Skip to content

Optimize column reader by bitmap counting definition levels #48732

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
aryan-cloud wants to merge 1 commit into
base: main
Choose a base branch
from
+22 −3
Open

Optimize column reader by bitmap counting definition levels #48732

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 22 additions & 3 deletions cpp/src/parquet/column_reader.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,6 +93,27 @@ inline void CheckNumberDecoded(int64_t number_decoded, int64_t expected) {
constexpr std::string_view kErrorRepDefLevelNotMatchesNumValues =
"Number of decoded rep / def levels do not match num_values in page header";

inline int64_t CountMaxDefLevels(const int16_t* levels, int64_t num_levels,
int16_t max_def_level) {
if (num_levels <= 0) {
return 0;
}
const int16_t rhs = static_cast<int16_t>(max_def_level - 1);
int64_t count = 0;
int64_t offset = 0;
for (; offset + 64 <= num_levels; offset += 64) {
const uint64_t bitmap =
internal::GreaterThanBitmap(levels + offset, 64, rhs);
count += static_cast<int64_t>(bit_util::PopCount(bitmap));
}
if (offset < num_levels) {
const uint64_t bitmap =
internal::GreaterThanBitmap(levels + offset, num_levels - offset, rhs);
count += static_cast<int64_t>(bit_util::PopCount(bitmap));
}
Comment on lines +103 to +113
Copy link
Contributor

@HuaHuaY HuaHuaY Jan 9, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
int64_t offset = 0;
for (; offset + 64 <= num_levels; offset += 64) {
const uint64_t bitmap =
internal::GreaterThanBitmap(levels + offset, 64, rhs);
count += static_cast<int64_t>(bit_util::PopCount(bitmap));
}
if (offset < num_levels) {
const uint64_t bitmap =
internal::GreaterThanBitmap(levels + offset, num_levels - offset, rhs);
count += static_cast<int64_t>(bit_util::PopCount(bitmap));
}
for (int64_t offset = 0; offset < num_levels; offset += 64) {
const int64_t chunk_size = std::min<int64_t>(64, num_levels - offset);
const uint64_t bitmap =
internal::GreaterThanBitmap(levels + offset, static_cast<int>(chunk_size), rhs);
count += static_cast<int64_t>(bit_util::PopCount(bitmap));
}

I suggest simplifying the two loops into one.

return count;
}

} // namespace

LevelDecoder::LevelDecoder() : num_values_remaining_(0) {}
Expand Down Expand Up @@ -1010,10 +1031,8 @@ class TypedColumnReaderImpl : public TypedColumnReader<DType>,
if (ARROW_PREDICT_FALSE(*num_def_levels != batch_size)) {
throw ParquetException(kErrorRepDefLevelNotMatchesNumValues);
}
// TODO(wesm): this tallying of values-to-decode can be performed with better
// cache-efficiency if fused with the level decoding.
*non_null_values_to_read +=
std::count(def_levels, def_levels + *num_def_levels, this->max_def_level_);
CountMaxDefLevels(def_levels, *num_def_levels, this->max_def_level_);
} else {
// Required field, read all values
if (num_def_levels != nullptr) {
Expand Down