/* $Id: CoinFactorization3.cpp 1373 2011-01-03 23:57:44Z lou $ */
// Copyright (C) 2002, International Business Machines
// Corporation and others.  All Rights Reserved.
// This code is licensed under the terms of the Eclipse Public License (EPL).

#if defined(_MSC_VER)
// Turn off compiler warning about long names
#  pragma warning(disable:4786)
#endif

#include "CoinUtilsConfig.h"

#include <cassert>
#include <cstdio>

#include "CoinFactorization.hpp"
#include "CoinIndexedVector.hpp"
#include "CoinHelperFunctions.hpp"
#include <stdio.h>
#include <iostream>
#if DENSE_CODE==1 
// using simple lapack interface
extern "C" 
{
  /** LAPACK Fortran subroutine DGETRS. */
  void F77_FUNC(dgetrs,DGETRS)(char *trans, cipfint *n,
                               cipfint *nrhs, const double *A, cipfint *ldA,
                               cipfint * ipiv, double *B, cipfint *ldB, ipfint *info,
			       int trans_len);
}
#endif
// For semi-sparse
#define BITS_PER_CHECK 8
#define CHECK_SHIFT 3
typedef unsigned char CoinCheckZero;

//:class CoinFactorization.  Deals with Factorization and Updates

/* Updates one column (FTRAN) from region2 and permutes.
   region1 starts as zero
   Note - if regionSparse2 packed on input - will be packed on output
   - returns un-permuted result in region2 and region1 is zero */
int CoinFactorization::updateColumn ( CoinIndexedVector * regionSparse,
				      CoinIndexedVector * regionSparse2,
				      bool noPermute) 
  const
{
  //permute and move indices into index array
  int * COIN_RESTRICT regionIndex = regionSparse->getIndices (  );
  int numberNonZero;
  const int *permute = permute_.array();
  double * COIN_RESTRICT region = regionSparse->denseVector();

#ifndef CLP_FACTORIZATION
  if (!noPermute) {
#endif
    numberNonZero = regionSparse2->getNumElements();
    int * COIN_RESTRICT index = regionSparse2->getIndices();
    double * COIN_RESTRICT array = regionSparse2->denseVector();
#ifndef CLP_FACTORIZATION
    bool packed = regionSparse2->packedMode();
    if (packed) {
      for (int j = 0; j < numberNonZero; j ++ ) {
	int iRow = index[j];
	double value = array[j];
	array[j]=0.0;
	iRow = permute[iRow];
	region[iRow] = value;
	regionIndex[j] = iRow;
      }
    } else {
#else
      assert (!regionSparse2->packedMode());
#endif
      for (int j = 0; j < numberNonZero; j ++ ) {
	int iRow = index[j];
	double value = array[iRow];
	array[iRow]=0.0;
	iRow = permute[iRow];
	region[iRow] = value;
	regionIndex[j] = iRow;
      }
#ifndef CLP_FACTORIZATION
    }
#endif
    regionSparse->setNumElements ( numberNonZero );
#ifndef CLP_FACTORIZATION
  } else {
    numberNonZero = regionSparse->getNumElements();
  }
#endif
  if (collectStatistics_) {
    numberFtranCounts_++;
    ftranCountInput_ += numberNonZero;
  }
    
  //  ******* L
  updateColumnL ( regionSparse, regionIndex );
  if (collectStatistics_) 
    ftranCountAfterL_ += regionSparse->getNumElements();
  //permute extra
  //row bits here
  updateColumnR ( regionSparse );
  if (collectStatistics_) 
    ftranCountAfterR_ += regionSparse->getNumElements();
  
  //update counts
  //  ******* U
  updateColumnU ( regionSparse, regionIndex);
  if (!doForrestTomlin_) {
    // Do PFI after everything else
    updateColumnPFI(regionSparse);
  }
  if (!noPermute) {
    permuteBack(regionSparse,regionSparse2);
    return regionSparse2->getNumElements (  );
  } else {
    return regionSparse->getNumElements (  );
  }
}
// Permutes back at end of updateColumn
void 
CoinFactorization::permuteBack ( CoinIndexedVector * regionSparse, 
				 CoinIndexedVector * outVector) const
{
  // permute back
  int oldNumber = regionSparse->getNumElements();
  const int * COIN_RESTRICT regionIndex = regionSparse->getIndices (  );
  double * COIN_RESTRICT region = regionSparse->denseVector();
  int * COIN_RESTRICT outIndex = outVector->getIndices (  );
  double * COIN_RESTRICT out = outVector->denseVector();
  const int * COIN_RESTRICT permuteBack = pivotColumnBack();
  int number=0;
  if (outVector->packedMode()) {
    for (int j = 0; j < oldNumber; j ++ ) {
      int iRow = regionIndex[j];
      double value = region[iRow];
      region[iRow]=0.0;
      if (fabs(value)>zeroTolerance_) {
	iRow = permuteBack[iRow];
	outIndex[number]=iRow;
	out[number++] = value;
      }
    }
  } else {
    for (int j = 0; j < oldNumber; j ++ ) {
      int iRow = regionIndex[j];
      double value = region[iRow];
      region[iRow]=0.0;
      if (fabs(value)>zeroTolerance_) {
	iRow = permuteBack[iRow];
	outIndex[number++]=iRow;
	out[iRow] = value;
      }
    }
  }
  outVector->setNumElements(number);
  regionSparse->setNumElements(0);
}
//  updateColumnL.  Updates part of column (FTRANL)
void
CoinFactorization::updateColumnL ( CoinIndexedVector * regionSparse,
					   int * COIN_RESTRICT regionIndex) const
{
  if (numberL_) {
    int number = regionSparse->getNumElements (  );
    int goSparse;
    // Guess at number at end
    if (sparseThreshold_>0) {
      if (ftranAverageAfterL_) {
	int newNumber = static_cast<int> (number*ftranAverageAfterL_);
	if (newNumber< sparseThreshold_&&(numberL_<<2)>newNumber)
	  goSparse = 2;
	else if (newNumber< sparseThreshold2_&&(numberL_<<1)>newNumber)
	  goSparse = 1;
	else
	  goSparse = 0;
      } else {
	if (number<sparseThreshold_&&(numberL_<<2)>number) 
	  goSparse = 2;
	else
	  goSparse = 0;
      }
    } else {
      goSparse=0;
    }
    switch (goSparse) {
    case 0: // densish
      updateColumnLDensish(regionSparse,regionIndex);
      break;
    case 1: // middling
      updateColumnLSparsish(regionSparse,regionIndex);
      break;
    case 2: // sparse
      updateColumnLSparse(regionSparse,regionIndex);
      break;
    }
  }
#ifdef DENSE_CODE
  if (numberDense_) {
    //take off list
    int lastSparse = numberRows_-numberDense_;
    int number = regionSparse->getNumElements();
    double * COIN_RESTRICT region = regionSparse->denseVector (  );
    int i=0;
    bool doDense=false;
    while (i<number) {
      int iRow = regionIndex[i];
      if (iRow>=lastSparse) {
	doDense=true;
	regionIndex[i] = regionIndex[--number];
      } else {
	i++;
      }
    }
    if (doDense) {
      char trans = 'N';
      int ione=1;
      int info;
      F77_FUNC(dgetrs,DGETRS)(&trans,&numberDense_,&ione,denseArea_,&numberDense_,
			      densePermute_,region+lastSparse,&numberDense_,&info,1);
      for (int i=lastSparse;i<numberRows_;i++) {
	double value = region[i];
	if (value) {
	  if (fabs(value)>=1.0e-15) 
	    regionIndex[number++] = i;
	  else
	    region[i]=0.0;
	}
      }
      regionSparse->setNumElements(number);
    }
  }
#endif
}
// Updates part of column (FTRANL) when densish
void 
CoinFactorization::updateColumnLDensish ( CoinIndexedVector * regionSparse ,
					  int * COIN_RESTRICT regionIndex)
  const
{
  double * COIN_RESTRICT region = regionSparse->denseVector (  );
  int number = regionSparse->getNumElements (  );
  int numberNonZero;
  double tolerance = zeroTolerance_;
  
  numberNonZero = 0;
  
  const CoinBigIndex * COIN_RESTRICT startColumn = startColumnL_.array();
  const int * COIN_RESTRICT indexRow = indexRowL_.array();
  const CoinFactorizationDouble * COIN_RESTRICT element = elementL_.array();
  int last = numberRows_;
  assert ( last == baseL_ + numberL_);
#if DENSE_CODE==1
  //can take out last bit of sparse L as empty
  last -= numberDense_;
#endif
  int smallestIndex = numberRowsExtra_;
  // do easy ones
  for (int k=0;k<number;k++) {
    int iPivot=regionIndex[k];
    if (iPivot>=baseL_) 
      smallestIndex = CoinMin(iPivot,smallestIndex);
    else
      regionIndex[numberNonZero++]=iPivot;
  }
  // now others
  for (int i = smallestIndex; i < last; i++ ) {
    CoinFactorizationDouble pivotValue = region[i];
    
    if ( fabs(pivotValue) > tolerance ) {
      CoinBigIndex start = startColumn[i];
      CoinBigIndex end = startColumn[i + 1];
      for (CoinBigIndex j = start; j < end; j ++ ) {
	int iRow = indexRow[j];
	CoinFactorizationDouble result = region[iRow];
	CoinFactorizationDouble value = element[j];

	region[iRow] = result - value * pivotValue;
      }     
      regionIndex[numberNonZero++] = i;
    } else {
      region[i] = 0.0;
    }       
  }     
  // and dense
  for (int i=last ; i < numberRows_; i++ ) {
    CoinFactorizationDouble pivotValue = region[i];
    if ( fabs(pivotValue) > tolerance ) {
      regionIndex[numberNonZero++] = i;
    } else {
      region[i] = 0.0;
    }       
  }     
  regionSparse->setNumElements ( numberNonZero );
} 
// Updates part of column (FTRANL) when sparsish
void 
CoinFactorization::updateColumnLSparsish ( CoinIndexedVector * regionSparse,
					   int * COIN_RESTRICT regionIndex)
  const
{
  double * COIN_RESTRICT region = regionSparse->denseVector (  );
  int number = regionSparse->getNumElements (  );
  int numberNonZero;
  double tolerance = zeroTolerance_;
  
  numberNonZero = 0;
  
  const CoinBigIndex *startColumn = startColumnL_.array();
  const int *indexRow = indexRowL_.array();
  const CoinFactorizationDouble *element = elementL_.array();
  int last = numberRows_;
  assert ( last == baseL_ + numberL_);
#if DENSE_CODE==1
  //can take out last bit of sparse L as empty
  last -= numberDense_;
#endif
  // mark known to be zero
  int nInBig = sizeof(CoinBigIndex)/sizeof(int);
  CoinCheckZero * COIN_RESTRICT mark = reinterpret_cast<CoinCheckZero *> (sparse_.array()+(2+nInBig)*maximumRowsExtra_);
  int smallestIndex = numberRowsExtra_;
  // do easy ones
  for (int k=0;k<number;k++) {
    int iPivot=regionIndex[k];
    if (iPivot<baseL_) { 
      regionIndex[numberNonZero++]=iPivot;
    } else {
      smallestIndex = CoinMin(iPivot,smallestIndex);
      int iWord = iPivot>>CHECK_SHIFT;
      int iBit = iPivot-(iWord<<CHECK_SHIFT);
      if (mark[iWord]) {
	mark[iWord] = static_cast<CoinCheckZero>(mark[iWord] | (1<<iBit));
      } else {
	mark[iWord] = static_cast<CoinCheckZero>(1<<iBit);
      }
    }
  }
  // now others
  // First do up to convenient power of 2
  int jLast = (smallestIndex+BITS_PER_CHECK-1)>>CHECK_SHIFT;
  jLast = CoinMin((jLast<<CHECK_SHIFT),last);
  int i;
  for ( i = smallestIndex; i < jLast; i++ ) {
    CoinFactorizationDouble pivotValue = region[i];
    CoinBigIndex start = startColumn[i];
    CoinBigIndex end = startColumn[i + 1];
    
    if ( fabs(pivotValue) > tolerance ) {
      for (CoinBigIndex j = start; j < end; j ++ ) {
	int iRow = indexRow[j];
	CoinFactorizationDouble result = region[iRow];
	CoinFactorizationDouble value = element[j];
	region[iRow] = result - value * pivotValue;
	int iWord = iRow>>CHECK_SHIFT;
	int iBit = iRow-(iWord<<CHECK_SHIFT);
	if (mark[iWord]) {
	  mark[iWord] = static_cast<CoinCheckZero>(mark[iWord] | (1<<iBit));
	} else {
	  mark[iWord] = static_cast<CoinCheckZero>(1<<iBit);
	}
      }     
      regionIndex[numberNonZero++] = i;
    } else {
      region[i] = 0.0;
    }       
  }
  
  int kLast = last>>CHECK_SHIFT;
  if (jLast<last) {
    // now do in chunks
    for (int k=(jLast>>CHECK_SHIFT);k<kLast;k++) {
      unsigned int iMark = mark[k];
      if (iMark) {
	// something in chunk - do all (as imark may change)
	i = k<<CHECK_SHIFT;
	int iLast = i+BITS_PER_CHECK;
	for ( ; i < iLast; i++ ) {
	  CoinFactorizationDouble pivotValue = region[i];
	  CoinBigIndex start = startColumn[i];
	  CoinBigIndex end = startColumn[i + 1];
	  
	  if ( fabs(pivotValue) > tolerance ) {
	    CoinBigIndex j;
	    for ( j = start; j < end; j ++ ) {
	      int iRow = indexRow[j];
	      CoinFactorizationDouble result = region[iRow];
	      CoinFactorizationDouble value = element[j];
	      region[iRow] = result - value * pivotValue;
	      int iWord = iRow>>CHECK_SHIFT;
	      int iBit = iRow-(iWord<<CHECK_SHIFT);
	      if (mark[iWord]) {
		mark[iWord] = static_cast<CoinCheckZero>(mark[iWord] | (1<<iBit));
	      } else {
		mark[iWord] = static_cast<CoinCheckZero>(1<<iBit);
	      }
	    }     
	    regionIndex[numberNonZero++] = i;
	  } else {
	    region[i] = 0.0;
	  }       
	}
	mark[k]=0; // zero out marked
      }
    }
    i = kLast<<CHECK_SHIFT;
  }
  for ( ; i < last; i++ ) {
    CoinFactorizationDouble pivotValue = region[i];
    CoinBigIndex start = startColumn[i];
    CoinBigIndex end = startColumn[i + 1];
    
    if ( fabs(pivotValue) > tolerance ) {
      for (CoinBigIndex j = start; j < end; j ++ ) {
	int iRow = indexRow[j];
	CoinFactorizationDouble result = region[iRow];
	CoinFactorizationDouble value = element[j];
	region[iRow] = result - value * pivotValue;
      }     
      regionIndex[numberNonZero++] = i;
    } else {
      region[i] = 0.0;
    }       
  }
  // Now dense part
  for ( ; i < numberRows_; i++ ) {
    double pivotValue = region[i];
    if ( fabs(pivotValue) > tolerance ) {
      regionIndex[numberNonZero++] = i;
    } else {
      region[i] = 0.0;
    }       
  }
  // zero out ones that might have been skipped
  mark[smallestIndex>>CHECK_SHIFT]=0;
  int kkLast = (numberRows_+BITS_PER_CHECK-1)>>CHECK_SHIFT;
  CoinZeroN(mark+kLast,kkLast-kLast);
  regionSparse->setNumElements ( numberNonZero );
}
// Updates part of column (FTRANL) when sparse
void 
CoinFactorization::updateColumnLSparse ( CoinIndexedVector * regionSparse ,
					   int * COIN_RESTRICT regionIndex)
  const
{
  double * COIN_RESTRICT region = regionSparse->denseVector (  );
  int number = regionSparse->getNumElements (  );
  int numberNonZero;
  double tolerance = zeroTolerance_;
  
  numberNonZero = 0;
  
  const CoinBigIndex *startColumn = startColumnL_.array();
  const int *indexRow = indexRowL_.array();
  const CoinFactorizationDouble *element = elementL_.array();
  // use sparse_ as temporary area
  // mark known to be zero
  int * COIN_RESTRICT stack = sparse_.array();  /* pivot */
  int * COIN_RESTRICT list = stack + maximumRowsExtra_;  /* final list */
  CoinBigIndex * COIN_RESTRICT next = reinterpret_cast<CoinBigIndex *> (list + maximumRowsExtra_);  /* jnext */
  char * COIN_RESTRICT mark = reinterpret_cast<char *> (next + maximumRowsExtra_);
  int nList;
#ifdef COIN_DEBUG
  for (int i=0;i<maximumRowsExtra_;i++) {
    assert (!mark[i]);
  }
#endif
  nList=0;
  for (int k=0;k<number;k++) {
    int kPivot=regionIndex[k];
    if (kPivot>=baseL_) {
      assert (kPivot<numberRowsExtra_);
      //if (kPivot>=numberRowsExtra_) abort();
      if(!mark[kPivot]) {
	stack[0]=kPivot;
	CoinBigIndex j=startColumn[kPivot+1]-1;
        int nStack=0;
	while (nStack>=0) {
	  /* take off stack */
	  if (j>=startColumn[kPivot]) {
	    int jPivot=indexRow[j--];
	    assert (jPivot>=baseL_&&jPivot<numberRowsExtra_);
	    //if (jPivot<baseL_||jPivot>=numberRowsExtra_) abort();
	    /* put back on stack */
	    next[nStack] =j;
	    if (!mark[jPivot]) {
	      /* and new one */
	      kPivot=jPivot;
	      j = startColumn[kPivot+1]-1;
	      stack[++nStack]=kPivot;
	      assert (kPivot<numberRowsExtra_);
	      //if (kPivot>=numberRowsExtra_) abort();
	      mark[kPivot]=1;
	      next[nStack]=j;
	    }
	  } else {
	    /* finished so mark */
	    list[nList++]=kPivot;
	    mark[kPivot]=1;
	    --nStack;
	    if (nStack>=0) {
	      kPivot=stack[nStack];
	      assert (kPivot<numberRowsExtra_);
	      j=next[nStack];
	    }
	  }
	}
      }
    } else {
      // just put on list
      regionIndex[numberNonZero++]=kPivot;
    }
  }
  for (int i=nList-1;i>=0;i--) {
    int iPivot = list[i];
    mark[iPivot]=0;
    CoinFactorizationDouble pivotValue = region[iPivot];
    if ( fabs ( pivotValue ) > tolerance ) {
      regionIndex[numberNonZero++]=iPivot;
      for (CoinBigIndex j = startColumn[iPivot]; 
	   j < startColumn[iPivot+1]; j ++ ) {
	int iRow = indexRow[j];
	CoinFactorizationDouble value = element[j];
	region[iRow] -= value * pivotValue;
      }
    } else {
      region[iPivot]=0.0;
    }
  }
  regionSparse->setNumElements ( numberNonZero );
}
/* Updates one column (FTRAN) from region2
   Tries to do FT update
   number returned is negative if no room.
   Also updates region3
   region1 starts as zero and is zero at end */
int 
CoinFactorization::updateTwoColumnsFT ( CoinIndexedVector * regionSparse1,
					CoinIndexedVector * regionSparse2,
					CoinIndexedVector * regionSparse3,
					bool noPermuteRegion3)
{
#if 1
  //#ifdef NDEBUG
  //#undef NDEBUG
  //#endif
  //#define COIN_DEBUG
#ifdef COIN_DEBUG
  regionSparse1->checkClean();
  CoinIndexedVector save2(*regionSparse2);
  CoinIndexedVector save3(*regionSparse3);
#endif
  CoinIndexedVector * regionFT ;
  CoinIndexedVector * regionUpdate ;
  int * COIN_RESTRICT regionIndex ;
  int numberNonZero ;
  const int *permute = permute_.array();
  int * COIN_RESTRICT index ;
  double * COIN_RESTRICT region ;
  if (!noPermuteRegion3) {
    regionFT = regionSparse3;
    regionUpdate = regionSparse1;
    //permute and move indices into index array
    regionIndex = regionUpdate->getIndices (  );
    //int numberNonZero;
    region = regionUpdate->denseVector();
    
    numberNonZero = regionSparse3->getNumElements();
    int * COIN_RESTRICT index = regionSparse3->getIndices();
    double * COIN_RESTRICT array = regionSparse3->denseVector();
    assert (!regionSparse3->packedMode());
    for (int j = 0; j < numberNonZero; j ++ ) {
      int iRow = index[j];
      double value = array[iRow];
      array[iRow]=0.0;
      iRow = permute[iRow];
      region[iRow] = value;
      regionIndex[j] = iRow;
    }
    regionUpdate->setNumElements ( numberNonZero );
  } else {
    regionFT = regionSparse1;
    regionUpdate = regionSparse3;
  }
  //permute and move indices into index array (in U)
  regionIndex = regionFT->getIndices (  );
  numberNonZero = regionSparse2->getNumElements();
  index = regionSparse2->getIndices();
  region = regionFT->denseVector();
  double * COIN_RESTRICT array = regionSparse2->denseVector();
  CoinBigIndex * COIN_RESTRICT startColumnU = startColumnU_.array();
  CoinBigIndex start = startColumnU[maximumColumnsExtra_];
  startColumnU[numberColumnsExtra_] = start;
  regionIndex = indexRowU_.array() + start;

  assert(regionSparse2->packedMode());
  for (int j = 0; j < numberNonZero; j ++ ) {
    int iRow = index[j];
    double value = array[j];
    array[j]=0.0;
    iRow = permute[iRow];
    region[iRow] = value;
    regionIndex[j] = iRow;
  }
  regionFT->setNumElements ( numberNonZero );
  if (collectStatistics_) {
    numberFtranCounts_+=2;
    ftranCountInput_ += regionFT->getNumElements()+
      regionUpdate->getNumElements();
  }
    
  //  ******* L
  updateColumnL ( regionFT, regionIndex );
  updateColumnL ( regionUpdate, regionUpdate->getIndices() );
  if (collectStatistics_) 
    ftranCountAfterL_ += regionFT->getNumElements()+
      regionUpdate->getNumElements();
  //permute extra
  //row bits here
  updateColumnRFT ( regionFT, regionIndex );
  updateColumnR ( regionUpdate );
  if (collectStatistics_) 
    ftranCountAfterR_ += regionFT->getNumElements()+
    regionUpdate->getNumElements();
  //  ******* U - see if densish
  // Guess at number at end
  int goSparse=0;
  if (sparseThreshold_>0) {
    int numberNonZero = (regionUpdate->getNumElements (  )+
			 regionFT->getNumElements())>>1;
    if (ftranAverageAfterR_) {
      int newNumber = static_cast<int> (numberNonZero*ftranAverageAfterU_);
      if (newNumber< sparseThreshold_)
	goSparse = 2;
      else if (newNumber< sparseThreshold2_)
	goSparse = 1;
    } else {
      if (numberNonZero<sparseThreshold_) 
	goSparse = 2;
    }
  }
#ifndef COIN_FAST_CODE
  assert (slackValue_==-1.0);
#endif
  if (!goSparse&&numberRows_<1000) {
    double * COIN_RESTRICT arrayFT = regionFT->denseVector (  );
    int * COIN_RESTRICT indexFT = regionFT->getIndices();
    int numberNonZeroFT;
    double * COIN_RESTRICT arrayUpdate = regionUpdate->denseVector (  );
    int * COIN_RESTRICT indexUpdate = regionUpdate->getIndices();
    int numberNonZeroUpdate;
    updateTwoColumnsUDensish(numberNonZeroFT,arrayFT,indexFT,
			     numberNonZeroUpdate,arrayUpdate,indexUpdate);
    regionFT->setNumElements ( numberNonZeroFT );
    regionUpdate->setNumElements ( numberNonZeroUpdate );
  } else {
    // sparse 
    updateColumnU ( regionFT, regionIndex);
    updateColumnU ( regionUpdate, regionUpdate->getIndices());
  }
  permuteBack(regionFT,regionSparse2);
  if (!noPermuteRegion3) {
    permuteBack(regionUpdate,regionSparse3);
  }
#ifdef COIN_DEBUG
  int n2=regionSparse2->getNumElements();
  regionSparse1->checkClean();
  int n2a= updateColumnFT(regionSparse1,&save2);
  assert(n2==n2a);
  {
    int j;
    double * regionA = save2.denseVector();
    int * indexA = save2.getIndices();
    double * regionB = regionSparse2->denseVector();
    int * indexB = regionSparse2->getIndices();
    for (j=0;j<n2;j++) {
      int k = indexA[j];
      assert (k==indexB[j]);
      CoinFactorizationDouble value = regionA[j];
      assert (value==regionB[j]);
    }
  }
  updateColumn(&save3,
	       &save3,
	       noPermuteRegion3);
  int n3=regionSparse3->getNumElements();
  assert (n3==save3.getNumElements());
  {
    int j;
    double * regionA = save3.denseVector();
    int * indexA = save3.getIndices();
    double * regionB = regionSparse3->denseVector();
    int * indexB = regionSparse3->getIndices();
    for (j=0;j<n3;j++) {
      int k = indexA[j];
      assert (k==indexB[j]);
      CoinFactorizationDouble value = regionA[k];
      assert (value==regionB[k]);
    }
  }
  //*regionSparse2=save2;
  //*regionSparse3=save3;
  printf("REGION2 %d els\n",regionSparse2->getNumElements());
  regionSparse2->print();
  printf("REGION3 %d els\n",regionSparse3->getNumElements());
  regionSparse3->print();
#endif
  return regionSparse2->getNumElements();
#else
  int returnCode= updateColumnFT(regionSparse1,
				 regionSparse2);
  assert (noPermuteRegion3);
  updateColumn(regionSparse3,
	       regionSparse3,
	       noPermuteRegion3);
  //printf("REGION2 %d els\n",regionSparse2->getNumElements());
  //regionSparse2->print();
  //printf("REGION3 %d els\n",regionSparse3->getNumElements());
  //regionSparse3->print();
  return returnCode;
#endif
}
// Updates part of 2 columns (FTRANU) real work
void
CoinFactorization::updateTwoColumnsUDensish (
					     int & numberNonZero1,
					     double * COIN_RESTRICT region1, 
					     int * COIN_RESTRICT index1,
					     int & numberNonZero2,
					     double * COIN_RESTRICT region2, 
					     int * COIN_RESTRICT index2) const
{
  double tolerance = zeroTolerance_;
  const CoinBigIndex * COIN_RESTRICT startColumn = startColumnU_.array();
  const int * COIN_RESTRICT indexRow = indexRowU_.array();
  const CoinFactorizationDouble * COIN_RESTRICT element = elementU_.array();
  int numberNonZeroA = 0;
  int numberNonZeroB = 0;
  const int *numberInColumn = numberInColumn_.array();
  const CoinFactorizationDouble *pivotRegion = pivotRegion_.array();
  
  for (int i = numberU_-1 ; i >= numberSlacks_; i-- ) {
    CoinFactorizationDouble pivotValue2 = region2[i];
    region2[i] = 0.0;
    CoinFactorizationDouble pivotValue1 = region1[i];
    region1[i] = 0.0;
    if ( fabs ( pivotValue2 ) > tolerance ) {
      CoinBigIndex start = startColumn[i];
      const CoinFactorizationDouble * COIN_RESTRICT thisElement = element+start;
      const int * COIN_RESTRICT thisIndex = indexRow+start;
      if ( fabs ( pivotValue1 ) <= tolerance ) {
	// just region 2
	for (CoinBigIndex j=numberInColumn[i]-1 ; j >=0; j-- ) {
	  int iRow = thisIndex[j];
	  CoinFactorizationDouble value = thisElement[j];
#ifdef NO_LOAD
	  region2[iRow] -= value * pivotValue2;
#else
	  CoinFactorizationDouble regionValue2 = region2[iRow];
	  region2[iRow] = regionValue2 - value * pivotValue2;
#endif
	}
	pivotValue2 *= pivotRegion[i];
	region2[i]=pivotValue2;
	index2[numberNonZeroB++]=i;
      } else {
	// both
	for (CoinBigIndex j=numberInColumn[i]-1 ; j >=0; j-- ) {
	  int iRow = thisIndex[j];
	  CoinFactorizationDouble value = thisElement[j];
#ifdef NO_LOAD
	  region1[iRow] -= value * pivotValue1;
	  region2[iRow] -= value * pivotValue2;
#else
	  CoinFactorizationDouble regionValue1 = region1[iRow];
	  CoinFactorizationDouble regionValue2 = region2[iRow];
	  region1[iRow] = regionValue1 - value * pivotValue1;
	  region2[iRow] = regionValue2 - value * pivotValue2;
#endif
	}
	pivotValue1 *= pivotRegion[i];
	pivotValue2 *= pivotRegion[i];
	region1[i]=pivotValue1;
	index1[numberNonZeroA++]=i;
	region2[i]=pivotValue2;
	index2[numberNonZeroB++]=i;
      }
    } else if ( fabs ( pivotValue1 ) > tolerance ) {
      CoinBigIndex start = startColumn[i];
      const CoinFactorizationDouble * COIN_RESTRICT thisElement = element+start;
      const int * COIN_RESTRICT thisIndex = indexRow+start;
      // just region 1
      for (CoinBigIndex j=numberInColumn[i]-1 ; j >=0; j-- ) {
	int iRow = thisIndex[j];
	CoinFactorizationDouble value = thisElement[j];
#ifdef NO_LOAD
	region1[iRow] -= value * pivotValue1;
#else
	CoinFactorizationDouble regionValue1 = region1[iRow];
	region1[iRow] = regionValue1 - value * pivotValue1;
#endif
      }
      pivotValue1 *= pivotRegion[i];
      region1[i]=pivotValue1;
      index1[numberNonZeroA++]=i;
    }
  }
  // Slacks 
    
  for (int i = numberSlacks_-1; i>=0;i--) {
    double value2 = region2[i];
    double value1 = region1[i];
    bool value1NonZero = (value1!=0.0);
    if ( fabs(value2) > tolerance ) {
      region2[i]=-value2;
      index2[numberNonZeroB++]=i;
    } else {
      region2[i]=0.0;
    }
    if ( value1NonZero ) {
      index1[numberNonZeroA]=i;
      if ( fabs(value1) > tolerance ) {
	region1[i]=-value1;
	numberNonZeroA++;
      } else {
	region1[i]=0.0;
      }
    }
  }
  numberNonZero1=numberNonZeroA;
  numberNonZero2=numberNonZeroB;
}
//  updateColumnU.  Updates part of column (FTRANU)
void
CoinFactorization::updateColumnU ( CoinIndexedVector * regionSparse,
				   int * indexIn) const
{
  int numberNonZero = regionSparse->getNumElements (  );

  int goSparse;
  // Guess at number at end
  if (sparseThreshold_>0) {
    if (ftranAverageAfterR_) {
      int newNumber = static_cast<int> (numberNonZero*ftranAverageAfterU_);
      if (newNumber< sparseThreshold_)
	goSparse = 2;
      else if (newNumber< sparseThreshold2_)
	goSparse = 1;
      else
	goSparse = 0;
    } else {
      if (numberNonZero<sparseThreshold_) 
	goSparse = 2;
      else
	goSparse = 0;
    }
  } else {
    goSparse=0;
  }
  switch (goSparse) {
  case 0: // densish
    {
      double *region = regionSparse->denseVector (  );
      int * regionIndex = regionSparse->getIndices();
      int numberNonZero=updateColumnUDensish(region,regionIndex);
      regionSparse->setNumElements ( numberNonZero );
    }
    break;
  case 1: // middling
    updateColumnUSparsish(regionSparse,indexIn);
    break;
  case 2: // sparse
    updateColumnUSparse(regionSparse,indexIn);
    break;
  }
  if (collectStatistics_) 
    ftranCountAfterU_ += regionSparse->getNumElements (  );
}
#ifdef COIN_DEVELOP
double ncall_DZ=0.0;
double nrow_DZ=0.0;
double nslack_DZ=0.0;
double nU_DZ=0.0;
double nnz_DZ=0.0;
double nDone_DZ=0.0;
#endif
// Updates part of column (FTRANU) real work
int 
CoinFactorization::updateColumnUDensish ( double * COIN_RESTRICT region, 
					  int * COIN_RESTRICT regionIndex) const
{
  double tolerance = zeroTolerance_;
  const CoinBigIndex *startColumn = startColumnU_.array();
  const int *indexRow = indexRowU_.array();
  const CoinFactorizationDouble *element = elementU_.array();
  int numberNonZero = 0;
  const int *numberInColumn = numberInColumn_.array();
  const CoinFactorizationDouble *pivotRegion = pivotRegion_.array();
#ifdef COIN_DEVELOP
  ncall_DZ++;
  nrow_DZ += numberRows_;
  nslack_DZ += numberSlacks_;
  nU_DZ += numberU_;
#endif
  
  for (int i = numberU_-1 ; i >= numberSlacks_; i-- ) {
    CoinFactorizationDouble pivotValue = region[i];
    if (pivotValue) {
#ifdef COIN_DEVELOP
      nnz_DZ++;
#endif
      region[i] = 0.0;
      if ( fabs ( pivotValue ) > tolerance ) {
	CoinBigIndex start = startColumn[i];
	const CoinFactorizationDouble * thisElement = element+start;
	const int * thisIndex = indexRow+start;
#ifdef COIN_DEVELOP
	nDone_DZ += numberInColumn[i];
#endif
	for (CoinBigIndex j=numberInColumn[i]-1 ; j >=0; j-- ) {
	  int iRow = thisIndex[j];
	  CoinFactorizationDouble regionValue = region[iRow];
	  CoinFactorizationDouble value = thisElement[j];
	  region[iRow] = regionValue - value * pivotValue;
	}
	pivotValue *= pivotRegion[i];
	region[i]=pivotValue;
	regionIndex[numberNonZero++]=i;
      }
    }
  }
    
  // now do slacks
#ifndef COIN_FAST_CODE
  if (slackValue_==-1.0) {
#endif
#if 0
    // Could skew loop to pick up next one earlier
    // might improve pipelining
    for (int i = numberSlacks_-1; i>2;i-=2) {
      double value0 = region[i];
      double absValue0 = fabs ( value0 );
      double value1 = region[i-1];
      double absValue1 = fabs ( value1 );
      if ( value0 ) {
	if ( absValue0 > tolerance ) {
	  region[i]=-value0;
	  regionIndex[numberNonZero++]=i;
	} else {
	  region[i]=0.0;
	}
      }
      if ( value1 ) {
	if ( absValue1 > tolerance ) {
	  region[i-1]=-value1;
	  regionIndex[numberNonZero++]=i-1;
	} else {
	  region[i-1]=0.0;
	}
      }
    }
    for ( ; i>=0;i--) {
      double value = region[i];
      double absValue = fabs ( value );
      if ( value ) {
	if ( absValue > tolerance ) {
	  region[i]=-value;
	  regionIndex[numberNonZero++]=i;
	} else {
	  region[i]=0.0;
	}
      }
    }
#else
    for (int i = numberSlacks_-1; i>=0;i--) {
      double value = region[i];
      if ( value ) {
	region[i]=-value;
	regionIndex[numberNonZero]=i;
	if ( fabs(value) > tolerance ) 
	  numberNonZero++;
	else 
	  region[i]=0.0;
      }
    }
#endif
#ifndef COIN_FAST_CODE
  } else {
    assert (slackValue_==1.0);
    for (int i = numberSlacks_-1; i>=0;i--) {
      double value = region[i];
      double absValue = fabs ( value );
      if ( value ) {
	region[i]=0.0;
	if ( absValue > tolerance ) {
	  region[i]=value;
	  regionIndex[numberNonZero++]=i;
	}
      }
    }
  }
#endif
  return numberNonZero;
}
//  updateColumnU.  Updates part of column (FTRANU)
/*
  Since everything is in order I should be able to do a better job of
  marking stuff - think.  Also as L is static maybe I can do something
  better there (I know I could if I marked the depth of every element
  but that would lead to other inefficiencies.
*/
void
CoinFactorization::updateColumnUSparse ( CoinIndexedVector * regionSparse,
					 int * COIN_RESTRICT indexIn) const
{
  int numberNonZero = regionSparse->getNumElements (  );
  int * COIN_RESTRICT regionIndex = regionSparse->getIndices (  );
  double * COIN_RESTRICT region = regionSparse->denseVector (  );
  double tolerance = zeroTolerance_;
  const CoinBigIndex *startColumn = startColumnU_.array();
  const int *indexRow = indexRowU_.array();
  const CoinFactorizationDouble *element = elementU_.array();
  const CoinFactorizationDouble *pivotRegion = pivotRegion_.array();
  // use sparse_ as temporary area
  // mark known to be zero
  int * COIN_RESTRICT stack = sparse_.array();  /* pivot */
  int * COIN_RESTRICT list = stack + maximumRowsExtra_;  /* final list */
  CoinBigIndex * COIN_RESTRICT next = reinterpret_cast<CoinBigIndex *> (list + maximumRowsExtra_);  /* jnext */
  char * COIN_RESTRICT mark = reinterpret_cast<char *> (next + maximumRowsExtra_);
#ifdef COIN_DEBUG
  for (int i=0;i<maximumRowsExtra_;i++) {
    assert (!mark[i]);
  }
#endif

  // move slacks to end of stack list
  int * COIN_RESTRICT putLast = stack+maximumRowsExtra_;
  int * COIN_RESTRICT put = putLast;

  const int *numberInColumn = numberInColumn_.array();
  int nList = 0;
  for (int i=0;i<numberNonZero;i++) {
    int kPivot=indexIn[i];
    stack[0]=kPivot;
    CoinBigIndex j=startColumn[kPivot]+numberInColumn[kPivot]-1;
    int nStack=1;
    next[0]=j;
    while (nStack) {
      /* take off stack */
      int kPivot=stack[--nStack];
      if (mark[kPivot]!=1) {
	j=next[nStack];
	if (j>=startColumn[kPivot]) {
	  kPivot=indexRow[j--];
	  /* put back on stack */
	  next[nStack++] =j;
	  if (!mark[kPivot]) {
	    /* and new one */
	    int numberIn = numberInColumn[kPivot];
	    if (numberIn) {
	      j = startColumn[kPivot]+numberIn-1;
	      stack[nStack]=kPivot;
	      mark[kPivot]=2;
	      next[nStack++]=j;
	    } else {
	      // can do immediately
	      /* finished so mark */
	      mark[kPivot]=1;
	      if (kPivot>=numberSlacks_) {
		list[nList++]=kPivot;
	      } else {
		// slack - put at end
		--put;
		*put=kPivot;
	      }
	    }
	  }
	} else {
	  /* finished so mark */
	  mark[kPivot]=1;
	  if (kPivot>=numberSlacks_) {
	    list[nList++]=kPivot;
	  } else {
	    // slack - put at end
	    assert (!numberInColumn[kPivot]);
	    --put;
	    *put=kPivot;
	  }
	}
      }
    }
  }
#if 0
  {
    std::sort(list,list+nList);
    int i;
    int last;
    last =-1;
    for (i=0;i<nList;i++) {
      int k = list[i];
      assert (k>last);
      last=k;
    }
    std::sort(put,putLast);
    int n = putLast-put;
    last =-1;
    for (i=0;i<n;i++) {
      int k = put[i];
      assert (k>last);
      last=k;
    }
  }
#endif
  numberNonZero=0;
  for (int i=nList-1;i>=0;i--) {
    int iPivot = list[i];
    mark[iPivot]=0;
    CoinFactorizationDouble pivotValue = region[iPivot];
    region[iPivot]=0.0;
    if ( fabs ( pivotValue ) > tolerance ) {
      CoinBigIndex start = startColumn[iPivot];
      int number = numberInColumn[iPivot];
      
      CoinBigIndex j;
      for ( j = start; j < start+number; j++ ) {
	CoinFactorizationDouble value = element[j];
	int iRow = indexRow[j];
	region[iRow] -=  value * pivotValue;
      }
      pivotValue *= pivotRegion[iPivot];
      region[iPivot]=pivotValue;
      regionIndex[numberNonZero++]=iPivot;
    }
  }
  // slacks
#ifndef COIN_FAST_CODE
  if (slackValue_==1.0) {
    for (;put<putLast;put++) {
      int iPivot = *put;
      mark[iPivot]=0;
      CoinFactorizationDouble pivotValue = region[iPivot];
      region[iPivot]=0.0;
      if ( fabs ( pivotValue ) > tolerance ) {
	region[iPivot]=pivotValue;
	regionIndex[numberNonZero++]=iPivot;
      }
    }
  } else {
#endif
    for (;put<putLast;put++) {
      int iPivot = *put;
      mark[iPivot]=0;
      CoinFactorizationDouble pivotValue = region[iPivot];
      region[iPivot]=0.0;
      if ( fabs ( pivotValue ) > tolerance ) {
	region[iPivot]=-pivotValue;
	regionIndex[numberNonZero++]=iPivot;
      }
    }
#ifndef COIN_FAST_CODE
  }
#endif
  regionSparse->setNumElements ( numberNonZero );
}
//  updateColumnU.  Updates part of column (FTRANU)
/*
  Since everything is in order I should be able to do a better job of
  marking stuff - think.  Also as L is static maybe I can do something
  better there (I know I could if I marked the depth of every element
  but that would lead to other inefficiencies.
*/
#ifdef COIN_DEVELOP
double ncall_SZ=0.0;
double nrow_SZ=0.0;
double nslack_SZ=0.0;
double nU_SZ=0.0;
double nnz_SZ=0.0;
double nDone_SZ=0.0;
#endif
void
CoinFactorization::updateColumnUSparsish ( CoinIndexedVector * regionSparse,
					   int * COIN_RESTRICT indexIn) const
{
  int * COIN_RESTRICT regionIndex = regionSparse->getIndices (  );
  // mark known to be zero
  int * COIN_RESTRICT stack = sparse_.array();  /* pivot */
  int * COIN_RESTRICT list = stack + maximumRowsExtra_;  /* final list */
  CoinBigIndex * COIN_RESTRICT next = reinterpret_cast<CoinBigIndex *> (list + maximumRowsExtra_);  /* jnext */
  CoinCheckZero * COIN_RESTRICT mark = reinterpret_cast<CoinCheckZero *> (next + maximumRowsExtra_);
  const int *numberInColumn = numberInColumn_.array();
#ifdef COIN_DEBUG
  for (int i=0;i<maximumRowsExtra_;i++) {
    assert (!mark[i]);
  }
#endif

  int nMarked=0;
  int numberNonZero = regionSparse->getNumElements (  );
  double * COIN_RESTRICT region = regionSparse->denseVector (  );
  double tolerance = zeroTolerance_;
  const CoinBigIndex *startColumn = startColumnU_.array();
  const int *indexRow = indexRowU_.array();
  const CoinFactorizationDouble *element = elementU_.array();
  const CoinFactorizationDouble *pivotRegion = pivotRegion_.array();
#ifdef COIN_DEVELOP
  ncall_SZ++;
  nrow_SZ += numberRows_;
  nslack_SZ += numberSlacks_;
  nU_SZ += numberU_;
#endif

  for (int ii=0;ii<numberNonZero;ii++) {
    int iPivot=indexIn[ii];
    int iWord = iPivot>>CHECK_SHIFT;
    int iBit = iPivot-(iWord<<CHECK_SHIFT);
    if (mark[iWord]) {
      mark[iWord] = static_cast<CoinCheckZero>(mark[iWord] | (1<<iBit));
    } else {
      mark[iWord] = static_cast<CoinCheckZero>(1<<iBit);
      stack[nMarked++]=iWord;
    }
  }
  numberNonZero = 0;
  // First do down to convenient power of 2
  CoinBigIndex jLast = (numberU_-1)>>CHECK_SHIFT;
  jLast = CoinMax((jLast<<CHECK_SHIFT),static_cast<CoinBigIndex> (numberSlacks_));
  int i;
  for ( i = numberU_-1 ; i >= jLast; i-- ) {
    CoinFactorizationDouble pivotValue = region[i];
    region[i] = 0.0;
    if ( fabs ( pivotValue ) > tolerance ) {
#ifdef COIN_DEVELOP
      nnz_SZ ++;
#endif
      CoinBigIndex start = startColumn[i];
      const CoinFactorizationDouble * thisElement = element+start;
      const int * thisIndex = indexRow+start;
      
#ifdef COIN_DEVELOP
      nDone_SZ += numberInColumn[i];
#endif
      for (int j=numberInColumn[i]-1 ; j >=0; j-- ) {
	int iRow0 = thisIndex[j];
	CoinFactorizationDouble regionValue0 = region[iRow0];
	CoinFactorizationDouble value0 = thisElement[j];
	int iWord = iRow0>>CHECK_SHIFT;
	int iBit = iRow0-(iWord<<CHECK_SHIFT);
	if (mark[iWord]) {
	  mark[iWord] = static_cast<CoinCheckZero>(mark[iWord] | (1<<iBit));
	} else {
	  mark[iWord] = static_cast<CoinCheckZero>(1<<iBit);
	  stack[nMarked++]=iWord;
	}
	region[iRow0] = regionValue0 - value0 * pivotValue;
      }
      pivotValue *= pivotRegion[i];
      region[i]=pivotValue;
      regionIndex[numberNonZero++]=i;
    }
  }
  int kLast = (numberSlacks_+BITS_PER_CHECK-1)>>CHECK_SHIFT;
  if (jLast>numberSlacks_) {
    // now do in chunks
    for (int k=(jLast>>CHECK_SHIFT)-1;k>=kLast;k--) {
      unsigned int iMark = mark[k];
      if (iMark) {
	// something in chunk - do all (as imark may change)
	int iLast = k<<CHECK_SHIFT;
	for ( i = iLast+BITS_PER_CHECK-1 ; i >= iLast; i-- ) {
	  CoinFactorizationDouble pivotValue = region[i];
	  if (pivotValue) {
#ifdef COIN_DEVELOP
	    nnz_SZ ++;
#endif
	    region[i] = 0.0;
	    if ( fabs ( pivotValue ) > tolerance ) {
	      CoinBigIndex start = startColumn[i];
	      const CoinFactorizationDouble * thisElement = element+start;
	      const int * thisIndex = indexRow+start;
#ifdef COIN_DEVELOP
	      nDone_SZ += numberInColumn[i];
#endif
	      for (int j=numberInColumn[i]-1 ; j >=0; j-- ) {
		int iRow0 = thisIndex[j];
		CoinFactorizationDouble regionValue0 = region[iRow0];
		CoinFactorizationDouble value0 = thisElement[j];
		int iWord = iRow0>>CHECK_SHIFT;
		int iBit = iRow0-(iWord<<CHECK_SHIFT);
		if (mark[iWord]) {
		  mark[iWord] = static_cast<CoinCheckZero>(mark[iWord] | (1<<iBit));
		} else {
		  mark[iWord] = static_cast<CoinCheckZero>(1<<iBit);
		  stack[nMarked++]=iWord;
		}
		region[iRow0] = regionValue0 - value0 * pivotValue;
	      }
	      pivotValue *= pivotRegion[i];
	      region[i]=pivotValue;
	      regionIndex[numberNonZero++]=i;
	    }
	  }
	}
	mark[k]=0;
      }
    }
    i = (kLast<<CHECK_SHIFT)-1;
  }
  for ( ; i >= numberSlacks_; i-- ) {
    CoinFactorizationDouble pivotValue = region[i];
    region[i] = 0.0;
    if ( fabs ( pivotValue ) > tolerance ) {
#ifdef COIN_DEVELOP
      nnz_SZ ++;
#endif
      CoinBigIndex start = startColumn[i];
      const CoinFactorizationDouble * thisElement = element+start;
      const int * thisIndex = indexRow+start;
#ifdef COIN_DEVELOP
      nDone_SZ += numberInColumn[i];
#endif
      for (int j=numberInColumn[i]-1 ; j >=0; j-- ) {
	int iRow0 = thisIndex[j];
	CoinFactorizationDouble regionValue0 = region[iRow0];
	CoinFactorizationDouble value0 = thisElement[j];
	int iWord = iRow0>>CHECK_SHIFT;
	int iBit = iRow0-(iWord<<CHECK_SHIFT);
	if (mark[iWord]) {
	  mark[iWord] = static_cast<CoinCheckZero>(mark[iWord] | (1<<iBit));
	} else {
	  mark[iWord] = static_cast<CoinCheckZero>(1<<iBit);
	  stack[nMarked++]=iWord;
	}
	region[iRow0] = regionValue0 - value0 * pivotValue;
      }
      pivotValue *= pivotRegion[i];
      region[i]=pivotValue;
      regionIndex[numberNonZero++]=i;
    }
  }
  
  if (numberSlacks_) {
    // now do slacks
#ifndef COIN_FAST_CODE
    double factor = slackValue_;
    if (factor==1.0) {
      // First do down to convenient power of 2
      CoinBigIndex jLast = (numberSlacks_-1)>>CHECK_SHIFT;
      jLast = jLast<<CHECK_SHIFT;
      for ( i = numberSlacks_-1; i>=jLast;i--) {
	double value = region[i];
	double absValue = fabs ( value );
	if ( value ) {
	  region[i]=0.0;
	  if ( absValue > tolerance ) {
	  region[i]=value;
	  regionIndex[numberNonZero++]=i;
	  }
	}
      }
      mark[jLast]=0;
      // now do in chunks
      for (int k=(jLast>>CHECK_SHIFT)-1;k>=0;k--) {
	unsigned int iMark = mark[k];
	if (iMark) {
	  // something in chunk - do all (as imark may change)
	  int iLast = k<<CHECK_SHIFT;
	  i = iLast+BITS_PER_CHECK-1;
	  for ( ; i >= iLast; i-- ) {
	    double value = region[i];
	    double absValue = fabs ( value );
	    if ( value ) {
	      region[i]=0.0;
	      if ( absValue > tolerance ) {
		region[i]=value;
		regionIndex[numberNonZero++]=i;
	      }
	    }
	  }
	  mark[k]=0;
	}
      }
    } else {
      assert (factor==-1.0);
#endif
      // First do down to convenient power of 2
      CoinBigIndex jLast = (numberSlacks_-1)>>CHECK_SHIFT;
      jLast = jLast<<CHECK_SHIFT;
      for ( i = numberSlacks_-1; i>=jLast;i--) {
	double value = region[i];
	double absValue = fabs ( value );
	if ( value ) {
	  region[i]=0.0;
	  if ( absValue > tolerance ) {
	    region[i]=-value;
	    regionIndex[numberNonZero++]=i;
	  }
	}
      }
      mark[jLast]=0;
      // now do in chunks
      for (int k=(jLast>>CHECK_SHIFT)-1;k>=0;k--) {
	unsigned int iMark = mark[k];
	if (iMark) {
	  // something in chunk - do all (as imark may change)
	  int iLast = k<<CHECK_SHIFT;
	  i = iLast+BITS_PER_CHECK-1;
	  for ( ; i >= iLast; i-- ) {
	    double value = region[i];
	    double absValue = fabs ( value );
	    if ( value ) {
	      region[i]=0.0;
	      if ( absValue > tolerance ) {
		region[i]=-value;
		regionIndex[numberNonZero++]=i;
	      }
	    }
	  }
	  mark[k]=0;
	}
      }
#ifndef COIN_FAST_CODE
    }
#endif
  }
  regionSparse->setNumElements ( numberNonZero );
  mark[(numberU_-1)>>CHECK_SHIFT]=0;
  mark[numberSlacks_>>CHECK_SHIFT]=0;
  if (numberSlacks_)
    mark[(numberSlacks_-1)>>CHECK_SHIFT]=0;
#ifdef COIN_DEBUG
  for (i=0;i<maximumRowsExtra_;i++) {
    assert (!mark[i]);
  }
#endif
}
//  updateColumnR.  Updates part of column (FTRANR)
void
CoinFactorization::updateColumnR ( CoinIndexedVector * regionSparse ) const
{
  double * COIN_RESTRICT region = regionSparse->denseVector (  );
  int * COIN_RESTRICT regionIndex = regionSparse->getIndices (  );
  int numberNonZero = regionSparse->getNumElements (  );

  if ( !numberR_ )
    return;	//return if nothing to do
  double tolerance = zeroTolerance_;

  const CoinBigIndex * startColumn = startColumnR_.array()-numberRows_;
  const int * indexRow = indexRowR_;
  const CoinFactorizationDouble * element = elementR_;
  const int * permute = permute_.array();

  // Work out very dubious idea of what would be fastest
  int method=-1;
  // Size of R
  double sizeR=startColumnR_.array()[numberR_];
  // Average
  double averageR = sizeR/(static_cast<double> (numberRowsExtra_));
  // weights (relative to actual work)
  double setMark = 0.1; // setting mark
  double test1= 1.0; // starting ftran (without testPivot)
  double testPivot = 2.0; // Seeing if zero etc
  double startDot=2.0; // For starting dot product version
  // For final scan
  double final = numberNonZero*1.0;
  double methodTime[3];
  // For second type
  methodTime[1] = numberPivots_ * (testPivot + ((static_cast<double> (numberNonZero))/(static_cast<double> (numberRows_))
						* averageR));
  methodTime[1] += numberNonZero *(test1 + averageR);
  // For first type
  methodTime[0] = methodTime[1] + (numberNonZero+numberPivots_)*setMark;
  methodTime[1] += numberNonZero*final;
  // third
  methodTime[2] = sizeR + numberPivots_*startDot + numberNonZero*final;
  // switch off if necessary
  if (!numberInColumnPlus_.array()) {
    methodTime[0]=1.0e100;
    methodTime[1]=1.0e100;
  } else if (!sparse_.array()) {
    methodTime[0]=1.0e100;
  }
  double best=1.0e100;
  for (int i=0;i<3;i++) {
    if (methodTime[i]<best) {
      best=methodTime[i];
      method=i;
    }
  }
  assert (method>=0);
  const int * numberInColumnPlus = numberInColumnPlus_.array();
  //if (method==1)
  //printf(" methods %g %g %g - chosen %d\n",methodTime[0],methodTime[1],methodTime[2],method);

  switch (method) {
  case 0:
#ifdef STACK
    {
      // use sparse_ as temporary area
      // mark known to be zero
      int * COIN_RESTRICT stack = sparse_.array();  /* pivot */
      int * COIN_RESTRICT list = stack + maximumRowsExtra_;  /* final list */
      CoinBigIndex * COIN_RESTRICT next = (CoinBigIndex *) (list + maximumRowsExtra_);  /* jnext */
      char * COIN_RESTRICT mark = (char *) (next + maximumRowsExtra_);
      // we have another copy of R in R
      const CoinFactorizationDouble * elementR = elementR_ + lengthAreaR_;
      const int * indexRowR = indexRowR_ + lengthAreaR_;
      const CoinBigIndex * startR = startColumnR_.array()+maximumPivots_+1;
      int nList=0;
      const int * permuteBack = permuteBack_.array();
      for (int k=0;k<numberNonZero;k++) {
	int kPivot=regionIndex[k];
	if(!mark[kPivot]) {
	  stack[0]=kPivot;
	  CoinBigIndex j=-10;
	  next[0]=j;
	  int nStack=0;
	  while (nStack>=0) {
	    /* take off stack */
	    if (j>=startR[kPivot]) {
	      int jPivot=indexRowR[j--];
	      /* put back on stack */
	      next[nStack] =j;
	      if (!mark[jPivot]) {
		/* and new one */
		kPivot=jPivot;
		j=-10;
		stack[++nStack]=kPivot;
		mark[kPivot]=1;
		next[nStack]=j;
	      }
	    } else if (j==-10) {
	      // before first - see if followon
	      int jPivot = permuteBack[kPivot];
	      if (jPivot<numberRows_) {
		// no
		j=startR[kPivot]+numberInColumnPlus[kPivot]-1;
		next[nStack]=j;
	      } else {
		// add to list
		if (!mark[jPivot]) {
		  /* and new one */
		  kPivot=jPivot;
		  j=-10;
		  stack[++nStack]=kPivot;
		  mark[kPivot]=1;
		  next[nStack]=j;
		} else {
		  j=startR[kPivot]+numberInColumnPlus[kPivot]-1;
		  next[nStack]=j;
		}
	      }
	    } else {
	      // finished
	      list[nList++]=kPivot;
	      mark[kPivot]=1;
	      --nStack;
	      if (nStack>=0) {
		kPivot=stack[nStack];
		j=next[nStack];
	      }
	    }
	  }
	}
      }
      numberNonZero=0;
      for (int i=nList-1;i>=0;i--) {
	int iPivot = list[i];
	mark[iPivot]=0;
	CoinFactorizationDouble pivotValue;
	if (iPivot<numberRows_) {
	  pivotValue = region[iPivot];
	} else {
	  int before = permute[iPivot];
	  pivotValue = region[iPivot] + region[before];
	  region[before]=0.0;
	}
	if ( fabs ( pivotValue ) > tolerance ) {
	  region[iPivot] = pivotValue;
	  CoinBigIndex start = startR[iPivot];
	  int number = numberInColumnPlus[iPivot];
	  CoinBigIndex end = start + number;
	  CoinBigIndex j;
	  for (j=start ; j < end; j ++ ) {
	    int iRow = indexRowR[j];
	    CoinFactorizationDouble value = elementR[j];
	    region[iRow] -= value * pivotValue;
	  }     
	  regionIndex[numberNonZero++] = iPivot;
	} else {
	  region[iPivot] = 0.0;
	}       
      }
    }
#else
    {
      
      // use sparse_ as temporary area
      // mark known to be zero
      int * COIN_RESTRICT stack = sparse_.array();  /* pivot */
      int * COIN_RESTRICT list = stack + maximumRowsExtra_;  /* final list */
      CoinBigIndex * COIN_RESTRICT next = reinterpret_cast<CoinBigIndex *> (list + maximumRowsExtra_);  /* jnext */
      char * COIN_RESTRICT mark = reinterpret_cast<char *> (next + maximumRowsExtra_);
      // mark all rows which will be permuted
      for (int i = numberRows_; i < numberRowsExtra_; i++ ) {
	int iRow = permute[i];
	mark[iRow]=1;
      }
      // we have another copy of R in R
      const CoinFactorizationDouble * elementR = elementR_ + lengthAreaR_;
      const int * indexRowR = indexRowR_ + lengthAreaR_;
      const CoinBigIndex * startR = startColumnR_.array()+maximumPivots_+1;
      // For current list order does not matter as
      // only affects end
      int newNumber=0;
      for (int i = 0; i < numberNonZero; i++ ) {
	int iRow = regionIndex[i];
	assert (region[iRow]);
	if (!mark[iRow])
	  regionIndex[newNumber++]=iRow;
	int number = numberInColumnPlus[iRow];
	if (number) {
	  CoinFactorizationDouble pivotValue = region[iRow];
	  CoinBigIndex start=startR[iRow];
	  CoinBigIndex end = start+number;
	  for (CoinBigIndex j = start; j < end; j ++ ) {
	    CoinFactorizationDouble value = elementR[j];
	    int jRow = indexRowR[j];
	    region[jRow] -= pivotValue*value;
	  }
	}
      }
      numberNonZero = newNumber;
      for (int i = numberRows_; i < numberRowsExtra_; i++ ) {
	//move using permute_ (stored in inverse fashion)
	int iRow = permute[i];
	CoinFactorizationDouble pivotValue = region[iRow]+region[i];
	//zero out pre-permuted
	region[iRow] = 0.0;
	if ( fabs ( pivotValue ) > tolerance ) {
	  region[i] = pivotValue;
	  if (!mark[i])
	    regionIndex[numberNonZero++] = i;
	  int number = numberInColumnPlus[i];
	  CoinBigIndex start=startR[i];
	  CoinBigIndex end = start+number;
	  for (CoinBigIndex j = start; j < end; j ++ ) {
	    CoinFactorizationDouble value = elementR[j];
	    int jRow = indexRowR[j];
	    region[jRow] -= pivotValue*value;
	  }
	} else {
	  region[i] = 0.0;
	}
	mark[iRow]=0;
      }
    }
#endif
    break;
  case 1:
    {
      // no sparse region
      // we have another copy of R in R
      const CoinFactorizationDouble * elementR = elementR_ + lengthAreaR_;
      const int * indexRowR = indexRowR_ + lengthAreaR_;
      const CoinBigIndex * startR = startColumnR_.array()+maximumPivots_+1;
      // For current list order does not matter as
      // only affects end
      for (int i = 0; i < numberNonZero; i++ ) {
	int iRow = regionIndex[i];
	assert (region[iRow]);
	int number = numberInColumnPlus[iRow];
	if (number) {
	  CoinFactorizationDouble pivotValue = region[iRow];
	  CoinBigIndex start=startR[iRow];
	  CoinBigIndex end = start+number;
	  for (CoinBigIndex j = start; j < end; j ++ ) {
	    CoinFactorizationDouble value = elementR[j];
	    int jRow = indexRowR[j];
	    region[jRow] -= pivotValue*value;
	  }
	}
      }
      for (int i = numberRows_; i < numberRowsExtra_; i++ ) {
	//move using permute_ (stored in inverse fashion)
	int iRow = permute[i];
	CoinFactorizationDouble pivotValue = region[iRow]+region[i];
	//zero out pre-permuted
	region[iRow] = 0.0;
	if ( fabs ( pivotValue ) > tolerance ) {
	  region[i] = pivotValue;
	  regionIndex[numberNonZero++] = i;
	  int number = numberInColumnPlus[i];
	  CoinBigIndex start=startR[i];
	  CoinBigIndex end = start+number;
	  for (CoinBigIndex j = start; j < end; j ++ ) {
	    CoinFactorizationDouble value = elementR[j];
	    int jRow = indexRowR[j];
	    region[jRow] -= pivotValue*value;
	  }
	} else {
	  region[i] = 0.0;
	}
      }
    }
    break;
  case 2:
    {
      CoinBigIndex start = startColumn[numberRows_];
      for (int i = numberRows_; i < numberRowsExtra_; i++ ) {
	//move using permute_ (stored in inverse fashion)
	CoinBigIndex end = startColumn[i+1];
	int iRow = permute[i];
	CoinFactorizationDouble pivotValue = region[iRow];
	//zero out pre-permuted
	region[iRow] = 0.0;

	for (CoinBigIndex j = start; j < end; j ++ ) {
	  CoinFactorizationDouble value = element[j];
	  int jRow = indexRow[j];
	  value *= region[jRow];
	  pivotValue -= value;
	}
	start=end;
	if ( fabs ( pivotValue ) > tolerance ) {
	  region[i] = pivotValue;
	  regionIndex[numberNonZero++] = i;
	} else {
	region[i] = 0.0;
	}
      }
    }
    break;
  }
  if (method) {
    // pack down
    int n=numberNonZero;
    numberNonZero=0;
    for (int i=0;i<n;i++) {
      int indexValue = regionIndex[i];
      double value = region[indexValue];
      if (value) 
	regionIndex[numberNonZero++]=indexValue;
    }
  }
  //set counts
  regionSparse->setNumElements ( numberNonZero );
}
//  updateColumnR.  Updates part of column (FTRANR)
void
CoinFactorization::updateColumnRFT ( CoinIndexedVector * regionSparse,
				     int * COIN_RESTRICT regionIndex) 
{
  double * COIN_RESTRICT region = regionSparse->denseVector (  );
  //int *regionIndex = regionSparse->getIndices (  );
  CoinBigIndex * COIN_RESTRICT startColumnU = startColumnU_.array();
  int numberNonZero = regionSparse->getNumElements (  );

  if ( numberR_ ) {
    double tolerance = zeroTolerance_;
    
    const CoinBigIndex * startColumn = startColumnR_.array()-numberRows_;
    const int * indexRow = indexRowR_;
    const CoinFactorizationDouble * element = elementR_;
    const int * permute = permute_.array();
    

    // Work out very dubious idea of what would be fastest
    int method=-1;
    // Size of R
    double sizeR=startColumnR_.array()[numberR_];
    // Average
    double averageR = sizeR/(static_cast<double> (numberRowsExtra_));
    // weights (relative to actual work)
    double setMark = 0.1; // setting mark
    double test1= 1.0; // starting ftran (without testPivot)
    double testPivot = 2.0; // Seeing if zero etc
    double startDot=2.0; // For starting dot product version
    // For final scan
    double final = numberNonZero*1.0;
    double methodTime[3];
    // For second type
    methodTime[1] = numberPivots_ * (testPivot + ((static_cast<double> (numberNonZero))/(static_cast<double> (numberRows_))
						  * averageR));
    methodTime[1] += numberNonZero *(test1 + averageR);
    // For first type
    methodTime[0] = methodTime[1] + (numberNonZero+numberPivots_)*setMark;
    methodTime[1] += numberNonZero*final;
    // third
    methodTime[2] = sizeR + numberPivots_*startDot + numberNonZero*final;
    // switch off if necessary
    if (!numberInColumnPlus_.array()) {
      methodTime[0]=1.0e100;
      methodTime[1]=1.0e100;
    } else if (!sparse_.array()) {
      methodTime[0]=1.0e100;
    }
    const int * numberInColumnPlus = numberInColumnPlus_.array();
    int * numberInColumn = numberInColumn_.array();
    // adjust for final scan
    methodTime[1] += final;
    double best=1.0e100;
    for (int i=0;i<3;i++) {
      if (methodTime[i]<best) {
	best=methodTime[i];
	method=i;
      }
    }
    assert (method>=0);
    
    switch (method) {
    case 0:
      {
	// use sparse_ as temporary area
	// mark known to be zero
	int * COIN_RESTRICT stack = sparse_.array();  /* pivot */
	int * COIN_RESTRICT list = stack + maximumRowsExtra_;  /* final list */
	CoinBigIndex * COIN_RESTRICT next = reinterpret_cast<CoinBigIndex *> (list + maximumRowsExtra_);  /* jnext */
	char * COIN_RESTRICT mark = reinterpret_cast<char *> (next + maximumRowsExtra_);
	// mark all rows which will be permuted
	for (int i = numberRows_; i < numberRowsExtra_; i++ ) {
	  int iRow = permute[i];
	  mark[iRow]=1;
	}
	// we have another copy of R in R
	const CoinFactorizationDouble * elementR = elementR_ + lengthAreaR_;
	const int * indexRowR = indexRowR_ + lengthAreaR_;
	const CoinBigIndex * startR = startColumnR_.array()+maximumPivots_+1;
	//save in U
	//in at end
	int iColumn = numberColumnsExtra_;

	startColumnU[iColumn] = startColumnU[maximumColumnsExtra_];
	CoinBigIndex start = startColumnU[iColumn];
  
	//int * putIndex = indexRowU_ + start;
	CoinFactorizationDouble * COIN_RESTRICT putElement = elementU_.array() + start;
	// For current list order does not matter as
	// only affects end
	int newNumber=0;
	for (int i = 0; i < numberNonZero; i++ ) {
	  int iRow = regionIndex[i];
	  CoinFactorizationDouble pivotValue = region[iRow];
	  assert (region[iRow]);
	  if (!mark[iRow]) {
	    //putIndex[newNumber]=iRow;
	    putElement[newNumber]=pivotValue;;
	    regionIndex[newNumber++]=iRow;
	  }
	  int number = numberInColumnPlus[iRow];
	  if (number) {
	    CoinBigIndex start=startR[iRow];
	    CoinBigIndex end = start+number;
	    for (CoinBigIndex j = start; j < end; j ++ ) {
	      CoinFactorizationDouble value = elementR[j];
	      int jRow = indexRowR[j];
	      region[jRow] -= pivotValue*value;
	    }
	  }
	}
	numberNonZero = newNumber;
	for (int i = numberRows_; i < numberRowsExtra_; i++ ) {
	  //move using permute_ (stored in inverse fashion)
	  int iRow = permute[i];
	  CoinFactorizationDouble pivotValue = region[iRow]+region[i];
	  //zero out pre-permuted
	  region[iRow] = 0.0;
	  if ( fabs ( pivotValue ) > tolerance ) {
	    region[i] = pivotValue;
	    if (!mark[i]) {
	      //putIndex[numberNonZero]=i;
	      putElement[numberNonZero]=pivotValue;;
	      regionIndex[numberNonZero++]=i;
	    }
	    int number = numberInColumnPlus[i];
	    CoinBigIndex start=startR[i];
	    CoinBigIndex end = start+number;
	    for (CoinBigIndex j = start; j < end; j ++ ) {
	      CoinFactorizationDouble value = elementR[j];
	      int jRow = indexRowR[j];
	      region[jRow] -= pivotValue*value;
	    }
	  } else {
	    region[i] = 0.0;
	  }
	  mark[iRow]=0;
	}
	numberInColumn[iColumn] = numberNonZero;
	startColumnU[maximumColumnsExtra_] = start + numberNonZero;
      }
      break;
    case 1:
      {
	// no sparse region
	// we have another copy of R in R
	const CoinFactorizationDouble * elementR = elementR_ + lengthAreaR_;
	const int * indexRowR = indexRowR_ + lengthAreaR_;
	const CoinBigIndex * startR = startColumnR_.array()+maximumPivots_+1;
	// For current list order does not matter as
	// only affects end
	for (int i = 0; i < numberNonZero; i++ ) {
	  int iRow = regionIndex[i];
	  assert (region[iRow]);
	  int number = numberInColumnPlus[iRow];
	  if (number) {
	    CoinFactorizationDouble pivotValue = region[iRow];
	    CoinBigIndex start=startR[iRow];
	    CoinBigIndex end = start+number;
	    for (CoinBigIndex j = start; j < end; j ++ ) {
	      CoinFactorizationDouble value = elementR[j];
	      int jRow = indexRowR[j];
	      region[jRow] -= pivotValue*value;
	    }
	  }
	}
	for (int i = numberRows_; i < numberRowsExtra_; i++ ) {
	  //move using permute_ (stored in inverse fashion)
	  int iRow = permute[i];
	  CoinFactorizationDouble pivotValue = region[iRow]+region[i];
	  //zero out pre-permuted
	  region[iRow] = 0.0;
	  if ( fabs ( pivotValue ) > tolerance ) {
	    region[i] = pivotValue;
	    regionIndex[numberNonZero++] = i;
	    int number = numberInColumnPlus[i];
	    CoinBigIndex start=startR[i];
	    CoinBigIndex end = start+number;
	    for (CoinBigIndex j = start; j < end; j ++ ) {
	      CoinFactorizationDouble value = elementR[j];
	      int jRow = indexRowR[j];
	      region[jRow] -= pivotValue*value;
	    }
	  } else {
	    region[i] = 0.0;
	  }
	}
      }
      break;
    case 2:
      {
	CoinBigIndex start = startColumn[numberRows_];
	for (int i = numberRows_; i < numberRowsExtra_; i++ ) {
	  //move using permute_ (stored in inverse fashion)
	  CoinBigIndex end = startColumn[i+1];
	  int iRow = permute[i];
	  CoinFactorizationDouble pivotValue = region[iRow];
	  //zero out pre-permuted
	  region[iRow] = 0.0;
	  
	  for (CoinBigIndex j = start; j < end; j ++ ) {
	    CoinFactorizationDouble value = element[j];
	    int jRow = indexRow[j];
	    value *= region[jRow];
	    pivotValue -= value;
	  }
	  start=end;
	  if ( fabs ( pivotValue ) > tolerance ) {
	    region[i] = pivotValue;
	    regionIndex[numberNonZero++] = i;
	  } else {
	    region[i] = 0.0;
	  }
	}
      }
      break;
    }
    if (method) {
      // pack down
      int n=numberNonZero;
      numberNonZero=0;
      //save in U
      //in at end
      int iColumn = numberColumnsExtra_;
      
      assert(startColumnU[iColumn] == startColumnU[maximumColumnsExtra_]);
      CoinBigIndex start = startColumnU[iColumn];
      
      int * COIN_RESTRICT putIndex = indexRowU_.array() + start;
      CoinFactorizationDouble * COIN_RESTRICT putElement = elementU_.array() + start;
      for (int i=0;i<n;i++) {
	int indexValue = regionIndex[i];
	double value = region[indexValue];
	if (value) {
	  putIndex[numberNonZero]=indexValue;
	  putElement[numberNonZero]=value;
	  regionIndex[numberNonZero++]=indexValue;
	}
      }
      numberInColumn[iColumn] = numberNonZero;
      startColumnU[maximumColumnsExtra_] = start + numberNonZero;
    }
    //set counts
    regionSparse->setNumElements ( numberNonZero );
  } else {
    // No R but we still need to save column
    //save in U
    //in at end
    int * COIN_RESTRICT numberInColumn = numberInColumn_.array();
    numberNonZero = regionSparse->getNumElements (  );
    int iColumn = numberColumnsExtra_;
    
    assert(startColumnU[iColumn] == startColumnU[maximumColumnsExtra_]);
    CoinBigIndex start = startColumnU[iColumn];
    numberInColumn[iColumn] = numberNonZero;
    startColumnU[maximumColumnsExtra_] = start + numberNonZero;
    
    int * COIN_RESTRICT putIndex = indexRowU_.array() + start;
    CoinFactorizationDouble * COIN_RESTRICT putElement = elementU_.array() + start;
    for (int i=0;i<numberNonZero;i++) {
      int indexValue = regionIndex[i];
      double value = region[indexValue];
      putIndex[i]=indexValue;
      putElement[i]=value;
    }
  }
}
/* Updates one column (FTRAN) from region2 and permutes.
   region1 starts as zero
   Note - if regionSparse2 packed on input - will be packed on output
   - returns un-permuted result in region2 and region1 is zero */
int CoinFactorization::updateColumnFT ( CoinIndexedVector * regionSparse,
					CoinIndexedVector * regionSparse2)
{
  //permute and move indices into index array
  int * COIN_RESTRICT regionIndex = regionSparse->getIndices (  );
  int numberNonZero = regionSparse2->getNumElements();
  const int *permute = permute_.array();
  int * COIN_RESTRICT index = regionSparse2->getIndices();
  double * COIN_RESTRICT region = regionSparse->denseVector();
  double * COIN_RESTRICT array = regionSparse2->denseVector();
  CoinBigIndex * COIN_RESTRICT startColumnU = startColumnU_.array();
  bool doFT=doForrestTomlin_;
  // see if room
  if (doFT) {
    int iColumn = numberColumnsExtra_;
    
    startColumnU[iColumn] = startColumnU[maximumColumnsExtra_];
    CoinBigIndex start = startColumnU[iColumn];
    CoinBigIndex space = lengthAreaU_ - ( start + numberRowsExtra_ );
    doFT = space>=0;
    if (doFT) {
      regionIndex = indexRowU_.array() + start;
    } else {
      startColumnU[maximumColumnsExtra_] = lengthAreaU_+1;
    }
  }

#ifndef CLP_FACTORIZATION
  bool packed = regionSparse2->packedMode();
  if (packed) {
#else
    assert (regionSparse2->packedMode());
#endif
    for (int j = 0; j < numberNonZero; j ++ ) {
      int iRow = index[j];
      double value = array[j];
      array[j]=0.0;
      iRow = permute[iRow];
      region[iRow] = value;
      regionIndex[j] = iRow;
    }
#ifndef CLP_FACTORIZATION
  } else {
    for (int j = 0; j < numberNonZero; j ++ ) {
      int iRow = index[j];
      double value = array[iRow];
      array[iRow]=0.0;
      iRow = permute[iRow];
      region[iRow] = value;
      regionIndex[j] = iRow;
    }
  }
#endif
  regionSparse->setNumElements ( numberNonZero );
  if (collectStatistics_) {
    numberFtranCounts_++;
    ftranCountInput_ += numberNonZero;
  }
    
  //  ******* L
#if 0
  {
    double *region = regionSparse->denseVector (  );
    //int *regionIndex = regionSparse->getIndices (  );
    int numberNonZero = regionSparse->getNumElements (  );
    for (int i=0;i<numberNonZero;i++) {
      int iRow = regionIndex[i];
      assert (region[iRow]);
    }
  }
#endif
  updateColumnL ( regionSparse, regionIndex );
#if 0
  {
    double *region = regionSparse->denseVector (  );
    //int *regionIndex = regionSparse->getIndices (  );
    int numberNonZero = regionSparse->getNumElements (  );
    for (int i=0;i<numberNonZero;i++) {
      int iRow = regionIndex[i];
      assert (region[iRow]);
    }
  }
#endif
  if (collectStatistics_) 
    ftranCountAfterL_ += regionSparse->getNumElements();
  //permute extra
  //row bits here
  if ( doFT ) 
    updateColumnRFT ( regionSparse, regionIndex );
  else
    updateColumnR ( regionSparse );
  if (collectStatistics_) 
    ftranCountAfterR_ += regionSparse->getNumElements();
  //  ******* U
  updateColumnU ( regionSparse, regionIndex);
  if (!doForrestTomlin_) {
    // Do PFI after everything else
    updateColumnPFI(regionSparse);
  }
  permuteBack(regionSparse,regionSparse2);
  // will be negative if no room
  if ( doFT ) 
    return regionSparse2->getNumElements();
  else 
    return -regionSparse2->getNumElements();
}