GH-45201: [C++][Parquet] Improve performance of generating size statistics #45202

pitrou · 2025-01-08T11:28:42Z

Rationale for this change

We found out in #45085 that there is a non-trivial overhead when writing size statistics is enabled.

What changes are included in this PR?

Dramatically reduce overhead by speeding up def/rep levels histogram updates.

Performance results on the author's machine:

------------------------------------------------------------------------------------------------------------------------------------------------
Benchmark                                                                                      Time             CPU   Iterations UserCounters...
------------------------------------------------------------------------------------------------------------------------------------------------
BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::Int64Type>                   8103053 ns      8098569 ns           86 bytes_per_second=1003.26Mi/s items_per_second=129.477M/s output_size=537.472k page_index_size=33
BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type>            8153499 ns      8148492 ns           86 bytes_per_second=997.117Mi/s items_per_second=128.683M/s output_size=537.488k page_index_size=33
BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type>     8212560 ns      8207754 ns           83 bytes_per_second=989.918Mi/s items_per_second=127.754M/s output_size=537.502k page_index_size=47

BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::StringType>                 10405020 ns     10400775 ns           67 bytes_per_second=444.142Mi/s items_per_second=100.817M/s output_size=848.305k page_index_size=34
BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType>          10464784 ns     10460778 ns           66 bytes_per_second=441.594Mi/s items_per_second=100.239M/s output_size=848.325k page_index_size=34
BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType>   10469832 ns     10465739 ns           67 bytes_per_second=441.385Mi/s items_per_second=100.191M/s output_size=848.344k page_index_size=48

BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::Int64Type>                       13004962 ns     12992678 ns           52 bytes_per_second=657.101Mi/s items_per_second=80.7052M/s output_size=617.464k page_index_size=34
BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type>                13718352 ns     13705599 ns           50 bytes_per_second=622.921Mi/s items_per_second=76.5071M/s output_size=617.486k page_index_size=34
BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type>         13845553 ns     13832138 ns           52 bytes_per_second=617.222Mi/s items_per_second=75.8072M/s output_size=617.506k page_index_size=54

BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::StringType>                      15715263 ns     15702707 ns           44 bytes_per_second=320.449Mi/s items_per_second=66.7768M/s output_size=927.326k page_index_size=35
BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType>               16507328 ns     16493800 ns           43 bytes_per_second=305.079Mi/s items_per_second=63.5739M/s output_size=927.352k page_index_size=35
BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType>        16575359 ns     16561311 ns           42 bytes_per_second=303.836Mi/s items_per_second=63.3148M/s output_size=927.377k page_index_size=55

Performance results without this PR:

------------------------------------------------------------------------------------------------------------------------------------------------
Benchmark                                                                                      Time             CPU   Iterations UserCounters...
------------------------------------------------------------------------------------------------------------------------------------------------
BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::Int64Type>                   8042576 ns      8037678 ns           87 bytes_per_second=1010.86Mi/s items_per_second=130.458M/s output_size=537.472k page_index_size=33
BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type>            9576627 ns      9571279 ns           73 bytes_per_second=848.894Mi/s items_per_second=109.554M/s output_size=537.488k page_index_size=33
BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type>     9570204 ns      9563595 ns           73 bytes_per_second=849.576Mi/s items_per_second=109.642M/s output_size=537.502k page_index_size=47

BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::StringType>                 10165397 ns     10160868 ns           69 bytes_per_second=454.628Mi/s items_per_second=103.197M/s output_size=848.305k page_index_size=34
BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType>          11662568 ns     11657396 ns           60 bytes_per_second=396.265Mi/s items_per_second=89.9494M/s output_size=848.325k page_index_size=34
BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType>   11657135 ns     11653063 ns           60 bytes_per_second=396.412Mi/s items_per_second=89.9829M/s output_size=848.344k page_index_size=48

BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::Int64Type>                       13182006 ns     13168704 ns           51 bytes_per_second=648.318Mi/s items_per_second=79.6264M/s output_size=617.464k page_index_size=34
BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type>                16438205 ns     16421762 ns           43 bytes_per_second=519.89Mi/s items_per_second=63.8528M/s output_size=617.486k page_index_size=34
BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type>         16424615 ns     16409032 ns           42 bytes_per_second=520.293Mi/s items_per_second=63.9024M/s output_size=617.506k page_index_size=54

BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::StringType>                      15387808 ns     15373086 ns           46 bytes_per_second=327.32Mi/s items_per_second=68.2086M/s output_size=927.326k page_index_size=35
BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType>               18319628 ns     18302938 ns           37 bytes_per_second=274.924Mi/s items_per_second=57.29M/s output_size=927.352k page_index_size=35
BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType>        18346665 ns     18329336 ns           37 bytes_per_second=274.528Mi/s items_per_second=57.2075M/s output_size=927.377k page_index_size=55

Are these changes tested?

Tested by existing tests, validated by existing benchmarks.

Are there any user-facing changes?

No.

GitHub Issue: [C++][Parquet] Improve performance of writing size statistics #45201

pitrou · 2025-01-08T11:29:02Z

cc @wgtmac

pitrou · 2025-01-08T11:29:13Z

@ursabot please benchmark lang=C++

github-actions · 2025-01-08T11:29:15Z

⚠️ GitHub issue #45201 has been automatically assigned in GitHub to PR creator.

pitrou · 2025-01-08T11:41:14Z

Oops, looks like the Continuous Benchmarking site is completely broken now.

conbench-apache-arrow · 2025-01-09T01:09:06Z

Thanks for your patience. Conbench analyzed the 2 benchmarking runs that have been run so far on PR commit e67472b.

There weren't enough matching historic benchmark results to make a call on whether there were regressions.

The full Conbench report has more details.

cpp/src/parquet/size_statistics.cc

mapleFU

General LGTM!

mapleFU · 2025-01-09T03:00:14Z

cpp/src/parquet/column_writer.cc

                      const int16_t* rep_levels) {
+    // Update histograms now, to maximize cache efficiency.


Don't fully understand why scaning a pass of def_levels and rep_levels first would maximize cache efficiency, would WriteDefinitionLevels first also update the L1D cache?

I'm not sure and I don't see any difference on our mini-benchmark. But WriteDefinitionLevels also writes to a destination buffer, so perhaps some of the levels will be evicted from the CPU cache.

pitrou · 2025-01-09T10:28:36Z

Do you want to take a look at the updated code @wgtmac @mapleFU ?

mapleFU · 2025-01-09T10:30:51Z

Does the benchmark result changed after the new code patch?

pitrou · 2025-01-09T10:31:43Z

Yes, it's even better, see the new numbers in the PR description.

wgtmac

LGTM. Thanks!

mapleFU · 2025-01-09T13:22:50Z

cpp/src/parquet/size_statistics.cc

+void UpdateLevelHistogram(::arrow::util::span<const int16_t> levels,
+                          ::arrow::util::span<int64_t> histogram) {
+  const int64_t num_levels = static_cast<int64_t>(levels.size());
+  const int16_t max_level = static_cast<int16_t>(histogram.size() - 1);


(Should we DCHECK_GE(histogram.size(), 1)?)

pitrou · 2025-01-09T13:50:51Z

@github-actions crossbow submit -g cpp

github-actions · 2025-01-09T13:53:22Z

Revision: 9f9bc94

Submitted crossbow builds: ursacomputing/crossbow @ actions-f6939022a3

Task	Status
example-cpp-minimal-build-static
example-cpp-minimal-build-static-system-dependency
example-cpp-tutorial
test-alpine-linux-cpp
test-build-cpp-fuzz
test-conda-cpp
test-conda-cpp-valgrind
test-cuda-cpp-ubuntu-20.04-cuda-11.2.2
test-cuda-cpp-ubuntu-22.04-cuda-11.7.1
test-debian-12-cpp-amd64
test-debian-12-cpp-i386
test-fedora-39-cpp
test-ubuntu-20.04-cpp
test-ubuntu-20.04-cpp-bundled
test-ubuntu-22.04-cpp
test-ubuntu-22.04-cpp-20
test-ubuntu-22.04-cpp-emscripten
test-ubuntu-22.04-cpp-no-threading
test-ubuntu-24.04-cpp
test-ubuntu-24.04-cpp-bundled-offline
test-ubuntu-24.04-cpp-gcc-13-bundled
test-ubuntu-24.04-cpp-gcc-14
test-ubuntu-24.04-cpp-minimal-with-formats
test-ubuntu-24.04-cpp-thread-sanitizer

pitrou · 2025-01-09T14:45:29Z

The Valgrind error is unrelated.

conbench-apache-arrow · 2025-01-15T07:41:05Z

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 2b5f56c.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 4 possible false positives for unstable benchmarks that are known to sometimes produce them.

…stics (#45202) We found out in #45085 that there is a non-trivial overhead when writing size statistics is enabled. Dramatically reduce overhead by speeding up def/rep levels histogram updates. Performance results on the author's machine: ``` ------------------------------------------------------------------------------------------------------------------------------------------------ Benchmark Time CPU Iterations UserCounters... ------------------------------------------------------------------------------------------------------------------------------------------------ BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::Int64Type> 8103053 ns 8098569 ns 86 bytes_per_second=1003.26Mi/s items_per_second=129.477M/s output_size=537.472k page_index_size=33 BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type> 8153499 ns 8148492 ns 86 bytes_per_second=997.117Mi/s items_per_second=128.683M/s output_size=537.488k page_index_size=33 BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type> 8212560 ns 8207754 ns 83 bytes_per_second=989.918Mi/s items_per_second=127.754M/s output_size=537.502k page_index_size=47 BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::StringType> 10405020 ns 10400775 ns 67 bytes_per_second=444.142Mi/s items_per_second=100.817M/s output_size=848.305k page_index_size=34 BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType> 10464784 ns 10460778 ns 66 bytes_per_second=441.594Mi/s items_per_second=100.239M/s output_size=848.325k page_index_size=34 BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType> 10469832 ns 10465739 ns 67 bytes_per_second=441.385Mi/s items_per_second=100.191M/s output_size=848.344k page_index_size=48 BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::Int64Type> 13004962 ns 12992678 ns 52 bytes_per_second=657.101Mi/s items_per_second=80.7052M/s output_size=617.464k page_index_size=34 BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type> 13718352 ns 13705599 ns 50 bytes_per_second=622.921Mi/s items_per_second=76.5071M/s output_size=617.486k page_index_size=34 BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type> 13845553 ns 13832138 ns 52 bytes_per_second=617.222Mi/s items_per_second=75.8072M/s output_size=617.506k page_index_size=54 BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::StringType> 15715263 ns 15702707 ns 44 bytes_per_second=320.449Mi/s items_per_second=66.7768M/s output_size=927.326k page_index_size=35 BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType> 16507328 ns 16493800 ns 43 bytes_per_second=305.079Mi/s items_per_second=63.5739M/s output_size=927.352k page_index_size=35 BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType> 16575359 ns 16561311 ns 42 bytes_per_second=303.836Mi/s items_per_second=63.3148M/s output_size=927.377k page_index_size=55 ``` Performance results without this PR: ``` ------------------------------------------------------------------------------------------------------------------------------------------------ Benchmark Time CPU Iterations UserCounters... ------------------------------------------------------------------------------------------------------------------------------------------------ BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::Int64Type> 8042576 ns 8037678 ns 87 bytes_per_second=1010.86Mi/s items_per_second=130.458M/s output_size=537.472k page_index_size=33 BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type> 9576627 ns 9571279 ns 73 bytes_per_second=848.894Mi/s items_per_second=109.554M/s output_size=537.488k page_index_size=33 BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type> 9570204 ns 9563595 ns 73 bytes_per_second=849.576Mi/s items_per_second=109.642M/s output_size=537.502k page_index_size=47 BM_WritePrimitiveColumn<SizeStatisticsLevel::None, ::arrow::StringType> 10165397 ns 10160868 ns 69 bytes_per_second=454.628Mi/s items_per_second=103.197M/s output_size=848.305k page_index_size=34 BM_WritePrimitiveColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType> 11662568 ns 11657396 ns 60 bytes_per_second=396.265Mi/s items_per_second=89.9494M/s output_size=848.325k page_index_size=34 BM_WritePrimitiveColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType> 11657135 ns 11653063 ns 60 bytes_per_second=396.412Mi/s items_per_second=89.9829M/s output_size=848.344k page_index_size=48 BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::Int64Type> 13182006 ns 13168704 ns 51 bytes_per_second=648.318Mi/s items_per_second=79.6264M/s output_size=617.464k page_index_size=34 BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::Int64Type> 16438205 ns 16421762 ns 43 bytes_per_second=519.89Mi/s items_per_second=63.8528M/s output_size=617.486k page_index_size=34 BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::Int64Type> 16424615 ns 16409032 ns 42 bytes_per_second=520.293Mi/s items_per_second=63.9024M/s output_size=617.506k page_index_size=54 BM_WriteListColumn<SizeStatisticsLevel::None, ::arrow::StringType> 15387808 ns 15373086 ns 46 bytes_per_second=327.32Mi/s items_per_second=68.2086M/s output_size=927.326k page_index_size=35 BM_WriteListColumn<SizeStatisticsLevel::ColumnChunk, ::arrow::StringType> 18319628 ns 18302938 ns 37 bytes_per_second=274.924Mi/s items_per_second=57.29M/s output_size=927.352k page_index_size=35 BM_WriteListColumn<SizeStatisticsLevel::PageAndColumnChunk, ::arrow::StringType> 18346665 ns 18329336 ns 37 bytes_per_second=274.528Mi/s items_per_second=57.2075M/s output_size=927.377k page_index_size=55 ``` Tested by existing tests, validated by existing benchmarks. No. * GitHub Issue: #45201 Authored-by: Antoine Pitrou <antoine@python.org> Signed-off-by: Antoine Pitrou <antoine@python.org>

github-actions bot added awaiting review Awaiting review Component: Parquet Component: C++ and removed awaiting review Awaiting review labels Jan 8, 2025

[C++][Parquet] Improve performance of generating size statistics

8a4ffbc

pitrou force-pushed the faster-size-statistics branch from e67472b to 8a4ffbc Compare January 8, 2025 11:38

github-actions bot added the awaiting review Awaiting review label Jan 8, 2025

Add tests

2a82065

pitrou marked this pull request as ready for review January 8, 2025 16:46

pitrou requested a review from wgtmac as a code owner January 8, 2025 16:46

wgtmac reviewed Jan 9, 2025

View reviewed changes

cpp/src/parquet/size_statistics.cc Outdated Show resolved Hide resolved

github-actions bot added awaiting committer review Awaiting committer review and removed awaiting review Awaiting review labels Jan 9, 2025

mapleFU approved these changes Jan 9, 2025

View reviewed changes

Make the fast path faster

b05549e

wgtmac approved these changes Jan 9, 2025

View reviewed changes

mapleFU approved these changes Jan 9, 2025

View reviewed changes

Add debug check suggestion

9f9bc94

pitrou merged commit 2b5f56c into apache:main Jan 9, 2025
34 checks passed

pitrou removed the awaiting committer review Awaiting committer review label Jan 9, 2025

pitrou mentioned this pull request Jan 9, 2025

[C++][Parquet] Improve performance of writing size statistics #45201

Closed

pitrou deleted the faster-size-statistics branch January 9, 2025 15:01

pitrou mentioned this pull request Jan 11, 2025

[C++][Parquet] Enable size statistics by default #45227

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

GH-45201: [C++][Parquet] Improve performance of generating size statistics #45202

GH-45201: [C++][Parquet] Improve performance of generating size statistics #45202

pitrou commented Jan 8, 2025 •

edited

Loading

pitrou commented Jan 8, 2025

pitrou commented Jan 8, 2025

github-actions bot commented Jan 8, 2025

pitrou commented Jan 8, 2025

conbench-apache-arrow bot commented Jan 9, 2025

mapleFU left a comment

mapleFU Jan 9, 2025 •

edited

Loading

pitrou Jan 9, 2025

pitrou commented Jan 9, 2025

mapleFU commented Jan 9, 2025

pitrou commented Jan 9, 2025

wgtmac left a comment

mapleFU Jan 9, 2025

pitrou Jan 9, 2025

pitrou commented Jan 9, 2025

github-actions bot commented Jan 9, 2025

pitrou commented Jan 9, 2025

conbench-apache-arrow bot commented Jan 15, 2025

		const int16_t* rep_levels) {
		// Update histograms now, to maximize cache efficiency.

GH-45201: [C++][Parquet] Improve performance of generating size statistics #45202

GH-45201: [C++][Parquet] Improve performance of generating size statistics #45202

Conversation

pitrou commented Jan 8, 2025 • edited Loading

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

pitrou commented Jan 8, 2025

pitrou commented Jan 8, 2025

github-actions bot commented Jan 8, 2025

pitrou commented Jan 8, 2025

conbench-apache-arrow bot commented Jan 9, 2025

mapleFU left a comment

Choose a reason for hiding this comment

mapleFU Jan 9, 2025 • edited Loading

Choose a reason for hiding this comment

pitrou Jan 9, 2025

Choose a reason for hiding this comment

pitrou commented Jan 9, 2025

mapleFU commented Jan 9, 2025

pitrou commented Jan 9, 2025

wgtmac left a comment

Choose a reason for hiding this comment

mapleFU Jan 9, 2025

Choose a reason for hiding this comment

pitrou Jan 9, 2025

Choose a reason for hiding this comment

pitrou commented Jan 9, 2025

github-actions bot commented Jan 9, 2025

pitrou commented Jan 9, 2025

conbench-apache-arrow bot commented Jan 15, 2025

pitrou commented Jan 8, 2025 •

edited

Loading

mapleFU Jan 9, 2025 •

edited

Loading