tustvold commented on a change in pull request #1228:
URL: https://github.com/apache/arrow-rs/pull/1228#discussion_r790286234
##########
File path: parquet/src/arrow/record_reader/definition_levels.rs
##########
@@ -353,34 +353,27 @@ impl PackedDecoder {
/// Counts the number of set bits in the provided range
pub fn count_set_bits(bytes: &[u8], range: Range<usize>) -> usize {
- let mut count = 0_usize;
- let chunks = BitChunks::new(bytes, range.start, range.end - range.start);
- chunks.iter().for_each(|chunk| {
- count += chunk.count_ones() as usize;
- });
- count += chunks.remainder_bits().count_ones() as usize;
- count
+ let unaligned = UnalignedBitChunk::new(bytes, range.start, range.end -
range.start);
+ unaligned.count_ones()
}
fn iter_set_bits_rev(bytes: &[u8]) -> impl Iterator<Item = usize> + '_ {
- let (mut byte_idx, mut in_progress) = match bytes.len() {
- 0 => (0, 0),
- len => (len - 1, bytes[len - 1]),
- };
-
- std::iter::from_fn(move || loop {
- if in_progress != 0 {
- let bit_pos = 7 - in_progress.leading_zeros();
- in_progress ^= 1 << bit_pos;
- return Some((byte_idx << 3) + (bit_pos as usize));
- }
-
- if byte_idx == 0 {
+ let bit_length = bytes.len() * 8;
Review comment:
This wasn't ever really a bottleneck in the parquet parsing, but still
sees a slight aggregate improvement of 2-3%. Updated for curiosity more than
necessity.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
