etseidl commented on code in PR #11285:
URL: https://github.com/apache/arrow-rs/pull/11285#discussion_r4148813298
##########
parquet/benches/arrow_statistics.rs:
##########
@@ -347,33 +342,18 @@ fn page_index_benchmark(c: &mut Criterion) {
(num_pages, &data[range.start as usize..range.end as usize])
})
.collect();
- let row_group_indices: Vec<usize> = (0..row_groups).collect();
- let num_columns = reader.parquet_schema().num_columns();
let mut group = c.benchmark_group(format!(
"Decode page index statistics for {data_type} ({} pages)",
column_indexes.iter().map(|(n, _)| n).sum::<usize>()
));
group.bench_function("full page index", |b| {
b.iter(|| {
- let mut builder = PageIndexBuilder::new(row_groups,
num_columns);
- for (rg, (_, bytes)) in column_indexes.iter().enumerate() {
- let index = decode_column_index(bytes,
physical_type).unwrap();
- builder.put_column_index(index, rg, column);
- }
- let page_index = builder.build();
- let _ = converter
- .data_page_mins(&page_index, &row_group_indices)
- .unwrap();
- let _ = converter
- .data_page_maxes(&page_index, &row_group_indices)
- .unwrap();
- let _ = converter
- .data_page_null_counts(&page_index, &row_group_indices)
- .unwrap();
- let _ = converter
- .data_page_nan_counts(&page_index, &row_group_indices)
- .unwrap();
+ converter
Review Comment:
I agree this is the comparison we want to make for this PR, but it does
change the meaning of the bench, making it hard to measure any gains we might
make in the original path. How about adding a third set of benches that
measure the performance of this new work. I know you can't run critcmp then,
but I think seeing the old and new in a single run suffices to prove impact.
--
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]
