#ifndef _TRISKELE_DEAL_THREADS_TPP
#define _TRISKELE_DEAL_THREADS_TPP

// ========================================
template<typename DimImg, typename FunctId>
inline void
dealThreadRange (const DimImg &maxId, const unsigned int &coreCount, const FunctId &functId/* functId (id) */) {
  dealThread (maxId, coreCount, [&functId] (const unsigned int &threadId, const DimImg &minVal, const DimImg &maxVal) {
      for (DimImg id = minVal; id < maxVal; ++id)
	functId (id);
    });
}

// ----------------------------------------
template<typename DimImg, typename FunctThreadId>
inline void
dealThreadThreadRange (const DimImg &maxId, const unsigned int &coreCount, const FunctThreadId &functThreadId/* functThreadId (threadId, id) */) {
  dealThread (maxId, coreCount, [&functThreadId] (const unsigned int &threadId, const DimImg &minVal, const DimImg &maxVal) {
      for (DimImg id = minVal; id < maxVal; ++id)
	functThreadId (threadId, id);
    });
}

// ----------------------------------------
template<typename DimImg, typename FunctMinMax>
inline void
dealThreadBound (const DimImg &maxId, const unsigned int &coreCount, const FunctMinMax &functMinMax/* functMinMax (minVal, maxVal) */) {
  dealThread (maxId, coreCount, [&functMinMax] (const unsigned int &threadId, const DimImg &minVal, const DimImg &maxVal) {
      functMinMax (minVal, maxVal);
    });
}

// ----------------------------------------
template<typename DimImg>
inline std::vector<DimImg>
getDealThreadBounds (const DimImg &maxId, const unsigned int &coreCount) {
  if (!maxId || !coreCount)
    return std::vector<DimImg> (0);
  DimImg average = maxId/coreCount;
  std::vector<DimImg> maxIds (coreCount+1, average);

  for (unsigned int core = 0; core < coreCount; ++core)
    maxIds[core] = DimImg (core*average);
  maxIds[coreCount] = maxId;
  return maxIds;
}

// ----------------------------------------
template<typename DimImg, typename FunctThreadMinMax>
inline void
dealThread (const DimImg &maxId, unsigned int coreCount, const FunctThreadMinMax &functThreadMinMax/* functThreadMinMax (threadId, minVal, maxVal) */) {
  //DEF_LOG ("dealThreadBound", "coreCount:" << coreCount << " maxId:" << maxId);
  if (!maxId || !coreCount)
    return;
  if (DimImg (coreCount) > maxId)
    coreCount = (unsigned int) maxId;
  if (coreCount == 1) {
    functThreadMinMax (0, 0, maxId);
    return;
  }
    
  std::vector<DimImg> maxIds = getDealThreadBounds (maxId, coreCount);
#ifdef THREAD_DISABLE
  for (unsigned int idCopyValInThread = 0; idCopyValInThread < coreCount; ++idCopyValInThread) {
    functThreadMinMax (idCopyValInThread, maxIds[idCopyValInThread], maxIds[idCopyValInThread+1]);
  }
#elif INTEL_TBB_THREAD
  using namespace tbb;
#pragma warning(disable: 588)
  parallel_for (size_t (0), size_t (coreCount), [&maxIds, &functThreadMinMax] (size_t idCopyValInThread) {
      functThreadMinMax (idCopyValInThread, maxIds[idCopyValInThread], maxIds[idCopyValInThread+1]);
    });
#else /* BOOST thread */
  std::vector<boost::thread> tasks;
  for (unsigned int idCopyValInThread = 0; idCopyValInThread < coreCount; ++idCopyValInThread) {
    tasks.push_back (boost::thread ([/*no ref!!!*/idCopyValInThread, &maxIds, &functThreadMinMax] () {
	  functThreadMinMax (idCopyValInThread, maxIds[idCopyValInThread], maxIds[idCopyValInThread+1]);
	}));
  }
  for (unsigned int i = 0; i < coreCount; ++i)
    tasks[i].join ();
#endif
}

// ========================================
template <typename DimImg, class OutputIterator, class T>
inline void dealThreadFill_n (const DimImg &maxId, unsigned int coreCount, OutputIterator first, const T& val) {
  dealThreadBound (maxId, coreCount, [&first, &val] (const DimImg &minVal, const DimImg &maxVal) {
      fill_n (first+minVal, maxVal-minVal, val);
    });
}

// ========================================
template<typename DimImg, typename WeightT, typename WeightFunct, typename CmpFunct, typename CallFunct>
inline void
callOnSortedSets (const std::vector<DimImg> &sizes,
		  const WeightFunct &getWeight/* getWeight (vectId, itemId) */,
		  CmpFunct isWeightInf/* isWeightInf (w1, w2) */,
		  const CallFunct &callIdId/* callIdId (vectId, itemId) */) {
  DimImg size = sizes.size ();
  DEF_LOG ("callOnSortedSets", "size:" << size);
  if (!size)
    return;
  std::vector<DimImg> vectIds (size, 0);
  std::vector<DimImg> vectCounts (sizes);
  // get min
  bool found = false;
  DimImg minVectIdx = 0;
  WeightT maxWeight = 0;
  for (DimImg vectId = 0; vectId < size; ++vectId) {
    if (!vectCounts [vectId])
      continue;
    WeightT tmpWeight = getWeight (vectId, 0);
    if (found && !isWeightInf (tmpWeight, maxWeight))
      continue;
    minVectIdx = vectId;
    maxWeight = tmpWeight;
    found = true;
  }
  LOG ("found:" << found << " minVectIdx:" << minVectIdx << " maxWeight:" << maxWeight);
  // loop
  for ( ; found; ) {
    // get next min
    found = false;
    DimImg nextMinVectIdx = 0;
    WeightT nextMaxWeight = 0;
    for (DimImg vectId = minVectIdx; ; ) {
      if (vectCounts [vectId]) {
	WeightT tmpWeight = getWeight (vectId, vectIds [vectId]);
	if (!isWeightInf (maxWeight, tmpWeight)) {
	  // maxWeight == tmpWeight
	  callIdId (vectId, vectIds [vectId]);
	  ++vectIds [vectId];
	  --vectCounts [vectId];
	  continue;
	}
	if (!found || isWeightInf (tmpWeight, nextMaxWeight)) {
	  nextMinVectIdx = vectId;
	  nextMaxWeight = tmpWeight;
	  found = true;
	}
      }
      vectId = (vectId+1)%size;
      if (vectId == minVectIdx)
	break;
    }
    minVectIdx = nextMinVectIdx;
    maxWeight = nextMaxWeight;
  }
}

#endif // _TRISKELE_DEAL_THREADS_TPP