【无标题】

template<typename FPType, bool do_prefetch, typename BinIdxType>
void BuildHistDenseKernel(const std::vector<GradientPair>& gpair,
                          const RowSetCollection::Elem row_indices,
                          const GHistIndexMatrix& gmat,
                          const size_t n_features,
                          GHistRow hist)
 {
  const size_t size = row_indices.Size();
  const size_t* rid = row_indices.begin;
  const float* pgh = reinterpret_cast<const float*>(gpair.data());
  const BinIdxType* gradient_index = gmat.index.data<BinIdxType>();//返回指向binidxtype类型的指针，默认为data_ptr_
  const uint32_t* offsets = gmat.index.Offset();
  FPType* hist_data = reinterpret_cast<FPType*>(hist.data());
  const uint32_t two {2};  // Each element from 'gpair' and 'hist' contains
                           // 2 FP values: gradient and hessian.
                           // So we need to multiply each row-index/bin-index by 2
                           // to work with gradient pairs as a singe row FP array

  for (size_t i = 0; i < size; ++i) {
    const size_t icol_start = rid[i] * n_features;
    const size_t idx_gh = two * rid[i];

    if (do_prefetch) {
      const size_t icol_start_prefetch = rid[i + Prefetch::kPrefetchOffset] * n_features;

      PREFETCH_READ_T0(pgh + two * rid[i + Prefetch::kPrefetchOffset]);
      for (size_t j = icol_start_prefetch; j < icol_start_prefetch + n_features;
           j += Prefetch::GetPrefetchStep<BinIdxType>()) {
        PREFETCH_READ_T0(gradient_index + j);
      }
    }
    const BinIdxType* gr_index_local = gradient_index + icol_start;
    for (size_t j = 0; j < n_features; ++j) {
      const uint32_t idx_bin = two * (static_cast<uint32_t>(gr_index_local[j]) +
                                      offsets[j]);

      hist_data[idx_bin]   += pgh[idx_gh];
      hist_data[idx_bin+1] += pgh[idx_gh+1];
    }
  }
}                         

函数模板参数分析

FPTtype//传入数据类型FPTtype：猜测是表示特征的类似，在spark的接口里传入的类型是double

\\在文件夹Spark-ml-algo-lib中
void BuildHistDenseKernel<double, false, uint8_t>(const std::vector<GradientPair>& gpair,
                                                  const RowSetCollection::Elem row_indices,
                                                  const GHistIndexMatrix& gmat,
                                                  const size_t n_features,
                                                  GHistRow hist)

• 传入Bool类型的数据 do_prefetch//判断是否提前抓取;

• 传入数据类型 BinIdexType//桶的下标数据类型，

函数参数分析

（const都是在BuildHistDenseKernel中保持不变的参数）

• const std::vector<GradientPair>& gpair//类型为GradientPair类型的向量容器
  其中using xgboost::GradientPair = typedef detail::GradientPairInternal<float> //存储grad和hess
  参考链接：RowSetCollection
• const RowSetCollection::Elem row_indices
  Elem：data structure to store an instance set, a subset of rows (instances) associated with a particular node in a decision tree
  存储实例集合的数据结构，实例的子集与决策树的特定节点有关
  参考链接：RowSetCollection
• const GHistIndexMatrix& gmat
  GHistIndexMatrix ：brief preprocessed global index matrix, in CSR format
  Transform floating values to integer index in histogram This is a global histogram index for CPU histogram. On GPU ellpack page is used.
  将全局的索引矩阵，以稀疏矩阵的格式存储，在直方图上将浮点数数据转化为整型的索引，在cpu直方图上，这是一个全局的直方图索引。
• const size_t n_features
  size_t 无符号的整型
• GHistRow hist
  /*brief histogram of gradient statistics for a single node.
  Consists of multiple GradStats, each entry showing total gradient statistics
  for that particular bin
  Uses global bin id so as to represent all features simultaneously
  */
  单个节点的梯度统计信息的简要直方图。由多个GradStats组成，对于该特定桶，每个条目显示总梯度统计数据使用全局 bin ID，以便同时表示所有特征。
  其中using xgboost::common::GHistRow = typedef Span<xgboost::GradientPairPrecise>
  参考链接：span