for(unsigned i=0; i < resources.size(); ++i)
-      for(unsigned j=0; j < resources[i].size(); ++j) {
-        if(!resourceUsageCount.count(resources[i][j])) {
-          resourceUsageCount[resources[i][j]] = 1;
-        }
-        else {
-          resourceUsageCount[resources[i][j]] =  resourceUsageCount[resources[i][j]] + 1;
-        }
-      }
-  }
-
-  //Find maximum usage count
-
-  //Get max number of instructions that can be issued at once. (FIXME)
-  int issueSlots = msi->maxNumIssueTotal;
-
-  for(std::map<unsigned,unsigned>::iterator RB = resourceUsageCount.begin(), RE = resourceUsageCount.end(); RB != RE; ++RB) {
-
-    //Get the total number of the resources in our cpu
-    int resourceNum = CPUResource::getCPUResource(RB->first)->maxNumUsers;
-
-    //Get total usage count for this resources
-    unsigned usageCount = RB->second;
-
-    //Divide the usage count by either the max number we can issue or the number of
-    //resources (whichever is its upper bound)
-    double finalUsageCount;
-    DEBUG(std::cerr << "Resource Num: " << RB->first << " Usage: " << usageCount << " TotalNum: " << resourceNum << "\n");
-
-    if( resourceNum <= issueSlots)
-      finalUsageCount = ceil(1.0 * usageCount / resourceNum);
-    else
-      finalUsageCount = ceil(1.0 * usageCount / issueSlots);
-
-
-    //Only keep track of the max
-    ResMII = std::max( (int) finalUsageCount, ResMII);
-
-  }
-
-  return ResMII;
-
-}
-
-/// calculateRecMII - Calculates the value of the highest recurrence
-/// By value we mean the total latency
-int ModuloSchedulingPass::calculateRecMII(MSchedGraph *graph, int MII) {
-  /*std::vector<MSchedGraphNode*> vNodes;
-  //Loop over all nodes in the graph
-  for(MSchedGraph::iterator I = graph->begin(), E = graph->end(); I != E; ++I) {
-    findAllReccurrences(I->second, vNodes, MII);
-    vNodes.clear();
-  }*/
-
-  TIME_REGION(X, "calculateRecMII");
-
-  findAllCircuits(graph, MII);
-  int RecMII = 0;
-
- for(std::set<std::pair<int, std::vector<MSchedGraphNode*> > >::iterator I = recurrenceList.begin(), E=recurrenceList.end(); I !=E; ++I) {
-    RecMII = std::max(RecMII, I->first);
-  }
-
-  return MII;
-}
-
-/// calculateNodeAttributes - The following properties are calculated for
-/// each node in the dependence graph: ASAP, ALAP, Depth, Height, and
-/// MOB.
-void ModuloSchedulingPass::calculateNodeAttributes(MSchedGraph *graph, int MII) {
-
-  TIME_REGION(X, "calculateNodeAttributes");
-
-  assert(nodeToAttributesMap.empty() && "Node attribute map was not cleared");
-
-  //Loop over the nodes and add them to the map
-  for(MSchedGraph::iterator I = graph->begin(), E = graph->end(); I != E; ++I) {
-
-    DEBUG(std::cerr << "Inserting node into attribute map: " << *I->second << "\n");
-
-    //Assert if its already in the map
-    assert(nodeToAttributesMap.count(I->second) == 0 &&
-           "Node attributes are already in the map");
-
-    //Put into the map with default attribute values
-    nodeToAttributesMap[I->second] = MSNodeAttributes();
-  }
-
-  //Create set to deal with reccurrences
-  std::set<MSchedGraphNode*> visitedNodes;
-
-  //Now Loop over map and calculate the node attributes
-  for(std::map<MSchedGraphNode*, MSNodeAttributes>::iterator I = nodeToAttributesMap.begin(), E = nodeToAttributesMap.end(); I != E; ++I) {
-    calculateASAP(I->first, MII, (MSchedGraphNode*) 0);
-    visitedNodes.clear();
-  }
-
-  int maxASAP = findMaxASAP();
-  //Calculate ALAP which depends on ASAP being totally calculated
-  for(std::map<MSchedGraphNode*, MSNodeAttributes>::iterator I = nodeToAttributesMap.begin(), E = nodeToAttributesMap.end(); I != E; ++I) {
-    calculateALAP(I->first, MII, maxASAP, (MSchedGraphNode*) 0);
-    visitedNodes.clear();
-  }
-
-  //Calculate MOB which depends on ASAP being totally calculated, also do depth and height
-  for(std::map<MSchedGraphNode*, MSNodeAttributes>::iterator I = nodeToAttributesMap.begin(), E = nodeToAttributesMap.end(); I != E; ++I) {
-    (I->second).MOB = std::max(0,(I->second).ALAP - (I->second).ASAP);
-
-    DEBUG(std::cerr << "MOB: " << (I->second).MOB << " (" << *(I->first) << ")\n");
-    calculateDepth(I->first, (MSchedGraphNode*) 0);
-    calculateHeight(I->first, (MSchedGraphNode*) 0);
-  }
-
-
-}
-
-/// ignoreEdge - Checks to see if this edge of a recurrence should be ignored or not
-bool ModuloSchedulingPass::ignoreEdge(MSchedGraphNode *srcNode, MSchedGraphNode *destNode) {
-  if(destNode == 0 || srcNode ==0)
-    return false;
-
-  bool findEdge = edgesToIgnore.count(std::make_pair(srcNode, destNode->getInEdgeNum(srcNode)));
-
-  DEBUG(std::cerr << "Ignoring edge? from: " << *srcNode << " to " << *destNode << "\n");
-
-  return findEdge;
-}
-
-
-/// calculateASAP - Calculates the
-int  ModuloSchedulingPass::calculateASAP(MSchedGraphNode *node, int MII, MSchedGraphNode *destNode) {
-
-  DEBUG(std::cerr << "Calculating ASAP for " << *node << "\n");
-
-  //Get current node attributes
-  MSNodeAttributes &attributes = nodeToAttributesMap.find(node)->second;
-
-  if(attributes.ASAP != -1)
-    return attributes.ASAP;
-
-  int maxPredValue = 0;
-
-  //Iterate over all of the predecessors and find max
-  for(MSchedGraphNode::pred_iterator P = node->pred_begin(), E = node->pred_end(); P != E; ++P) {
-
-    //Only process if we are not ignoring the edge
-    if(!ignoreEdge(*P, node)) {
-      int predASAP = -1;
-      predASAP = calculateASAP(*P, MII, node);
-
-      assert(predASAP != -1 && "ASAP has not been calculated");
-      int iteDiff = node->getInEdge(*P).getIteDiff();
-
-      int currentPredValue = predASAP + (*P)->getLatency() - (iteDiff * MII);
-      DEBUG(std::cerr << "pred ASAP: " << predASAP << ", iteDiff: " << iteDiff << ", PredLatency: " << (*P)->getLatency() << ", Current ASAP pred: " << currentPredValue << "\n");
-      maxPredValue = std::max(maxPredValue, currentPredValue);
-    }
-  }
-
-  attributes.ASAP = maxPredValue;
-
-  DEBUG(std::cerr << "ASAP: " << attributes.ASAP << " (" << *node << ")\n");
-
-  return maxPredValue;
-}
-
-
-int ModuloSchedulingPass::calculateALAP(MSchedGraphNode *node, int MII,
-                                        int maxASAP, MSchedGraphNode *srcNode) {
-
-  DEBUG(std::cerr << "Calculating ALAP for " << *node << "\n");
-
-  MSNodeAttributes &attributes = nodeToAttributesMap.find(node)->second;
-
-  if(attributes.ALAP != -1)
-    return attributes.ALAP;
-
-  if(node->hasSuccessors()) {
-
-    //Trying to deal with the issue where the node has successors, but
-    //we are ignoring all of the edges to them. So this is my hack for
-    //now.. there is probably a more elegant way of doing this (FIXME)
-    bool processedOneEdge = false;
-
-    //FIXME, set to something high to start
-    int minSuccValue = 9999999;
-
-    //Iterate over all of the predecessors and fine max
-    for(MSchedGraphNode::succ_iterator P = node->succ_begin(),
-          E = node->succ_end(); P != E; ++P) {
-
-      //Only process if we are not ignoring the edge
-      if(!ignoreEdge(node, *P)) {
-        processedOneEdge = true;
-        int succALAP = -1;
-        succALAP = calculateALAP(*P, MII, maxASAP, node);
-
-        assert(succALAP != -1 && "Successors ALAP should have been caclulated");
-
-        int iteDiff = P.getEdge().getIteDiff();
-
-        int currentSuccValue = succALAP - node->getLatency() + iteDiff * MII;
-
-        DEBUG(std::cerr << "succ ALAP: " << succALAP << ", iteDiff: " << iteDiff << ", SuccLatency: " << (*P)->getLatency() << ", Current ALAP succ: " << currentSuccValue << "\n");
-
-        minSuccValue = std::min(minSuccValue, currentSuccValue);
-      }
-    }
-
-    if(processedOneEdge)
-        attributes.ALAP = minSuccValue;
-
-    else
-      attributes.ALAP = maxASAP;
-  }
-  else
-    attributes.ALAP = maxASAP;
-
-  DEBUG(std::cerr << "ALAP: " << attributes.ALAP << " (" << *node << ")\n");
-
-  if(attributes.ALAP < 0)
-    attributes.ALAP = 0;
-
-  return attributes.ALAP;
-}
-
-int ModuloSchedulingPass::findMaxASAP() {
-  int maxASAP = 0;
-
-  for(std::map<MSchedGraphNode*, MSNodeAttributes>::iterator I = nodeToAttributesMap.begin(),
-        E = nodeToAttributesMap.end(); I != E; ++I)
-    maxASAP = std::max(maxASAP, I->second.ASAP);
-  return maxASAP;
-}
-
-
-int ModuloSchedulingPass::calculateHeight(MSchedGraphNode *node,MSchedGraphNode *srcNode) {
-
-  MSNodeAttributes &attributes = nodeToAttributesMap.find(node)->second;
-
-  if(attributes.height != -1)
-    return attributes.height;
-
-  int maxHeight = 0;
-
-  //Iterate over all of the predecessors and find max
-  for(MSchedGraphNode::succ_iterator P = node->succ_begin(),
-        E = node->succ_end(); P != E; ++P) {
-
-
-    if(!ignoreEdge(node, *P)) {
-      int succHeight = calculateHeight(*P, node);
-
-      assert(succHeight != -1 && "Successors Height should have been caclulated");
-
-      int currentHeight = succHeight + node->getLatency();
-      maxHeight = std::max(maxHeight, currentHeight);
-    }
-  }
-  attributes.height = maxHeight;
-  DEBUG(std::cerr << "Height: " << attributes.height << " (" << *node << ")\n");
-  return maxHeight;
-}
-
-
-int ModuloSchedulingPass::calculateDepth(MSchedGraphNode *node,
-                                          MSchedGraphNode *destNode) {
-
-  MSNodeAttributes &attributes = nodeToAttributesMap.find(node)->second;
-
-  if(attributes.depth != -1)
-    return attributes.depth;
-
-  int maxDepth = 0;
-
-  //Iterate over all of the predecessors and fine max
-  for(MSchedGraphNode::pred_iterator P = node->pred_begin(), E = node->pred_end(); P != E; ++P) {
-
-    if(!ignoreEdge(*P, node)) {
-      int predDepth = -1;
-      predDepth = calculateDepth(*P, node);
-
-      assert(predDepth != -1 && "Predecessors ASAP should have been caclulated");
-
-      int currentDepth = predDepth + (*P)->getLatency();
-      maxDepth = std::max(maxDepth, currentDepth);
-    }
-  }
-  attributes.depth = maxDepth;
-
-  DEBUG(std::cerr << "Depth: " << attributes.depth << " (" << *node << "*)\n");
-  return maxDepth;
-}
-
-
-
-void ModuloSchedulingPass::addReccurrence(std::vector<MSchedGraphNode*> &recurrence, int II, MSchedGraphNode *srcBENode, MSchedGraphNode *destBENode) {
-  //Check to make sure that this recurrence is unique
-  bool same = false;
-
-
-  //Loop over all recurrences already in our list
-  for(std::set<std::pair<int, std::vector<MSchedGraphNode*> > >::iterator R = recurrenceList.begin(), RE = recurrenceList.end(); R != RE; ++R) {
-
-    bool all_same = true;
-     //First compare size
-    if(R->second.size() == recurrence.size()) {
-
-      for(std::vector<MSchedGraphNode*>::const_iterator node = R->second.begin(), end = R->second.end(); node != end; ++node) {
-        if(std::find(recurrence.begin(), recurrence.end(), *node) == recurrence.end()) {
-          all_same = all_same && false;
-          break;
-        }
-        else
-          all_same = all_same && true;
-      }
-      if(all_same) {
-        same = true;
-        break;
-      }
-    }
-  }
-
-  if(!same) {
-    srcBENode = recurrence.back();
-    destBENode = recurrence.front();
-
-    //FIXME
-    if(destBENode->getInEdge(srcBENode).getIteDiff() == 0) {
-      //DEBUG(std::cerr << "NOT A BACKEDGE\n");
-      //find actual backedge HACK HACK
-      for(unsigned i=0; i< recurrence.size()-1; ++i) {
-        if(recurrence[i+1]->getInEdge(recurrence[i]).getIteDiff() == 1) {
-          srcBENode = recurrence[i];
-          destBENode = recurrence[i+1];
-          break;
-        }
-
-      }
-
-    }
-    DEBUG(std::cerr << "Back Edge to Remove: " << *srcBENode << " to " << *destBENode << "\n");
-    edgesToIgnore.insert(std::make_pair(srcBENode, destBENode->getInEdgeNum(srcBENode)));
-    recurrenceList.insert(std::make_pair(II, recurrence));
-  }
-
-}
-
-int CircCount;
-
-void ModuloSchedulingPass::unblock(MSchedGraphNode *u, std::set<MSchedGraphNode*> &blocked,
-             std::map<MSchedGraphNode*, std::set<MSchedGraphNode*> > &B) {
-
-  //Unblock u
-  DEBUG(std::cerr << "Unblocking: " << *u << "\n");
-  blocked.erase(u);
-
-  //std::set<MSchedGraphNode*> toErase;
-  while (!B[u].empty()) {
-    MSchedGraphNode *W = *B[u].begin();
-    B[u].erase(W);
-    //toErase.insert(*W);
-    DEBUG(std::cerr << "Removed: " << *W << "from B-List\n");
-    if(blocked.count(W))
-      unblock(W, blocked, B);
-  }
-
-}
-
-bool ModuloSchedulingPass::circuit(MSchedGraphNode *v, std::vector<MSchedGraphNode*> &stack,
-             std::set<MSchedGraphNode*> &blocked, std::vector<MSchedGraphNode*> &SCC,
-             MSchedGraphNode *s, std::map<MSchedGraphNode*, std::set<MSchedGraphNode*> > &B,
-                                   int II, std::map<MSchedGraphNode*, MSchedGraphNode*> &newNodes) {
-  bool f = false;
-
-  DEBUG(std::cerr << "Finding Circuits Starting with: ( " << v << ")"<< *v << "\n");
-
-  //Push node onto the stack
-  stack.push_back(v);
-
-  //block this node
-  blocked.insert(v);
-
-  //Loop over all successors of node v that are in the scc, create Adjaceny list
-  std::set<MSchedGraphNode*> AkV;
-  for(MSchedGraphNode::succ_iterator I = v->succ_begin(), E = v->succ_end(); I != E; ++I) {
-    if((std::find(SCC.begin(), SCC.end(), *I) != SCC.end())) {
-      AkV.insert(*I);
-    }
-  }
-
-  for(std::set<MSchedGraphNode*>::iterator I = AkV.begin(), E = AkV.end(); I != E; ++I) {
-    if(*I == s) {
-      //We have a circuit, so add it to our list
-      addRecc(stack, newNodes);
-      f = true;
-    }
-    else if(!blocked.count(*I)) {
-      if(circuit(*I, stack, blocked, SCC, s, B, II, newNodes))
-        f = true;
-    }
-    else
-      DEBUG(std::cerr << "Blocked: " << **I << "\n");
-  }
-
-
-  if(f) {
-    unblock(v, blocked, B);
-  }
-  else {
-    for(std::set<MSchedGraphNode*>::iterator I = AkV.begin(), E = AkV.end(); I != E; ++I)
-      B[*I].insert(v);
-
-  }
-
-  //Pop v
-  stack.pop_back();
-
-  return f;
-
-}
-
-void ModuloSchedulingPass::addRecc(std::vector<MSchedGraphNode*> &stack, std::map<MSchedGraphNode*, MSchedGraphNode*> &newNodes) {
-  std::vector<MSchedGraphNode*> recc;
-  //Dump recurrence for now
-  DEBUG(std::cerr << "Starting Recc\n");
-
-  int totalDelay = 0;
-  int totalDistance = 0;
-  MSchedGraphNode *lastN = 0;
-  MSchedGraphNode *start = 0;
-  MSchedGraphNode *end = 0;
-
-  //Loop over recurrence, get delay and distance
-  for(std::vector<MSchedGraphNode*>::iterator N = stack.begin(), NE = stack.end(); N != NE; ++N) {
-    DEBUG(std::cerr << **N << "\n");
-    totalDelay += (*N)->getLatency();
-    if(lastN) {
-      int iteDiff = (*N)->getInEdge(lastN).getIteDiff();
-      totalDistance += iteDiff;
-
-      if(iteDiff > 0) {
-        start = lastN;
-        end = *N;
-      }
-    }
-    //Get the original node
-    lastN = *N;
-    recc.push_back(newNodes[*N]);
-
-
-  }
-
-  //Get the loop edge
-  totalDistance += lastN->getIteDiff(*stack.begin());
-
-  DEBUG(std::cerr << "End Recc\n");
-  CircCount++;
-
-  if(start && end) {
-    //Insert reccurrence into the list
-    DEBUG(std::cerr << "Ignore Edge from!!: " << *start << " to " << *end << "\n");
-    edgesToIgnore.insert(std::make_pair(newNodes[start], (newNodes[end])->getInEdgeNum(newNodes[start])));
-  }
-  else {
-    //Insert reccurrence into the list
-    DEBUG(std::cerr << "Ignore Edge from: " << *lastN << " to " << **stack.begin() << "\n");
-    edgesToIgnore.insert(std::make_pair(newNodes[lastN], newNodes[(*stack.begin())]->getInEdgeNum(newNodes[lastN])));
-
-  }
-  //Adjust II until we get close to the inequality delay - II*distance <= 0
-  int RecMII = II; //Starting value
-  int value = totalDelay-(RecMII * totalDistance);
-  int lastII = II;
-  while(value < 0) {
-
-    lastII = RecMII;
-    RecMII--;
-    value = totalDelay-(RecMII * totalDistance);
-  }
-
-  recurrenceList.insert(std::make_pair(lastII, recc));
-
-}
-
-void ModuloSchedulingPass::addSCC(std::vector<MSchedGraphNode*> &SCC, std::map<MSchedGraphNode*, MSchedGraphNode*> &newNodes) {
-
-  int totalDelay = 0;
-  int totalDistance = 0;
-  std::vector<MSchedGraphNode*> recc;
-  MSchedGraphNode *start = 0;
-  MSchedGraphNode *end = 0;
-
-  //Loop over recurrence, get delay and distance
-  for(std::vector<MSchedGraphNode*>::iterator N = SCC.begin(), NE = SCC.end(); N != NE; ++N) {
-    DEBUG(std::cerr << **N << "\n");
-    totalDelay += (*N)->getLatency();
-
-    for(unsigned i = 0; i < (*N)->succ_size(); ++i) {
-      MSchedGraphEdge *edge = (*N)->getSuccessor(i);
-      if(find(SCC.begin(), SCC.end(), edge->getDest()) != SCC.end()) {
-        totalDistance += edge->getIteDiff();
-        if(edge->getIteDiff() > 0)
-          if(!start && !end) {
-            start = *N;
-            end = edge->getDest();
-          }
-
-      }
-    }
-
-
-    //Get the original node
-    recc.push_back(newNodes[*N]);
-
-
-  }
-
-  DEBUG(std::cerr << "End Recc\n");
-  CircCount++;
-
-  assert( (start && end) && "Must have start and end node to ignore edge for SCC");
-
-  if(start && end) {
-    //Insert reccurrence into the list
-    DEBUG(std::cerr << "Ignore Edge from!!: " << *start << " to " << *end << "\n");
-    edgesToIgnore.insert(std::make_pair(newNodes[start], (newNodes[end])->getInEdgeNum(newNodes[start])));
-  }
-
-  int lastII = totalDelay / totalDistance;
-
-
-  recurrenceList.insert(std::make_pair(lastII, recc));
-
-}
-
-void ModuloSchedulingPass::findAllCircuits(MSchedGraph *g, int II) {
-
-  CircCount = 0;
-
-  //Keep old to new node mapping information
-  std::map<MSchedGraphNode*, MSchedGraphNode*> newNodes;
-
-  //copy the graph
-  MSchedGraph *MSG = new MSchedGraph(*g, newNodes);
-
-  DEBUG(std::cerr << "Finding All Circuits\n");
-
-  //Set of blocked nodes
-  std::set<MSchedGraphNode*> blocked;
-
-  //Stack holding current circuit
-  std::vector<MSchedGraphNode*> stack;
-
-  //Map for B Lists
-  std::map<MSchedGraphNode*, std::set<MSchedGraphNode*> > B;
-
-  //current node
-  MSchedGraphNode *s;
-
-
-  //Iterate over the graph until its down to one node or empty
-  while(MSG->size() > 1) {
-
-    //Write Graph out to file
-    //WriteGraphToFile(std::cerr, "Graph" + utostr(MSG->size()), MSG);
-
-    DEBUG(std::cerr << "Graph Size: " << MSG->size() << "\n");
-    DEBUG(std::cerr << "Finding strong component Vk with least vertex\n");
-
-    //Iterate over all the SCCs in the graph
-    std::set<MSchedGraphNode*> Visited;
-    std::vector<MSchedGraphNode*> Vk;
-    MSchedGraphNode* s = 0;
-    int numEdges = 0;
-
-    //Find scc with the least vertex
-    for (MSchedGraph::iterator GI = MSG->begin(), E = MSG->end(); GI != E; ++GI)
-      if (Visited.insert(GI->second).second) {
-        for (scc_iterator<MSchedGraphNode*> SCCI = scc_begin(GI->second),
-               E = scc_end(GI->second); SCCI != E; ++SCCI) {
-          std::vector<MSchedGraphNode*> &nextSCC = *SCCI;
-
-          if (Visited.insert(nextSCC[0]).second) {
-            Visited.insert(nextSCC.begin()+1, nextSCC.end());
-
-            if(nextSCC.size() > 1) {
-              std::cerr << "SCC size: " << nextSCC.size() << "\n";
-
-              for(unsigned i = 0; i < nextSCC.size(); ++i) {
-                //Loop over successor and see if in scc, then count edge
-                MSchedGraphNode *node = nextSCC[i];
-                for(MSchedGraphNode::succ_iterator S = node->succ_begin(), SE = node->succ_end(); S != SE; ++S) {
-                  if(find(nextSCC.begin(), nextSCC.end(), *S) != nextSCC.end())
-                    numEdges++;
-                }
-              }
-              std::cerr << "Num Edges: " << numEdges << "\n";
-            }
-
-            //Ignore self loops
-            if(nextSCC.size() > 1) {
-
-              //Get least vertex in Vk
-              if(!s) {
-                s = nextSCC[0];
-                Vk = nextSCC;
-              }
-
-              for(unsigned i = 0; i < nextSCC.size(); ++i) {
-                if(nextSCC[i] < s) {
-                  s = nextSCC[i];
-                  Vk = nextSCC;
-                }
-              }
-            }
-          }
-        }
-      }
-
-
-
-    //Process SCC
-    DEBUG(for(std::vector<MSchedGraphNode*>::iterator N = Vk.begin(), NE = Vk.end();
-              N != NE; ++N) { std::cerr << *((*N)->getInst()); });
-
-    //Iterate over all nodes in this scc
-    for(std::vector<MSchedGraphNode*>::iterator N = Vk.begin(), NE = Vk.end();
-        N != NE; ++N) {
-      blocked.erase(*N);
-      B[*N].clear();
-    }
-    if(Vk.size() > 1) {
-      if(numEdges < 98)
-        circuit(s, stack, blocked, Vk, s, B, II, newNodes);
-      else
-        addSCC(Vk, newNodes);
-
-      //Delete nodes from the graph
-      //Find all nodes up to s and delete them
-      std::vector<MSchedGraphNode*> nodesToRemove;
-      nodesToRemove.push_back(s);
-      for(MSchedGraph::iterator N = MSG->begin(), NE = MSG->end(); N != NE; ++N) {
-        if(N->second < s )
-            nodesToRemove.push_back(N->second);
-      }
-      for(std::vector<MSchedGraphNode*>::iterator N = nodesToRemove.begin(), NE = nodesToRemove.end(); N != NE; ++N) {
-        DEBUG(std::cerr << "Deleting Node: " << **N << "\n");
-        MSG->deleteNode(*N);
-      }
-    }
-    else
-      break;
-  }
-  DEBUG(std::cerr << "Num Circuits found: " << CircCount << "\n");
-}
-
-
-void ModuloSchedulingPass::findAllReccurrences(MSchedGraphNode *node,
-                                               std::vector<MSchedGraphNode*> &visitedNodes,
-                                               int II) {
-
-
-  if(std::find(visitedNodes.begin(), visitedNodes.end(), node) != visitedNodes.end()) {
-    std::vector<MSchedGraphNode*> recurrence;
-    bool first = true;
-    int delay = 0;
-    int distance = 0;
-    int RecMII = II; //Starting value
-    MSchedGraphNode *last = node;
-    MSchedGraphNode *srcBackEdge = 0;
-    MSchedGraphNode *destBackEdge = 0;
-
-
-
-    for(std::vector<MSchedGraphNode*>::iterator I = visitedNodes.begin(), E = visitedNodes.end();
-        I !=E; ++I) {
-
-      if(*I == node)
-        first = false;
-      if(first)
-        continue;
-
-      delay = delay + (*I)->getLatency();
-
-      if(*I != node) {
-        int diff = (*I)->getInEdge(last).getIteDiff();
-        distance += diff;
-        if(diff > 0) {
-          srcBackEdge = last;
-          destBackEdge = *I;
-        }
-      }
-
-      recurrence.push_back(*I);
-      last = *I;
-    }
-
-
-
-    //Get final distance calc
-    distance += node->getInEdge(last).getIteDiff();
-    DEBUG(std::cerr << "Reccurrence Distance: " << distance << "\n");
-
-    //Adjust II until we get close to the inequality delay - II*distance <= 0
-
-    int value = delay-(RecMII * distance);
-    int lastII = II;
-    while(value <= 0) {
-
-      lastII = RecMII;
-      RecMII--;
-      value = delay-(RecMII * distance);
-    }
-
-
-    DEBUG(std::cerr << "Final II for this recurrence: " << lastII << "\n");
-    addReccurrence(recurrence, lastII, srcBackEdge, destBackEdge);
-    assert(distance != 0 && "Recurrence distance should not be zero");
-    return;
-  }
-
-  unsigned count = 0;
-  for(MSchedGraphNode::succ_iterator I = node->succ_begin(), E = node->succ_end(); I != E; ++I) {
-    visitedNodes.push_back(node);
-    //if(!edgesToIgnore.count(std::make_pair(node, count)))
-    findAllReccurrences(*I, visitedNodes, II);
-    visitedNodes.pop_back();
-    count++;
-  }
-}
-
-void ModuloSchedulingPass::searchPath(MSchedGraphNode *node,
-                                      std::vector<MSchedGraphNode*> &path,
-                                      std::set<MSchedGraphNode*> &nodesToAdd,
-                                     std::set<MSchedGraphNode*> &new_reccurrence) {
-  //Push node onto the path
-  path.push_back(node);
-
-  //Loop over all successors and see if there is a path from this node to
-  //a recurrence in the partial order, if so.. add all nodes to be added to recc
-  for(MSchedGraphNode::succ_iterator S = node->succ_begin(), SE = node->succ_end(); S != SE;
-      ++S) {
-
-    //Check if we should ignore this edge first
-    if(ignoreEdge(node,*S))
-      continue;
-
-    //check if successor is in this recurrence, we will get to it eventually
-    if(new_reccurrence.count(*S))
-      continue;
-
-    //If this node exists in a recurrence already in the partial
-    //order, then add all nodes in the path to the set of nodes to add
-     //Check if its already in our partial order, if not add it to the
-     //final vector
-    bool found = false;
-    for(std::vector<std::set<MSchedGraphNode*> >::iterator PO = partialOrder.begin(),
-          PE = partialOrder.end(); PO != PE; ++PO) {
-
-      if(PO->count(*S)) {
-        found = true;
-        break;
-      }
-    }
-
-    if(!found) {
-      nodesToAdd.insert(*S);
-      searchPath(*S, path, nodesToAdd, new_reccurrence);
-    }
-  }
-
-  //Pop Node off the path
-  path.pop_back();
-}
-
-void ModuloSchedulingPass::pathToRecc(MSchedGraphNode *node,
-                                      std::vector<MSchedGraphNode*> &path,
-                                      std::set<MSchedGraphNode*> &poSet,
-                                      std::set<MSchedGraphNode*> &lastNodes) {
-  //Push node onto the path
-  path.push_back(node);
-
-  DEBUG(std::cerr << "Current node: " << *node << "\n");
-
-  //Loop over all successors and see if there is a path from this node to
-  //a recurrence in the partial order, if so.. add all nodes to be added to recc
-  for(MSchedGraphNode::succ_iterator S = node->succ_begin(), SE = node->succ_end(); S != SE;
-      ++S) {
-    DEBUG(std::cerr << "Succ:" << **S << "\n");
-    //Check if we should ignore this edge first
-    if(ignoreEdge(node,*S))
-      continue;
-
-    if(poSet.count(*S)) {
-      DEBUG(std::cerr << "Found path to recc from no pred\n");
-      //Loop over path, if it exists in lastNodes, then add to poset, and remove from lastNodes
-      for(std::vector<MSchedGraphNode*>::iterator I = path.begin(), IE = path.end(); I != IE; ++I) {
-        if(lastNodes.count(*I)) {
-          DEBUG(std::cerr << "Inserting node into recc: " << **I << "\n");
-          poSet.insert(*I);
-          lastNodes.erase(*I);
-        }
-      }
-    }
-    else
-      pathToRecc(*S, path, poSet, lastNodes);
-  }
-
-  //Pop Node off the path
-  path.pop_back();
-}
-
-void ModuloSchedulingPass::computePartialOrder() {
-
-  TIME_REGION(X, "calculatePartialOrder");
-
-  DEBUG(std::cerr << "Computing Partial Order\n");
-
-  //Only push BA branches onto the final node order, we put other
-  //branches after it FIXME: Should we really be pushing branches on
-  //it a specific order instead of relying on BA being there?
-
-  std::vector<MSchedGraphNode*> branches;
-
-  //Steps to add a recurrence to the partial order 1) Find reccurrence
-  //with the highest RecMII. Add it to the partial order.  2) For each
-  //recurrence with decreasing RecMII, add it to the partial order
-  //along with any nodes that connect this recurrence to recurrences
-  //already in the partial order
-  for(std::set<std::pair<int, std::vector<MSchedGraphNode*> > >::reverse_iterator
-        I = recurrenceList.rbegin(), E=recurrenceList.rend(); I !=E; ++I) {
-
-    std::set<MSchedGraphNode*> new_recurrence;
-
-    //Loop through recurrence and remove any nodes already in the partial order
-    for(std::vector<MSchedGraphNode*>::const_iterator N = I->second.begin(),
-          NE = I->second.end(); N != NE; ++N) {
-
-      bool found = false;
-      for(std::vector<std::set<MSchedGraphNode*> >::iterator PO = partialOrder.begin(),
-            PE = partialOrder.end(); PO != PE; ++PO) {
-        if(PO->count(*N))
-          found = true;
-      }
-
-      //Check if its a branch, and remove to handle special
-      if(!found) {
-        if((*N)->isBranch() && !(*N)->hasPredecessors()) {
-          branches.push_back(*N);
-        }
-        else
-          new_recurrence.insert(*N);
-      }
-
-    }
-
-
-    if(new_recurrence.size() > 0) {
-
-      std::vector<MSchedGraphNode*> path;
-      std::set<MSchedGraphNode*> nodesToAdd;
-
-      //Dump recc we are dealing with (minus nodes already in PO)
-      DEBUG(std::cerr << "Recc: ");
-      DEBUG(for(std::set<MSchedGraphNode*>::iterator R = new_recurrence.begin(), RE = new_recurrence.end(); R != RE; ++R) { std::cerr << **R ; });
-
-      //Add nodes that connect this recurrence to recurrences in the partial path
-      for(std::set<MSchedGraphNode*>::iterator N = new_recurrence.begin(),
-          NE = new_recurrence.end(); N != NE; ++N)
-        searchPath(*N, path, nodesToAdd, new_recurrence);
-
-      //Add nodes to this recurrence if they are not already in the partial order
-      for(std::set<MSchedGraphNode*>::iterator N = nodesToAdd.begin(), NE = nodesToAdd.end();
-          N != NE; ++N) {
-        bool found = false;
-        for(std::vector<std::set<MSchedGraphNode*> >::iterator PO = partialOrder.begin(),
-              PE = partialOrder.end(); PO != PE; ++PO) {
-          if(PO->count(*N))
-            found = true;
-        }
-        if(!found) {
-          assert("FOUND CONNECTOR");
-          new_recurrence.insert(*N);
-        }
-      }
-
-      partialOrder.push_back(new_recurrence);
-
-
-      //Dump out partial order
-      DEBUG(for(std::vector<std::set<MSchedGraphNode*> >::iterator I = partialOrder.begin(),
-                  E = partialOrder.end(); I !=E; ++I) {
-              std::cerr << "Start set in PO\n";
-              for(std::set<MSchedGraphNode*>::iterator J = I->begin(), JE = I->end(); J != JE; ++J)
-                std::cerr << "PO:" << **J << "\n";
-            });
-
-    }
-  }
-
-  //Add any nodes that are not already in the partial order
-  //Add them in a set, one set per connected component
-  std::set<MSchedGraphNode*> lastNodes;
-  std::set<MSchedGraphNode*> noPredNodes;
-  for(std::map<MSchedGraphNode*, MSNodeAttributes>::iterator I = nodeToAttributesMap.begin(),
-        E = nodeToAttributesMap.end(); I != E; ++I) {
-
-    bool found = false;
-
-    //Check if its already in our partial order, if not add it to the final vector
-    for(std::vector<std::set<MSchedGraphNode*> >::iterator PO = partialOrder.begin(),
-          PE = partialOrder.end(); PO != PE; ++PO) {
-      if(PO->count(I->first))
-        found = true;
-    }
-    if(!found)
-      lastNodes.insert(I->first);
-  }
-
-  //For each node w/out preds, see if there is a path to one of the
-  //recurrences, and if so add them to that current recc
-  /*for(std::set<MSchedGraphNode*>::iterator N = noPredNodes.begin(), NE = noPredNodes.end();
-      N != NE; ++N) {
-    DEBUG(std::cerr << "No Pred Path from: " << **N << "\n");
-    for(std::vector<std::set<MSchedGraphNode*> >::iterator PO = partialOrder.begin(),
-          PE = partialOrder.end(); PO != PE; ++PO) {
-      std::vector<MSchedGraphNode*> path;
-      pathToRecc(*N, path, *PO, lastNodes);
-    }
-    }*/
-
-
-  //Break up remaining nodes that are not in the partial order
-  ///into their connected compoenents
-    while(lastNodes.size() > 0) {
-      std::set<MSchedGraphNode*> ccSet;
-      connectedComponentSet(*(lastNodes.begin()),ccSet, lastNodes);
-      if(ccSet.size() > 0)
-        partialOrder.push_back(ccSet);
-    }
-
-
-  //Clean up branches by putting them in final order
-    assert(branches.size() == 0 && "We should not have any branches in our graph");
-}
-
-
-void ModuloSchedulingPass::connectedComponentSet(MSchedGraphNode *node, std::set<MSchedGraphNode*> &ccSet, std::set<MSchedGraphNode*> &lastNodes) {
-
-//Add to final set
-  if( !ccSet.count(node) && lastNodes.count(node)) {
-    lastNodes.erase(node);
-    ccSet.insert(node);
-  }
-  else
-    return;
-
-  //Loop over successors and recurse if we have not seen this node before
-  for(MSchedGraphNode::succ_iterator node_succ = node->succ_begin(), end=node->succ_end(); node_succ != end; ++node_succ) {
-    connectedComponentSet(*node_succ, ccSet, lastNodes);
-  }
-
-}
-
-void ModuloSchedulingPass::predIntersect(std::set<MSchedGraphNode*> &CurrentSet, std::set<MSchedGraphNode*> &IntersectResult) {
-
-  for(unsigned j=0; j < FinalNodeOrder.size(); ++j) {
-    for(MSchedGraphNode::pred_iterator P = FinalNodeOrder[j]->pred_begin(),
-          E = FinalNodeOrder[j]->pred_end(); P != E; ++P) {
-
-      //Check if we are supposed to ignore this edge or not
-      if(ignoreEdge(*P,FinalNodeOrder[j]))
-        continue;
-
-      if(CurrentSet.count(*P))
-        if(std::find(FinalNodeOrder.begin(), FinalNodeOrder.end(), *P) == FinalNodeOrder.end())
-          IntersectResult.insert(*P);
-    }
-  }
-}
-
-
-
-
-
-void ModuloSchedulingPass::succIntersect(std::set<MSchedGraphNode*> &CurrentSet, std::set<MSchedGraphNode*> &IntersectResult) {
-
-  for(unsigned j=0; j < FinalNodeOrder.size(); ++j) {
-    for(MSchedGraphNode::succ_iterator P = FinalNodeOrder[j]->succ_begin(),
-          E = FinalNodeOrder[j]->succ_end(); P != E; ++P) {
-
-      //Check if we are supposed to ignore this edge or not
-      if(ignoreEdge(FinalNodeOrder[j],*P))
-        continue;
-
-      if(CurrentSet.count(*P))
-        if(std::find(FinalNodeOrder.begin(), FinalNodeOrder.end(), *P) == FinalNodeOrder.end())
-          IntersectResult.insert(*P);
-    }
-  }
-}
-
-void dumpIntersection(std::set<MSchedGraphNode*> &IntersectCurrent) {
-  std::cerr << "Intersection (";
-  for(std::set<MSchedGraphNode*>::iterator I = IntersectCurrent.begin(), E = IntersectCurrent.end(); I != E; ++I)
-    std::cerr << **I << ", ";
-  std::cerr << ")\n";
-}
-
-
-
-void ModuloSchedulingPass::orderNodes() {
-
-  TIME_REGION(X, "orderNodes");
-
-  int BOTTOM_UP = 0;
-  int TOP_DOWN = 1;
-
-  //Set default order
-  int order = BOTTOM_UP;
-
-
-  //Loop over all the sets and place them in the final node order
-  for(std::vector<std::set<MSchedGraphNode*> >::iterator CurrentSet = partialOrder.begin(), E= partialOrder.end(); CurrentSet != E; ++CurrentSet) {
-
-    DEBUG(std::cerr << "Processing set in S\n");
-    DEBUG(dumpIntersection(*CurrentSet));
-
-    //Result of intersection
-    std::set<MSchedGraphNode*> IntersectCurrent;
-
-    predIntersect(*CurrentSet, IntersectCurrent);
-
-    //If the intersection of predecessor and current set is not empty
-    //sort nodes bottom up
-    if(IntersectCurrent.size() != 0) {
-      DEBUG(std::cerr << "Final Node Order Predecessors and Current Set interesection is NOT empty\n");
-      order = BOTTOM_UP;
-    }
-    //If empty, use successors
-    else {
-      DEBUG(std::cerr << "Final Node Order Predecessors and Current Set interesection is empty\n");
-
-      succIntersect(*CurrentSet, IntersectCurrent);
-
-      //sort top-down
-      if(IntersectCurrent.size() != 0) {
-         DEBUG(std::cerr << "Final Node Order Successors and Current Set interesection is NOT empty\n");
-        order = TOP_DOWN;
-      }
-      else {
-        DEBUG(std::cerr << "Final Node Order Successors and Current Set interesection is empty\n");
-        //Find node with max ASAP in current Set
-        MSchedGraphNode *node;
-        int maxASAP = 0;
-        DEBUG(std::cerr << "Using current set of size " << CurrentSet->size() << "to find max ASAP\n");
-        for(std::set<MSchedGraphNode*>::iterator J = CurrentSet->begin(), JE = CurrentSet->end(); J != JE; ++J) {
-          //Get node attributes
-          MSNodeAttributes nodeAttr= nodeToAttributesMap.find(*J)->second;
-          //assert(nodeAttr != nodeToAttributesMap.end() && "Node not in attributes map!");
-
-          if(maxASAP <= nodeAttr.ASAP) {
-            maxASAP = nodeAttr.ASAP;
-            node = *J;
-          }
-        }
-        assert(node != 0 && "In node ordering node should not be null");
-        IntersectCurrent.insert(node);
-        order = BOTTOM_UP;
-      }
-    }
-
-    //Repeat until all nodes are put into the final order from current set
-    while(IntersectCurrent.size() > 0) {
-
-      if(order == TOP_DOWN) {
-        DEBUG(std::cerr << "Order is TOP DOWN\n");
-
-        while(IntersectCurrent.size() > 0) {
-          DEBUG(std::cerr << "Intersection is not empty, so find heighest height\n");
-
-          int MOB = 0;
-          int height = 0;
-          MSchedGraphNode *highestHeightNode = *(IntersectCurrent.begin());
-
-          //Find node in intersection with highest heigh and lowest MOB
-          for(std::set<MSchedGraphNode*>::iterator I = IntersectCurrent.begin(),
-                E = IntersectCurrent.end(); I != E; ++I) {
-
-            //Get current nodes properties
-            MSNodeAttributes nodeAttr= nodeToAttributesMap.find(*I)->second;
-
-            if(height < nodeAttr.height) {
-              highestHeightNode = *I;
-              height = nodeAttr.height;
-              MOB = nodeAttr.MOB;
-            }
-            else if(height ==  nodeAttr.height) {
-              if(MOB > nodeAttr.height) {
-                highestHeightNode = *I;
-                height =  nodeAttr.height;
-                MOB = nodeAttr.MOB;
-              }
-            }
-          }
-
-          //Append our node with greatest height to the NodeOrder
-          if(std::find(FinalNodeOrder.begin(), FinalNodeOrder.end(), highestHeightNode) == FinalNodeOrder.end()) {
-            DEBUG(std::cerr << "Adding node to Final Order: " << *highestHeightNode << "\n");
-            FinalNodeOrder.push_back(highestHeightNode);
-          }
-
-          //Remove V from IntersectOrder
-          IntersectCurrent.erase(std::find(IntersectCurrent.begin(),
-                                      IntersectCurrent.end(), highestHeightNode));
-
-
-          //Intersect V's successors with CurrentSet
-          for(MSchedGraphNode::succ_iterator P = highestHeightNode->succ_begin(),
-                E = highestHeightNode->succ_end(); P != E; ++P) {
-            //if(lower_bound(CurrentSet->begin(),
-            //     CurrentSet->end(), *P) != CurrentSet->end()) {
-            if(std::find(CurrentSet->begin(), CurrentSet->end(), *P) != CurrentSet->end()) {
-              if(ignoreEdge(highestHeightNode, *P))
-                continue;
-              //If not already in Intersect, add
-              if(!IntersectCurrent.count(*P))
-                IntersectCurrent.insert(*P);
-            }
-          }
-        } //End while loop over Intersect Size
-
-        //Change direction
-        order = BOTTOM_UP;
-
-        //Reset Intersect to reflect changes in OrderNodes
-        IntersectCurrent.clear();
-        predIntersect(*CurrentSet, IntersectCurrent);
-
-      } //End If TOP_DOWN
-
-        //Begin if BOTTOM_UP
-      else {
-        DEBUG(std::cerr << "Order is BOTTOM UP\n");
-        while(IntersectCurrent.size() > 0) {
-          DEBUG(std::cerr << "Intersection of size " << IntersectCurrent.size() << ", finding highest depth\n");
-
-          //dump intersection
-          DEBUG(dumpIntersection(IntersectCurrent));
-          //Get node with highest depth, if a tie, use one with lowest
-          //MOB
-          int MOB = 0;
-          int depth = 0;
-          MSchedGraphNode *highestDepthNode = *(IntersectCurrent.begin());
-
-          for(std::set<MSchedGraphNode*>::iterator I = IntersectCurrent.begin(),
-                E = IntersectCurrent.end(); I != E; ++I) {
-            //Find node attribute in graph
-            MSNodeAttributes nodeAttr= nodeToAttributesMap.find(*I)->second;
-
-            if(depth < nodeAttr.depth) {
-              highestDepthNode = *I;
-              depth = nodeAttr.depth;
-              MOB = nodeAttr.MOB;
-            }
-            else if(depth == nodeAttr.depth) {
-              if(MOB > nodeAttr.MOB) {
-                highestDepthNode = *I;
-                depth = nodeAttr.depth;
-                MOB = nodeAttr.MOB;
-              }
-            }
-          }
-
-
-
-          //Append highest depth node to the NodeOrder
-           if(std::find(FinalNodeOrder.begin(), FinalNodeOrder.end(), highestDepthNode) == FinalNodeOrder.end()) {
-             DEBUG(std::cerr << "Adding node to Final Order: " << *highestDepthNode << "\n");
-             FinalNodeOrder.push_back(highestDepthNode);
-           }
-          //Remove heightestDepthNode from IntersectOrder
-           IntersectCurrent.erase(highestDepthNode);
-
-
-          //Intersect heightDepthNode's pred with CurrentSet
-          for(MSchedGraphNode::pred_iterator P = highestDepthNode->pred_begin(),
-                E = highestDepthNode->pred_end(); P != E; ++P) {
-            if(CurrentSet->count(*P)) {
-              if(ignoreEdge(*P, highestDepthNode))
-                continue;
-
-            //If not already in Intersect, add
-            if(!IntersectCurrent.count(*P))
-              IntersectCurrent.insert(*P);
-            }
-          }
-
-        } //End while loop over Intersect Size
-
-          //Change order
-        order = TOP_DOWN;
-
-        //Reset IntersectCurrent to reflect changes in OrderNodes
-        IntersectCurrent.clear();
-        succIntersect(*CurrentSet, IntersectCurrent);
-        } //End if BOTTOM_DOWN
-
-      DEBUG(std::cerr << "Current Intersection Size: " << IntersectCurrent.size() << "\n");
-    }
-    //End Wrapping while loop
-    DEBUG(std::cerr << "Ending Size of Current Set: " << CurrentSet->size() << "\n");
-  }//End for over all sets of nodes
-
-  //FIXME: As the algorithm stands it will NEVER add an instruction such as ba (with no
-  //data dependencies) to the final order. We add this manually. It will always be
-  //in the last set of S since its not part of a recurrence
-    //Loop over all the sets and place them in the final node order
-  std::vector<std::set<MSchedGraphNode*> > ::reverse_iterator LastSet = partialOrder.rbegin();
-  for(std::set<MSchedGraphNode*>::iterator CurrentNode = LastSet->begin(), LastNode = LastSet->end();
-      CurrentNode != LastNode; ++CurrentNode) {
-    if((*CurrentNode)->getInst()->getOpcode() == V9::BA)
-      FinalNodeOrder.push_back(*CurrentNode);
-  }
-  //Return final Order
-  //return FinalNodeOrder;
-}
-
-bool ModuloSchedulingPass::computeSchedule(const MachineBasicBlock *BB, MSchedGraph *MSG) {
-
-  TIME_REGION(X, "computeSchedule");
-
-  bool success = false;
-
-  //FIXME: Should be set to max II of the original loop
-  //Cap II in order to prevent infinite loop
-  int capII = MSG->totalDelay();
-
-  while(!success) {
-
-    //Keep track of branches, but do not insert into the schedule
-    std::vector<MSchedGraphNode*> branches;
-
-    //Loop over the final node order and process each node
-    for(std::vector<MSchedGraphNode*>::iterator I = FinalNodeOrder.begin(),
-          E = FinalNodeOrder.end(); I != E; ++I) {
-
-      //CalculateEarly and Late start
-      bool initialLSVal = false;
-      bool initialESVal = false;
-      int EarlyStart = 0;
-      int LateStart = 0;
-      bool hasSucc = false;
-      bool hasPred = false;
-      bool sched;
-
-      if((*I)->isBranch())
-        if((*I)->hasPredecessors())
-          sched = true;
-        else
-          sched = false;
-      else
-        sched = true;
-
-      if(sched) {
-        //Loop over nodes in the schedule and determine if they are predecessors
-        //or successors of the node we are trying to schedule
-        for(MSSchedule::schedule_iterator nodesByCycle = schedule.begin(), nodesByCycleEnd = schedule.end();
-            nodesByCycle != nodesByCycleEnd; ++nodesByCycle) {
-
-          //For this cycle, get the vector of nodes schedule and loop over it
-          for(std::vector<MSchedGraphNode*>::iterator schedNode = nodesByCycle->second.begin(), SNE = nodesByCycle->second.end(); schedNode != SNE; ++schedNode) {
-
-            if((*I)->isPredecessor(*schedNode)) {
-              int diff = (*I)->getInEdge(*schedNode).getIteDiff();
-              int ES_Temp = nodesByCycle->first + (*schedNode)->getLatency() - diff * II;
-              DEBUG(std::cerr << "Diff: " << diff << " Cycle: " << nodesByCycle->first << "\n");
-              DEBUG(std::cerr << "Temp EarlyStart: " << ES_Temp << " Prev EarlyStart: " << EarlyStart << "\n");
-              if(initialESVal)
-                EarlyStart = std::max(EarlyStart, ES_Temp);
-              else {
-                EarlyStart = ES_Temp;
-                initialESVal = true;
-              }
-              hasPred = true;
-            }
-            if((*I)->isSuccessor(*schedNode)) {
-              int diff = (*schedNode)->getInEdge(*I).getIteDiff();
-              int LS_Temp = nodesByCycle->first - (*I)->getLatency() + diff * II;
-              DEBUG(std::cerr << "Diff: " << diff << " Cycle: " << nodesByCycle->first << "\n");
-              DEBUG(std::cerr << "Temp LateStart: " << LS_Temp << " Prev LateStart: " << LateStart << "\n");
-              if(initialLSVal)
-                LateStart = std::min(LateStart, LS_Temp);
-              else {
-                LateStart = LS_Temp;
-                initialLSVal = true;
-              }
-              hasSucc = true;
-            }
-          }
-        }
-      }
-      else {
-        branches.push_back(*I);
-        continue;
-      }
-
-      //Check if this node is a pred or succ to a branch, and restrict its placement
-      //even though the branch is not in the schedule
-      /*int count = branches.size();
-      for(std::vector<MSchedGraphNode*>::iterator B = branches.begin(), BE = branches.end();
-          B != BE; ++B) {
-        if((*I)->isPredecessor(*B)) {
-          int diff = (*I)->getInEdge(*B).getIteDiff();
-          int ES_Temp = (II+count-1) + (*B)->getLatency() - diff * II;
-          DEBUG(std::cerr << "Diff: " << diff << " Cycle: " << (II+count)-1 << "\n");
-          DEBUG(std::cerr << "Temp EarlyStart: " << ES_Temp << " Prev EarlyStart: " << EarlyStart << "\n");
-          EarlyStart = std::max(EarlyStart, ES_Temp);
-          hasPred = true;
-        }
-
-        if((*I)->isSuccessor(*B)) {
-          int diff = (*B)->getInEdge(*I).getIteDiff();
-          int LS_Temp = (II+count-1) - (*I)->getLatency() + diff * II;
-          DEBUG(std::cerr << "Diff: " << diff << " Cycle: " << (II+count-1) << "\n");
-          DEBUG(std::cerr << "Temp LateStart: " << LS_Temp << " Prev LateStart: " << LateStart << "\n");
-          LateStart = std::min(LateStart, LS_Temp);
-          hasSucc = true;
-        }
-
-        count--;
-      }*/
-
-      //Check if the node has no pred or successors and set Early Start to its ASAP
-      if(!hasSucc && !hasPred)
-        EarlyStart = nodeToAttributesMap.find(*I)->second.ASAP;
-
-      DEBUG(std::cerr << "Has Successors: " << hasSucc << ", Has Pred: " << hasPred << "\n");
-      DEBUG(std::cerr << "EarlyStart: " << EarlyStart << ", LateStart: " << LateStart << "\n");
-
-      //Now, try to schedule this node depending upon its pred and successor in the schedule
-      //already
-      if(!hasSucc && hasPred)
-        success = scheduleNode(*I, EarlyStart, (EarlyStart + II -1));
-      else if(!hasPred && hasSucc)
-        success = scheduleNode(*I, LateStart, (LateStart - II +1));
-      else if(hasPred && hasSucc) {
-        if(EarlyStart > LateStart) {
-        success = false;
-          //LateStart = EarlyStart;
-          DEBUG(std::cerr << "Early Start can not be later then the late start cycle, schedule fails\n");
-        }
-        else
-          success = scheduleNode(*I, EarlyStart, std::min(LateStart, (EarlyStart + II -1)));
-      }
-      else
-        success = scheduleNode(*I, EarlyStart, EarlyStart + II - 1);
-
-      if(!success) {
-        ++II;
-        schedule.clear();
-        break;
-      }
-
-    }
-
-    if(success) {
-      DEBUG(std::cerr << "Constructing Schedule Kernel\n");
-      success = schedule.constructKernel(II, branches, indVarInstrs[BB]);
-      DEBUG(std::cerr << "Done Constructing Schedule Kernel\n");
-      if(!success) {
-        ++II;
-        schedule.clear();
-      }
-      DEBUG(std::cerr << "Final II: " << II << "\n");
-    }
-
-
-    if(II >= capII) {
-      DEBUG(std::cerr << "Maximum II reached, giving up\n");
-      return false;
-    }
-
-    assert(II < capII && "The II should not exceed the original loop number of cycles");
-  }
-  return true;
-}
-
-
-bool ModuloSchedulingPass::scheduleNode(MSchedGraphNode *node,
-                                      int start, int end) {
-  bool success = false;
-
-  DEBUG(std::cerr << *node << " (Start Cycle: " << start << ", End Cycle: " << end << ")\n");
-
-  //Make sure start and end are not negative
-  //if(start < 0) {
-  //start = 0;
-
-  //}
-  //if(end < 0)
-  //end = 0;
-
-  bool forward = true;
-  if(start > end)
-    forward = false;
-
-  bool increaseSC = true;
-  int cycle = start ;
-
-
-  while(increaseSC) {
-
-    increaseSC = false;
-
-    increaseSC = schedule.insert(node, cycle, II);
-
-    if(!increaseSC)
-      return true;
-
-    //Increment cycle to try again
-    if(forward) {
-      ++cycle;
-      DEBUG(std::cerr << "Increase cycle: " << cycle << "\n");
-      if(cycle > end)
-        return false;
-    }
-    else {
-      --cycle;
-      DEBUG(std::cerr << "Decrease cycle: " << cycle << "\n");
-      if(cycle < end)
-        return false;
-    }
-  }
-
-  return success;
-}
-
-void ModuloSchedulingPass::writePrologues(std::vector<MachineBasicBlock *> &prologues, MachineBasicBlock *origBB, std::vector<BasicBlock*> &llvm_prologues, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave, std::map<Value*, std::map<int, Value*> > &newValues, std::map<Value*, MachineBasicBlock*> &newValLocation) {
-
-  //Keep a map to easily know whats in the kernel
-  std::map<int, std::set<const MachineInstr*> > inKernel;
-  int maxStageCount = 0;
-
-  //Keep a map of new values we consumed in case they need to be added back
-  std::map<Value*, std::map<int, Value*> > consumedValues;
-
-  MSchedGraphNode *branch = 0;
-  MSchedGraphNode *BAbranch = 0;
-
-  DEBUG(schedule.print(std::cerr));
-
-  std::vector<MSchedGraphNode*> branches;
-
-  for(MSSchedule::kernel_iterator I = schedule.kernel_begin(), E = schedule.kernel_end(); I != E; ++I) {
-    maxStageCount = std::max(maxStageCount, I->second);
-
-    //Put int the map so we know what instructions in each stage are in the kernel
-    DEBUG(std::cerr << "Inserting instruction " << *(I->first) << " into map at stage " << I->second << "\n");
-    inKernel[I->second].insert(I->first);
-  }
-
-  //Get target information to look at machine operands
-  const TargetInstrInfo *mii = target.getInstrInfo();
-
- //Now write the prologues
-  for(int i = 0; i < maxStageCount; ++i) {
-    BasicBlock *llvmBB = new BasicBlock("PROLOGUE", (Function*) (origBB->getBasicBlock()->getParent()));
-    MachineBasicBlock *machineBB = new MachineBasicBlock(llvmBB);
-
-    DEBUG(std::cerr << "i=" << i << "\n");
-    for(int j = i; j >= 0; --j) {
-      for(MachineBasicBlock::const_iterator MI = origBB->begin(), ME = origBB->end(); ME != MI; ++MI) {
-        if(inKernel[j].count(&*MI)) {
-          MachineInstr *instClone = MI->clone();
-          machineBB->push_back(instClone);
-
-          //If its a branch, insert a nop
-          if(mii->isBranch(instClone->getOpcode()))
-            BuildMI(machineBB, V9::NOP, 0);
-
-
-          DEBUG(std::cerr << "Cloning: " << *MI << "\n");
-
-          //After cloning, we may need to save the value that this instruction defines
-          for(unsigned opNum=0; opNum < MI->getNumOperands(); ++opNum) {
-            Instruction *tmp;
-
-            //get machine operand
-            MachineOperand &mOp = instClone->getOperand(opNum);
-            if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef()) {
-
-              //Check if this is a value we should save
-              if(valuesToSave.count(mOp.getVRegValue())) {
-                //Save copy in tmpInstruction
-                tmp = new TmpInstruction(mOp.getVRegValue());
-
-                //Add TmpInstruction to safe LLVM Instruction MCFI
-                MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-                tempMvec.addTemp((Value*) tmp);
-
-                DEBUG(std::cerr << "Value: " << *(mOp.getVRegValue()) << " New Value: " << *tmp << " Stage: " << i << "\n");
-
-                newValues[mOp.getVRegValue()][i]= tmp;
-                newValLocation[tmp] = machineBB;
-
-                DEBUG(std::cerr << "Machine Instr Operands: " << *(mOp.getVRegValue()) << ", 0, " << *tmp << "\n");
-
-                //Create machine instruction and put int machineBB
-                MachineInstr *saveValue;
-                if(mOp.getVRegValue()->getType() == Type::FloatTy)
-                  saveValue = BuildMI(machineBB, V9::FMOVS, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-                else if(mOp.getVRegValue()->getType() == Type::DoubleTy)
-                  saveValue = BuildMI(machineBB, V9::FMOVD, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-                else
-                  saveValue = BuildMI(machineBB, V9::ORr, 3).addReg(mOp.getVRegValue()).addImm(0).addRegDef(tmp);
-
-
-                DEBUG(std::cerr << "Created new machine instr: " << *saveValue << "\n");
-              }
-            }
-
-            //We may also need to update the value that we use if its from an earlier prologue
-            if(j != 0) {
-              if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isUse()) {
-                if(newValues.count(mOp.getVRegValue())) {
-                  if(newValues[mOp.getVRegValue()].count(i-1)) {
-                    Value *oldV =  mOp.getVRegValue();
-                    DEBUG(std::cerr << "Replaced this value: " << mOp.getVRegValue() << " With:" << (newValues[mOp.getVRegValue()][i-1]) << "\n");
-                    //Update the operand with the right value
-                    mOp.setValueReg(newValues[mOp.getVRegValue()][i-1]);
-
-                    //Remove this value since we have consumed it
-                    //NOTE: Should this only be done if j != maxStage?
-                    consumedValues[oldV][i-1] = (newValues[oldV][i-1]);
-                    DEBUG(std::cerr << "Deleted value: " << consumedValues[oldV][i-1] << "\n");
-                    newValues[oldV].erase(i-1);
-                  }
-                }
-                else
-                  if(consumedValues.count(mOp.getVRegValue()))
-                    assert(!consumedValues[mOp.getVRegValue()].count(i-1) && "Found a case where we need the value");
-              }
-            }
-          }
-        }
-      }
-    }
-
-    MachineFunction *F = (((MachineBasicBlock*)origBB)->getParent());
-    MachineFunction::BasicBlockListType &BL = F->getBasicBlockList();
-    MachineFunction::BasicBlockListType::iterator BLI = origBB;
-    assert(BLI != BL.end() && "Must find original BB in machine function\n");
-    BL.insert(BLI,machineBB);
-    prologues.push_back(machineBB);
-    llvm_prologues.push_back(llvmBB);
-  }
-}
-
-void ModuloSchedulingPass::writeEpilogues(std::vector<MachineBasicBlock *> &epilogues, const MachineBasicBlock *origBB, std::vector<BasicBlock*> &llvm_epilogues, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave, std::map<Value*, std::map<int, Value*> > &newValues,std::map<Value*, MachineBasicBlock*> &newValLocation, std::map<Value*, std::map<int, Value*> > &kernelPHIs ) {
-
-  std::map<int, std::set<const MachineInstr*> > inKernel;
-
-  for(MSSchedule::kernel_iterator I = schedule.kernel_begin(), E = schedule.kernel_end(); I != E; ++I) {
-
-    //Ignore the branch, we will handle this separately
-    //if(I->first->isBranch())
-    //continue;
-
-    //Put int the map so we know what instructions in each stage are in the kernel
-    inKernel[I->second].insert(I->first);
-  }
-
-  std::map<Value*, Value*> valPHIs;
-
-  //some debug stuff, will remove later
-  DEBUG(for(std::map<Value*, std::map<int, Value*> >::iterator V = newValues.begin(), E = newValues.end(); V !=E; ++V) {
-    std::cerr << "Old Value: " << *(V->first) << "\n";
-    for(std::map<int, Value*>::iterator I = V->second.begin(), IE = V->second.end(); I != IE; ++I)
-      std::cerr << "Stage: " << I->first << " Value: " << *(I->second) << "\n";
-  });
-
-  //some debug stuff, will remove later
-  DEBUG(for(std::map<Value*, std::map<int, Value*> >::iterator V = kernelPHIs.begin(), E = kernelPHIs.end(); V !=E; ++V) {
-    std::cerr << "Old Value: " << *(V->first) << "\n";
-    for(std::map<int, Value*>::iterator I = V->second.begin(), IE = V->second.end(); I != IE; ++I)
-      std::cerr << "Stage: " << I->first << " Value: " << *(I->second) << "\n";
-  });
-
-  //Now write the epilogues
-  for(int i = schedule.getMaxStage()-1; i >= 0; --i) {
-    BasicBlock *llvmBB = new BasicBlock("EPILOGUE", (Function*) (origBB->getBasicBlock()->getParent()));
-    MachineBasicBlock *machineBB = new MachineBasicBlock(llvmBB);
-
-    DEBUG(std::cerr << " Epilogue #: " << i << "\n");
-
-
-    std::map<Value*, int> inEpilogue;
-
-     for(MachineBasicBlock::const_iterator MI = origBB->begin(), ME = origBB->end(); ME != MI; ++MI) {
-      for(int j=schedule.getMaxStage(); j > i; --j) {
-        if(inKernel[j].count(&*MI)) {
-          DEBUG(std::cerr << "Cloning instruction " << *MI << "\n");
-          MachineInstr *clone = MI->clone();
-
-          //Update operands that need to use the result from the phi
-          for(unsigned opNum=0; opNum < clone->getNumOperands(); ++opNum) {
-            //get machine operand
-            const MachineOperand &mOp = clone->getOperand(opNum);
-
-            if((mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isUse())) {
-
-              DEBUG(std::cerr << "Writing PHI for " << (mOp.getVRegValue()) << "\n");
-
-              //If this is the last instructions for the max iterations ago, don't update operands
-              if(inEpilogue.count(mOp.getVRegValue()))
-                if(inEpilogue[mOp.getVRegValue()] == i)
-                  continue;
-
-              //Quickly write appropriate phis for this operand
-              if(newValues.count(mOp.getVRegValue())) {
-                if(newValues[mOp.getVRegValue()].count(i)) {
-                  Instruction *tmp = new TmpInstruction(newValues[mOp.getVRegValue()][i]);
-
-                  //Get machine code for this instruction
-                  MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-                  tempMvec.addTemp((Value*) tmp);
-
-                  //assert of no kernelPHI for this value
-                  assert(kernelPHIs[mOp.getVRegValue()][i] !=0 && "Must have final kernel phi to construct epilogue phi");
-
-                  MachineInstr *saveValue = BuildMI(machineBB, V9::PHI, 3).addReg(newValues[mOp.getVRegValue()][i]).addReg(kernelPHIs[mOp.getVRegValue()][i]).addRegDef(tmp);
-                  DEBUG(std::cerr << "Resulting PHI: " << *saveValue << "\n");
-                  valPHIs[mOp.getVRegValue()] = tmp;
-                }
-              }
-
-              if(valPHIs.count(mOp.getVRegValue())) {
-                //Update the operand in the cloned instruction
-                clone->getOperand(opNum).setValueReg(valPHIs[mOp.getVRegValue()]);
-              }
-            }
-            else if((mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef())) {
-              inEpilogue[mOp.getVRegValue()] = i;
-            }
-          }
-          machineBB->push_back(clone);
-        }
-      }
-     }
-
-     MachineFunction *F = (((MachineBasicBlock*)origBB)->getParent());
-     MachineFunction::BasicBlockListType &BL = F->getBasicBlockList();
-     MachineFunction::BasicBlockListType::iterator BLI = (MachineBasicBlock*) origBB;
-     assert(BLI != BL.end() && "Must find original BB in machine function\n");
-     BL.insert(BLI,machineBB);
-     epilogues.push_back(machineBB);
-     llvm_epilogues.push_back(llvmBB);
-
-     DEBUG(std::cerr << "EPILOGUE #" << i << "\n");
-     DEBUG(machineBB->print(std::cerr));
-  }
-}
-
-void ModuloSchedulingPass::writeKernel(BasicBlock *llvmBB, MachineBasicBlock *machineBB, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave, std::map<Value*, std::map<int, Value*> > &newValues, std::map<Value*, MachineBasicBlock*> &newValLocation, std::map<Value*, std::map<int, Value*> > &kernelPHIs) {
-
-  //Keep track of operands that are read and saved from a previous iteration. The new clone
-  //instruction will use the result of the phi instead.
-  std::map<Value*, Value*> finalPHIValue;
-  std::map<Value*, Value*> kernelValue;
-
-  //Branches are a special case
-  std::vector<MachineInstr*> branches;
-
-  //Get target information to look at machine operands
-  const TargetInstrInfo *mii = target.getInstrInfo();
-
-  //Create TmpInstructions for the final phis
-  for(MSSchedule::kernel_iterator I = schedule.kernel_begin(), E = schedule.kernel_end(); I != E; ++I) {
-
-   DEBUG(std::cerr << "Stage: " << I->second << " Inst: " << *(I->first) << "\n";);
-
-   //Clone instruction
-   const MachineInstr *inst = I->first;
-   MachineInstr *instClone = inst->clone();
-
-   //Insert into machine basic block
-   machineBB->push_back(instClone);
-
-   if(mii->isBranch(instClone->getOpcode()))
-     BuildMI(machineBB, V9::NOP, 0);
-
-   DEBUG(std::cerr <<  "Cloned Inst: " << *instClone << "\n");
-
-   //Loop over Machine Operands
-   for(unsigned i=0; i < inst->getNumOperands(); ++i) {
-     //get machine operand
-     const MachineOperand &mOp = inst->getOperand(i);
-
-     if(I->second != 0) {
-       if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isUse()) {
-
-         //Check to see where this operand is defined if this instruction is from max stage
-         if(I->second == schedule.getMaxStage()) {
-           DEBUG(std::cerr << "VREG: " << *(mOp.getVRegValue()) << "\n");
-         }
-
-         //If its in the value saved, we need to create a temp instruction and use that instead
-         if(valuesToSave.count(mOp.getVRegValue())) {
-
-           //Check if we already have a final PHI value for this
-           if(!finalPHIValue.count(mOp.getVRegValue())) {
-             //Only create phi if the operand def is from a stage before this one
-             if(schedule.defPreviousStage(mOp.getVRegValue(), I->second)) {
-             TmpInstruction *tmp = new TmpInstruction(mOp.getVRegValue());
-
-             //Get machine code for this instruction
-             MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-             tempMvec.addTemp((Value*) tmp);
-
-             //Update the operand in the cloned instruction
-             instClone->getOperand(i).setValueReg(tmp);
-
-             //save this as our final phi
-             finalPHIValue[mOp.getVRegValue()] = tmp;
-             newValLocation[tmp] = machineBB;
-             }
-           }
-           else {
-             //Use the previous final phi value
-             instClone->getOperand(i).setValueReg(finalPHIValue[mOp.getVRegValue()]);
-           }
-         }
-       }
-     }
-     if(I->second != schedule.getMaxStage()) {
-       if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef()) {
-         if(valuesToSave.count(mOp.getVRegValue())) {
-
-           TmpInstruction *tmp = new TmpInstruction(mOp.getVRegValue());
-
-           //Get machine code for this instruction
-           MachineCodeForInstruction & tempVec = MachineCodeForInstruction::get(defaultInst);
-           tempVec.addTemp((Value*) tmp);
-
-           //Create new machine instr and put in MBB
-           MachineInstr *saveValue;
-           if(mOp.getVRegValue()->getType() == Type::FloatTy)
-             saveValue = BuildMI(machineBB, V9::FMOVS, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-           else if(mOp.getVRegValue()->getType() == Type::DoubleTy)
-             saveValue = BuildMI(machineBB, V9::FMOVD, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-           else
-             saveValue = BuildMI(machineBB, V9::ORr, 3).addReg(mOp.getVRegValue()).addImm(0).addRegDef(tmp);
-
-
-           //Save for future cleanup
-           kernelValue[mOp.getVRegValue()] = tmp;
-           newValLocation[tmp] = machineBB;
-           kernelPHIs[mOp.getVRegValue()][schedule.getMaxStage()-1] = tmp;
-         }
-       }
-     }
-   }
-
- }
-
- //Add branches
- for(std::vector<MachineInstr*>::iterator I = branches.begin(), E = branches.end(); I != E; ++I) {
-   machineBB->push_back(*I);
-   BuildMI(machineBB, V9::NOP, 0);
- }
-
-
-  DEBUG(std::cerr << "KERNEL before PHIs\n");
-  DEBUG(machineBB->print(std::cerr));
-
-
- //Loop over each value we need to generate phis for
- for(std::map<Value*, std::map<int, Value*> >::iterator V = newValues.begin(),
-       E = newValues.end(); V != E; ++V) {
-
-
-   DEBUG(std::cerr << "Writing phi for" << *(V->first));
-   DEBUG(std::cerr << "\nMap of Value* for this phi\n");
-   DEBUG(for(std::map<int, Value*>::iterator I = V->second.begin(),
-               IE = V->second.end(); I != IE; ++I) {
-     std::cerr << "Stage: " << I->first;
-     std::cerr << " Value: " << *(I->second) << "\n";
-   });
-
-   //If we only have one current iteration live, its safe to set lastPhi = to kernel value
-   if(V->second.size() == 1) {
-     assert(kernelValue[V->first] != 0 && "Kernel value* must exist to create phi");
-     MachineInstr *saveValue = BuildMI(*machineBB, machineBB->begin(),V9::PHI, 3).addReg(V->second.begin()->second).addReg(kernelValue[V->first]).addRegDef(finalPHIValue[V->first]);
-     DEBUG(std::cerr << "Resulting PHI (one live): " << *saveValue << "\n");
-     kernelPHIs[V->first][V->second.begin()->first] = kernelValue[V->first];
-     DEBUG(std::cerr << "Put kernel phi in at stage: " << schedule.getMaxStage()-1 << " (map stage = " << V->second.begin()->first << ")\n");
-    }
-   else {
-
-     //Keep track of last phi created.
-     Instruction *lastPhi = 0;
-
-     unsigned count = 1;
-     //Loop over the the map backwards to generate phis
-     for(std::map<int, Value*>::reverse_iterator I = V->second.rbegin(), IE = V->second.rend();
-         I != IE; ++I) {
-
-       if(count < (V->second).size()) {
-         if(lastPhi == 0) {
-           lastPhi = new TmpInstruction(I->second);
-
-           //Get machine code for this instruction
-           MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-           tempMvec.addTemp((Value*) lastPhi);
-
-           MachineInstr *saveValue = BuildMI(*machineBB, machineBB->begin(), V9::PHI, 3).addReg(kernelValue[V->first]).addReg(I->second).addRegDef(lastPhi);
-           DEBUG(std::cerr << "Resulting PHI: " << *saveValue << "\n");
-           newValLocation[lastPhi] = machineBB;
-         }
-         else {
-           Instruction *tmp = new TmpInstruction(I->second);
-
-           //Get machine code for this instruction
-           MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-           tempMvec.addTemp((Value*) tmp);
-
-
-           MachineInstr *saveValue = BuildMI(*machineBB, machineBB->begin(), V9::PHI, 3).addReg(lastPhi).addReg(I->second).addRegDef(tmp);
-           DEBUG(std::cerr << "Resulting PHI: " << *saveValue << "\n");
-           lastPhi = tmp;
-           kernelPHIs[V->first][I->first] = lastPhi;
-           newValLocation[lastPhi] = machineBB;
-         }
-       }
-       //Final phi value
-       else {
-         //The resulting value must be the Value* we created earlier
-         assert(lastPhi != 0 && "Last phi is NULL!\n");
-         MachineInstr *saveValue = BuildMI(*machineBB, machineBB->begin(), V9::PHI, 3).addReg(lastPhi).addReg(I->second).addRegDef(finalPHIValue[V->first]);
-         DEBUG(std::cerr << "Resulting PHI: " << *saveValue << "\n");
-         kernelPHIs[V->first][I->first] = finalPHIValue[V->first];
-       }
-
-       ++count;
-     }
-
-   }
- }
-
-  DEBUG(std::cerr << "KERNEL after PHIs\n");
-  DEBUG(machineBB->print(std::cerr));
-}
-
-
-void ModuloSchedulingPass::removePHIs(const MachineBasicBlock *origBB, std::vector<MachineBasicBlock *> &prologues, std::vector<MachineBasicBlock *> &epilogues, MachineBasicBlock *kernelBB, std::map<Value*, MachineBasicBlock*> &newValLocation) {
-
-  //Worklist to delete things
-  std::vector<std::pair<MachineBasicBlock*, MachineBasicBlock::iterator> > worklist;
-
-  //Worklist of TmpInstructions that need to be added to a MCFI
-  std::vector<Instruction*> addToMCFI;
-
-  //Worklist to add OR instructions to end of kernel so not to invalidate the iterator
-  //std::vector<std::pair<Instruction*, Value*> > newORs;
-
-  const TargetInstrInfo *TMI = target.getInstrInfo();
-
-  //Start with the kernel and for each phi insert a copy for the phi def and for each arg
-  for(MachineBasicBlock::iterator I = kernelBB->begin(), E = kernelBB->end(); I != E; ++I) {
-
-    DEBUG(std::cerr << "Looking at Instr: " << *I << "\n");
-    //Get op code and check if its a phi
-    if(I->getOpcode() == V9::PHI) {
-
-      DEBUG(std::cerr << "Replacing PHI: " << *I << "\n");
-      Instruction *tmp = 0;
-
-      for(unsigned i = 0; i < I->getNumOperands(); ++i) {
-        //Get Operand
-        const MachineOperand &mOp = I->getOperand(i);
-        assert(mOp.getType() == MachineOperand::MO_VirtualRegister && "Should be a Value*\n");
-
-        if(!tmp) {
-          tmp = new TmpInstruction(mOp.getVRegValue());
-          addToMCFI.push_back(tmp);
-        }
-
-        //Now for all our arguments we read, OR to the new TmpInstruction that we created
-        if(mOp.isUse()) {
-          DEBUG(std::cerr << "Use: " << mOp << "\n");
-          //Place a copy at the end of its BB but before the branches
-          assert(newValLocation.count(mOp.getVRegValue()) && "We must know where this value is located\n");
-          //Reverse iterate to find the branches, we can safely assume no instructions have been
-          //put in the nop positions
-          for(MachineBasicBlock::iterator inst = --(newValLocation[mOp.getVRegValue()])->end(), endBB = (newValLocation[mOp.getVRegValue()])->begin(); inst != endBB; --inst) {
-            MachineOpCode opc = inst->getOpcode();
-            if(TMI->isBranch(opc) || TMI->isNop(opc))
-              continue;
-            else {
-              if(mOp.getVRegValue()->getType() == Type::FloatTy)
-                BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::FMOVS, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-              else if(mOp.getVRegValue()->getType() == Type::DoubleTy)
-                BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::FMOVD, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-              else
-                BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::ORr, 3).addReg(mOp.getVRegValue()).addImm(0).addRegDef(tmp);
-
-              break;
-            }
-
-          }
-
-        }
-        else {
-          //Remove the phi and replace it with an OR
-          DEBUG(std::cerr << "Def: " << mOp << "\n");
-          //newORs.push_back(std::make_pair(tmp, mOp.getVRegValue()));
-          if(tmp->getType() == Type::FloatTy)
-            BuildMI(*kernelBB, I, V9::FMOVS, 3).addReg(tmp).addRegDef(mOp.getVRegValue());
-          else if(tmp->getType() == Type::DoubleTy)
-            BuildMI(*kernelBB, I, V9::FMOVD, 3).addReg(tmp).addRegDef(mOp.getVRegValue());
-          else
-            BuildMI(*kernelBB, I, V9::ORr, 3).addReg(tmp).addImm(0).addRegDef(mOp.getVRegValue());
-
-
-          worklist.push_back(std::make_pair(kernelBB, I));
-        }
-
-      }
-
-    }
-
-
-  }
-
-  //Add TmpInstructions to some MCFI
-  if(addToMCFI.size() > 0) {
-    MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-    for(unsigned x = 0; x < addToMCFI.size(); ++x) {
-      tempMvec.addTemp(addToMCFI[x]);
-    }
-    addToMCFI.clear();
-  }
-
-
-  //Remove phis from epilogue
-  for(std::vector<MachineBasicBlock*>::iterator MB = epilogues.begin(), ME = epilogues.end(); MB != ME; ++MB) {
-    for(MachineBasicBlock::iterator I = (*MB)->begin(), E = (*MB)->end(); I != E; ++I) {
-
-      DEBUG(std::cerr << "Looking at Instr: " << *I << "\n");
-      //Get op code and check if its a phi
-      if(I->getOpcode() == V9::PHI) {
-        Instruction *tmp = 0;
-
-        for(unsigned i = 0; i < I->getNumOperands(); ++i) {
-          //Get Operand
-          const MachineOperand &mOp = I->getOperand(i);
-          assert(mOp.getType() == MachineOperand::MO_VirtualRegister && "Should be a Value*\n");
-
-          if(!tmp) {
-            tmp = new TmpInstruction(mOp.getVRegValue());
-            addToMCFI.push_back(tmp);
-          }
-
-          //Now for all our arguments we read, OR to the new TmpInstruction that we created
-          if(mOp.isUse()) {
-            DEBUG(std::cerr << "Use: " << mOp << "\n");
-            //Place a copy at the end of its BB but before the branches
-            assert(newValLocation.count(mOp.getVRegValue()) && "We must know where this value is located\n");
-            //Reverse iterate to find the branches, we can safely assume no instructions have been
-            //put in the nop positions
-            for(MachineBasicBlock::iterator inst = --(newValLocation[mOp.getVRegValue()])->end(), endBB = (newValLocation[mOp.getVRegValue()])->begin(); inst != endBB; --inst) {
-              MachineOpCode opc = inst->getOpcode();
-              if(TMI->isBranch(opc) || TMI->isNop(opc))
-                continue;
-              else {
-                if(mOp.getVRegValue()->getType() == Type::FloatTy)
-                  BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::FMOVS, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-                else if(mOp.getVRegValue()->getType() == Type::DoubleTy)
-                  BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::FMOVD, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-                else
-                  BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::ORr, 3).addReg(mOp.getVRegValue()).addImm(0).addRegDef(tmp);
-
-
-                break;
-              }
-
-            }
-
-          }
-          else {
-            //Remove the phi and replace it with an OR
-            DEBUG(std::cerr << "Def: " << mOp << "\n");
-             if(tmp->getType() == Type::FloatTy)
-               BuildMI(**MB, I, V9::FMOVS, 3).addReg(tmp).addRegDef(mOp.getVRegValue());
-             else if(tmp->getType() == Type::DoubleTy)
-               BuildMI(**MB, I, V9::FMOVD, 3).addReg(tmp).addRegDef(mOp.getVRegValue());
-             else
-               BuildMI(**MB, I, V9::ORr, 3).addReg(tmp).addImm(0).addRegDef(mOp.getVRegValue());
-
-            worklist.push_back(std::make_pair(*MB,I));
-          }
-
-        }
-      }
-
-
-    }
-  }
-
-
-  if(addToMCFI.size() > 0) {
-    MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-    for(unsigned x = 0; x < addToMCFI.size(); ++x) {
-      tempMvec.addTemp(addToMCFI[x]);
-    }
-    addToMCFI.clear();
-  }
-
-    //Delete the phis
-  for(std::vector<std::pair<MachineBasicBlock*, MachineBasicBlock::iterator> >::iterator I =  worklist.begin(), E = worklist.end(); I != E; ++I) {
-
-    DEBUG(std::cerr << "Deleting PHI " << *I->second << "\n");
-    I->first->erase(I->second);
-
-  }
-
-
-  assert((addToMCFI.size() == 0) && "We should have added all TmpInstructions to some MachineCodeForInstruction");
-}
-
-
-void ModuloSchedulingPass::reconstructLoop(MachineBasicBlock *BB) {
-
-  TIME_REGION(X, "reconstructLoop");
-
-
-  DEBUG(std::cerr << "Reconstructing Loop\n");
-
-  //First find the value *'s that we need to "save"
-  std::map<const Value*, std::pair<const MachineInstr*, int> > valuesToSave;
-
-  //Keep track of instructions we have already seen and their stage because
-  //we don't want to "save" values if they are used in the kernel immediately
-  std::map<const MachineInstr*, int> lastInstrs;
-  std::map<const Value*, int> phiUses;
-
-  //Loop over kernel and only look at instructions from a stage > 0
-  //Look at its operands and save values *'s that are read
-  for(MSSchedule::kernel_iterator I = schedule.kernel_begin(), E = schedule.kernel_end(); I != E; ++I) {
-
-    if(I->second !=0) {
-      //For this instruction, get the Value*'s that it reads and put them into the set.
-      //Assert if there is an operand of another type that we need to save
-      const MachineInstr *inst = I->first;
-      lastInstrs[inst] = I->second;
-
-      for(unsigned i=0; i < inst->getNumOperands(); ++i) {
-        //get machine operand
-        const MachineOperand &mOp = inst->getOperand(i);
-
-        if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isUse()) {
-          //find the value in the map
-          if (const Value* srcI = mOp.getVRegValue()) {
-
-            if(isa<Constant>(srcI) || isa<Argument>(srcI))
-              continue;
-
-            //Before we declare this Value* one that we should save
-            //make sure its def is not of the same stage as this instruction
-            //because it will be consumed before its used
-            Instruction *defInst = (Instruction*) srcI;
-
-            //Should we save this value?
-            bool save = true;
-
-            //Continue if not in the def map, loop invariant code does not need to be saved
-            if(!defMap.count(srcI))
-              continue;
-
-            MachineInstr *defInstr = defMap[srcI];
-
-
-            if(lastInstrs.count(defInstr)) {
-              if(lastInstrs[defInstr] == I->second) {
-                save = false;
-
-              }
-            }
-
-            if(save) {
-              assert(!phiUses.count(srcI) && "Did not expect to see phi use twice");
-              if(isa<PHINode>(srcI))
-                phiUses[srcI] = I->second;
-
-              valuesToSave[srcI] = std::make_pair(I->first, i);
-
-            }
-          }
-        }
-        else if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef()) {
-          if (const Value* destI = mOp.getVRegValue()) {
-            if(!isa<PHINode>(destI))
-              continue;
-            if(phiUses.count(destI)) {
-              if(phiUses[destI] == I->second) {
-                //remove from save list
-                valuesToSave.erase(destI);
-              }
-            }
-          }
-        }
-
-        if(mOp.getType() != MachineOperand::MO_VirtualRegister && mOp.isUse()) {
-          assert("Our assumption is wrong. We have another type of register that needs to be saved\n");
-        }
-      }
-    }
-  }
-
-  //The new loop will consist of one or more prologues, the kernel, and one or more epilogues.
-
-  //Map to keep track of old to new values
-  std::map<Value*, std::map<int, Value*> > newValues;
-
-  //Map to keep track of old to new values in kernel
-  std::map<Value*, std::map<int, Value*> > kernelPHIs;
-
-  //Another map to keep track of what machine basic blocks these new value*s are in since
-  //they have no llvm instruction equivalent
-  std::map<Value*, MachineBasicBlock*> newValLocation;
-
-  std::vector<MachineBasicBlock*> prologues;
-  std::vector<BasicBlock*> llvm_prologues;
-
-
-  //Write prologue
-  if(schedule.getMaxStage() != 0)
-    writePrologues(prologues, BB, llvm_prologues, valuesToSave, newValues, newValLocation);
-
-  //Print out epilogues and prologue
-  DEBUG(for(std::vector<MachineBasicBlock*>::iterator I = prologues.begin(), E = prologues.end();
-      I != E; ++I) {
-    std::cerr << "PROLOGUE\n";
-    (*I)->print(std::cerr);
-  });
-
-  BasicBlock *llvmKernelBB = new BasicBlock("Kernel", (Function*) (BB->getBasicBlock()->getParent()));
-  MachineBasicBlock *machineKernelBB = new MachineBasicBlock(llvmKernelBB);
-
-  MachineFunction *F = (((MachineBasicBlock*)BB)->getParent());
-  MachineFunction::BasicBlockListType &BL = F->getBasicBlockList();
-  MachineFunction::BasicBlockListType::iterator BLI = BB;
-  assert(BLI != BL.end() && "Must find original BB in machine function\n");
-  BL.insert(BLI,machineKernelBB);
-
-  //(((MachineBasicBlock*)BB)->getParent())->getBasicBlockList().push_back(machineKernelBB);
-  writeKernel(llvmKernelBB, machineKernelBB, valuesToSave, newValues, newValLocation, kernelPHIs);
-
-
-  std::vector<MachineBasicBlock*> epilogues;
-  std::vector<BasicBlock*> llvm_epilogues;
-
-  //Write epilogues
-  if(schedule.getMaxStage() != 0)
-    writeEpilogues(epilogues, BB, llvm_epilogues, valuesToSave, newValues, newValLocation, kernelPHIs);
-
-
-  //Fix our branches
-  fixBranches(prologues, llvm_prologues, machineKernelBB, llvmKernelBB, epilogues, llvm_epilogues, BB);
-
-  //Remove phis
-  removePHIs(BB, prologues, epilogues, machineKernelBB, newValLocation);
-
-  //Print out epilogues and prologue
-  DEBUG(for(std::vector<MachineBasicBlock*>::iterator I = prologues.begin(), E = prologues.end();
-      I != E; ++I) {
-    std::cerr << "PROLOGUE\n";
-    (*I)->print(std::cerr);
-  });
-
-  DEBUG(std::cerr << "KERNEL\n");
-  DEBUG(machineKernelBB->print(std::cerr));
-
-  DEBUG(for(std::vector<MachineBasicBlock*>::iterator I = epilogues.begin(), E = epilogues.end();
-      I != E; ++I) {
-    std::cerr << "EPILOGUE\n";
-    (*I)->print(std::cerr);
-  });
-
-
-  DEBUG(std::cerr << "New Machine Function" << "\n");
-  DEBUG(std::cerr << BB->getParent() << "\n");
-
-
-}
-
-void ModuloSchedulingPass::fixBranches(std::vector<MachineBasicBlock *> &prologues, std::vector<BasicBlock*> &llvm_prologues, MachineBasicBlock *machineKernelBB, BasicBlock *llvmKernelBB, std::vector<MachineBasicBlock *> &epilogues, std::vector<BasicBlock*> &llvm_epilogues, MachineBasicBlock *BB) {
-
-  const TargetInstrInfo *TMI = target.getInstrInfo();
-
-  if(schedule.getMaxStage() != 0) {
-    //Fix prologue branches
-    for(unsigned I = 0; I <  prologues.size(); ++I) {
-
-      //Find terminator since getFirstTerminator does not work!
-      for(MachineBasicBlock::reverse_iterator mInst = prologues[I]->rbegin(), mInstEnd = prologues[I]->rend(); mInst != mInstEnd; ++mInst) {
-        MachineOpCode OC = mInst->getOpcode();
-        //If its a branch update its branchto
-        if(TMI->isBranch(OC)) {
-          for(unsigned opNum = 0; opNum < mInst->getNumOperands(); ++opNum) {
-            MachineOperand &mOp = mInst->getOperand(opNum);
-            if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-              //Check if we are branching to the kernel, if not branch to epilogue
-              if(mOp.getVRegValue() == BB->getBasicBlock()) {
-                if(I == prologues.size()-1)
-                  mOp.setValueReg(llvmKernelBB);
-                else
-                  mOp.setValueReg(llvm_prologues[I+1]);
-              }
-              else {
-                mOp.setValueReg(llvm_epilogues[(llvm_epilogues.size()-1-I)]);
-              }
-            }
-          }
-
-          DEBUG(std::cerr << "New Prologue Branch: " << *mInst << "\n");
-        }
-      }
-
-
-      //Update llvm basic block with our new branch instr
-      DEBUG(std::cerr << BB->getBasicBlock()->getTerminator() << "\n");
-      const BranchInst *branchVal = dyn_cast<BranchInst>(BB->getBasicBlock()->getTerminator());
-
-      if(I == prologues.size()-1) {
-        TerminatorInst *newBranch = new BranchInst(llvmKernelBB,
-                                                   llvm_epilogues[(llvm_epilogues.size()-1-I)],
-                                                   branchVal->getCondition(),
-                                                   llvm_prologues[I]);
-      }
-      else
-        TerminatorInst *newBranch = new BranchInst(llvm_prologues[I+1],
-                                                   llvm_epilogues[(llvm_epilogues.size()-1-I)],
-                                                   branchVal->getCondition(),
-                                                   llvm_prologues[I]);
-
-    }
-  }
-
-  Value *origBranchExit = 0;
-
-  //Fix up kernel machine branches
-  for(MachineBasicBlock::reverse_iterator mInst = machineKernelBB->rbegin(), mInstEnd = machineKernelBB->rend(); mInst != mInstEnd; ++mInst) {
-    MachineOpCode OC = mInst->getOpcode();
-    if(TMI->isBranch(OC)) {
-      for(unsigned opNum = 0; opNum < mInst->getNumOperands(); ++opNum) {
-        MachineOperand &mOp = mInst->getOperand(opNum);
-
-        if(mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-          if(mOp.getVRegValue() == BB->getBasicBlock())
-            mOp.setValueReg(llvmKernelBB);
-          else
-            if(llvm_epilogues.size() > 0) {
-              assert(origBranchExit == 0 && "There should only be one branch out of the loop");
-
-              origBranchExit = mOp.getVRegValue();
-              mOp.setValueReg(llvm_epilogues[0]);
-            }
-            else
-              origBranchExit = mOp.getVRegValue();
-        }
-      }
-    }
-  }
-
-  //Update kernelLLVM branches
-  const BranchInst *branchVal = dyn_cast<BranchInst>(BB->getBasicBlock()->getTerminator());
-
-  assert(origBranchExit != 0 && "We must have the original bb the kernel exits to!");
-
-  if(epilogues.size() > 0) {
-    TerminatorInst *newBranch = new BranchInst(llvmKernelBB,
-                                               llvm_epilogues[0],
-                                               branchVal->getCondition(),
-                                               llvmKernelBB);
-  }
-  else {
-    BasicBlock *origBBExit = dyn_cast<BasicBlock>(origBranchExit);
-    assert(origBBExit !=0 && "Original exit basic block must be set");
-    TerminatorInst *newBranch = new BranchInst(llvmKernelBB,
-                                               origBBExit,
-                                               branchVal->getCondition(),
-                                               llvmKernelBB);
-  }
-
-  if(schedule.getMaxStage() != 0) {
-   //Lastly add unconditional branches for the epilogues
-   for(unsigned I = 0; I <  epilogues.size(); ++I) {
-
-    //Now since we don't have fall throughs, add a unconditional branch to the next prologue
-     if(I != epilogues.size()-1) {
-       BuildMI(epilogues[I], V9::BA, 1).addPCDisp(llvm_epilogues[I+1]);
-       //Add unconditional branch to end of epilogue
-       TerminatorInst *newBranch = new BranchInst(llvm_epilogues[I+1],
-                                                  llvm_epilogues[I]);
-
-     }
-     else {
-       BuildMI(epilogues[I], V9::BA, 1).addPCDisp(origBranchExit);
-
-
-       //Update last epilogue exit branch
-       BranchInst *branchVal = (BranchInst*) dyn_cast<BranchInst>(BB->getBasicBlock()->getTerminator());
-       //Find where we are supposed to branch to
-       BasicBlock *nextBlock = 0;
-       for(unsigned j=0; j <branchVal->getNumSuccessors(); ++j) {
-         if(branchVal->getSuccessor(j) != BB->getBasicBlock())
-           nextBlock = branchVal->getSuccessor(j);
-       }
-
-       assert((nextBlock != 0) && "Next block should not be null!");
-       TerminatorInst *newBranch = new BranchInst(nextBlock, llvm_epilogues[I]);
-     }
-     //Add one more nop!
-     BuildMI(epilogues[I], V9::NOP, 0);
-
-   }
-  }
-
-   //FIX UP Machine BB entry!!
-   //We are looking at the predecesor of our loop basic block and we want to change its ba instruction
-
-
-   //Find all llvm basic blocks that branch to the loop entry and change to our first prologue.
-   const BasicBlock *llvmBB = BB->getBasicBlock();
-
-   std::vector<const BasicBlock*>Preds (pred_begin(llvmBB), pred_end(llvmBB));
-
-   //for(pred_const_iterator P = pred_begin(llvmBB), PE = pred_end(llvmBB); P != PE; ++PE) {
-   for(std::vector<const BasicBlock*>::iterator P = Preds.begin(), PE = Preds.end(); P != PE; ++P) {
-     if(*P == llvmBB)
-       continue;
-     else {
-       DEBUG(std::cerr << "Found our entry BB\n");
-       //Get the Terminator instruction for this basic block and print it out
-       DEBUG(std::cerr << *((*P)->getTerminator()) << "\n");
-       //Update the terminator
-       TerminatorInst *term = ((BasicBlock*)*P)->getTerminator();
-       for(unsigned i=0; i < term->getNumSuccessors(); ++i) {
-         if(term->getSuccessor(i) == llvmBB) {
-           DEBUG(std::cerr << "Replacing successor bb\n");
-           if(llvm_prologues.size() > 0) {
-             term->setSuccessor(i, llvm_prologues[0]);
-             //Also update its corresponding machine instruction
-             MachineCodeForInstruction & tempMvec =
-               MachineCodeForInstruction::get(term);
-             for (unsigned j = 0; j < tempMvec.size(); j++) {
-               MachineInstr *temp = tempMvec[j];
-               MachineOpCode opc = temp->getOpcode();
-               if(TMI->isBranch(opc)) {
-                 DEBUG(std::cerr << *temp << "\n");
-                 //Update branch
-                 for(unsigned opNum = 0; opNum < temp->getNumOperands(); ++opNum) {
-                   MachineOperand &mOp = temp->getOperand(opNum);
-                   if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-                     if(mOp.getVRegValue() == llvmBB)
-                       mOp.setValueReg(llvm_prologues[0]);
-                   }
-                 }
-               }
-             }
-           }
-           else {
-             term->setSuccessor(i, llvmKernelBB);
-           //Also update its corresponding machine instruction
-             MachineCodeForInstruction & tempMvec =
-               MachineCodeForInstruction::get(term);
-             for (unsigned j = 0; j < tempMvec.size(); j++) {
-               MachineInstr *temp = tempMvec[j];
-               MachineOpCode opc = temp->getOpcode();
-               if(TMI->isBranch(opc)) {
-                 DEBUG(std::cerr << *temp << "\n");
-                 //Update branch
-                 for(unsigned opNum = 0; opNum < temp->getNumOperands(); ++opNum) {
-                   MachineOperand &mOp = temp->getOperand(opNum);
-                   if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-                     if(mOp.getVRegValue() == llvmBB)
-                       mOp.setValueReg(llvmKernelBB);
-                   }
-                 }
-               }
-             }
-           }
-         }
-       }
-       break;
-     }
-   }
-
-
-  //BB->getParent()->getBasicBlockList().erase(BB);
-
-}
-
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloScheduling.h b/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloScheduling.h
deleted file mode 100644
index 840623e02d7..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloScheduling.h
+++ /dev/null
@@ -1,171 +0,0 @@
-//===-- ModuloScheduling.h - Swing Modulo Scheduling------------*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_MODULOSCHEDULING_H
-#define LLVM_MODULOSCHEDULING_H
-
-#include "MSchedGraph.h"
-#include "MSSchedule.h"
-#include "llvm/Function.h"
-#include "llvm/Pass.h"
-#include "DependenceAnalyzer.h"
-#include "llvm/Target/TargetData.h"
-#include "llvm/Analysis/LoopInfo.h"
-#include "llvm/Analysis/ScalarEvolution.h"
-#include <set>
-
-namespace llvm {
-
-
-  //Struct to contain ModuloScheduling Specific Information for each node
-  struct MSNodeAttributes {
-    int ASAP; //Earliest time at which the opreation can be scheduled
-    int ALAP; //Latest time at which the operation can be scheduled.
-    int MOB;
-    int depth;
-    int height;
-    MSNodeAttributes(int asap=-1, int alap=-1, int mob=-1,
-                             int d=-1, int h=-1) : ASAP(asap), ALAP(alap),
-                                                   MOB(mob), depth(d),
-                                                   height(h) {}
-  };
-
-
-  class ModuloSchedulingPass : public FunctionPass {
-    const TargetMachine &target;
-
-    //Map to hold Value* defs
-    std::map<const Value*, MachineInstr*> defMap;
-
-    //Map to hold list of instructions associate to the induction var for each BB
-    std::map<const MachineBasicBlock*, std::map<const MachineInstr*, unsigned> > indVarInstrs;
-
-    //Map to hold machine to  llvm instrs for each valid BB
-    std::map<const MachineBasicBlock*, std::map<MachineInstr*, Instruction*> > machineTollvm;
-
-    //LLVM Instruction we know we can add TmpInstructions to its MCFI
-    Instruction *defaultInst;
-
-    //Map that holds node to node attribute information
-    std::map<MSchedGraphNode*, MSNodeAttributes> nodeToAttributesMap;
-
-    //Map to hold all reccurrences
-    std::set<std::pair<int, std::vector<MSchedGraphNode*> > > recurrenceList;
-
-    //Set of edges to ignore, stored as src node and index into vector of successors
-    std::set<std::pair<MSchedGraphNode*, unsigned> > edgesToIgnore;
-
-    //Vector containing the partial order
-    std::vector<std::set<MSchedGraphNode*> > partialOrder;
-
-    //Vector containing the final node order
-    std::vector<MSchedGraphNode*> FinalNodeOrder;
-
-    //Schedule table, key is the cycle number and the vector is resource, node pairs
-    MSSchedule schedule;
-
-    //Current initiation interval
-    int II;
-
-    //Internal functions
-    bool CreateDefMap(MachineBasicBlock *BI);
-    bool MachineBBisValid(const MachineBasicBlock *BI);
-    bool assocIndVar(Instruction *I, std::set<Instruction*> &indVar,
-                     std::vector<Instruction*> &stack, BasicBlock *BB);
-    int calculateResMII(const MachineBasicBlock *BI);
-    int calculateRecMII(MSchedGraph *graph, int MII);
-    void calculateNodeAttributes(MSchedGraph *graph, int MII);
-
-    bool ignoreEdge(MSchedGraphNode *srcNode, MSchedGraphNode *destNode);
-
-    int calculateASAP(MSchedGraphNode *node, int MII,MSchedGraphNode *destNode);
-    int calculateALAP(MSchedGraphNode *node, int MII, int maxASAP, MSchedGraphNode *srcNode);
-
-    int calculateHeight(MSchedGraphNode *node,MSchedGraphNode *srcNode);
-    int calculateDepth(MSchedGraphNode *node, MSchedGraphNode *destNode);
-
-    int findMaxASAP();
-    void orderNodes();
-    void findAllReccurrences(MSchedGraphNode *node,
-                             std::vector<MSchedGraphNode*> &visitedNodes, int II);
-    void addReccurrence(std::vector<MSchedGraphNode*> &recurrence, int II, MSchedGraphNode*, MSchedGraphNode*);
-    void addSCC(std::vector<MSchedGraphNode*> &SCC, std::map<MSchedGraphNode*, MSchedGraphNode*> &newNodes);
-
-    void findAllCircuits(MSchedGraph *MSG, int II);
-    bool circuit(MSchedGraphNode *v, std::vector<MSchedGraphNode*> &stack,
-                 std::set<MSchedGraphNode*> &blocked,
-                 std::vector<MSchedGraphNode*> &SCC, MSchedGraphNode *s,
-                 std::map<MSchedGraphNode*, std::set<MSchedGraphNode*> > &B, int II,
-                 std::map<MSchedGraphNode*, MSchedGraphNode*> &newNodes);
-
-    void unblock(MSchedGraphNode *u, std::set<MSchedGraphNode*> &blocked,
-                 std::map<MSchedGraphNode*, std::set<MSchedGraphNode*> > &B);
-
-    void addRecc(std::vector<MSchedGraphNode*> &stack, std::map<MSchedGraphNode*, MSchedGraphNode*> &newNodes);
-
-    void searchPath(MSchedGraphNode *node,
-                    std::vector<MSchedGraphNode*> &path,
-                    std::set<MSchedGraphNode*> &nodesToAdd,
-                    std::set<MSchedGraphNode*> &new_reccurence);
-
-    void pathToRecc(MSchedGraphNode *node,
-                    std::vector<MSchedGraphNode*> &path,
-                    std::set<MSchedGraphNode*> &poSet, std::set<MSchedGraphNode*> &lastNodes);
-
-    void computePartialOrder();
-
-    bool computeSchedule(const MachineBasicBlock *BB, MSchedGraph *MSG);
-    bool scheduleNode(MSchedGraphNode *node,
-                      int start, int end);
-
-    void predIntersect(std::set<MSchedGraphNode*> &CurrentSet, std::set<MSchedGraphNode*> &IntersectResult);
-    void succIntersect(std::set<MSchedGraphNode*> &CurrentSet, std::set<MSchedGraphNode*> &IntersectResult);
-
-    void reconstructLoop(MachineBasicBlock*);
-
-    //void saveValue(const MachineInstr*, const std::set<Value*>&, std::vector<Value*>*);
-
-    void fixBranches(std::vector<MachineBasicBlock *> &prologues, std::vector<BasicBlock*> &llvm_prologues, MachineBasicBlock *machineBB, BasicBlock *llvmBB, std::vector<MachineBasicBlock *> &epilogues, std::vector<BasicBlock*> &llvm_epilogues, MachineBasicBlock*);
-
-    void writePrologues(std::vector<MachineBasicBlock *> &prologues, MachineBasicBlock *origBB, std::vector<BasicBlock*> &llvm_prologues, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave, std::map<Value*, std::map<int, Value*> > &newValues, std::map<Value*, MachineBasicBlock*> &newValLocation);
-
-    void writeEpilogues(std::vector<MachineBasicBlock *> &epilogues, const MachineBasicBlock *origBB, std::vector<BasicBlock*> &llvm_epilogues, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave,std::map<Value*, std::map<int, Value*> > &newValues, std::map<Value*, MachineBasicBlock*> &newValLocation,  std::map<Value*, std::map<int, Value*> > &kernelPHIs);
-
-
-    void writeKernel(BasicBlock *llvmBB, MachineBasicBlock *machineBB, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave, std::map<Value*, std::map<int, Value*> > &newValues, std::map<Value*, MachineBasicBlock*> &newValLocation, std::map<Value*, std::map<int, Value*> > &kernelPHIs);
-
-    void removePHIs(const MachineBasicBlock* SB, std::vector<MachineBasicBlock*> &prologues, std::vector<MachineBasicBlock *> &epilogues, MachineBasicBlock *kernelBB, std::map<Value*, MachineBasicBlock*> &newValLocation);
-
-    void connectedComponentSet(MSchedGraphNode *node, std::set<MSchedGraphNode*> &ccSet, std::set<MSchedGraphNode*> &lastNodes);
-
-  public:
-    ModuloSchedulingPass(TargetMachine &targ) : target(targ) {}
-    virtual bool runOnFunction(Function &F);
-    virtual const char* getPassName() const { return "ModuloScheduling"; }
-
-    // getAnalysisUsage
-    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
-      /// HACK: We don't actually need loopinfo or scev, but we have
-      /// to say we do so that the pass manager does not delete it
-      /// before we run.
-      AU.addRequired<LoopInfo>();
-      AU.addRequired<ScalarEvolution>();
-
-      AU.addRequired<DependenceAnalyzer>();
-    }
-
-  };
-
-}
-
-
-#endif
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloSchedulingSuperBlock.cpp b/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloSchedulingSuperBlock.cpp
deleted file mode 100644
index 8b3185155e1..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloSchedulingSuperBlock.cpp
+++ /dev/null
@@ -1,3155 +0,0 @@
-//===-- ModuloSchedulingSuperBlock.cpp - ModuloScheduling--------*- C++ -*-===//
-//
-//                     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 ModuloScheduling pass is based on the Swing Modulo Scheduling
-//  algorithm, but has been extended to support SuperBlocks (multiple
-//  basic block, single entry, multipl exit loops).
-//
-//===----------------------------------------------------------------------===//
-
-#define DEBUG_TYPE "ModuloSchedSB"
-
-#include "DependenceAnalyzer.h"
-#include "ModuloSchedulingSuperBlock.h"
-#include "llvm/Constants.h"
-#include "llvm/ADT/Statistic.h"
-#include "llvm/CodeGen/MachineFunction.h"
-#include "llvm/CodeGen/Passes.h"
-#include "llvm/Support/CFG.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/Support/GraphWriter.h"
-#include "llvm/Support/Timer.h"
-#include "llvm/ADT/StringExtras.h"
-#include "llvm/ADT/SCCIterator.h"
-#include "llvm/Instructions.h"
-#include "../MachineCodeForInstruction.h"
-#include "../SparcV9RegisterInfo.h"
-#include "../SparcV9Internals.h"
-#include "../SparcV9TmpInstr.h"
-#include <fstream>
-#include <sstream>
-#include <cmath>
-#include <utility>
-
-using namespace llvm;
-/// Create ModuloSchedulingSBPass
-///
-FunctionPass *llvm::createModuloSchedulingSBPass(TargetMachine & targ) {
-  DEBUG(std::cerr << "Created ModuloSchedulingSBPass\n");
-  return new ModuloSchedulingSBPass(targ);
-}
-
-
-#if 1
-#define TIME_REGION(VARNAME, DESC) \
-   NamedRegionTimer VARNAME(DESC)
-#else
-#define TIME_REGION(VARNAME, DESC)
-#endif
-
-
-//Graph Traits for printing out the dependence graph
-template<typename GraphType>
-static void WriteGraphToFileSB(std::ostream &O, const std::string &GraphName,
-                             const GraphType &GT) {
-  std::string Filename = GraphName + ".dot";
-  O << "Writing '" << Filename << "'...";
-  std::ofstream F(Filename.c_str());
-
-  if (F.good())
-    WriteGraph(F, GT);
-  else
-    O << "  error opening file for writing!";
-  O << "\n";
-};
-
-namespace llvm {
-  Statistic<> NumLoops("moduloschedSB-numLoops", "Total Number of Loops");
-  Statistic<> NumSB("moduloschedSB-numSuperBlocks", "Total Number of SuperBlocks");
-  Statistic<> BBWithCalls("modulosched-BBCalls", "Basic Blocks rejected due to calls");
-  Statistic<> BBWithCondMov("modulosched-loopCondMov",
-                            "Basic Blocks rejected due to conditional moves");
-  Statistic<> SBResourceConstraint("modulosched-resourceConstraint",
-                                 "Loops constrained by resources");
-  Statistic<> SBRecurrenceConstraint("modulosched-recurrenceConstraint",
-                                   "Loops constrained by recurrences");
-  Statistic<> SBFinalIISum("modulosched-finalIISum", "Sum of all final II");
-  Statistic<> SBIISum("modulosched-IISum", "Sum of all theoretical II");
-  Statistic<> SBMSLoops("modulosched-schedLoops", "Number of loops successfully modulo-scheduled");
-  Statistic<> SBNoSched("modulosched-noSched", "No schedule");
-  Statistic<> SBSameStage("modulosched-sameStage", "Max stage is 0");
-  Statistic<> SBBLoops("modulosched-SBBLoops", "Number single basic block loops");
-  Statistic<> SBInvalid("modulosched-SBInvalid", "Number invalid superblock loops");
-  Statistic<> SBValid("modulosched-SBValid", "Number valid superblock loops");
- Statistic<> SBSize("modulosched-SBSize", "Total size of all valid superblocks");
-
-  template<>
-  struct DOTGraphTraits<MSchedGraphSB*> : public DefaultDOTGraphTraits {
-    static std::string getGraphName(MSchedGraphSB *F) {
-      return "Dependence Graph";
-    }
-
-    static std::string getNodeLabel(MSchedGraphSBNode *Node, MSchedGraphSB *Graph) {
-      if(!Node->isPredicate()) {
-        if (Node->getInst()) {
-          std::stringstream ss;
-          ss << *(Node->getInst());
-          return ss.str(); //((MachineInstr*)Node->getInst());
-        }
-        else
-          return "No Inst";
-      }
-      else
-        return "Pred Node";
-    }
-    static std::string getEdgeSourceLabel(MSchedGraphSBNode *Node,
-                                          MSchedGraphSBNode::succ_iterator I) {
-      //Label each edge with the type of dependence
-      std::string edgelabel = "";
-      switch (I.getEdge().getDepOrderType()) {
-
-      case MSchedGraphSBEdge::TrueDep:
-        edgelabel = "True";
-        break;
-
-      case MSchedGraphSBEdge::AntiDep:
-        edgelabel =  "Anti";
-        break;
-
-      case MSchedGraphSBEdge::OutputDep:
-        edgelabel = "Output";
-        break;
-
-      case MSchedGraphSBEdge::NonDataDep:
-        edgelabel = "Pred";
-        break;
-
-      default:
-        edgelabel = "Unknown";
-        break;
-      }
-
-      //FIXME
-      int iteDiff = I.getEdge().getIteDiff();
-      std::string intStr = "(IteDiff: ";
-      intStr += itostr(iteDiff);
-
-      intStr += ")";
-      edgelabel += intStr;
-
-      return edgelabel;
-    }
-  };
-
-  bool ModuloSchedulingSBPass::runOnFunction(Function &F) {
-    bool Changed = false;
-
-    //Get MachineFunction
-    MachineFunction &MF = MachineFunction::get(&F);
-
-    //Get Loop Info & Dependence Anaysis info
-    LoopInfo &LI = getAnalysis<LoopInfo>();
-    DependenceAnalyzer &DA = getAnalysis<DependenceAnalyzer>();
-
-    //Worklist of superblocks
-    std::vector<std::vector<const MachineBasicBlock*> > Worklist;
-    FindSuperBlocks(F, LI, Worklist);
-
-    DEBUG(if(Worklist.size() == 0) std::cerr << "No superblocks in function to ModuloSchedule\n");
-
-    //Loop over worklist and ModuloSchedule each SuperBlock
-    for(std::vector<std::vector<const MachineBasicBlock*> >::iterator SB = Worklist.begin(),
-          SBE = Worklist.end(); SB != SBE; ++SB) {
-
-      //Print out Superblock
-      DEBUG(std::cerr << "ModuloScheduling SB: \n";
-            for(std::vector<const MachineBasicBlock*>::const_iterator BI = SB->begin(),
-                  BE = SB->end(); BI != BE; ++BI) {
-              (*BI)->print(std::cerr);});
-
-      if(!CreateDefMap(*SB)) {
-        defaultInst = 0;
-        defMap.clear();
-        continue;
-      }
-
-      MSchedGraphSB *MSG = new MSchedGraphSB(*SB, target, indVarInstrs[*SB], DA,
-                                         machineTollvm[*SB]);
-
-      //Write Graph out to file
-      DEBUG(WriteGraphToFileSB(std::cerr, F.getName(), MSG));
-
-      //Calculate Resource II
-      int ResMII = calculateResMII(*SB);
-
-      //Calculate Recurrence II
-      int RecMII = calculateRecMII(MSG, ResMII);
-
-      DEBUG(std::cerr << "Number of reccurrences found: " << recurrenceList.size() << "\n");
-
-      //Our starting initiation interval is the maximum of RecMII and ResMII
-      if(RecMII < ResMII)
-        ++SBRecurrenceConstraint;
-      else
-        ++SBResourceConstraint;
-
-      II = std::max(RecMII, ResMII);
-      int mII = II;
-
-
-      //Print out II, RecMII, and ResMII
-      DEBUG(std::cerr << "II starts out as " << II << " ( RecMII=" << RecMII << " and ResMII=" << ResMII << ")\n");
-
-      //Calculate Node Properties
-      calculateNodeAttributes(MSG, ResMII);
-
-      //Dump node properties if in debug mode
-      DEBUG(for(std::map<MSchedGraphSBNode*, MSNodeSBAttributes>::iterator I =  nodeToAttributesMap.begin(),
-                  E = nodeToAttributesMap.end(); I !=E; ++I) {
-              std::cerr << "Node: " << *(I->first) << " ASAP: " << I->second.ASAP << " ALAP: "
-                        << I->second.ALAP << " MOB: " << I->second.MOB << " Depth: " << I->second.depth
-                        << " Height: " << I->second.height << "\n";
-            });
-
-
-      //Put nodes in order to schedule them
-      computePartialOrder();
-
-      //Dump out partial order
-      DEBUG(for(std::vector<std::set<MSchedGraphSBNode*> >::iterator I = partialOrder.begin(),
-                  E = partialOrder.end(); I !=E; ++I) {
-              std::cerr << "Start set in PO\n";
-              for(std::set<MSchedGraphSBNode*>::iterator J = I->begin(), JE = I->end(); J != JE; ++J)
-                std::cerr << "PO:" << **J << "\n";
-            });
-
-      //Place nodes in final order
-      orderNodes();
-
-      //Dump out order of nodes
-      DEBUG(for(std::vector<MSchedGraphSBNode*>::iterator I = FinalNodeOrder.begin(), E = FinalNodeOrder.end(); I != E; ++I) {
-              std::cerr << "FO:" << **I << "\n";
-            });
-
-
-      //Finally schedule nodes
-      bool haveSched = computeSchedule(*SB, MSG);
-
-      //Print out final schedule
-      DEBUG(schedule.print(std::cerr));
-
-      //Final scheduling step is to reconstruct the loop only if we actual have
-      //stage > 0
-      if(haveSched) {
-        //schedule.printSchedule(std::cerr);
-        reconstructLoop(*SB);
-        ++SBMSLoops;
-        //Changed = true;
-        SBIISum += mII;
-        SBFinalIISum += II;
-
-      if(schedule.getMaxStage() == 0)
-        ++SBSameStage;
-      }
-      else
-        ++SBNoSched;
-
-      //Clear out our maps for the next basic block that is processed
-      nodeToAttributesMap.clear();
-      partialOrder.clear();
-      recurrenceList.clear();
-      FinalNodeOrder.clear();
-      schedule.clear();
-      defMap.clear();
-
-    }
-    return Changed;
-  }
-
-  void ModuloSchedulingSBPass::FindSuperBlocks(Function &F, LoopInfo &LI,
-                      std::vector<std::vector<const MachineBasicBlock*> > &Worklist) {
-
-    //Get MachineFunction
-    MachineFunction &MF = MachineFunction::get(&F);
-
-    //Map of LLVM BB to machine BB
-    std::map<BasicBlock*, MachineBasicBlock*> bbMap;
-
-    for (MachineFunction::iterator BI = MF.begin(); BI != MF.end(); ++BI) {
-      BasicBlock *llvmBB = (BasicBlock*) BI->getBasicBlock();
-      assert(!bbMap.count(llvmBB) && "LLVM BB already in map!");
-      bbMap[llvmBB] = &*BI;
-    }
-
-    //Iterate over the loops, and find super blocks
-    for(LoopInfo::iterator LB = LI.begin(), LE = LI.end(); LB != LE; ++LB) {
-      Loop *L = *LB;
-      ++NumLoops;
-
-      //If loop is not single entry, try the next one
-      if(!L->getLoopPreheader())
-        continue;
-
-      //Check size of this loop, we don't want SBB loops
-      if(L->getBlocks().size() == 1)
-        continue;
-
-      //Check if this loop contains no sub loops
-      if(L->getSubLoops().size() == 0) {
-
-        std::vector<const MachineBasicBlock*> superBlock;
-
-        //Get Loop Headers
-        BasicBlock *header = L->getHeader();
-
-        //Follow the header and make sure each BB only has one entry and is valid
-        BasicBlock *current = header;
-        assert(bbMap.count(current) && "LLVM BB must have corresponding Machine BB\n");
-        MachineBasicBlock *currentMBB = bbMap[header];
-        bool done = false;
-        bool success = true;
-        unsigned offset = 0;
-        std::map<const MachineInstr*, unsigned> indexMap;
-
-        while(!done) {
-          //Loop over successors of this BB, they should be in the
-          //loop block and be valid
-          BasicBlock *next = 0;
-          for(succ_iterator I = succ_begin(current), E = succ_end(current);
-              I != E; ++I) {
-            if(L->contains(*I)) {
-              if(!next)
-                next = *I;
-              else {
-                done = true;
-                success = false;
-                break;
-              }
-            }
-          }
-
-          if(success) {
-            superBlock.push_back(currentMBB);
-            if(next == header)
-              done = true;
-            else if(!next->getSinglePredecessor()) {
-              done = true;
-              success = false;
-            }
-            else {
-              //Check that the next BB only has one entry
-              current = next;
-              assert(bbMap.count(current) && "LLVM BB must have corresponding Machine BB");
-              currentMBB = bbMap[current];
-            }
-          }
-        }
-
-
-
-
-
-        if(success) {
-          ++NumSB;
-
-          //Loop over all the blocks in the superblock
-          for(std::vector<const MachineBasicBlock*>::iterator currentMBB = superBlock.begin(), MBBEnd = superBlock.end(); currentMBB != MBBEnd; ++currentMBB) {
-            if(!MachineBBisValid(*currentMBB, indexMap, offset)) {
-              success = false;
-              break;
-            }
-          }
-        }
-
-        if(success) {
-          if(getIndVar(superBlock, bbMap, indexMap)) {
-            ++SBValid;
-            Worklist.push_back(superBlock);
-            SBSize += superBlock.size();
-          }
-          else
-            ++SBInvalid;
-        }
-      }
-    }
-  }
-
-
-  bool ModuloSchedulingSBPass::getIndVar(std::vector<const MachineBasicBlock*> &superBlock, std::map<BasicBlock*, MachineBasicBlock*> &bbMap,
-                                  std::map<const MachineInstr*, unsigned> &indexMap) {
-    //See if we can get induction var instructions
-    std::set<const BasicBlock*> llvmSuperBlock;
-
-    for(unsigned i =0; i < superBlock.size(); ++i)
-      llvmSuperBlock.insert(superBlock[i]->getBasicBlock());
-
-    //Get Target machine instruction info
-    const TargetInstrInfo *TMI = target.getInstrInfo();
-
-    //Get the loop back branch
-    BranchInst *b = dyn_cast<BranchInst>(((BasicBlock*) (superBlock[superBlock.size()-1])->getBasicBlock())->getTerminator());
-     std::set<Instruction*> indVar;
-
-    if(b->isConditional()) {
-      //Get the condition for the branch
-      Value *cond = b->getCondition();
-
-      DEBUG(std::cerr << "Condition: " << *cond << "\n");
-
-      //List of instructions associated with induction variable
-      std::vector<Instruction*> stack;
-
-      //Add branch
-      indVar.insert(b);
-
-      if(Instruction *I = dyn_cast<Instruction>(cond))
-        if(bbMap.count(I->getParent())) {
-          if (!assocIndVar(I, indVar, stack, bbMap, superBlock[(superBlock.size()-1)]->getBasicBlock(), llvmSuperBlock))
-            return false;
-        }
-        else
-          return false;
-      else
-        return false;
-    }
-    else {
-      indVar.insert(b);
-    }
-
-    //Dump out instructions associate with indvar for debug reasons
-    DEBUG(for(std::set<Instruction*>::iterator N = indVar.begin(), NE = indVar.end();
-              N != NE; ++N) {
-            std::cerr << **N << "\n";
-          });
-
-    //Create map of machine instr to llvm instr
-    std::map<MachineInstr*, Instruction*> mllvm;
-    for(std::vector<const MachineBasicBlock*>::iterator MBB = superBlock.begin(), MBE = superBlock.end(); MBB != MBE; ++MBB) {
-      BasicBlock *BB = (BasicBlock*) (*MBB)->getBasicBlock();
-      for(BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
-        MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(I);
-        for (unsigned j = 0; j < tempMvec.size(); j++) {
-          mllvm[tempMvec[j]] = I;
-        }
-      }
-    }
-
-      //Convert list of LLVM Instructions to list of Machine instructions
-      std::map<const MachineInstr*, unsigned> mIndVar;
-      for(std::set<Instruction*>::iterator N = indVar.begin(),
-            NE = indVar.end(); N != NE; ++N) {
-
-        //If we have a load, we can't handle this loop because
-        //there is no way to preserve dependences between loads
-        //and stores
-        if(isa<LoadInst>(*N))
-          return false;
-
-        MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(*N);
-        for (unsigned j = 0; j < tempMvec.size(); j++) {
-          MachineOpCode OC = (tempMvec[j])->getOpcode();
-          if(TMI->isNop(OC))
-            continue;
-          if(!indexMap.count(tempMvec[j]))
-            continue;
-          mIndVar[(MachineInstr*) tempMvec[j]] = indexMap[(MachineInstr*) tempMvec[j]];
-          DEBUG(std::cerr << *(tempMvec[j]) << " at index " << indexMap[(MachineInstr*) tempMvec[j]] << "\n");
-        }
-      }
-
-      //Put into a map for future access
-      indVarInstrs[superBlock] = mIndVar;
-      machineTollvm[superBlock] = mllvm;
-
-      return true;
-
-  }
-
-  bool ModuloSchedulingSBPass::assocIndVar(Instruction *I,
-                                           std::set<Instruction*> &indVar,
-                                           std::vector<Instruction*> &stack,
-                                       std::map<BasicBlock*, MachineBasicBlock*> &bbMap,
-                                           const BasicBlock *last, std::set<const BasicBlock*> &llvmSuperBlock) {
-
-    stack.push_back(I);
-
-    //If this is a phi node, check if its the canonical indvar
-    if(PHINode *PN = dyn_cast<PHINode>(I)) {
-      if(llvmSuperBlock.count(PN->getParent())) {
-        if (Instruction *Inc =
-            dyn_cast<Instruction>(PN->getIncomingValueForBlock(last)))
-          if (Inc->getOpcode() == Instruction::Add && Inc->getOperand(0) == PN)
-            if (ConstantInt *CI = dyn_cast<ConstantInt>(Inc->getOperand(1)))
-              if (CI->equalsInt(1)) {
-                //We have found the indvar, so add the stack, and inc instruction to the set
-                indVar.insert(stack.begin(), stack.end());
-                indVar.insert(Inc);
-                stack.pop_back();
-                return true;
-              }
-        return false;
-      }
-    }
-    else {
-      //Loop over each of the instructions operands, check if they are an instruction and in this BB
-      for(unsigned i = 0; i < I->getNumOperands(); ++i) {
-        if(Instruction *N =  dyn_cast<Instruction>(I->getOperand(i))) {
-          if(bbMap.count(N->getParent()))
-            if(!assocIndVar(N, indVar, stack, bbMap, last, llvmSuperBlock))
-              return false;
-        }
-      }
-    }
-
-    stack.pop_back();
-    return true;
-  }
-
-
-  /// This function checks if a Machine Basic Block is valid for modulo
-  /// scheduling. This means that it has no control flow (if/else or
-  /// calls) in the block.  Currently ModuloScheduling only works on
-  /// single basic block loops.
-  bool ModuloSchedulingSBPass::MachineBBisValid(const MachineBasicBlock *BI,
-                        std::map<const MachineInstr*, unsigned> &indexMap,
-                                                unsigned &offset) {
-
-    //Check size of our basic block.. make sure we have more then just the terminator in it
-    if(BI->getBasicBlock()->size() == 1)
-      return false;
-
-    //Get Target machine instruction info
-    const TargetInstrInfo *TMI = target.getInstrInfo();
-
-    unsigned count = 0;
-    for(MachineBasicBlock::const_iterator I = BI->begin(), E = BI->end(); I != E; ++I) {
-      //Get opcode to check instruction type
-      MachineOpCode OC = I->getOpcode();
-
-      //Look for calls
-      if(TMI->isCall(OC)) {
-        ++BBWithCalls;
-        return false;
-      }
-
-      //Look for conditional move
-      if(OC == V9::MOVRZr || OC == V9::MOVRZi || OC == V9::MOVRLEZr || OC == V9::MOVRLEZi
-         || OC == V9::MOVRLZr || OC == V9::MOVRLZi || OC == V9::MOVRNZr || OC == V9::MOVRNZi
-         || OC == V9::MOVRGZr || OC == V9::MOVRGZi || OC == V9::MOVRGEZr
-         || OC == V9::MOVRGEZi || OC == V9::MOVLEr || OC == V9::MOVLEi || OC == V9::MOVLEUr
-         || OC == V9::MOVLEUi || OC == V9::MOVFLEr || OC == V9::MOVFLEi
-         || OC == V9::MOVNEr || OC == V9::MOVNEi || OC == V9::MOVNEGr || OC == V9::MOVNEGi
-         || OC == V9::MOVFNEr || OC == V9::MOVFNEi) {
-        ++BBWithCondMov;
-        return false;
-      }
-
-      indexMap[I] = count + offset;
-
-      if(TMI->isNop(OC))
-        continue;
-
-      ++count;
-    }
-
-    offset += count;
-
-    return true;
-  }
-}
-
-bool ModuloSchedulingSBPass::CreateDefMap(std::vector<const MachineBasicBlock*> &SB) {
-  defaultInst = 0;
-
-  for(std::vector<const MachineBasicBlock*>::iterator BI = SB.begin(),
-        BE = SB.end(); BI != BE; ++BI) {
-
-    for(MachineBasicBlock::const_iterator I = (*BI)->begin(), E = (*BI)->end(); I != E; ++I) {
-      for(unsigned opNum = 0; opNum < I->getNumOperands(); ++opNum) {
-        const MachineOperand &mOp = I->getOperand(opNum);
-        if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef()) {
-          Value *V = mOp.getVRegValue();
-          //assert if this is the second def we have seen
-          if(defMap.count(V) && isa<PHINode>(V))
-            DEBUG(std::cerr << "FIXME: Dup def for phi!\n");
-          else {
-            //assert(!defMap.count(V) && "Def already in the map");
-            if(defMap.count(V))
-              return false;
-            defMap[V] = (MachineInstr*) &*I;
-          }
-        }
-
-        //See if we can use this Value* as our defaultInst
-        if(!defaultInst && mOp.getType() == MachineOperand::MO_VirtualRegister) {
-          Value *V = mOp.getVRegValue();
-          if(!isa<TmpInstruction>(V) && !isa<Argument>(V) && !isa<Constant>(V) && !isa<PHINode>(V))
-            defaultInst = (Instruction*) V;
-        }
-      }
-    }
-  }
-
-  if(!defaultInst)
-    return false;
-
-  return true;
-
-}
-
-
-//ResMII is calculated by determining the usage count for each resource
-//and using the maximum.
-//FIXME: In future there should be a way to get alternative resources
-//for each instruction
-int ModuloSchedulingSBPass::calculateResMII(std::vector<const MachineBasicBlock*> &superBlock) {
-
-  TIME_REGION(X, "calculateResMII");
-
-  const TargetInstrInfo *mii = target.getInstrInfo();
-  const TargetSchedInfo *msi = target.getSchedInfo();
-
-  int ResMII = 0;
-
-  //Map to keep track of usage count of each resource
-  std::map<unsigned, unsigned> resourceUsageCount;
-
-  for(std::vector<const MachineBasicBlock*>::iterator BI = superBlock.begin(), BE = superBlock.end(); BI != BE; ++BI) {
-    for(MachineBasicBlock::const_iterator I = (*BI)->begin(), E = (*BI)->end(); I != E; ++I) {
-
-      //Get resource usage for this instruction
-      InstrRUsage rUsage = msi->getInstrRUsage(I->getOpcode());
-      std::vector<std::vector<resourceId_t> > resources = rUsage.resourcesByCycle;
-
-      //Loop over resources in each cycle and increments their usage count
-      for(unsigned i=0; i < resources.size(); ++i)
-        for(unsigned j=0; j < resources[i].size(); ++j) {
-          if(!resourceUsageCount.count(resources[i][j])) {
-            resourceUsageCount[resources[i][j]] = 1;
-          }
-          else {
-            resourceUsageCount[resources[i][j]] =  resourceUsageCount[resources[i][j]] + 1;
-          }
-        }
-    }
-  }
-
-  //Find maximum usage count
-
-  //Get max number of instructions that can be issued at once. (FIXME)
-  int issueSlots = msi->maxNumIssueTotal;
-
-  for(std::map<unsigned,unsigned>::iterator RB = resourceUsageCount.begin(), RE = resourceUsageCount.end(); RB != RE; ++RB) {
-
-    //Get the total number of the resources in our cpu
-    int resourceNum = CPUResource::getCPUResource(RB->first)->maxNumUsers;
-
-    //Get total usage count for this resources
-    unsigned usageCount = RB->second;
-
-    //Divide the usage count by either the max number we can issue or the number of
-    //resources (whichever is its upper bound)
-    double finalUsageCount;
-    DEBUG(std::cerr << "Resource Num: " << RB->first << " Usage: " << usageCount << " TotalNum: " << resourceNum << "\n");
-
-    if( resourceNum <= issueSlots)
-      finalUsageCount = ceil(1.0 * usageCount / resourceNum);
-    else
-      finalUsageCount = ceil(1.0 * usageCount / issueSlots);
-
-
-    //Only keep track of the max
-    ResMII = std::max( (int) finalUsageCount, ResMII);
-
-  }
-
-  return ResMII;
-
-}
-
-/// calculateRecMII - Calculates the value of the highest recurrence
-/// By value we mean the total latency/distance
-int ModuloSchedulingSBPass::calculateRecMII(MSchedGraphSB *graph, int MII) {
-
-  TIME_REGION(X, "calculateRecMII");
-
-  findAllCircuits(graph, MII);
-  int RecMII = 0;
-
-  for(std::set<std::pair<int, std::vector<MSchedGraphSBNode*> > >::iterator I = recurrenceList.begin(), E=recurrenceList.end(); I !=E; ++I) {
-    RecMII = std::max(RecMII, I->first);
-  }
-
-  return MII;
-}
-
-int CircCountSB;
-
-void ModuloSchedulingSBPass::unblock(MSchedGraphSBNode *u, std::set<MSchedGraphSBNode*> &blocked,
-             std::map<MSchedGraphSBNode*, std::set<MSchedGraphSBNode*> > &B) {
-
-  //Unblock u
-  DEBUG(std::cerr << "Unblocking: " << *u << "\n");
-  blocked.erase(u);
-
-  //std::set<MSchedGraphSBNode*> toErase;
-  while (!B[u].empty()) {
-    MSchedGraphSBNode *W = *B[u].begin();
-    B[u].erase(W);
-    //toErase.insert(*W);
-    DEBUG(std::cerr << "Removed: " << *W << "from B-List\n");
-    if(blocked.count(W))
-      unblock(W, blocked, B);
-  }
-
-}
-
-void ModuloSchedulingSBPass::addSCC(std::vector<MSchedGraphSBNode*> &SCC, std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> &newNodes) {
-
-  int totalDelay = 0;
-  int totalDistance = 0;
-  std::vector<MSchedGraphSBNode*> recc;
-  MSchedGraphSBNode *start = 0;
-  MSchedGraphSBNode *end = 0;
-
-  //Loop over recurrence, get delay and distance
-  for(std::vector<MSchedGraphSBNode*>::iterator N = SCC.begin(), NE = SCC.end(); N != NE; ++N) {
-    DEBUG(std::cerr << **N << "\n");
-    totalDelay += (*N)->getLatency();
-
-    for(unsigned i = 0; i < (*N)->succ_size(); ++i) {
-      MSchedGraphSBEdge *edge = (*N)->getSuccessor(i);
-      if(find(SCC.begin(), SCC.end(), edge->getDest()) != SCC.end()) {
-        totalDistance += edge->getIteDiff();
-        if(edge->getIteDiff() > 0)
-          if(!start && !end) {
-            start = *N;
-            end = edge->getDest();
-          }
-
-      }
-    }
-
-
-    //Get the original node
-    recc.push_back(newNodes[*N]);
-
-
-  }
-
-  DEBUG(std::cerr << "End Recc\n");
-
-
-  assert( (start && end) && "Must have start and end node to ignore edge for SCC");
-
-  if(start && end) {
-    //Insert reccurrence into the list
-    DEBUG(std::cerr << "Ignore Edge from!!: " << *start << " to " << *end << "\n");
-    edgesToIgnore.insert(std::make_pair(newNodes[start], (newNodes[end])->getInEdgeNum(newNodes[start])));
-  }
-
-  int lastII = totalDelay / totalDistance;
-
-
-  recurrenceList.insert(std::make_pair(lastII, recc));
-
-}
-
-bool ModuloSchedulingSBPass::circuit(MSchedGraphSBNode *v, std::vector<MSchedGraphSBNode*> &stack,
-             std::set<MSchedGraphSBNode*> &blocked, std::vector<MSchedGraphSBNode*> &SCC,
-             MSchedGraphSBNode *s, std::map<MSchedGraphSBNode*, std::set<MSchedGraphSBNode*> > &B,
-                                   int II, std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> &newNodes) {
-  bool f = false;
-
-  DEBUG(std::cerr << "Finding Circuits Starting with: ( " << v << ")"<< *v << "\n");
-
-  //Push node onto the stack
-  stack.push_back(v);
-
-  //block this node
-  blocked.insert(v);
-
-  //Loop over all successors of node v that are in the scc, create Adjaceny list
-  std::set<MSchedGraphSBNode*> AkV;
-  for(MSchedGraphSBNode::succ_iterator I = v->succ_begin(), E = v->succ_end(); I != E; ++I) {
-    if((std::find(SCC.begin(), SCC.end(), *I) != SCC.end())) {
-      AkV.insert(*I);
-    }
-  }
-
-  for(std::set<MSchedGraphSBNode*>::iterator I = AkV.begin(), E = AkV.end(); I != E; ++I) {
-    if(*I == s) {
-      //We have a circuit, so add it to our list
-      addRecc(stack, newNodes);
-      f = true;
-    }
-    else if(!blocked.count(*I)) {
-      if(circuit(*I, stack, blocked, SCC, s, B, II, newNodes))
-        f = true;
-    }
-    else
-      DEBUG(std::cerr << "Blocked: " << **I << "\n");
-  }
-
-
-  if(f) {
-    unblock(v, blocked, B);
-  }
-  else {
-    for(std::set<MSchedGraphSBNode*>::iterator I = AkV.begin(), E = AkV.end(); I != E; ++I)
-      B[*I].insert(v);
-
-  }
-
-  //Pop v
-  stack.pop_back();
-
-  return f;
-
-}
-
-void ModuloSchedulingSBPass::addRecc(std::vector<MSchedGraphSBNode*> &stack, std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> &newNodes) {
-  std::vector<MSchedGraphSBNode*> recc;
-  //Dump recurrence for now
-  DEBUG(std::cerr << "Starting Recc\n");
-
-  int totalDelay = 0;
-  int totalDistance = 0;
-  MSchedGraphSBNode *lastN = 0;
-  MSchedGraphSBNode *start = 0;
-  MSchedGraphSBNode *end = 0;
-
-  //Loop over recurrence, get delay and distance
-  for(std::vector<MSchedGraphSBNode*>::iterator N = stack.begin(), NE = stack.end(); N != NE; ++N) {
-    DEBUG(std::cerr << **N << "\n");
-    totalDelay += (*N)->getLatency();
-    if(lastN) {
-      int iteDiff = (*N)->getInEdge(lastN).getIteDiff();
-      totalDistance += iteDiff;
-
-      if(iteDiff > 0) {
-        start = lastN;
-        end = *N;
-      }
-    }
-    //Get the original node
-    lastN = *N;
-    recc.push_back(newNodes[*N]);
-
-
-  }
-
-  //Get the loop edge
-  totalDistance += lastN->getIteDiff(*stack.begin());
-
-  DEBUG(std::cerr << "End Recc\n");
-  CircCountSB++;
-
-  if(start && end) {
-    //Insert reccurrence into the list
-    DEBUG(std::cerr << "Ignore Edge from!!: " << *start << " to " << *end << "\n");
-    edgesToIgnore.insert(std::make_pair(newNodes[start], (newNodes[end])->getInEdgeNum(newNodes[start])));
-  }
-  else {
-    //Insert reccurrence into the list
-    DEBUG(std::cerr << "Ignore Edge from: " << *lastN << " to " << **stack.begin() << "\n");
-    edgesToIgnore.insert(std::make_pair(newNodes[lastN], newNodes[(*stack.begin())]->getInEdgeNum(newNodes[lastN])));
-
-  }
-  //Adjust II until we get close to the inequality delay - II*distance <= 0
-  int RecMII = II; //Starting value
-  int value = totalDelay-(RecMII * totalDistance);
-  int lastII = II;
-  while(value < 0) {
-
-    lastII = RecMII;
-    RecMII--;
-    value = totalDelay-(RecMII * totalDistance);
-  }
-
-  recurrenceList.insert(std::make_pair(lastII, recc));
-
-}
-
-
-void ModuloSchedulingSBPass::findAllCircuits(MSchedGraphSB *g, int II) {
-
-  CircCountSB = 0;
-
-  //Keep old to new node mapping information
-  std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> newNodes;
-
-  //copy the graph
-  MSchedGraphSB *MSG = new MSchedGraphSB(*g, newNodes);
-
-  DEBUG(std::cerr << "Finding All Circuits\n");
-
-  //Set of blocked nodes
-  std::set<MSchedGraphSBNode*> blocked;
-
-  //Stack holding current circuit
-  std::vector<MSchedGraphSBNode*> stack;
-
-  //Map for B Lists
-  std::map<MSchedGraphSBNode*, std::set<MSchedGraphSBNode*> > B;
-
-  //current node
-  MSchedGraphSBNode *s;
-
-
-  //Iterate over the graph until its down to one node or empty
-  while(MSG->size() > 1) {
-
-    //Write Graph out to file
-    //WriteGraphToFile(std::cerr, "Graph" + utostr(MSG->size()), MSG);
-
-    DEBUG(std::cerr << "Graph Size: " << MSG->size() << "\n");
-    DEBUG(std::cerr << "Finding strong component Vk with least vertex\n");
-
-    //Iterate over all the SCCs in the graph
-    std::set<MSchedGraphSBNode*> Visited;
-    std::vector<MSchedGraphSBNode*> Vk;
-    MSchedGraphSBNode* s = 0;
-    int numEdges = 0;
-
-    //Find scc with the least vertex
-    for (MSchedGraphSB::iterator GI = MSG->begin(), E = MSG->end(); GI != E; ++GI)
-      if (Visited.insert(GI->second).second) {
-        for (scc_iterator<MSchedGraphSBNode*> SCCI = scc_begin(GI->second),
-               E = scc_end(GI->second); SCCI != E; ++SCCI) {
-          std::vector<MSchedGraphSBNode*> &nextSCC = *SCCI;
-
-          if (Visited.insert(nextSCC[0]).second) {
-            Visited.insert(nextSCC.begin()+1, nextSCC.end());
-
-            if(nextSCC.size() > 1) {
-              DEBUG(std::cerr << "SCC size: " << nextSCC.size() << "\n");
-
-              for(unsigned i = 0; i < nextSCC.size(); ++i) {
-                //Loop over successor and see if in scc, then count edge
-                MSchedGraphSBNode *node = nextSCC[i];
-                for(MSchedGraphSBNode::succ_iterator S = node->succ_begin(), SE = node->succ_end(); S != SE; ++S) {
-                  if(find(nextSCC.begin(), nextSCC.end(), *S) != nextSCC.end())
-                    numEdges++;
-                }
-              }
-              DEBUG(std::cerr << "Num Edges: " << numEdges << "\n");
-            }
-
-            //Ignore self loops
-            if(nextSCC.size() > 1) {
-
-              //Get least vertex in Vk
-              if(!s) {
-                s = nextSCC[0];
-                Vk = nextSCC;
-              }
-
-              for(unsigned i = 0; i < nextSCC.size(); ++i) {
-                if(nextSCC[i] < s) {
-                  s = nextSCC[i];
-                  Vk = nextSCC;
-                }
-              }
-            }
-          }
-        }
-      }
-
-
-
-    //Process SCC
-    DEBUG(for(std::vector<MSchedGraphSBNode*>::iterator N = Vk.begin(), NE = Vk.end();
-              N != NE; ++N) { std::cerr << *((*N)->getInst()); });
-
-    //Iterate over all nodes in this scc
-    for(std::vector<MSchedGraphSBNode*>::iterator N = Vk.begin(), NE = Vk.end();
-        N != NE; ++N) {
-      blocked.erase(*N);
-      B[*N].clear();
-    }
-    if(Vk.size() > 1) {
-      if(numEdges < 98)
-        circuit(s, stack, blocked, Vk, s, B, II, newNodes);
-      else
-        addSCC(Vk, newNodes);
-
-
-      //Delete nodes from the graph
-      //Find all nodes up to s and delete them
-      std::vector<MSchedGraphSBNode*> nodesToRemove;
-      nodesToRemove.push_back(s);
-      for(MSchedGraphSB::iterator N = MSG->begin(), NE = MSG->end(); N != NE; ++N) {
-        if(N->second < s )
-            nodesToRemove.push_back(N->second);
-      }
-      for(std::vector<MSchedGraphSBNode*>::iterator N = nodesToRemove.begin(), NE = nodesToRemove.end(); N != NE; ++N) {
-        DEBUG(std::cerr << "Deleting Node: " << **N << "\n");
-        MSG->deleteNode(*N);
-      }
-    }
-    else
-      break;
-  }
-  DEBUG(std::cerr << "Num Circuits found: " << CircCountSB << "\n");
-}
-/// calculateNodeAttributes - The following properties are calculated for
-/// each node in the dependence graph: ASAP, ALAP, Depth, Height, and
-/// MOB.
-void ModuloSchedulingSBPass::calculateNodeAttributes(MSchedGraphSB *graph, int MII) {
-
-  TIME_REGION(X, "calculateNodeAttributes");
-
-  assert(nodeToAttributesMap.empty() && "Node attribute map was not cleared");
-
-  //Loop over the nodes and add them to the map
-  for(MSchedGraphSB::iterator I = graph->begin(), E = graph->end(); I != E; ++I) {
-
-    DEBUG(std::cerr << "Inserting node into attribute map: " << *I->second << "\n");
-
-    //Assert if its already in the map
-    assert(nodeToAttributesMap.count(I->second) == 0 &&
-           "Node attributes are already in the map");
-
-    //Put into the map with default attribute values
-    nodeToAttributesMap[I->second] = MSNodeSBAttributes();
-  }
-
-  //Create set to deal with reccurrences
-  std::set<MSchedGraphSBNode*> visitedNodes;
-
-  //Now Loop over map and calculate the node attributes
-  for(std::map<MSchedGraphSBNode*, MSNodeSBAttributes>::iterator I = nodeToAttributesMap.begin(), E = nodeToAttributesMap.end(); I != E; ++I) {
-    calculateASAP(I->first, MII, (MSchedGraphSBNode*) 0);
-    visitedNodes.clear();
-  }
-
-  int maxASAP = findMaxASAP();
-  //Calculate ALAP which depends on ASAP being totally calculated
-  for(std::map<MSchedGraphSBNode*, MSNodeSBAttributes>::iterator I = nodeToAttributesMap.begin(), E = nodeToAttributesMap.end(); I != E; ++I) {
-    calculateALAP(I->first, MII, maxASAP, (MSchedGraphSBNode*) 0);
-    visitedNodes.clear();
-  }
-
-  //Calculate MOB which depends on ASAP being totally calculated, also do depth and height
-  for(std::map<MSchedGraphSBNode*, MSNodeSBAttributes>::iterator I = nodeToAttributesMap.begin(), E = nodeToAttributesMap.end(); I != E; ++I) {
-    (I->second).MOB = std::max(0,(I->second).ALAP - (I->second).ASAP);
-
-    DEBUG(std::cerr << "MOB: " << (I->second).MOB << " (" << *(I->first) << ")\n");
-    calculateDepth(I->first, (MSchedGraphSBNode*) 0);
-    calculateHeight(I->first, (MSchedGraphSBNode*) 0);
-  }
-
-
-}
-
-/// ignoreEdge - Checks to see if this edge of a recurrence should be ignored or not
-bool ModuloSchedulingSBPass::ignoreEdge(MSchedGraphSBNode *srcNode, MSchedGraphSBNode *destNode) {
-  if(destNode == 0 || srcNode ==0)
-    return false;
-
-  bool findEdge = edgesToIgnore.count(std::make_pair(srcNode, destNode->getInEdgeNum(srcNode)));
-
-  DEBUG(std::cerr << "Ignoring edge? from: " << *srcNode << " to " << *destNode << "\n");
-
-  return findEdge;
-}
-
-
-/// calculateASAP - Calculates the
-int  ModuloSchedulingSBPass::calculateASAP(MSchedGraphSBNode *node, int MII, MSchedGraphSBNode *destNode) {
-
-  DEBUG(std::cerr << "Calculating ASAP for " << *node << "\n");
-
-  //Get current node attributes
-  MSNodeSBAttributes &attributes = nodeToAttributesMap.find(node)->second;
-
-  if(attributes.ASAP != -1)
-    return attributes.ASAP;
-
-  int maxPredValue = 0;
-
-  //Iterate over all of the predecessors and find max
-  for(MSchedGraphSBNode::pred_iterator P = node->pred_begin(), E = node->pred_end(); P != E; ++P) {
-
-    //Only process if we are not ignoring the edge
-    if(!ignoreEdge(*P, node)) {
-      int predASAP = -1;
-      predASAP = calculateASAP(*P, MII, node);
-
-      assert(predASAP != -1 && "ASAP has not been calculated");
-      int iteDiff = node->getInEdge(*P).getIteDiff();
-
-      int currentPredValue = predASAP + (*P)->getLatency() - (iteDiff * MII);
-      DEBUG(std::cerr << "pred ASAP: " << predASAP << ", iteDiff: " << iteDiff << ", PredLatency: " << (*P)->getLatency() << ", Current ASAP pred: " << currentPredValue << "\n");
-      maxPredValue = std::max(maxPredValue, currentPredValue);
-    }
-  }
-
-  attributes.ASAP = maxPredValue;
-
-  DEBUG(std::cerr << "ASAP: " << attributes.ASAP << " (" << *node << ")\n");
-
-  return maxPredValue;
-}
-
-
-int ModuloSchedulingSBPass::calculateALAP(MSchedGraphSBNode *node, int MII,
-                                        int maxASAP, MSchedGraphSBNode *srcNode) {
-
-  DEBUG(std::cerr << "Calculating ALAP for " << *node << "\n");
-
-  MSNodeSBAttributes &attributes = nodeToAttributesMap.find(node)->second;
-
-  if(attributes.ALAP != -1)
-    return attributes.ALAP;
-
-  if(node->hasSuccessors()) {
-
-    //Trying to deal with the issue where the node has successors, but
-    //we are ignoring all of the edges to them. So this is my hack for
-    //now.. there is probably a more elegant way of doing this (FIXME)
-    bool processedOneEdge = false;
-
-    //FIXME, set to something high to start
-    int minSuccValue = 9999999;
-
-    //Iterate over all of the predecessors and fine max
-    for(MSchedGraphSBNode::succ_iterator P = node->succ_begin(),
-          E = node->succ_end(); P != E; ++P) {
-
-      //Only process if we are not ignoring the edge
-      if(!ignoreEdge(node, *P)) {
-        processedOneEdge = true;
-        int succALAP = -1;
-        succALAP = calculateALAP(*P, MII, maxASAP, node);
-
-        assert(succALAP != -1 && "Successors ALAP should have been caclulated");
-
-        int iteDiff = P.getEdge().getIteDiff();
-
-        int currentSuccValue = succALAP - node->getLatency() + iteDiff * MII;
-
-        DEBUG(std::cerr << "succ ALAP: " << succALAP << ", iteDiff: " << iteDiff << ", SuccLatency: " << (*P)->getLatency() << ", Current ALAP succ: " << currentSuccValue << "\n");
-
-        minSuccValue = std::min(minSuccValue, currentSuccValue);
-      }
-    }
-
-    if(processedOneEdge)
-        attributes.ALAP = minSuccValue;
-
-    else
-      attributes.ALAP = maxASAP;
-  }
-  else
-    attributes.ALAP = maxASAP;
-
-  DEBUG(std::cerr << "ALAP: " << attributes.ALAP << " (" << *node << ")\n");
-
-  if(attributes.ALAP < 0)
-    attributes.ALAP = 0;
-
-  return attributes.ALAP;
-}
-
-int ModuloSchedulingSBPass::findMaxASAP() {
-  int maxASAP = 0;
-
-  for(std::map<MSchedGraphSBNode*, MSNodeSBAttributes>::iterator I = nodeToAttributesMap.begin(),
-        E = nodeToAttributesMap.end(); I != E; ++I)
-    maxASAP = std::max(maxASAP, I->second.ASAP);
-  return maxASAP;
-}
-
-
-int ModuloSchedulingSBPass::calculateHeight(MSchedGraphSBNode *node,MSchedGraphSBNode *srcNode) {
-
-  MSNodeSBAttributes &attributes = nodeToAttributesMap.find(node)->second;
-
-  if(attributes.height != -1)
-    return attributes.height;
-
-  int maxHeight = 0;
-
-  //Iterate over all of the predecessors and find max
-  for(MSchedGraphSBNode::succ_iterator P = node->succ_begin(),
-        E = node->succ_end(); P != E; ++P) {
-
-
-    if(!ignoreEdge(node, *P)) {
-      int succHeight = calculateHeight(*P, node);
-
-      assert(succHeight != -1 && "Successors Height should have been caclulated");
-
-      int currentHeight = succHeight + node->getLatency();
-      maxHeight = std::max(maxHeight, currentHeight);
-    }
-  }
-  attributes.height = maxHeight;
-  DEBUG(std::cerr << "Height: " << attributes.height << " (" << *node << ")\n");
-  return maxHeight;
-}
-
-
-int ModuloSchedulingSBPass::calculateDepth(MSchedGraphSBNode *node,
-                                          MSchedGraphSBNode *destNode) {
-
-  MSNodeSBAttributes &attributes = nodeToAttributesMap.find(node)->second;
-
-  if(attributes.depth != -1)
-    return attributes.depth;
-
-  int maxDepth = 0;
-
-  //Iterate over all of the predecessors and fine max
-  for(MSchedGraphSBNode::pred_iterator P = node->pred_begin(), E = node->pred_end(); P != E; ++P) {
-
-    if(!ignoreEdge(*P, node)) {
-      int predDepth = -1;
-      predDepth = calculateDepth(*P, node);
-
-      assert(predDepth != -1 && "Predecessors ASAP should have been caclulated");
-
-      int currentDepth = predDepth + (*P)->getLatency();
-      maxDepth = std::max(maxDepth, currentDepth);
-    }
-  }
-  attributes.depth = maxDepth;
-
-  DEBUG(std::cerr << "Depth: " << attributes.depth << " (" << *node << "*)\n");
-  return maxDepth;
-}
-
-void ModuloSchedulingSBPass::computePartialOrder() {
-
-  TIME_REGION(X, "calculatePartialOrder");
-
-  DEBUG(std::cerr << "Computing Partial Order\n");
-
-  //Steps to add a recurrence to the partial order 1) Find reccurrence
-  //with the highest RecMII. Add it to the partial order.  2) For each
-  //recurrence with decreasing RecMII, add it to the partial order
-  //along with any nodes that connect this recurrence to recurrences
-  //already in the partial order
-  for(std::set<std::pair<int, std::vector<MSchedGraphSBNode*> > >::reverse_iterator
-        I = recurrenceList.rbegin(), E=recurrenceList.rend(); I !=E; ++I) {
-
-    std::set<MSchedGraphSBNode*> new_recurrence;
-
-    //Loop through recurrence and remove any nodes already in the partial order
-    for(std::vector<MSchedGraphSBNode*>::const_iterator N = I->second.begin(),
-          NE = I->second.end(); N != NE; ++N) {
-
-      bool found = false;
-      for(std::vector<std::set<MSchedGraphSBNode*> >::iterator PO = partialOrder.begin(),
-            PE = partialOrder.end(); PO != PE; ++PO) {
-        if(PO->count(*N))
-          found = true;
-      }
-
-      //Check if its a branch, and remove to handle special
-      if(!found) {
-        new_recurrence.insert(*N);
-      }
-
-    }
-
-
-    if(new_recurrence.size() > 0) {
-
-      std::vector<MSchedGraphSBNode*> path;
-      std::set<MSchedGraphSBNode*> nodesToAdd;
-
-      //Dump recc we are dealing with (minus nodes already in PO)
-      DEBUG(std::cerr << "Recc: ");
-      DEBUG(for(std::set<MSchedGraphSBNode*>::iterator R = new_recurrence.begin(), RE = new_recurrence.end(); R != RE; ++R) { std::cerr << **R ; });
-
-      //Add nodes that connect this recurrence to recurrences in the partial path
-      for(std::set<MSchedGraphSBNode*>::iterator N = new_recurrence.begin(),
-          NE = new_recurrence.end(); N != NE; ++N)
-        searchPath(*N, path, nodesToAdd, new_recurrence);
-
-      //Add nodes to this recurrence if they are not already in the partial order
-      for(std::set<MSchedGraphSBNode*>::iterator N = nodesToAdd.begin(), NE = nodesToAdd.end();
-          N != NE; ++N) {
-        bool found = false;
-        for(std::vector<std::set<MSchedGraphSBNode*> >::iterator PO = partialOrder.begin(),
-              PE = partialOrder.end(); PO != PE; ++PO) {
-          if(PO->count(*N))
-            found = true;
-        }
-        if(!found) {
-          assert("FOUND CONNECTOR");
-          new_recurrence.insert(*N);
-        }
-      }
-
-      partialOrder.push_back(new_recurrence);
-    }
-  }
-
-  //Add any nodes that are not already in the partial order
-  //Add them in a set, one set per connected component
-  std::set<MSchedGraphSBNode*> lastNodes;
-  std::set<MSchedGraphSBNode*> noPredNodes;
-  for(std::map<MSchedGraphSBNode*, MSNodeSBAttributes>::iterator I = nodeToAttributesMap.begin(),
-        E = nodeToAttributesMap.end(); I != E; ++I) {
-
-    bool found = false;
-
-    //Check if its already in our partial order, if not add it to the final vector
-    for(std::vector<std::set<MSchedGraphSBNode*> >::iterator PO = partialOrder.begin(),
-          PE = partialOrder.end(); PO != PE; ++PO) {
-      if(PO->count(I->first))
-        found = true;
-    }
-    if(!found)
-      lastNodes.insert(I->first);
-  }
-
-  //For each node w/out preds, see if there is a path to one of the
-  //recurrences, and if so add them to that current recc
-  /*for(std::set<MSchedGraphSBNode*>::iterator N = noPredNodes.begin(), NE = noPredNodes.end();
-      N != NE; ++N) {
-    DEBUG(std::cerr << "No Pred Path from: " << **N << "\n");
-    for(std::vector<std::set<MSchedGraphSBNode*> >::iterator PO = partialOrder.begin(),
-          PE = partialOrder.end(); PO != PE; ++PO) {
-      std::vector<MSchedGraphSBNode*> path;
-      pathToRecc(*N, path, *PO, lastNodes);
-    }
-    }*/
-
-
-  //Break up remaining nodes that are not in the partial order
-  ///into their connected compoenents
-    while(lastNodes.size() > 0) {
-      std::set<MSchedGraphSBNode*> ccSet;
-      connectedComponentSet(*(lastNodes.begin()),ccSet, lastNodes);
-      if(ccSet.size() > 0)
-        partialOrder.push_back(ccSet);
-    }
-
-}
-
-void ModuloSchedulingSBPass::connectedComponentSet(MSchedGraphSBNode *node, std::set<MSchedGraphSBNode*> &ccSet, std::set<MSchedGraphSBNode*> &lastNodes) {
-
-  //Add to final set
-  if( !ccSet.count(node) && lastNodes.count(node)) {
-    lastNodes.erase(node);
-    ccSet.insert(node);
-  }
-  else
-    return;
-
-  //Loop over successors and recurse if we have not seen this node before
-  for(MSchedGraphSBNode::succ_iterator node_succ = node->succ_begin(), end=node->succ_end(); node_succ != end; ++node_succ) {
-    connectedComponentSet(*node_succ, ccSet, lastNodes);
-  }
-
-}
-
-void ModuloSchedulingSBPass::searchPath(MSchedGraphSBNode *node,
-                                      std::vector<MSchedGraphSBNode*> &path,
-                                      std::set<MSchedGraphSBNode*> &nodesToAdd,
-                                     std::set<MSchedGraphSBNode*> &new_reccurrence) {
-  //Push node onto the path
-  path.push_back(node);
-
-  //Loop over all successors and see if there is a path from this node to
-  //a recurrence in the partial order, if so.. add all nodes to be added to recc
-  for(MSchedGraphSBNode::succ_iterator S = node->succ_begin(), SE = node->succ_end(); S != SE;
-      ++S) {
-
-    //Check if we should ignore this edge first
-    if(ignoreEdge(node,*S))
-      continue;
-
-    //check if successor is in this recurrence, we will get to it eventually
-    if(new_reccurrence.count(*S))
-      continue;
-
-    //If this node exists in a recurrence already in the partial
-    //order, then add all nodes in the path to the set of nodes to add
-     //Check if its already in our partial order, if not add it to the
-     //final vector
-    bool found = false;
-    for(std::vector<std::set<MSchedGraphSBNode*> >::iterator PO = partialOrder.begin(),
-          PE = partialOrder.end(); PO != PE; ++PO) {
-
-      if(PO->count(*S)) {
-        found = true;
-        break;
-      }
-    }
-
-    if(!found) {
-      nodesToAdd.insert(*S);
-      searchPath(*S, path, nodesToAdd, new_reccurrence);
-    }
-  }
-
-  //Pop Node off the path
-  path.pop_back();
-}
-
-void dumpIntersection(std::set<MSchedGraphSBNode*> &IntersectCurrent) {
-  std::cerr << "Intersection (";
-  for(std::set<MSchedGraphSBNode*>::iterator I = IntersectCurrent.begin(), E = IntersectCurrent.end(); I != E; ++I)
-    std::cerr << **I << ", ";
-  std::cerr << ")\n";
-}
-
-void ModuloSchedulingSBPass::orderNodes() {
-
-  TIME_REGION(X, "orderNodes");
-
-  int BOTTOM_UP = 0;
-  int TOP_DOWN = 1;
-
-  //Set default order
-  int order = BOTTOM_UP;
-
-  //Loop over and find all pred nodes and schedule them first
-  /*for(std::vector<std::set<MSchedGraphSBNode*> >::iterator CurrentSet = partialOrder.begin(), E= partialOrder.end(); CurrentSet != E; ++CurrentSet) {
-    for(std::set<MSchedGraphSBNode*>::iterator N = CurrentSet->begin(), NE = CurrentSet->end(); N != NE; ++N)
-      if((*N)->isPredicate()) {
-        FinalNodeOrder.push_back(*N);
-        CurrentSet->erase(*N);
-      }
-      }*/
-
-
-
-  //Loop over all the sets and place them in the final node order
-  for(std::vector<std::set<MSchedGraphSBNode*> >::iterator CurrentSet = partialOrder.begin(), E= partialOrder.end(); CurrentSet != E; ++CurrentSet) {
-
-    DEBUG(std::cerr << "Processing set in S\n");
-    DEBUG(dumpIntersection(*CurrentSet));
-
-    //Result of intersection
-    std::set<MSchedGraphSBNode*> IntersectCurrent;
-
-    predIntersect(*CurrentSet, IntersectCurrent);
-
-    //If the intersection of predecessor and current set is not empty
-    //sort nodes bottom up
-    if(IntersectCurrent.size() != 0) {
-      DEBUG(std::cerr << "Final Node Order Predecessors and Current Set interesection is NOT empty\n");
-      order = BOTTOM_UP;
-    }
-    //If empty, use successors
-    else {
-      DEBUG(std::cerr << "Final Node Order Predecessors and Current Set interesection is empty\n");
-
-      succIntersect(*CurrentSet, IntersectCurrent);
-
-      //sort top-down
-      if(IntersectCurrent.size() != 0) {
-         DEBUG(std::cerr << "Final Node Order Successors and Current Set interesection is NOT empty\n");
-        order = TOP_DOWN;
-      }
-      else {
-        DEBUG(std::cerr << "Final Node Order Successors and Current Set interesection is empty\n");
-        //Find node with max ASAP in current Set
-        MSchedGraphSBNode *node;
-        int maxASAP = 0;
-        DEBUG(std::cerr << "Using current set of size " << CurrentSet->size() << "to find max ASAP\n");
-        for(std::set<MSchedGraphSBNode*>::iterator J = CurrentSet->begin(), JE = CurrentSet->end(); J != JE; ++J) {
-          //Get node attributes
-          MSNodeSBAttributes nodeAttr= nodeToAttributesMap.find(*J)->second;
-          //assert(nodeAttr != nodeToAttributesMap.end() && "Node not in attributes map!");
-
-          if(maxASAP <= nodeAttr.ASAP) {
-            maxASAP = nodeAttr.ASAP;
-            node = *J;
-          }
-        }
-        assert(node != 0 && "In node ordering node should not be null");
-        IntersectCurrent.insert(node);
-        order = BOTTOM_UP;
-      }
-    }
-
-    //Repeat until all nodes are put into the final order from current set
-    while(IntersectCurrent.size() > 0) {
-
-      if(order == TOP_DOWN) {
-        DEBUG(std::cerr << "Order is TOP DOWN\n");
-
-        while(IntersectCurrent.size() > 0) {
-          DEBUG(std::cerr << "Intersection is not empty, so find heighest height\n");
-
-          int MOB = 0;
-          int height = 0;
-          MSchedGraphSBNode *highestHeightNode = *(IntersectCurrent.begin());
-
-          //Find node in intersection with highest heigh and lowest MOB
-          for(std::set<MSchedGraphSBNode*>::iterator I = IntersectCurrent.begin(),
-                E = IntersectCurrent.end(); I != E; ++I) {
-
-            //Get current nodes properties
-            MSNodeSBAttributes nodeAttr= nodeToAttributesMap.find(*I)->second;
-
-            if(height < nodeAttr.height) {
-              highestHeightNode = *I;
-              height = nodeAttr.height;
-              MOB = nodeAttr.MOB;
-            }
-            else if(height ==  nodeAttr.height) {
-              if(MOB > nodeAttr.height) {
-                highestHeightNode = *I;
-                height =  nodeAttr.height;
-                MOB = nodeAttr.MOB;
-              }
-            }
-          }
-
-          //Append our node with greatest height to the NodeOrder
-          if(std::find(FinalNodeOrder.begin(), FinalNodeOrder.end(), highestHeightNode) == FinalNodeOrder.end()) {
-            DEBUG(std::cerr << "Adding node to Final Order: " << *highestHeightNode << "\n");
-            FinalNodeOrder.push_back(highestHeightNode);
-          }
-
-          //Remove V from IntersectOrder
-          IntersectCurrent.erase(std::find(IntersectCurrent.begin(),
-                                      IntersectCurrent.end(), highestHeightNode));
-
-
-          //Intersect V's successors with CurrentSet
-          for(MSchedGraphSBNode::succ_iterator P = highestHeightNode->succ_begin(),
-                E = highestHeightNode->succ_end(); P != E; ++P) {
-            //if(lower_bound(CurrentSet->begin(),
-            //     CurrentSet->end(), *P) != CurrentSet->end()) {
-            if(std::find(CurrentSet->begin(), CurrentSet->end(), *P) != CurrentSet->end()) {
-              if(ignoreEdge(highestHeightNode, *P))
-                continue;
-              //If not already in Intersect, add
-              if(!IntersectCurrent.count(*P))
-                IntersectCurrent.insert(*P);
-            }
-          }
-        } //End while loop over Intersect Size
-
-        //Change direction
-        order = BOTTOM_UP;
-
-        //Reset Intersect to reflect changes in OrderNodes
-        IntersectCurrent.clear();
-        predIntersect(*CurrentSet, IntersectCurrent);
-
-      } //End If TOP_DOWN
-
-        //Begin if BOTTOM_UP
-      else {
-        DEBUG(std::cerr << "Order is BOTTOM UP\n");
-        while(IntersectCurrent.size() > 0) {
-          DEBUG(std::cerr << "Intersection of size " << IntersectCurrent.size() << ", finding highest depth\n");
-
-          //dump intersection
-          DEBUG(dumpIntersection(IntersectCurrent));
-          //Get node with highest depth, if a tie, use one with lowest
-          //MOB
-          int MOB = 0;
-          int depth = 0;
-          MSchedGraphSBNode *highestDepthNode = *(IntersectCurrent.begin());
-
-          for(std::set<MSchedGraphSBNode*>::iterator I = IntersectCurrent.begin(),
-                E = IntersectCurrent.end(); I != E; ++I) {
-            //Find node attribute in graph
-            MSNodeSBAttributes nodeAttr= nodeToAttributesMap.find(*I)->second;
-
-            if(depth < nodeAttr.depth) {
-              highestDepthNode = *I;
-              depth = nodeAttr.depth;
-              MOB = nodeAttr.MOB;
-            }
-            else if(depth == nodeAttr.depth) {
-              if(MOB > nodeAttr.MOB) {
-                highestDepthNode = *I;
-                depth = nodeAttr.depth;
-                MOB = nodeAttr.MOB;
-              }
-            }
-          }
-
-
-
-          //Append highest depth node to the NodeOrder
-           if(std::find(FinalNodeOrder.begin(), FinalNodeOrder.end(), highestDepthNode) == FinalNodeOrder.end()) {
-             DEBUG(std::cerr << "Adding node to Final Order: " << *highestDepthNode << "\n");
-             FinalNodeOrder.push_back(highestDepthNode);
-           }
-          //Remove heightestDepthNode from IntersectOrder
-           IntersectCurrent.erase(highestDepthNode);
-
-
-          //Intersect heightDepthNode's pred with CurrentSet
-          for(MSchedGraphSBNode::pred_iterator P = highestDepthNode->pred_begin(),
-                E = highestDepthNode->pred_end(); P != E; ++P) {
-            if(CurrentSet->count(*P)) {
-              if(ignoreEdge(*P, highestDepthNode))
-                continue;
-
-            //If not already in Intersect, add
-            if(!IntersectCurrent.count(*P))
-              IntersectCurrent.insert(*P);
-            }
-          }
-
-        } //End while loop over Intersect Size
-
-          //Change order
-        order = TOP_DOWN;
-
-        //Reset IntersectCurrent to reflect changes in OrderNodes
-        IntersectCurrent.clear();
-        succIntersect(*CurrentSet, IntersectCurrent);
-        } //End if BOTTOM_DOWN
-
-      DEBUG(std::cerr << "Current Intersection Size: " << IntersectCurrent.size() << "\n");
-    }
-    //End Wrapping while loop
-    DEBUG(std::cerr << "Ending Size of Current Set: " << CurrentSet->size() << "\n");
-  }//End for over all sets of nodes
-
-  //FIXME: As the algorithm stands it will NEVER add an instruction such as ba (with no
-  //data dependencies) to the final order. We add this manually. It will always be
-  //in the last set of S since its not part of a recurrence
-    //Loop over all the sets and place them in the final node order
-  std::vector<std::set<MSchedGraphSBNode*> > ::reverse_iterator LastSet = partialOrder.rbegin();
-  for(std::set<MSchedGraphSBNode*>::iterator CurrentNode = LastSet->begin(), LastNode = LastSet->end();
-      CurrentNode != LastNode; ++CurrentNode) {
-    if((*CurrentNode)->getInst()->getOpcode() == V9::BA)
-      FinalNodeOrder.push_back(*CurrentNode);
-  }
-  //Return final Order
-  //return FinalNodeOrder;
-}
-
-
-void ModuloSchedulingSBPass::predIntersect(std::set<MSchedGraphSBNode*> &CurrentSet, std::set<MSchedGraphSBNode*> &IntersectResult) {
-
-  for(unsigned j=0; j < FinalNodeOrder.size(); ++j) {
-    for(MSchedGraphSBNode::pred_iterator P = FinalNodeOrder[j]->pred_begin(),
-          E = FinalNodeOrder[j]->pred_end(); P != E; ++P) {
-
-      //Check if we are supposed to ignore this edge or not
-      if(ignoreEdge(*P,FinalNodeOrder[j]))
-        continue;
-
-      if(CurrentSet.count(*P))
-        if(std::find(FinalNodeOrder.begin(), FinalNodeOrder.end(), *P) == FinalNodeOrder.end())
-          IntersectResult.insert(*P);
-    }
-  }
-}
-
-void ModuloSchedulingSBPass::succIntersect(std::set<MSchedGraphSBNode*> &CurrentSet, std::set<MSchedGraphSBNode*> &IntersectResult) {
-
-  for(unsigned j=0; j < FinalNodeOrder.size(); ++j) {
-    for(MSchedGraphSBNode::succ_iterator P = FinalNodeOrder[j]->succ_begin(),
-          E = FinalNodeOrder[j]->succ_end(); P != E; ++P) {
-
-      //Check if we are supposed to ignore this edge or not
-      if(ignoreEdge(FinalNodeOrder[j],*P))
-        continue;
-
-      if(CurrentSet.count(*P))
-        if(std::find(FinalNodeOrder.begin(), FinalNodeOrder.end(), *P) == FinalNodeOrder.end())
-          IntersectResult.insert(*P);
-    }
-  }
-}
-
-
-
-bool ModuloSchedulingSBPass::computeSchedule(std::vector<const MachineBasicBlock*> &SB, MSchedGraphSB *MSG) {
-
-  TIME_REGION(X, "computeSchedule");
-
-  bool success = false;
-
-  //FIXME: Should be set to max II of the original loop
-  //Cap II in order to prevent infinite loop
-  int capII = MSG->totalDelay();
-
-  while(!success) {
-
-    //Keep track of branches, but do not insert into the schedule
-    std::vector<MSchedGraphSBNode*> branches;
-
-    //Loop over the final node order and process each node
-    for(std::vector<MSchedGraphSBNode*>::iterator I = FinalNodeOrder.begin(),
-          E = FinalNodeOrder.end(); I != E; ++I) {
-
-      //CalculateEarly and Late start
-      bool initialLSVal = false;
-      bool initialESVal = false;
-      int EarlyStart = 0;
-      int LateStart = 0;
-      bool hasSucc = false;
-      bool hasPred = false;
-      bool sched;
-
-      if((*I)->isBranch())
-        if((*I)->hasPredecessors())
-          sched = true;
-        else
-          sched = false;
-      else
-        sched = true;
-
-      if(sched) {
-        //Loop over nodes in the schedule and determine if they are predecessors
-        //or successors of the node we are trying to schedule
-        for(MSScheduleSB::schedule_iterator nodesByCycle = schedule.begin(), nodesByCycleEnd = schedule.end();
-            nodesByCycle != nodesByCycleEnd; ++nodesByCycle) {
-
-          //For this cycle, get the vector of nodes schedule and loop over it
-          for(std::vector<MSchedGraphSBNode*>::iterator schedNode = nodesByCycle->second.begin(), SNE = nodesByCycle->second.end(); schedNode != SNE; ++schedNode) {
-
-            if((*I)->isPredecessor(*schedNode)) {
-              int diff = (*I)->getInEdge(*schedNode).getIteDiff();
-              int ES_Temp = nodesByCycle->first + (*schedNode)->getLatency() - diff * II;
-              DEBUG(std::cerr << "Diff: " << diff << " Cycle: " << nodesByCycle->first << "\n");
-              DEBUG(std::cerr << "Temp EarlyStart: " << ES_Temp << " Prev EarlyStart: " << EarlyStart << "\n");
-              if(initialESVal)
-                EarlyStart = std::max(EarlyStart, ES_Temp);
-              else {
-                EarlyStart = ES_Temp;
-                initialESVal = true;
-              }
-              hasPred = true;
-            }
-            if((*I)->isSuccessor(*schedNode)) {
-              int diff = (*schedNode)->getInEdge(*I).getIteDiff();
-              int LS_Temp = nodesByCycle->first - (*I)->getLatency() + diff * II;
-              DEBUG(std::cerr << "Diff: " << diff << " Cycle: " << nodesByCycle->first << "\n");
-              DEBUG(std::cerr << "Temp LateStart: " << LS_Temp << " Prev LateStart: " << LateStart << "\n");
-              if(initialLSVal)
-                LateStart = std::min(LateStart, LS_Temp);
-              else {
-                LateStart = LS_Temp;
-                initialLSVal = true;
-              }
-              hasSucc = true;
-            }
-          }
-        }
-      }
-      else {
-        branches.push_back(*I);
-        continue;
-      }
-
-      //Check if the node has no pred or successors and set Early Start to its ASAP
-      if(!hasSucc && !hasPred)
-        EarlyStart = nodeToAttributesMap.find(*I)->second.ASAP;
-
-      DEBUG(std::cerr << "Has Successors: " << hasSucc << ", Has Pred: " << hasPred << "\n");
-      DEBUG(std::cerr << "EarlyStart: " << EarlyStart << ", LateStart: " << LateStart << "\n");
-
-      //Now, try to schedule this node depending upon its pred and successor in the schedule
-      //already
-      if(!hasSucc && hasPred)
-        success = scheduleNode(*I, EarlyStart, (EarlyStart + II -1));
-      else if(!hasPred && hasSucc)
-        success = scheduleNode(*I, LateStart, (LateStart - II +1));
-      else if(hasPred && hasSucc) {
-        if(EarlyStart > LateStart) {
-        success = false;
-          //LateStart = EarlyStart;
-          DEBUG(std::cerr << "Early Start can not be later then the late start cycle, schedule fails\n");
-        }
-        else
-          success = scheduleNode(*I, EarlyStart, std::min(LateStart, (EarlyStart + II -1)));
-      }
-      else
-        success = scheduleNode(*I, EarlyStart, EarlyStart + II - 1);
-
-      if(!success) {
-        ++II;
-        schedule.clear();
-        break;
-      }
-
-    }
-
-    if(success) {
-      DEBUG(std::cerr << "Constructing Schedule Kernel\n");
-      success = schedule.constructKernel(II, branches, indVarInstrs[SB]);
-      DEBUG(std::cerr << "Done Constructing Schedule Kernel\n");
-      if(!success) {
-        ++II;
-        schedule.clear();
-      }
-      DEBUG(std::cerr << "Final II: " << II << "\n");
-
-    }
-
-    if(II >= capII) {
-      DEBUG(std::cerr << "Maximum II reached, giving up\n");
-      return false;
-    }
-
-    assert(II < capII && "The II should not exceed the original loop number of cycles");
-  }
-  return true;
-}
-
-
-bool ModuloSchedulingSBPass::scheduleNode(MSchedGraphSBNode *node,
-                                      int start, int end) {
-  bool success = false;
-
-  DEBUG(std::cerr << *node << " (Start Cycle: " << start << ", End Cycle: " << end << ")\n");
-
-  //Make sure start and end are not negative
-  //if(start < 0) {
-  //start = 0;
-
-  //}
-  //if(end < 0)
-  //end = 0;
-
-  bool forward = true;
-  if(start > end)
-    forward = false;
-
-  bool increaseSC = true;
-  int cycle = start ;
-
-
-  while(increaseSC) {
-
-    increaseSC = false;
-
-    increaseSC = schedule.insert(node, cycle, II);
-
-    if(!increaseSC)
-      return true;
-
-    //Increment cycle to try again
-    if(forward) {
-      ++cycle;
-      DEBUG(std::cerr << "Increase cycle: " << cycle << "\n");
-      if(cycle > end)
-        return false;
-    }
-    else {
-      --cycle;
-      DEBUG(std::cerr << "Decrease cycle: " << cycle << "\n");
-      if(cycle < end)
-        return false;
-    }
-  }
-
-  return success;
-}
-
-void ModuloSchedulingSBPass::reconstructLoop(std::vector<const MachineBasicBlock*> &SB) {
-
-  TIME_REGION(X, "reconstructLoop");
-
-
-  DEBUG(std::cerr << "Reconstructing Loop\n");
-
-  //First find the value *'s that we need to "save"
-  std::map<const Value*, std::pair<const MachineInstr*, int> > valuesToSave;
-
-  //Keep track of instructions we have already seen and their stage because
-  //we don't want to "save" values if they are used in the kernel immediately
-  std::map<const MachineInstr*, int> lastInstrs;
-
-
-  std::set<MachineBasicBlock*> seenBranchesBB;
-  const TargetInstrInfo *MTI = target.getInstrInfo();
-  std::map<MachineBasicBlock*, std::vector<std::pair<MachineInstr*, int> > > instrsMovedDown;
-  std::map<MachineBasicBlock*, int> branchStage;
-
-  //Loop over kernel and only look at instructions from a stage > 0
-  //Look at its operands and save values *'s that are read
-  for(MSScheduleSB::kernel_iterator I = schedule.kernel_begin(), E = schedule.kernel_end(); I != E; ++I) {
-
-    if(I->second !=0) {
-      //For this instruction, get the Value*'s that it reads and put them into the set.
-      //Assert if there is an operand of another type that we need to save
-      const MachineInstr *inst = I->first;
-      lastInstrs[inst] = I->second;
-
-      for(unsigned i=0; i < inst->getNumOperands(); ++i) {
-        //get machine operand
-        const MachineOperand &mOp = inst->getOperand(i);
-
-        if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isUse()) {
-          //find the value in the map
-          if (const Value* srcI = mOp.getVRegValue()) {
-
-            if(isa<Constant>(srcI) || isa<Argument>(srcI))
-              continue;
-
-            //Before we declare this Value* one that we should save
-            //make sure its def is not of the same stage as this instruction
-            //because it will be consumed before its used
-            Instruction *defInst = (Instruction*) srcI;
-
-            //Should we save this value?
-            bool save = true;
-
-            //Continue if not in the def map, loop invariant code does not need to be saved
-            if(!defMap.count(srcI))
-              continue;
-
-            MachineInstr *defInstr = defMap[srcI];
-
-
-            if(lastInstrs.count(defInstr)) {
-              if(lastInstrs[defInstr] == I->second) {
-                save = false;
-
-              }
-            }
-
-            if(save)
-              valuesToSave[srcI] = std::make_pair(I->first, i);
-          }
-        }
-
-        if(mOp.getType() != MachineOperand::MO_VirtualRegister && mOp.isUse()) {
-          assert("Our assumption is wrong. We have another type of register that needs to be saved\n");
-        }
-      }
-    }
-
-
-    //Do a check to see if instruction was moved below its original branch
-    if(MTI->isBranch(I->first->getOpcode())) {
-      seenBranchesBB.insert(I->first->getParent());
-      branchStage[I->first->getParent()] = I->second;
-    }
-    else {
-      instrsMovedDown[I->first->getParent()].push_back(std::make_pair(I->first, I->second));
-      //assert(seenBranchesBB.count(I->first->getParent()) && "Instruction moved below branch\n");
-    }
-
-  }
-
-  //The new loop will consist of one or more prologues, the kernel, and one or more epilogues.
-
-  //Map to keep track of old to new values
-  std::map<Value*, std::map<int, Value*> > newValues;
-
-  //Map to keep track of old to new values in kernel
-  std::map<Value*, std::map<int, Value*> > kernelPHIs;
-
-  //Another map to keep track of what machine basic blocks these new value*s are in since
-  //they have no llvm instruction equivalent
-  std::map<Value*, MachineBasicBlock*> newValLocation;
-
-  std::vector<std::vector<MachineBasicBlock*> > prologues;
-  std::vector<std::vector<BasicBlock*> > llvm_prologues;
-
-  //Map to keep track of where the inner branches go
-  std::map<const MachineBasicBlock*, Value*> sideExits;
-
-
-  //Write prologue
-  if(schedule.getMaxStage() != 0)
-    writePrologues(prologues, SB, llvm_prologues, valuesToSave, newValues, newValLocation);
-
-  std::vector<BasicBlock*> llvmKernelBBs;
-  std::vector<MachineBasicBlock*> machineKernelBBs;
-  Function *parent = (Function*) SB[0]->getBasicBlock()->getParent();
-
-  for(unsigned i = 0; i < SB.size(); ++i) {
-    llvmKernelBBs.push_back(new BasicBlock("Kernel", parent));
-
-    machineKernelBBs.push_back(new MachineBasicBlock(llvmKernelBBs[i]));
-    (((MachineBasicBlock*)SB[0])->getParent())->getBasicBlockList().push_back(machineKernelBBs[i]);
-  }
-
-  writeKernel(llvmKernelBBs, machineKernelBBs, valuesToSave, newValues, newValLocation, kernelPHIs);
-
-
-  std::vector<std::vector<MachineBasicBlock*> > epilogues;
-  std::vector<std::vector<BasicBlock*> > llvm_epilogues;
-
-  //Write epilogues
-  if(schedule.getMaxStage() != 0)
-    writeEpilogues(epilogues, SB, llvm_epilogues, valuesToSave, newValues, newValLocation, kernelPHIs);
-
-
-  //Fix our branches
-  fixBranches(prologues, llvm_prologues, machineKernelBBs, llvmKernelBBs, epilogues, llvm_epilogues, SB, sideExits);
-
-  //Print out epilogues and prologue
-  DEBUG(for(std::vector<std::vector<MachineBasicBlock*> >::iterator PI = prologues.begin(), PE = prologues.end();
-      PI != PE; ++PI) {
-          std::cerr << "PROLOGUE\n";
-          for(std::vector<MachineBasicBlock*>::iterator I = PI->begin(), E = PI->end(); I != E; ++I)
-            (*I)->print(std::cerr);
-        });
-
-  DEBUG(std::cerr << "KERNEL\n");
-  DEBUG(for(std::vector<MachineBasicBlock*>::iterator I = machineKernelBBs.begin(), E = machineKernelBBs.end(); I != E; ++I) { (*I)->print(std::cerr);});
-
-  DEBUG(for(std::vector<std::vector<MachineBasicBlock*> >::iterator EI = epilogues.begin(), EE = epilogues.end();
-      EI != EE; ++EI) {
-    std::cerr << "EPILOGUE\n";
-    for(std::vector<MachineBasicBlock*>::iterator I = EI->begin(), E = EI->end(); I != E; ++I)
-      (*I)->print(std::cerr);
-  });
-
-
-  //Remove phis
-  removePHIs(SB, prologues, epilogues, machineKernelBBs, newValLocation);
-
-  //Print out epilogues and prologue
-  DEBUG(for(std::vector<std::vector<MachineBasicBlock*> >::iterator PI = prologues.begin(), PE = prologues.end();
-      PI != PE; ++PI) {
-          std::cerr << "PROLOGUE\n";
-          for(std::vector<MachineBasicBlock*>::iterator I = PI->begin(), E = PI->end(); I != E; ++I)
-            (*I)->print(std::cerr);
-        });
-
-  DEBUG(std::cerr << "KERNEL\n");
-  DEBUG(for(std::vector<MachineBasicBlock*>::iterator I = machineKernelBBs.begin(), E = machineKernelBBs.end(); I != E; ++I) { (*I)->print(std::cerr);});
-
-  DEBUG(for(std::vector<std::vector<MachineBasicBlock*> >::iterator EI = epilogues.begin(), EE = epilogues.end();
-      EI != EE; ++EI) {
-    std::cerr << "EPILOGUE\n";
-    for(std::vector<MachineBasicBlock*>::iterator I = EI->begin(), E = EI->end(); I != E; ++I)
-      (*I)->print(std::cerr);
-  });
-
-  writeSideExits(prologues, llvm_prologues, epilogues, llvm_epilogues, sideExits, instrsMovedDown, SB, machineKernelBBs, branchStage);
-
-
-  DEBUG(std::cerr << "New Machine Function" << "\n");
-}
-
-
-void ModuloSchedulingSBPass::fixBranches(std::vector<std::vector<MachineBasicBlock*> > &prologues, std::vector<std::vector<BasicBlock*> > &llvm_prologues, std::vector<MachineBasicBlock*> &machineKernelBB, std::vector<BasicBlock*> &llvmKernelBB, std::vector<std::vector<MachineBasicBlock*> > &epilogues, std::vector<std::vector<BasicBlock*> > &llvm_epilogues, std::vector<const MachineBasicBlock*> &SB, std::map<const MachineBasicBlock*, Value*> &sideExits) {
-
-  const TargetInstrInfo *TMI = target.getInstrInfo();
-
-  //Get exit BB
-  BasicBlock *last = (BasicBlock*) SB[SB.size()-1]->getBasicBlock();
-  BasicBlock *kernel_exit = 0;
-  bool sawFirst = false;
-
-  for(succ_iterator I = succ_begin(last),
-        E = succ_end(last); I != E; ++I) {
-    if (*I != SB[0]->getBasicBlock()) {
-      kernel_exit = *I;
-      break;
-    }
-    else
-      sawFirst = true;
-  }
-  if(!kernel_exit && sawFirst) {
-    kernel_exit = (BasicBlock*) SB[0]->getBasicBlock();
-  }
-
-  assert(kernel_exit && "Kernel Exit can not be null");
-
-  if(schedule.getMaxStage() != 0) {
-    //Fix prologue branches
-    for(unsigned i = 0; i <  prologues.size(); ++i) {
-
-      for(unsigned j = 0; j < prologues[i].size(); ++j) {
-
-        MachineBasicBlock *currentMBB = prologues[i][j];
-
-        //Find terminator since getFirstTerminator does not work!
-        for(MachineBasicBlock::reverse_iterator mInst = currentMBB->rbegin(), mInstEnd = currentMBB->rend(); mInst != mInstEnd; ++mInst) {
-          MachineOpCode OC = mInst->getOpcode();
-          //If its a branch update its branchto
-          if(TMI->isBranch(OC)) {
-            for(unsigned opNum = 0; opNum < mInst->getNumOperands(); ++opNum) {
-              MachineOperand &mOp = mInst->getOperand(opNum);
-              if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-                //Check if we are branching to the kernel, if not branch to epilogue
-                if(mOp.getVRegValue() == SB[0]->getBasicBlock()) {
-                  if(i >= prologues.size()-1)
-                    mOp.setValueReg(llvmKernelBB[0]);
-                  else
-                    mOp.setValueReg(llvm_prologues[i+1][0]);
-                }
-                else if( (mOp.getVRegValue() == kernel_exit) && (j == prologues[i].size()-1)) {
-                  mOp.setValueReg(llvm_epilogues[i][0]);
-                }
-                else if(mOp.getVRegValue() == SB[j+1]->getBasicBlock()) {
-                  mOp.setValueReg(llvm_prologues[i][j+1]);
-                }
-
-              }
-            }
-
-            DEBUG(std::cerr << "New Prologue Branch: " << *mInst << "\n");
-          }
-        }
-
-        //Update llvm basic block with our new branch instr
-        DEBUG(std::cerr << SB[i]->getBasicBlock()->getTerminator() << "\n");
-
-        const BranchInst *branchVal = dyn_cast<BranchInst>(SB[i]->getBasicBlock()->getTerminator());
-
-        //Check for inner branch
-        if(j < prologues[i].size()-1) {
-          //Find our side exit LLVM basic block
-          BasicBlock *sideExit = 0;
-          for(unsigned s = 0; s < branchVal->getNumSuccessors(); ++s) {
-            if(branchVal->getSuccessor(s) != SB[i+1]->getBasicBlock())
-              sideExit = branchVal->getSuccessor(s);
-          }
-          assert(sideExit && "Must have side exit llvm basic block");
-          TerminatorInst *newBranch = new BranchInst(sideExit,
-                                        llvm_prologues[i][j+1],
-                                        branchVal->getCondition(),
-                                        llvm_prologues[i][j]);
-        }
-        else {
-          //If last prologue
-          if(i == prologues.size()-1) {
-            TerminatorInst *newBranch = new BranchInst(llvmKernelBB[0],
-                                                       llvm_epilogues[i][0],
-                                                       branchVal->getCondition(),
-                                                       llvm_prologues[i][j]);
-          }
-          else {
-            TerminatorInst *newBranch = new BranchInst(llvm_prologues[i+1][0],
-                                                       llvm_epilogues[i][0],
-                                                       branchVal->getCondition(),
-                                                       llvm_prologues[i][j]);
-          }
-        }
-      }
-    }
-  }
-
-  //Fix up kernel machine branches
-  for(unsigned i = 0; i < machineKernelBB.size(); ++i) {
-    MachineBasicBlock *currentMBB = machineKernelBB[i];
-
-    for(MachineBasicBlock::reverse_iterator mInst = currentMBB->rbegin(), mInstEnd = currentMBB->rend(); mInst != mInstEnd; ++mInst) {
-      MachineOpCode OC = mInst->getOpcode();
-      if(TMI->isBranch(OC)) {
-        for(unsigned opNum = 0; opNum < mInst->getNumOperands(); ++opNum) {
-          MachineOperand &mOp = mInst->getOperand(opNum);
-
-          if(mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-            //Deal with inner kernel branches
-            if(i < machineKernelBB.size()-1) {
-              if(mOp.getVRegValue() == SB[i+1]->getBasicBlock())
-                mOp.setValueReg(llvmKernelBB[i+1]);
-              //Side exit!
-              else {
-                sideExits[SB[i]] = mOp.getVRegValue();
-              }
-            }
-            else {
-              if(mOp.getVRegValue() == SB[0]->getBasicBlock())
-                mOp.setValueReg(llvmKernelBB[0]);
-              else {
-                if(llvm_epilogues.size() > 0)
-                  mOp.setValueReg(llvm_epilogues[0][0]);
-              }
-            }
-          }
-        }
-      }
-    }
-
-    //Update kernelLLVM branches
-    const BranchInst *branchVal = dyn_cast<BranchInst>(SB[0]->getBasicBlock()->getTerminator());
-
-    //deal with inner branch
-    if(i < machineKernelBB.size()-1) {
-
-      //Find our side exit LLVM basic block
-      BasicBlock *sideExit = 0;
-      for(unsigned s = 0; s < branchVal->getNumSuccessors(); ++s) {
-        if(branchVal->getSuccessor(s) != SB[i+1]->getBasicBlock())
-          sideExit = branchVal->getSuccessor(s);
-      }
-      assert(sideExit && "Must have side exit llvm basic block");
-      TerminatorInst *newBranch = new BranchInst(sideExit,
-                                                 llvmKernelBB[i+1],
-                                                 branchVal->getCondition(),
-                                                 llvmKernelBB[i]);
-    }
-    else {
-      //Deal with outter branches
-      if(epilogues.size() > 0) {
-        TerminatorInst *newBranch = new BranchInst(llvmKernelBB[0],
-                                                   llvm_epilogues[0][0],
-                                                   branchVal->getCondition(),
-                                                   llvmKernelBB[i]);
-      }
-      else {
-        TerminatorInst *newBranch = new BranchInst(llvmKernelBB[0],
-                                                   kernel_exit,
-                                                   branchVal->getCondition(),
-                                                   llvmKernelBB[i]);
-      }
-    }
-  }
-
-  if(schedule.getMaxStage() != 0) {
-
-    //Lastly add unconditional branches for the epilogues
-    for(unsigned i = 0; i <  epilogues.size(); ++i) {
-
-      for(unsigned j=0; j < epilogues[i].size(); ++j) {
-        //Now since we don't have fall throughs, add a unconditional
-        //branch to the next prologue
-
-        //Before adding these, we need to check if the epilogue already has
-        //a branch in it
-        bool hasBranch = false;
-        /*if(j < epilogues[i].size()-1) {
-          MachineBasicBlock *currentMBB = epilogues[i][j];
-          for(MachineBasicBlock::reverse_iterator mInst = currentMBB->rbegin(), mInstEnd = currentMBB->rend(); mInst != mInstEnd; ++mInst) {
-
-            MachineOpCode OC = mInst->getOpcode();
-
-            //If its a branch update its branchto
-            if(TMI->isBranch(OC)) {
-              hasBranch = true;
-              for(unsigned opNum = 0; opNum < mInst->getNumOperands(); ++opNum) {
-                MachineOperand &mOp = mInst->getOperand(opNum);
-                if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-
-                  if(mOp.getVRegValue() != sideExits[SB[j]]) {
-                    mOp.setValueReg(llvm_epilogues[i][j+1]);
-                  }
-
-                }
-              }
-
-
-              DEBUG(std::cerr << "New Epilogue Branch: " << *mInst << "\n");
-            }
-          }
-          if(hasBranch) {
-            const BranchInst *branchVal = dyn_cast<BranchInst>(SB[j]->getBasicBlock()->getTerminator());
-            TerminatorInst *newBranch = new BranchInst((BasicBlock*)sideExits[SB[j]],
-                                                       llvm_epilogues[i][j+1],
-                                                       branchVal->getCondition(),
-                                                       llvm_epilogues[i][j]);
-          }
-          }*/
-
-        if(!hasBranch) {
-
-          //Handle inner branches
-          if(j < epilogues[i].size()-1) {
-            BuildMI(epilogues[i][j], V9::BA, 1).addPCDisp(llvm_epilogues[i][j+1]);
-            TerminatorInst *newBranch = new BranchInst(llvm_epilogues[i][j+1],
-                                                       llvm_epilogues[i][j]);
-          }
-          else {
-
-            //Check if this is the last epilogue
-            if(i != epilogues.size()-1) {
-              BuildMI(epilogues[i][j], V9::BA, 1).addPCDisp(llvm_epilogues[i+1][0]);
-              //Add unconditional branch to end of epilogue
-              TerminatorInst *newBranch = new BranchInst(llvm_epilogues[i+1][0],
-                                                         llvm_epilogues[i][j]);
-
-            }
-            else {
-              BuildMI(epilogues[i][j], V9::BA, 1).addPCDisp(kernel_exit);
-              TerminatorInst *newBranch = new BranchInst(kernel_exit, llvm_epilogues[i][j]);
-            }
-          }
-
-          //Add one more nop!
-          BuildMI(epilogues[i][j], V9::NOP, 0);
-
-        }
-      }
-    }
-  }
-
-  //Find all llvm basic blocks that branch to the loop entry and
-  //change to our first prologue.
-  const BasicBlock *llvmBB = SB[0]->getBasicBlock();
-
-  std::vector<const BasicBlock*>Preds (pred_begin(llvmBB), pred_end(llvmBB));
-
-  for(std::vector<const BasicBlock*>::iterator P = Preds.begin(),
-        PE = Preds.end(); P != PE; ++P) {
-    if(*P == SB[SB.size()-1]->getBasicBlock())
-       continue;
-     else {
-       DEBUG(std::cerr << "Found our entry BB\n");
-       DEBUG((*P)->print(std::cerr));
-       //Get the Terminator instruction for this basic block and print it out
-       //DEBUG(std::cerr << *((*P)->getTerminator()) << "\n");
-
-       //Update the terminator
-       TerminatorInst *term = ((BasicBlock*)*P)->getTerminator();
-       for(unsigned i=0; i < term->getNumSuccessors(); ++i) {
-         if(term->getSuccessor(i) == llvmBB) {
-           DEBUG(std::cerr << "Replacing successor bb\n");
-           if(llvm_prologues.size() > 0) {
-             term->setSuccessor(i, llvm_prologues[0][0]);
-
-             DEBUG(std::cerr << "New Term" << *((*P)->getTerminator()) << "\n");
-
-             //Also update its corresponding machine instruction
-             MachineCodeForInstruction & tempMvec =
-               MachineCodeForInstruction::get(term);
-             for (unsigned j = 0; j < tempMvec.size(); j++) {
-               MachineInstr *temp = tempMvec[j];
-               MachineOpCode opc = temp->getOpcode();
-               if(TMI->isBranch(opc)) {
-                 DEBUG(std::cerr << *temp << "\n");
-                 //Update branch
-                 for(unsigned opNum = 0; opNum < temp->getNumOperands(); ++opNum) {
-                   MachineOperand &mOp = temp->getOperand(opNum);
-                   if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-                     if(mOp.getVRegValue() == llvmBB)
-                       mOp.setValueReg(llvm_prologues[0][0]);
-                   }
-                 }
-               }
-             }
-           }
-           else {
-             term->setSuccessor(i, llvmKernelBB[0]);
-
-             //Also update its corresponding machine instruction
-             MachineCodeForInstruction & tempMvec =
-               MachineCodeForInstruction::get(term);
-             for(unsigned j = 0; j < tempMvec.size(); j++) {
-               MachineInstr *temp = tempMvec[j];
-               MachineOpCode opc = temp->getOpcode();
-               if(TMI->isBranch(opc)) {
-                 DEBUG(std::cerr << *temp << "\n");
-                 //Update branch
-                 for(unsigned opNum = 0; opNum < temp->getNumOperands(); ++opNum) {
-                   MachineOperand &mOp = temp->getOperand(opNum);
-                   if(mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-                     if(mOp.getVRegValue() == llvmBB)
-                       mOp.setValueReg(llvmKernelBB[0]);
-                   }
-                 }
-               }
-             }
-           }
-         }
-       }
-       break;
-     }
-  }
-
-}
-
-
-void ModuloSchedulingSBPass::writePrologues(std::vector<std::vector<MachineBasicBlock *> > &prologues, std::vector<const MachineBasicBlock*> &origSB, std::vector<std::vector<BasicBlock*> > &llvm_prologues, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave, std::map<Value*, std::map<int, Value*> > &newValues, std::map<Value*, MachineBasicBlock*> &newValLocation) {
-
-  //Keep a map to easily know whats in the kernel
-  std::map<int, std::set<const MachineInstr*> > inKernel;
-  int maxStageCount = 0;
-
-  //Keep a map of new values we consumed in case they need to be added back
-  std::map<Value*, std::map<int, Value*> > consumedValues;
-
-  DEBUG(schedule.print(std::cerr));
-
-  for(MSScheduleSB::kernel_iterator I = schedule.kernel_begin(), E = schedule.kernel_end(); I != E; ++I) {
-    maxStageCount = std::max(maxStageCount, I->second);
-
-    //Put int the map so we know what instructions in each stage are in the kernel
-    DEBUG(std::cerr << "Inserting instruction " << *(I->first) << " into map at stage " << I->second << "\n");
-    inKernel[I->second].insert(I->first);
-  }
-
-  //Get target information to look at machine operands
-  const TargetInstrInfo *mii = target.getInstrInfo();
-
- //Now write the prologues
-  for(int i = 0; i < maxStageCount; ++i) {
-    std::vector<MachineBasicBlock*> current_prologue;
-    std::vector<BasicBlock*> current_llvm_prologue;
-
-    for(std::vector<const MachineBasicBlock*>::iterator MB = origSB.begin(),
-          MBE = origSB.end(); MB != MBE; ++MB) {
-      const MachineBasicBlock *MBB = *MB;
-      //Create new llvm and machine bb
-      BasicBlock *llvmBB = new BasicBlock("PROLOGUE", (Function*) (MBB->getBasicBlock()->getParent()));
-      MachineBasicBlock *machineBB = new MachineBasicBlock(llvmBB);
-
-      DEBUG(std::cerr << "i=" << i << "\n");
-
-      for(int j = i; j >= 0; --j) {
-        //iterate over instructions in original bb
-        for(MachineBasicBlock::const_iterator MI = MBB->begin(),
-              ME = MBB->end(); ME != MI; ++MI) {
-          if(inKernel[j].count(&*MI)) {
-            MachineInstr *instClone = MI->clone();
-            machineBB->push_back(instClone);
-
-            //If its a branch, insert a nop
-            if(mii->isBranch(instClone->getOpcode()))
-              BuildMI(machineBB, V9::NOP, 0);
-
-
-            DEBUG(std::cerr << "Cloning: " << *MI << "\n");
-
-            //After cloning, we may need to save the value that this instruction defines
-            for(unsigned opNum=0; opNum < MI->getNumOperands(); ++opNum) {
-              Instruction *tmp;
-
-              //get machine operand
-              MachineOperand &mOp = instClone->getOperand(opNum);
-              if(mOp.getType() == MachineOperand::MO_VirtualRegister
-                 && mOp.isDef()) {
-
-                //Check if this is a value we should save
-                if(valuesToSave.count(mOp.getVRegValue())) {
-                  //Save copy in tmpInstruction
-                  tmp = new TmpInstruction(mOp.getVRegValue());
-
-                  //Add TmpInstruction to safe LLVM Instruction MCFI
-                  MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-                  tempMvec.addTemp((Value*) tmp);
-
-                  DEBUG(std::cerr << "Value: " << *(mOp.getVRegValue())
-                        << " New Value: " << *tmp << " Stage: " << i << "\n");
-
-                newValues[mOp.getVRegValue()][i]= tmp;
-                newValLocation[tmp] = machineBB;
-
-                DEBUG(std::cerr << "Machine Instr Operands: "
-                      << *(mOp.getVRegValue()) << ", 0, " << *tmp << "\n");
-
-                //Create machine instruction and put int machineBB
-                MachineInstr *saveValue;
-                if(mOp.getVRegValue()->getType() == Type::FloatTy)
-                  saveValue = BuildMI(machineBB, V9::FMOVS, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-                else if(mOp.getVRegValue()->getType() == Type::DoubleTy)
-                  saveValue = BuildMI(machineBB, V9::FMOVD, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-                else
-                  saveValue = BuildMI(machineBB, V9::ORr, 3).addReg(mOp.getVRegValue()).addImm(0).addRegDef(tmp);
-
-
-                DEBUG(std::cerr << "Created new machine instr: " << *saveValue << "\n");
-                }
-              }
-
-              //We may also need to update the value that we use if
-              //its from an earlier prologue
-              if(j != 0) {
-                if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isUse()) {
-                  if(newValues.count(mOp.getVRegValue())) {
-                    if(newValues[mOp.getVRegValue()].count(i-1)) {
-                      Value *oldV =  mOp.getVRegValue();
-                      DEBUG(std::cerr << "Replaced this value: " << mOp.getVRegValue() << " With:" << (newValues[mOp.getVRegValue()][i-1]) << "\n");
-                      //Update the operand with the right value
-                      mOp.setValueReg(newValues[mOp.getVRegValue()][i-1]);
-
-                      //Remove this value since we have consumed it
-                      //NOTE: Should this only be done if j != maxStage?
-                      consumedValues[oldV][i-1] = (newValues[oldV][i-1]);
-                      DEBUG(std::cerr << "Deleted value: " << consumedValues[oldV][i-1] << "\n");
-                      newValues[oldV].erase(i-1);
-                    }
-                  }
-                  else
-                    if(consumedValues.count(mOp.getVRegValue()))
-                      assert(!consumedValues[mOp.getVRegValue()].count(i-1) && "Found a case where we need the value");
-                }
-              }
-            }
-          }
-        }
-      }
-        (((MachineBasicBlock*)MBB)->getParent())->getBasicBlockList().push_back(machineBB);
-        current_prologue.push_back(machineBB);
-        current_llvm_prologue.push_back(llvmBB);
-    }
-    prologues.push_back(current_prologue);
-    llvm_prologues.push_back(current_llvm_prologue);
-
-  }
-}
-
-
-void ModuloSchedulingSBPass::writeEpilogues(std::vector<std::vector<MachineBasicBlock*> > &epilogues, std::vector<const MachineBasicBlock*> &origSB, std::vector<std::vector<BasicBlock*> > &llvm_epilogues, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave, std::map<Value*, std::map<int, Value*> > &newValues,std::map<Value*, MachineBasicBlock*> &newValLocation, std::map<Value*, std::map<int, Value*> > &kernelPHIs ) {
-
-  std::map<int, std::set<const MachineInstr*> > inKernel;
-   const TargetInstrInfo *MTI = target.getInstrInfo();
-
-  for(MSScheduleSB::kernel_iterator I = schedule.kernel_begin(), E = schedule.kernel_end(); I != E; ++I) {
-
-    //Put int the map so we know what instructions in each stage are in the kernel
-    inKernel[I->second].insert(I->first);
-  }
-
-  std::map<Value*, Value*> valPHIs;
-
-  //some debug stuff, will remove later
-  DEBUG(for(std::map<Value*, std::map<int, Value*> >::iterator V = newValues.begin(), E = newValues.end(); V !=E; ++V) {
-    std::cerr << "Old Value: " << *(V->first) << "\n";
-    for(std::map<int, Value*>::iterator I = V->second.begin(), IE = V->second.end(); I != IE; ++I)
-      std::cerr << "Stage: " << I->first << " Value: " << *(I->second) << "\n";
-  });
-
-
-  //Now write the epilogues
-  for(int i = schedule.getMaxStage()-1; i >= 0; --i) {
-    std::vector<MachineBasicBlock*> current_epilogue;
-    std::vector<BasicBlock*> current_llvm_epilogue;
-
-    for(std::vector<const MachineBasicBlock*>::iterator MB = origSB.begin(), MBE = origSB.end(); MB != MBE; ++MB) {
-      const MachineBasicBlock *MBB = *MB;
-
-      BasicBlock *llvmBB = new BasicBlock("EPILOGUE", (Function*) (MBB->getBasicBlock()->getParent()));
-      MachineBasicBlock *machineBB = new MachineBasicBlock(llvmBB);
-
-      DEBUG(std::cerr << " Epilogue #: " << i << "\n");
-
-      std::map<Value*, int> inEpilogue;
-
-      for(MachineBasicBlock::const_iterator MI = MBB->begin(), ME = MBB->end(); ME != MI; ++MI) {
-        for(int j=schedule.getMaxStage(); j > i; --j) {
-          if(inKernel[j].count(&*MI)) {
-            DEBUG(std::cerr << "Cloning instruction " << *MI << "\n");
-            MachineInstr *clone = MI->clone();
-
-            //Update operands that need to use the result from the phi
-            for(unsigned opNum=0; opNum < clone->getNumOperands(); ++opNum) {
-              //get machine operand
-              const MachineOperand &mOp = clone->getOperand(opNum);
-
-              if((mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isUse())) {
-
-                DEBUG(std::cerr << "Writing PHI for " << (mOp.getVRegValue()) << "\n");
-
-                //If this is the last instructions for the max iterations ago, don't update operands
-                if(inEpilogue.count(mOp.getVRegValue()))
-                  if(inEpilogue[mOp.getVRegValue()] == i)
-                    continue;
-
-                //Quickly write appropriate phis for this operand
-                if(newValues.count(mOp.getVRegValue())) {
-                  if(newValues[mOp.getVRegValue()].count(i)) {
-                    Instruction *tmp = new TmpInstruction(newValues[mOp.getVRegValue()][i]);
-
-                    //Get machine code for this instruction
-                    MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-                    tempMvec.addTemp((Value*) tmp);
-
-                    //assert of no kernelPHI for this value
-                    assert(kernelPHIs[mOp.getVRegValue()][i] !=0 && "Must have final kernel phi to construct epilogue phi");
-
-                    MachineInstr *saveValue = BuildMI(machineBB, V9::PHI, 3).addReg(newValues[mOp.getVRegValue()][i]).addReg(kernelPHIs[mOp.getVRegValue()][i]).addRegDef(tmp);
-                    DEBUG(std::cerr << "Resulting PHI: " << *saveValue << "\n");
-                    valPHIs[mOp.getVRegValue()] = tmp;
-                  }
-                }
-
-                if(valPHIs.count(mOp.getVRegValue())) {
-                  //Update the operand in the cloned instruction
-                  clone->getOperand(opNum).setValueReg(valPHIs[mOp.getVRegValue()]);
-                }
-              }
-              else if((mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef())) {
-                inEpilogue[mOp.getVRegValue()] = i;
-              }
-
-            }
-            machineBB->push_back(clone);
-            //if(MTI->isBranch(clone->getOpcode()))
-            //BuildMI(machineBB, V9::NOP, 0);
-          }
-        }
-      }
-      (((MachineBasicBlock*)MBB)->getParent())->getBasicBlockList().push_back(machineBB);
-      current_epilogue.push_back(machineBB);
-      current_llvm_epilogue.push_back(llvmBB);
-    }
-
-    DEBUG(std::cerr << "EPILOGUE #" << i << "\n");
-    DEBUG(for(std::vector<MachineBasicBlock*>::iterator B = current_epilogue.begin(), BE = current_epilogue.end(); B != BE; ++B) {
-            (*B)->print(std::cerr);});
-
-    epilogues.push_back(current_epilogue);
-    llvm_epilogues.push_back(current_llvm_epilogue);
-  }
-}
-
-void ModuloSchedulingSBPass::writeKernel(std::vector<BasicBlock*> &llvmBB, std::vector<MachineBasicBlock*> &machineBB, std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave, std::map<Value*, std::map<int, Value*> > &newValues, std::map<Value*, MachineBasicBlock*> &newValLocation, std::map<Value*, std::map<int, Value*> > &kernelPHIs) {
-
-  //Keep track of operands that are read and saved from a previous iteration. The new clone
-  //instruction will use the result of the phi instead.
-  std::map<Value*, Value*> finalPHIValue;
-  std::map<Value*, Value*> kernelValue;
-
-  //Branches are a special case
-  std::vector<MachineInstr*> branches;
-
-  //Get target information to look at machine operands
-  const TargetInstrInfo *mii = target.getInstrInfo();
-  unsigned index = 0;
-  int numBr = 0;
-  bool seenBranch = false;
-
-  //Create TmpInstructions for the final phis
-  for(MSScheduleSB::kernel_iterator I = schedule.kernel_begin(), E = schedule.kernel_end(); I != E; ++I) {
-
-   DEBUG(std::cerr << "Stage: " << I->second << " Inst: " << *(I->first) << "\n";);
-
-   //Clone instruction
-   const MachineInstr *inst = I->first;
-   MachineInstr *instClone = inst->clone();
-
-   if(seenBranch && !mii->isBranch(instClone->getOpcode())) {
-     index++;
-     seenBranch = false;
-     numBr = 0;
-   }
-   else if(seenBranch && (numBr == 2)) {
-     index++;
-     numBr = 0;
-   }
-
-   //Insert into machine basic block
-   assert(index < machineBB.size() && "Must have a valid index into kernel MBBs");
-   machineBB[index]->push_back(instClone);
-
-   if(mii->isBranch(instClone->getOpcode())) {
-     BuildMI(machineBB[index], V9::NOP, 0);
-
-     seenBranch = true;
-     numBr++;
-   }
-
-   DEBUG(std::cerr <<  "Cloned Inst: " << *instClone << "\n");
-
-   //Loop over Machine Operands
-   for(unsigned i=0; i < inst->getNumOperands(); ++i) {
-     //get machine operand
-     const MachineOperand &mOp = inst->getOperand(i);
-
-     if(I->second != 0) {
-       if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isUse()) {
-
-         //Check to see where this operand is defined if this instruction is from max stage
-         if(I->second == schedule.getMaxStage()) {
-           DEBUG(std::cerr << "VREG: " << *(mOp.getVRegValue()) << "\n");
-         }
-
-         //If its in the value saved, we need to create a temp instruction and use that instead
-         if(valuesToSave.count(mOp.getVRegValue())) {
-
-           //Check if we already have a final PHI value for this
-           if(!finalPHIValue.count(mOp.getVRegValue())) {
-             //Only create phi if the operand def is from a stage before this one
-             if(schedule.defPreviousStage(mOp.getVRegValue(), I->second)) {
-             TmpInstruction *tmp = new TmpInstruction(mOp.getVRegValue());
-
-             //Get machine code for this instruction
-             MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-             tempMvec.addTemp((Value*) tmp);
-
-             //Update the operand in the cloned instruction
-             instClone->getOperand(i).setValueReg(tmp);
-
-             //save this as our final phi
-             finalPHIValue[mOp.getVRegValue()] = tmp;
-             newValLocation[tmp] = machineBB[index];
-             }
-           }
-           else {
-             //Use the previous final phi value
-             instClone->getOperand(i).setValueReg(finalPHIValue[mOp.getVRegValue()]);
-           }
-         }
-       }
-     }
-     if(I->second != schedule.getMaxStage()) {
-       if(mOp.getType() == MachineOperand::MO_VirtualRegister && mOp.isDef()) {
-         if(valuesToSave.count(mOp.getVRegValue())) {
-
-           TmpInstruction *tmp = new TmpInstruction(mOp.getVRegValue());
-
-           //Get machine code for this instruction
-           MachineCodeForInstruction & tempVec = MachineCodeForInstruction::get(defaultInst);
-           tempVec.addTemp((Value*) tmp);
-
-           //Create new machine instr and put in MBB
-           MachineInstr *saveValue;
-           if(mOp.getVRegValue()->getType() == Type::FloatTy)
-             saveValue = BuildMI(machineBB[index], V9::FMOVS, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-           else if(mOp.getVRegValue()->getType() == Type::DoubleTy)
-             saveValue = BuildMI(machineBB[index], V9::FMOVD, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-           else
-             saveValue = BuildMI(machineBB[index], V9::ORr, 3).addReg(mOp.getVRegValue()).addImm(0).addRegDef(tmp);
-
-
-           //Save for future cleanup
-           kernelValue[mOp.getVRegValue()] = tmp;
-           newValLocation[tmp] = machineBB[index];
-           kernelPHIs[mOp.getVRegValue()][schedule.getMaxStage()-1] = tmp;
-         }
-       }
-     }
-   }
-
- }
-
-  //Loop over each value we need to generate phis for
- for(std::map<Value*, std::map<int, Value*> >::iterator V = newValues.begin(),
-       E = newValues.end(); V != E; ++V) {
-
-
-   DEBUG(std::cerr << "Writing phi for" << *(V->first));
-   DEBUG(std::cerr << "\nMap of Value* for this phi\n");
-   DEBUG(for(std::map<int, Value*>::iterator I = V->second.begin(),
-               IE = V->second.end(); I != IE; ++I) {
-     std::cerr << "Stage: " << I->first;
-     std::cerr << " Value: " << *(I->second) << "\n";
-   });
-
-   //If we only have one current iteration live, its safe to set
-   //lastPhi = to kernel value
-   if(V->second.size() == 1) {
-     assert(kernelValue[V->first] != 0 && "Kernel value* must exist to create phi");
-     MachineInstr *saveValue = BuildMI(*machineBB[0], machineBB[0]->begin(),V9::PHI, 3).addReg(V->second.begin()->second).addReg(kernelValue[V->first]).addRegDef(finalPHIValue[V->first]);
-     DEBUG(std::cerr << "Resulting PHI (one live): " << *saveValue << "\n");
-     kernelPHIs[V->first][V->second.begin()->first] = kernelValue[V->first];
-     DEBUG(std::cerr << "Put kernel phi in at stage: " << schedule.getMaxStage()-1 << " (map stage = " << V->second.begin()->first << ")\n");
-   }
-   else {
-
-     //Keep track of last phi created.
-     Instruction *lastPhi = 0;
-
-     unsigned count = 1;
-     //Loop over the the map backwards to generate phis
-     for(std::map<int, Value*>::reverse_iterator I = V->second.rbegin(), IE = V->second.rend();
-         I != IE; ++I) {
-
-       if(count < (V->second).size()) {
-         if(lastPhi == 0) {
-           lastPhi = new TmpInstruction(I->second);
-
-           //Get machine code for this instruction
-           MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-           tempMvec.addTemp((Value*) lastPhi);
-
-           MachineInstr *saveValue = BuildMI(*machineBB[0], machineBB[0]->begin(), V9::PHI, 3).addReg(kernelValue[V->first]).addReg(I->second).addRegDef(lastPhi);
-           DEBUG(std::cerr << "Resulting PHI: " << *saveValue << "\n");
-           newValLocation[lastPhi] = machineBB[0];
-         }
-         else {
-           Instruction *tmp = new TmpInstruction(I->second);
-
-           //Get machine code for this instruction
-           MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-           tempMvec.addTemp((Value*) tmp);
-
-
-           MachineInstr *saveValue = BuildMI(*machineBB[0], machineBB[0]->begin(), V9::PHI, 3).addReg(lastPhi).addReg(I->second).addRegDef(tmp);
-           DEBUG(std::cerr << "Resulting PHI: " << *saveValue << "\n");
-           lastPhi = tmp;
-           kernelPHIs[V->first][I->first] = lastPhi;
-           newValLocation[lastPhi] = machineBB[0];
-         }
-       }
-       //Final phi value
-       else {
-         //The resulting value must be the Value* we created earlier
-         assert(lastPhi != 0 && "Last phi is NULL!\n");
-         MachineInstr *saveValue = BuildMI(*machineBB[0], machineBB[0]->begin(), V9::PHI, 3).addReg(lastPhi).addReg(I->second).addRegDef(finalPHIValue[V->first]);
-         DEBUG(std::cerr << "Resulting PHI: " << *saveValue << "\n");
-         kernelPHIs[V->first][I->first] = finalPHIValue[V->first];
-       }
-
-       ++count;
-     }
-
-   }
- }
-}
-
-
-void ModuloSchedulingSBPass::removePHIs(std::vector<const MachineBasicBlock*> &SB, std::vector<std::vector<MachineBasicBlock*> > &prologues, std::vector<std::vector<MachineBasicBlock*> > &epilogues, std::vector<MachineBasicBlock*> &kernelBB, std::map<Value*, MachineBasicBlock*> &newValLocation) {
-
-  //Worklist to delete things
-  std::vector<std::pair<MachineBasicBlock*, MachineBasicBlock::iterator> > worklist;
-
-  //Worklist of TmpInstructions that need to be added to a MCFI
-  std::vector<Instruction*> addToMCFI;
-
-  //Worklist to add OR instructions to end of kernel so not to invalidate the iterator
-  //std::vector<std::pair<Instruction*, Value*> > newORs;
-
-  const TargetInstrInfo *TMI = target.getInstrInfo();
-
-  //Start with the kernel and for each phi insert a copy for the phi
-  //def and for each arg
-  //phis are only in the first BB in the kernel
-  for(MachineBasicBlock::iterator I = kernelBB[0]->begin(), E = kernelBB[0]->end();
-      I != E; ++I) {
-
-    DEBUG(std::cerr << "Looking at Instr: " << *I << "\n");
-
-    //Get op code and check if its a phi
-    if(I->getOpcode() == V9::PHI) {
-
-      DEBUG(std::cerr << "Replacing PHI: " << *I << "\n");
-      Instruction *tmp = 0;
-
-      for(unsigned i = 0; i < I->getNumOperands(); ++i) {
-
-        //Get Operand
-        const MachineOperand &mOp = I->getOperand(i);
-        assert(mOp.getType() == MachineOperand::MO_VirtualRegister
-               && "Should be a Value*\n");
-
-        if(!tmp) {
-          tmp = new TmpInstruction(mOp.getVRegValue());
-          addToMCFI.push_back(tmp);
-        }
-
-        //Now for all our arguments we read, OR to the new
-        //TmpInstruction that we created
-        if(mOp.isUse()) {
-          DEBUG(std::cerr << "Use: " << mOp << "\n");
-          //Place a copy at the end of its BB but before the branches
-          assert(newValLocation.count(mOp.getVRegValue()) && "We must know where this value is located\n");
-          //Reverse iterate to find the branches, we can safely assume no instructions have been
-          //put in the nop positions
-          for(MachineBasicBlock::iterator inst = --(newValLocation[mOp.getVRegValue()])->end(), endBB = (newValLocation[mOp.getVRegValue()])->begin(); inst != endBB; --inst) {
-            MachineOpCode opc = inst->getOpcode();
-            if(TMI->isBranch(opc) || TMI->isNop(opc))
-              continue;
-            else {
-              if(mOp.getVRegValue()->getType() == Type::FloatTy)
-                BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::FMOVS, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-              else if(mOp.getVRegValue()->getType() == Type::DoubleTy)
-                BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::FMOVD, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-              else
-                BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::ORr, 3).addReg(mOp.getVRegValue()).addImm(0).addRegDef(tmp);
-
-              break;
-            }
-
-          }
-
-        }
-        else {
-          //Remove the phi and replace it with an OR
-          DEBUG(std::cerr << "Def: " << mOp << "\n");
-          //newORs.push_back(std::make_pair(tmp, mOp.getVRegValue()));
-          if(tmp->getType() == Type::FloatTy)
-            BuildMI(*kernelBB[0], I, V9::FMOVS, 3).addReg(tmp).addRegDef(mOp.getVRegValue());
-          else if(tmp->getType() == Type::DoubleTy)
-            BuildMI(*kernelBB[0], I, V9::FMOVD, 3).addReg(tmp).addRegDef(mOp.getVRegValue());
-          else
-            BuildMI(*kernelBB[0], I, V9::ORr, 3).addReg(tmp).addImm(0).addRegDef(mOp.getVRegValue());
-
-
-          worklist.push_back(std::make_pair(kernelBB[0], I));
-        }
-
-      }
-
-    }
-
-
-  }
-
-  //Add TmpInstructions to some MCFI
-  if(addToMCFI.size() > 0) {
-    MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-    for(unsigned x = 0; x < addToMCFI.size(); ++x) {
-      tempMvec.addTemp(addToMCFI[x]);
-    }
-    addToMCFI.clear();
-  }
-
-
-  //Remove phis from epilogue
-  for(std::vector<std::vector<MachineBasicBlock*> >::iterator MB = epilogues.begin(),
-        ME = epilogues.end(); MB != ME; ++MB) {
-
-    for(std::vector<MachineBasicBlock*>::iterator currentMBB = MB->begin(), currentME = MB->end(); currentMBB != currentME; ++currentMBB) {
-
-      for(MachineBasicBlock::iterator I = (*currentMBB)->begin(),
-            E = (*currentMBB)->end(); I != E; ++I) {
-
-        DEBUG(std::cerr << "Looking at Instr: " << *I << "\n");
-        //Get op code and check if its a phi
-        if(I->getOpcode() == V9::PHI) {
-          Instruction *tmp = 0;
-
-          for(unsigned i = 0; i < I->getNumOperands(); ++i) {
-            //Get Operand
-            const MachineOperand &mOp = I->getOperand(i);
-            assert(mOp.getType() == MachineOperand::MO_VirtualRegister && "Should be a Value*\n");
-
-            if(!tmp) {
-              tmp = new TmpInstruction(mOp.getVRegValue());
-              addToMCFI.push_back(tmp);
-            }
-
-            //Now for all our arguments we read, OR to the new TmpInstruction that we created
-            if(mOp.isUse()) {
-              DEBUG(std::cerr << "Use: " << mOp << "\n");
-              //Place a copy at the end of its BB but before the branches
-              assert(newValLocation.count(mOp.getVRegValue()) && "We must know where this value is located\n");
-              //Reverse iterate to find the branches, we can safely assume no instructions have been
-              //put in the nop positions
-              for(MachineBasicBlock::iterator inst = --(newValLocation[mOp.getVRegValue()])->end(), endBB = (newValLocation[mOp.getVRegValue()])->begin(); inst != endBB; --inst) {
-                MachineOpCode opc = inst->getOpcode();
-                if(TMI->isBranch(opc) || TMI->isNop(opc))
-                  continue;
-                else {
-                  if(mOp.getVRegValue()->getType() == Type::FloatTy)
-                    BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::FMOVS, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-                  else if(mOp.getVRegValue()->getType() == Type::DoubleTy)
-                    BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::FMOVD, 3).addReg(mOp.getVRegValue()).addRegDef(tmp);
-                  else
-                    BuildMI(*(newValLocation[mOp.getVRegValue()]), ++inst, V9::ORr, 3).addReg(mOp.getVRegValue()).addImm(0).addRegDef(tmp);
-
-
-                  break;
-                }
-
-              }
-
-            }
-            else {
-              //Remove the phi and replace it with an OR
-              DEBUG(std::cerr << "Def: " << mOp << "\n");
-              if(tmp->getType() == Type::FloatTy)
-                BuildMI(**currentMBB, I, V9::FMOVS, 3).addReg(tmp).addRegDef(mOp.getVRegValue());
-              else if(tmp->getType() == Type::DoubleTy)
-                BuildMI(**currentMBB, I, V9::FMOVD, 3).addReg(tmp).addRegDef(mOp.getVRegValue());
-              else
-                BuildMI(**currentMBB, I, V9::ORr, 3).addReg(tmp).addImm(0).addRegDef(mOp.getVRegValue());
-
-              worklist.push_back(std::make_pair(*currentMBB,I));
-            }
-          }
-        }
-      }
-    }
-  }
-
-
-  if(addToMCFI.size() > 0) {
-    MachineCodeForInstruction & tempMvec = MachineCodeForInstruction::get(defaultInst);
-    for(unsigned x = 0; x < addToMCFI.size(); ++x) {
-      tempMvec.addTemp(addToMCFI[x]);
-    }
-    addToMCFI.clear();
-  }
-
-    //Delete the phis
-  for(std::vector<std::pair<MachineBasicBlock*, MachineBasicBlock::iterator> >::iterator I =  worklist.begin(), E = worklist.end(); I != E; ++I) {
-    DEBUG(std::cerr << "Deleting PHI " << *I->second << "\n");
-    I->first->erase(I->second);
-
-  }
-
-
-  assert((addToMCFI.size() == 0) && "We should have added all TmpInstructions to some MachineCodeForInstruction");
-}
-
-
-
-
-void ModuloSchedulingSBPass::writeSideExits(std::vector<std::vector<MachineBasicBlock *> > &prologues, std::vector<std::vector<BasicBlock*> > &llvm_prologues, std::vector<std::vector<MachineBasicBlock *> > &epilogues, std::vector<std::vector<BasicBlock*> > &llvm_epilogues, std::map<const MachineBasicBlock*, Value*> &sideExits, std::map<MachineBasicBlock*, std::vector<std::pair<MachineInstr*, int> > > &instrsMovedDown, std::vector<const MachineBasicBlock*> &SB, std::vector<MachineBasicBlock*> &kernelMBBs, std::map<MachineBasicBlock*, int> branchStage) {
-
-  const TargetInstrInfo *TMI = target.getInstrInfo();
-
-  //Repeat for each side exit
-  for(unsigned sideExitNum = 0; sideExitNum < SB.size()-1; ++sideExitNum) {
-
-    std::vector<std::vector<BasicBlock*> > side_llvm_epilogues;
-    std::vector<std::vector<MachineBasicBlock*> > side_epilogues;
-    MachineBasicBlock* sideMBB;
-    BasicBlock* sideBB;
-
-    //Create side exit blocks
-    //Get the LLVM basic block
-    BasicBlock *bb = (BasicBlock*) SB[sideExitNum]->getBasicBlock();
-    MachineBasicBlock *mbb = (MachineBasicBlock*) SB[sideExitNum];
-
-    int stage = branchStage[mbb];
-
-    //Create new basic blocks for our side exit instructios that were moved below the branch
-    sideBB = new BasicBlock("SideExit", (Function*) bb->getParent());
-    sideMBB = new MachineBasicBlock(sideBB);
-    (((MachineBasicBlock*)SB[0])->getParent())->getBasicBlockList().push_back(sideMBB);
-
-
-    if(instrsMovedDown.count(mbb)) {
-      for(std::vector<std::pair<MachineInstr*, int> >::iterator I = instrsMovedDown[mbb].begin(), E = instrsMovedDown[mbb].end(); I != E; ++I) {
-        if(branchStage[mbb] == I->second)
-          sideMBB->push_back((I->first)->clone());
-      }
-
-      //Add unconditional branches to original exits
-      BuildMI(sideMBB, V9::BA, 1).addPCDisp(sideExits[mbb]);
-      BuildMI(sideMBB, V9::NOP, 0);
-
-      //Add unconditioal branch to llvm BB
-      BasicBlock *extBB = dyn_cast<BasicBlock>(sideExits[mbb]);
-      assert(extBB && "Side exit basicblock can not be null");
-      TerminatorInst *newBranch = new BranchInst(extBB, sideBB);
-    }
-
-     //Clone epilogues and update their branches, one cloned epilogue set per side exit
-    //only clone epilogues that are from a greater stage!
-    for(unsigned i = 0; i < epilogues.size()-stage; ++i) {
-      std::vector<MachineBasicBlock*> MB = epilogues[i];
-
-      std::vector<MachineBasicBlock*> newEp;
-      std::vector<BasicBlock*> newLLVMEp;
-
-      for(std::vector<MachineBasicBlock*>::iterator currentMBB = MB.begin(),
-            lastMBB = MB.end(); currentMBB != lastMBB; ++currentMBB) {
-        BasicBlock *tmpBB = new BasicBlock("SideEpilogue", (Function*) (*currentMBB)->getBasicBlock()->getParent());
-        MachineBasicBlock *tmp = new MachineBasicBlock(tmpBB);
-
-        //Clone instructions and insert into new MBB
-        for(MachineBasicBlock::iterator I = (*currentMBB)->begin(),
-              E = (*currentMBB)->end(); I != E; ++I) {
-
-          MachineInstr *clone = I->clone();
-          if(clone->getOpcode() == V9::BA && (currentMBB+1 == lastMBB)) {
-            //update branch to side exit
-            for(unsigned i = 0; i < clone->getNumOperands(); ++i) {
-              MachineOperand &mOp = clone->getOperand(i);
-              if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-                mOp.setValueReg(sideBB);
-              }
-            }
-          }
-
-          tmp->push_back(clone);
-
-        }
-
-        //Add llvm branch
-        TerminatorInst *newBranch = new BranchInst(sideBB, tmpBB);
-
-        newEp.push_back(tmp);
-        (((MachineBasicBlock*)SB[0])->getParent())->getBasicBlockList().push_back(tmp);
-
-        newLLVMEp.push_back(tmpBB);
-
-      }
-      side_llvm_epilogues.push_back(newLLVMEp);
-      side_epilogues.push_back(newEp);
-    }
-
-  //Now stich up all the branches
-
-  //Loop over prologues, and if its an inner branch and branches to our original side exit
-  //then have it branch to the appropriate epilogue first (if it exists)
-    for(unsigned P = 0; P < prologues.size(); ++P) {
-
-      //Get BB side exit we are dealing with
-      MachineBasicBlock *currentMBB = prologues[P][sideExitNum];
-      if(P >= (unsigned) stage) {
-        //Iterate backwards of machine instructions to find the branch we need to update
-        for(MachineBasicBlock::reverse_iterator mInst = currentMBB->rbegin(), mInstEnd = currentMBB->rend(); mInst != mInstEnd; ++mInst) {
-          MachineOpCode OC = mInst->getOpcode();
-
-          //If its a branch update its branchto
-          if(TMI->isBranch(OC)) {
-            for(unsigned opNum = 0; opNum < mInst->getNumOperands(); ++opNum) {
-              MachineOperand &mOp = mInst->getOperand(opNum);
-              if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-                //Check if we branch to side exit
-                if(mOp.getVRegValue() == sideExits[mbb]) {
-                  mOp.setValueReg(side_llvm_epilogues[P][0]);
-                }
-              }
-            }
-            DEBUG(std::cerr << "New Prologue Branch: " << *mInst << "\n");
-          }
-        }
-
-        //Update llvm branch
-        TerminatorInst *branchVal = ((BasicBlock*) currentMBB->getBasicBlock())->getTerminator();
-        DEBUG(std::cerr << *branchVal << "\n");
-
-        for(unsigned i=0; i < branchVal->getNumSuccessors(); ++i) {
-          if(branchVal->getSuccessor(i) == sideExits[mbb]) {
-            DEBUG(std::cerr << "Replacing successor bb\n");
-            branchVal->setSuccessor(i, side_llvm_epilogues[P][0]);
-          }
-        }
-      }
-      else {
-        //must add BA branch because another prologue or kernel has the actual side exit branch
-         //Add unconditional branches to original exits
-        assert( (sideExitNum+1) < prologues[P].size() && "must have valid prologue to branch to");
-        BuildMI(prologues[P][sideExitNum], V9::BA, 1).addPCDisp((BasicBlock*)(prologues[P][sideExitNum+1])->getBasicBlock());
-        BuildMI(prologues[P][sideExitNum], V9::NOP, 0);
-
-        TerminatorInst *newBranch = new BranchInst((BasicBlock*) (prologues[P][sideExitNum+1])->getBasicBlock(), (BasicBlock*) (prologues[P][sideExitNum])->getBasicBlock());
-
-      }
-    }
-
-
-    //Update side exits in kernel
-    MachineBasicBlock *currentMBB = kernelMBBs[sideExitNum];
-    //Iterate backwards of machine instructions to find the branch we need to update
-    for(MachineBasicBlock::reverse_iterator mInst = currentMBB->rbegin(), mInstEnd = currentMBB->rend(); mInst != mInstEnd; ++mInst) {
-      MachineOpCode OC = mInst->getOpcode();
-
-      //If its a branch update its branchto
-      if(TMI->isBranch(OC)) {
-        for(unsigned opNum = 0; opNum < mInst->getNumOperands(); ++opNum) {
-          MachineOperand &mOp = mInst->getOperand(opNum);
-          if (mOp.getType() == MachineOperand::MO_PCRelativeDisp) {
-            //Check if we branch to side exit
-            if(mOp.getVRegValue() == sideExits[mbb]) {
-              if(side_llvm_epilogues.size() > 0)
-                mOp.setValueReg(side_llvm_epilogues[0][0]);
-              else
-                mOp.setValueReg(sideBB);
-            }
-          }
-        }
-        DEBUG(std::cerr << "New Prologue Branch: " << *mInst << "\n");
-      }
-    }
-
-    //Update llvm branch
-    //Update llvm branch
-    TerminatorInst *branchVal = ((BasicBlock*)currentMBB->getBasicBlock())->getTerminator();
-    DEBUG(std::cerr << *branchVal << "\n");
-
-    for(unsigned i=0; i < branchVal->getNumSuccessors(); ++i) {
-      if(branchVal->getSuccessor(i) == sideExits[mbb]) {
-        DEBUG(std::cerr << "Replacing successor bb\n");
-        if(side_llvm_epilogues.size() > 0)
-          branchVal->setSuccessor(i, side_llvm_epilogues[0][0]);
-        else
-          branchVal->setSuccessor(i, sideBB);
-      }
-    }
-  }
-}
-
diff --git a/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloSchedulingSuperBlock.h b/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloSchedulingSuperBlock.h
deleted file mode 100644
index 669611e9261..00000000000
--- a/llvm/lib/Target/SparcV9/ModuloScheduling/ModuloSchedulingSuperBlock.h
+++ /dev/null
@@ -1,192 +0,0 @@
-//===-- ModuloSchedulingSuperBlock.h -Swing Modulo Scheduling-----*- C++ -*-===//
-//
-//                     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.
-//
-//===----------------------------------------------------------------------===//
-//Swing Modulo Scheduling done on Superblocks ( entry, multiple exit,
-//multiple basic block loops).
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_MODULOSCHEDULINGSB_H
-#define LLVM_MODULOSCHEDULINGSB_H
-
-#include "llvm/Analysis/LoopInfo.h"
-#include "llvm/Analysis/ScalarEvolution.h"
-#include "llvm/Function.h"
-#include "llvm/Pass.h"
-#include "llvm/CodeGen/MachineBasicBlock.h"
-#include "MSScheduleSB.h"
-#include "MSchedGraphSB.h"
-
-
-namespace llvm {
-
-  //Struct to contain ModuloScheduling Specific Information for each node
-  struct MSNodeSBAttributes {
-    int ASAP; //Earliest time at which the opreation can be scheduled
-    int ALAP; //Latest time at which the operation can be scheduled.
-    int MOB;
-    int depth;
-    int height;
-    MSNodeSBAttributes(int asap=-1, int alap=-1, int mob=-1,
-                             int d=-1, int h=-1) : ASAP(asap), ALAP(alap),
-                                                   MOB(mob), depth(d),
-                                                   height(h) {}
-  };
-
-
-  typedef std::vector<const MachineBasicBlock*> SuperBlock;
-
-  class ModuloSchedulingSBPass : public FunctionPass {
-    const TargetMachine &target;
-
-    //Map to hold Value* defs
-    std::map<const Value*, MachineInstr*> defMap;
-
-    //Map to hold list of instructions associate to the induction var for each BB
-    std::map<SuperBlock, std::map<const MachineInstr*, unsigned> > indVarInstrs;
-
-    //Map to hold machine to  llvm instrs for each valid BB
-    std::map<SuperBlock, std::map<MachineInstr*, Instruction*> > machineTollvm;
-
-    //LLVM Instruction we know we can add TmpInstructions to its MCFI
-    Instruction *defaultInst;
-
-    //Map that holds node to node attribute information
-    std::map<MSchedGraphSBNode*, MSNodeSBAttributes> nodeToAttributesMap;
-
-    //Map to hold all reccurrences
-    std::set<std::pair<int, std::vector<MSchedGraphSBNode*> > > recurrenceList;
-
-    //Set of edges to ignore, stored as src node and index into vector of successors
-    std::set<std::pair<MSchedGraphSBNode*, unsigned> > edgesToIgnore;
-
-    //Vector containing the partial order
-    std::vector<std::set<MSchedGraphSBNode*> > partialOrder;
-
-    //Vector containing the final node order
-    std::vector<MSchedGraphSBNode*> FinalNodeOrder;
-
-    //Schedule table, key is the cycle number and the vector is resource, node pairs
-    MSScheduleSB schedule;
-
-    //Current initiation interval
-    int II;
-
-    //Internal Functions
-    void FindSuperBlocks(Function &F, LoopInfo &LI,
-                         std::vector<std::vector<const MachineBasicBlock*> > &Worklist);
-    bool MachineBBisValid(const MachineBasicBlock *B,
-                          std::map<const MachineInstr*, unsigned> &indexMap,
-                          unsigned &offset);
-    bool CreateDefMap(std::vector<const MachineBasicBlock*> &SB);
-    bool getIndVar(std::vector<const MachineBasicBlock*> &superBlock,
-                   std::map<BasicBlock*, MachineBasicBlock*> &bbMap,
-                   std::map<const MachineInstr*, unsigned> &indexMap);
-    bool assocIndVar(Instruction *I, std::set<Instruction*> &indVar,
-                     std::vector<Instruction*> &stack,
-                     std::map<BasicBlock*, MachineBasicBlock*> &bbMap,
-                     const BasicBlock *first,
-                     std::set<const BasicBlock*> &llvmSuperBlock);
-    int calculateResMII(std::vector<const MachineBasicBlock*> &superBlock);
-    int calculateRecMII(MSchedGraphSB *graph, int MII);
-    void findAllCircuits(MSchedGraphSB *g, int II);
-    void addRecc(std::vector<MSchedGraphSBNode*> &stack,
-                 std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> &newNodes);
-    bool circuit(MSchedGraphSBNode *v, std::vector<MSchedGraphSBNode*> &stack,
-                 std::set<MSchedGraphSBNode*> &blocked, std::vector<MSchedGraphSBNode*> &SCC,
-                 MSchedGraphSBNode *s, std::map<MSchedGraphSBNode*,
-                 std::set<MSchedGraphSBNode*> > &B,
-                 int II, std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> &newNodes);
-    void unblock(MSchedGraphSBNode *u, std::set<MSchedGraphSBNode*> &blocked,
-                 std::map<MSchedGraphSBNode*, std::set<MSchedGraphSBNode*> > &B);
-    void addSCC(std::vector<MSchedGraphSBNode*> &SCC, std::map<MSchedGraphSBNode*, MSchedGraphSBNode*> &newNodes);
-    void calculateNodeAttributes(MSchedGraphSB *graph, int MII);
-    bool ignoreEdge(MSchedGraphSBNode *srcNode, MSchedGraphSBNode *destNode);
-    int  calculateASAP(MSchedGraphSBNode *node, int MII, MSchedGraphSBNode *destNode);
-    int calculateALAP(MSchedGraphSBNode *node, int MII,
-                      int maxASAP, MSchedGraphSBNode *srcNode);
-    int findMaxASAP();
-    int calculateHeight(MSchedGraphSBNode *node,MSchedGraphSBNode *srcNode);
-    int calculateDepth(MSchedGraphSBNode *node, MSchedGraphSBNode *destNode);
-    void computePartialOrder();
-    void connectedComponentSet(MSchedGraphSBNode *node, std::set<MSchedGraphSBNode*> &ccSet,
-                               std::set<MSchedGraphSBNode*> &lastNodes);
-    void searchPath(MSchedGraphSBNode *node,
-                    std::vector<MSchedGraphSBNode*> &path,
-                    std::set<MSchedGraphSBNode*> &nodesToAdd,
-                    std::set<MSchedGraphSBNode*> &new_reccurrence);
-    void orderNodes();
-    bool computeSchedule(std::vector<const MachineBasicBlock*> &BB, MSchedGraphSB *MSG);
-    bool scheduleNode(MSchedGraphSBNode *node, int start, int end);
-      void predIntersect(std::set<MSchedGraphSBNode*> &CurrentSet, std::set<MSchedGraphSBNode*> &IntersectResult);
-    void succIntersect(std::set<MSchedGraphSBNode*> &CurrentSet, std::set<MSchedGraphSBNode*> &IntersectResult);
-    void reconstructLoop(std::vector<const MachineBasicBlock*> &SB);
-    void fixBranches(std::vector<std::vector<MachineBasicBlock*> > &prologues,
-                     std::vector<std::vector<BasicBlock*> > &llvm_prologues,
-                     std::vector<MachineBasicBlock*> &machineKernelBB,
-                     std::vector<BasicBlock*> &llvmKernelBB,
-                     std::vector<std::vector<MachineBasicBlock*> > &epilogues,
-                     std::vector<std::vector<BasicBlock*> > &llvm_epilogues,
-                     std::vector<const MachineBasicBlock*> &SB,
-                     std::map<const MachineBasicBlock*, Value*> &sideExits);
-
-    void writePrologues(std::vector<std::vector<MachineBasicBlock *> > &prologues,
-                        std::vector<const MachineBasicBlock*> &origBB,
-                        std::vector<std::vector<BasicBlock*> > &llvm_prologues,
-                        std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave,
-                        std::map<Value*, std::map<int, Value*> > &newValues,
-                        std::map<Value*, MachineBasicBlock*> &newValLocation);
-
-    void writeKernel(std::vector<BasicBlock*> &llvmBB, std::vector<MachineBasicBlock*> &machineBB,
-                     std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave,
-                     std::map<Value*, std::map<int, Value*> > &newValues,
-                     std::map<Value*, MachineBasicBlock*> &newValLocation,
-                     std::map<Value*, std::map<int, Value*> > &kernelPHIs);
-
-    void removePHIs(std::vector<const MachineBasicBlock*> &SB,
-                    std::vector<std::vector<MachineBasicBlock*> > &prologues,
-                    std::vector<std::vector<MachineBasicBlock*> > &epilogues,
-                    std::vector<MachineBasicBlock*> &kernelBB,
-                    std::map<Value*, MachineBasicBlock*> &newValLocation);
-
-    void writeEpilogues(std::vector<std::vector<MachineBasicBlock*> > &epilogues,
-                        std::vector<const MachineBasicBlock*> &origSB,
-                        std::vector<std::vector<BasicBlock*> > &llvm_epilogues,
-                        std::map<const Value*, std::pair<const MachineInstr*, int> > &valuesToSave,
-                        std::map<Value*, std::map<int, Value*> > &newValues,
-                        std::map<Value*, MachineBasicBlock*> &newValLocation,
-                        std::map<Value*, std::map<int, Value*> > &kernelPHIs);
-
-    void writeSideExits(std::vector<std::vector<MachineBasicBlock *> > &prologues,
-                        std::vector<std::vector<BasicBlock*> > &llvm_prologues,
-                        std::vector<std::vector<MachineBasicBlock *> > &epilogues,
-                        std::vector<std::vector<BasicBlock*> > &llvm_epilogues,
-                        std::map<const MachineBasicBlock*, Value*> &sideExits,
-                        std::map<MachineBasicBlock*, std::vector<std::pair<MachineInstr*, int> > > &instrsMovedDown,
-                        std::vector<const MachineBasicBlock*> &SB,
-                        std::vector<MachineBasicBlock*> &kernelMBBs,
-                          std::map<MachineBasicBlock*, int> branchStage);
-
- public:
-    ModuloSchedulingSBPass(TargetMachine &targ) : target(targ) {}
-      virtual bool runOnFunction(Function &F);
-      virtual const char* getPassName() const { return "ModuloScheduling-SuperBlock"; }
-
-
-      // getAnalysisUsage
-      virtual void getAnalysisUsage(AnalysisUsage &AU) const {
-        /// HACK: We don't actually need scev, but we have
-        /// to say we do so that the pass manager does not delete it
-        /// before we run.
-        AU.addRequired<LoopInfo>();
-        AU.addRequired<ScalarEvolution>();
-        AU.addRequired<DependenceAnalyzer>();
-      }
-  };
-}
-#endif