fix refactoring bug in voting parallel (microsoft#3089)

src/treelearner/feature_histogram.hpp

@@ -1147,32 +1147,15 @@ class HistogramPool {
  }
  }
 
- void DynamicChangeSize(const Dataset* train_data, bool is_hist_colwise,
- const Config* config, int cache_size, int total_size) {
- if (feature_metas_.empty()) {
- SetFeatureInfo<true, true>(train_data, config, &feature_metas_);
- uint64_t bin_cnt_over_features = 0;
- for (int i = 0; i < train_data->num_features(); ++i) {
- bin_cnt_over_features +=
- static_cast<uint64_t>(feature_metas_[i].num_bin);
- }
- Log::Info("Total Bins %d", bin_cnt_over_features);
- }
- int old_cache_size = static_cast<int>(pool_.size());
- Reset(cache_size, total_size);
-
- if (cache_size > old_cache_size) {
- pool_.resize(cache_size);
- data_.resize(cache_size);
- }
+ static int GetNumTotalHistogramBins(const Dataset* train_data,
+ bool is_hist_colwise, std::vector<int>* offsets) {
  int num_total_bin = static_cast<int>(train_data->NumTotalBin());
-
- std::vector<int> offsets;
+ offsets->clear();
  if (is_hist_colwise) {
  int offset = 0;
  for (int j = 0; j < train_data->num_features(); ++j) {
  offset += train_data->SubFeatureBinOffset(j);
- offsets.push_back(offset);
+ offsets->push_back(offset);
  auto num_bin = train_data->FeatureNumBin(j);
  if (train_data->FeatureBinMapper(j)->GetMostFreqBin() == 0) {
  num_bin -= 1;
@@ -1182,13 +1165,37 @@ class HistogramPool {
  } else {
  num_total_bin = 1;
  for (int j = 0; j < train_data->num_features(); ++j) {
- offsets.push_back(num_total_bin);
+ offsets->push_back(num_total_bin);
  num_total_bin += train_data->FeatureBinMapper(j)->num_bin();
  if (train_data->FeatureBinMapper(j)->GetMostFreqBin() == 0) {
  num_total_bin -= 1;
  }
  }
  }
+ return num_total_bin;
+ }
+
+ void DynamicChangeSize(const Dataset* train_data, bool is_hist_colwise,
+ const Config* config, int cache_size, int total_size) {
+ if (feature_metas_.empty()) {
+ SetFeatureInfo<true, true>(train_data, config, &feature_metas_);
+ uint64_t bin_cnt_over_features = 0;
+ for (int i = 0; i < train_data->num_features(); ++i) {
+ bin_cnt_over_features +=
+ static_cast<uint64_t>(feature_metas_[i].num_bin);
+ }
+ Log::Info("Total Bins %d", bin_cnt_over_features);
+ }
+ int old_cache_size = static_cast<int>(pool_.size());
+ Reset(cache_size, total_size);
+
+ if (cache_size > old_cache_size) {
+ pool_.resize(cache_size);
+ data_.resize(cache_size);
+ }
+ std::vector<int> offsets;
+ int num_total_bin =
+ this->GetNumTotalHistogramBins(train_data, is_hist_colwise, &offsets);
  OMP_INIT_EX();
 #pragma omp parallel for schedule(static)
  for (int i = old_cache_size; i < cache_size; ++i) {

src/treelearner/voting_parallel_tree_learner.cpp

@@ -63,20 +63,15 @@ void VotingParallelTreeLearner<TREELEARNER_T>::Init(const Dataset* train_data, b
  // initialize histograms for global
  smaller_leaf_histogram_array_global_.reset(new FeatureHistogram[this->num_features_]);
  larger_leaf_histogram_array_global_.reset(new FeatureHistogram[this->num_features_]);
- auto num_total_bin = train_data->NumTotalBin();
- smaller_leaf_histogram_data_.resize(num_total_bin);
- larger_leaf_histogram_data_.resize(num_total_bin);
+ std::vector<int> offsets;
+ int num_total_bin = HistogramPool::GetNumTotalHistogramBins(
+ train_data, this->share_state_->is_colwise, &offsets);
+ smaller_leaf_histogram_data_.resize(num_total_bin * 2);
+ larger_leaf_histogram_data_.resize(num_total_bin * 2);
  HistogramPool::SetFeatureInfo<true, true>(train_data, this->config_, &feature_metas_);
- uint64_t offset = 0;
  for (int j = 0; j < train_data->num_features(); ++j) {
- offset += static_cast<uint64_t>(train_data->SubFeatureBinOffset(j));
- smaller_leaf_histogram_array_global_[j].Init(smaller_leaf_histogram_data_.data() + offset, &feature_metas_[j]);
- larger_leaf_histogram_array_global_[j].Init(larger_leaf_histogram_data_.data() + offset, &feature_metas_[j]);
- auto num_bin = train_data->FeatureNumBin(j);
- if (train_data->FeatureBinMapper(j)->GetMostFreqBin() == 0) {
- num_bin -= 1;
- }
- offset += static_cast<uint64_t>(num_bin);
+ smaller_leaf_histogram_array_global_[j].Init(smaller_leaf_histogram_data_.data() + offsets[j] * 2, &feature_metas_[j]);
+ larger_leaf_histogram_array_global_[j].Init(larger_leaf_histogram_data_.data() + offsets[j] * 2, &feature_metas_[j]);
  }
 }