Lower RAM implementation of slice_columns for BRWT#226
Draft
hmusta wants to merge 2 commits into
Draft
Conversation
DBGMode used in build_anno_graph
karasikov
requested changes
Nov 13, 2020
karasikov
left a comment
karasikov
left a comment
Member
There was a problem hiding this comment.
What are the results of the benchmarks?
| const ColumnCallback &callback) const { | ||
| uint64_t nrows = num_rows(); | ||
| sdsl::bit_vector code_column(reduced_matrix_.num_rows()); | ||
| reduced_matrix_.slice_columns(columns, [&](Column j, bitmap&& rows) { |
Member
There was a problem hiding this comment.
Suggested change
| reduced_matrix_.slice_columns(columns, [&](Column j, bitmap&& rows) { | |
| reduced_matrix_.slice_columns(columns, [&](Column j, bitmap&& reduced_column) { |
| rows.add_to(&code_column); | ||
|
|
||
| callback(j, bitmap_generator([&](const auto &index_callback) { | ||
| for (uint64_t i = 0; i < nrows; ++i) { |
Member
There was a problem hiding this comment.
This will take forever. Make it parallel
Suggested change
| for (uint64_t i = 0; i < nrows; ++i) { | |
| #pragma parallel num_threads(get_num_threads()) | |
| for (uint64_t i = 0; i < nrows; ++i) { |
Comment on lines
+43
to
+44
| void slice_columns(const std::vector<Column> &columns, | ||
| const ColumnCallback &callback) const override; |
Comment on lines
+196
to
+197
| void BRWT::slice_columns(const std::vector<Column> &column_ids, | ||
| const ColumnCallback &callback) const { |
| if (column_ids.empty()) | ||
| return; | ||
|
|
||
| auto num_nonzero_rows = nonzero_rows_->num_set_bits(); |
Member
There was a problem hiding this comment.
Suggested change
| auto num_nonzero_rows = nonzero_rows_->num_set_bits(); | |
| uint64_t num_nonzero_rows = nonzero_rows_->num_set_bits(); |
Comment on lines
+203
to
+205
| // check if the column is empty | ||
| if (!num_nonzero_rows) | ||
| return; |
Member
There was a problem hiding this comment.
Even if they are empty, you still need to call them. Add unit tests?
| if (!child_nodes_.size()) { | ||
| // return the index column | ||
| for (size_t k = 0; k < column_ids.size(); ++k) { | ||
| callback(column_ids[k], std::move(*nonzero_rows_->copy())); |
Member
There was a problem hiding this comment.
Better call a const reference, so the column can be copied by the caller if it's needed, and otherwise, there is no overhead.
Suggested change
| callback(column_ids[k], std::move(*nonzero_rows_->copy())); | |
| callback(column_ids[k], *nonzero_rows_); |
Comment on lines
+210
to
+211
| for (size_t k = 0; k < column_ids.size(); ++k) { | ||
| callback(column_ids[k], std::move(*nonzero_rows_->copy())); |
Member
There was a problem hiding this comment.
Why not range-based loop?
for (size_t col_id : column_ids) {
...
}
Comment on lines
+217
to
+263
| tsl::hopscotch_map<uint32_t, std::vector<Column>> child_columns_map; | ||
| for (size_t i = 0; i < column_ids.size(); ++i) { | ||
| assert(column_ids[i] < num_columns()); | ||
| auto child_node = assignments_.group(column_ids[i]); | ||
| auto child_column = assignments_.rank(column_ids[i]); | ||
|
|
||
| auto it = child_columns_map.find(child_node); | ||
| if (it == child_columns_map.end()) | ||
| it = child_columns_map.emplace(child_node, std::vector<Column>{}).first; | ||
|
|
||
| it.value().push_back(child_column); | ||
| } | ||
|
|
||
| auto process = [&](auto child_node, auto *child_columns_ptr) { | ||
| if (num_nonzero_rows == nonzero_rows_->size()) { | ||
| child_nodes_[child_node]->slice_columns(*child_columns_ptr, | ||
| [&](Column j, bitmap&& rows) { | ||
| callback(assignments_.get(child_node, j), std::move(rows)); | ||
| } | ||
| ); | ||
| } else { | ||
| const BRWT *child_node_brwt = dynamic_cast<const BRWT*>( | ||
| child_nodes_[child_node].get() | ||
| ); | ||
| if (child_node_brwt | ||
| && child_columns_ptr->size() > 1 | ||
| && !child_node_brwt->child_nodes_.size()) { | ||
| // if there are multiple column ids corresponding to the same leaf | ||
| // node, then this branch avoids doing redundant select1 calls | ||
| const auto *nonzero_rows = child_node_brwt->nonzero_rows_.get(); | ||
| size_t num_nonzero_rows = nonzero_rows->num_set_bits(); | ||
| if (num_nonzero_rows) { | ||
| std::vector<uint64_t> set_bits; | ||
| set_bits.reserve(num_nonzero_rows); | ||
| nonzero_rows->call_ones([&](auto i) { | ||
| set_bits.push_back(nonzero_rows->select1(i + 1)); | ||
| }); | ||
|
|
||
| for (size_t k = 0; k < child_columns_ptr->size() - 1; ++k) { | ||
| callback(assignments_.get(child_node, (*child_columns_ptr)[k]), | ||
| bitmap_generator(std::move(set_bits), num_rows())); | ||
| } | ||
|
|
||
| callback(assignments_.get(child_node, child_columns_ptr->back()), | ||
| bitmap_generator(std::move(set_bits), num_rows())); | ||
| } | ||
| } else { |
Member
There was a problem hiding this comment.
Could you add some comments to explain why this is going to make things faster than the basic call?
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.