Improve ColumnDecimal, ColumnVector getPermutation performance using pdqsort with RadixSort

[kitaisreal]

Changelog category (leave one):

• Performance Improvement

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Improve performance of sorting for decimal columns. Improve performance of insertion into MergeTree if ORDER BY contains Decimal column. Improve performance of sorting when data is already sorted or almost sorted.

[kitaisreal]

Insertion into MergeTree. Probably not common case.
Setup table:

CREATE TABLE test_table_insert (id Decimal64(8), value1 UInt64, value2 UInt64, value3 UInt64) ENGINE=MergeTree ORDER BY id;

Before:

INSERT INTO test_table_insert SELECT (rand64() % 50000) as a, a, a, a FROM system.numbers LIMIT 50000000;
0 rows in set. Elapsed: 6.039 sec. Processed 50.33 million rows, 402.64 MB (8.33 million rows/s., 66.68 MB/s.)

After:

INSERT INTO test_table_insert SELECT (rand64() % 50000) as a, a, a, a FROM system.numbers LIMIT 50000000;
0 rows in set. Elapsed: 2.893 sec. Processed 50.33 million rows, 402.64 MB (17.40 million rows/s., 139.17 MB/s.)

[kitaisreal]

Sort by decimal column.
Table setup:

CREATE TABLE test_table_sort (id UInt64, value Decimal64(4)) ENGINE=MergeTree ORDER BY id;
INSERT INTO test_table_sort SELECT number, rand64() % 50000 FROM system.numbers LIMIT 50000000;

Before:

SELECT * FROM test_table_sort ORDER BY value FORMAT Null;
0 rows in set. Elapsed: 1.404 sec. Processed 50.00 million rows, 800.00 MB (35.61 million rows/s., 569.68 MB/s.)

After:

SELECT * FROM test_table_sort ORDER BY value FORMAT Null;
0 rows in set. Elapsed: 1.321 sec. Processed 50.00 million rows, 800.00 MB (37.85 million rows/s., 605.62 MB/s.)

[antonio2368]

you can reverse the 2 ifs and do the if (!limit) check only for arithmetic types.

I assume compiler is smart enough to remove the branch with no code, but I think it's good to be explicit in any case.

[antonio2368]

same things as above, if you reverse the checks, you can eliminate a branch for arithmetic type.

[alexey-milovidov]

Did not speed up:

[alexey-milovidov]

But the numbers in #35961 (comment) look good.

[zhanglistar]

@kitaisreal May I ask what happened to this PR for not merging?

[kitaisreal]

@zhanglistar it can provide degradation in a couple of cases, for example when data is completely sorted.

[alexey-milovidov]

Continued in #46559.

[robot-clickhouse]

This is an automated comment for commit f7494a5 with description of existing statuses. It's updated for the latest CI running

❌ Click here to open a full report in a separate page

Successful checks

Check name	Description	Status
AST fuzzer	Runs randomly generated queries to catch program errors. The build type is optionally given in parenthesis. If it fails, ask a maintainer for help	✅ success
CI running	A meta-check that indicates the running CI. Normally, it's in success or pending state. The failed status indicates some problems with the PR	✅ success
ClickHouse build check	Builds ClickHouse in various configurations for use in further steps. You have to fix the builds that fail. Build logs often has enough information to fix the error, but you might have to reproduce the failure locally. The cmake options can be found in the build log, grepping for cmake. Use these options and follow the general build process	✅ success
Compatibility check	Checks that clickhouse binary runs on distributions with old libc versions. If it fails, ask a maintainer for help	✅ success
Docker image for servers	The check to build and optionally push the mentioned image to docker hub	✅ success
Fast test	Normally this is the first check that is ran for a PR. It builds ClickHouse and runs most of stateless functional tests, omitting some. If it fails, further checks are not started until it is fixed. Look at the report to see which tests fail, then reproduce the failure locally as described here	✅ success
Flaky tests	Checks if new added or modified tests are flaky by running them repeatedly, in parallel, with more randomization. Functional tests are run 100 times with address sanitizer, and additional randomization of thread scheduling. Integrational tests are run up to 10 times. If at least once a new test has failed, or was too long, this check will be red. We don't allow flaky tests, read the doc	✅ success
Install packages	Checks that the built packages are installable in a clear environment	✅ success
Mergeable Check	Checks if all other necessary checks are successful	✅ success
Push to Dockerhub	The check for building and pushing the CI related docker images to docker hub	✅ success
SQLTest	There's no description for the check yet, please add it to tests/ci/ci_config.py:CHECK_DESCRIPTIONS	✅ success
SQLancer	Fuzzing tests that detect logical bugs with SQLancer tool	✅ success
Sqllogic	Run clickhouse on the sqllogic test set against sqlite and checks that all statements are passed	✅ success
Stateful tests	Runs stateful functional tests for ClickHouse binaries built in various configurations -- release, debug, with sanitizers, etc	✅ success
Stateless tests	Runs stateless functional tests for ClickHouse binaries built in various configurations -- release, debug, with sanitizers, etc	✅ success
Stress test	Runs stateless functional tests concurrently from several clients to detect concurrency-related errors	✅ success
Style Check	Runs a set of checks to keep the code style clean. If some of tests failed, see the related log from the report	✅ success
Unit tests	Runs the unit tests for different release types	✅ success

Check name	Description	Status
Integration tests	The integration tests report. In parenthesis the package type is given, and in square brackets are the optional part/total tests	❌ failure
Performance Comparison	Measure changes in query performance. The performance test report is described in detail here. In square brackets are the optional part/total tests	❌ failure
Upgrade check	Runs stress tests on server version from last release and then tries to upgrade it to the version from the PR. It checks if the new server can successfully startup without any errors, crashes or sanitizer asserts	❌ failure

[alexey-milovidov]

@kitaisreal it does not build.

[kitaisreal]

Performance improvements:
X86-64
https://s3.amazonaws.com/clickhouse-test-reports/35961/17776005ca8b903330a9b79eb9c3807dd22acf46/performance_comparison_[1_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/17776005ca8b903330a9b79eb9c3807dd22acf46/performance_comparison_[2_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/17776005ca8b903330a9b79eb9c3807dd22acf46/performance_comparison_[3_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/17776005ca8b903330a9b79eb9c3807dd22acf46/performance_comparison_[4_4]/report.html
AARCH64
https://s3.amazonaws.com/clickhouse-test-reports/35961/17776005ca8b903330a9b79eb9c3807dd22acf46/performance_comparison_aarch64_[1_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/17776005ca8b903330a9b79eb9c3807dd22acf46/performance_comparison_aarch64_[2_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/17776005ca8b903330a9b79eb9c3807dd22acf46/performance_comparison_aarch64_[3_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/17776005ca8b903330a9b79eb9c3807dd22acf46/performance_comparison_aarch64_[4_4]/report.html

[kitaisreal]

Performance improvements:
X86-64
https://s3.amazonaws.com/clickhouse-test-reports/35961/64f5032e5881f0001b4b68e58057db538199b76c/performance_comparison_[1_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/64f5032e5881f0001b4b68e58057db538199b76c/performance_comparison_[2_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/64f5032e5881f0001b4b68e58057db538199b76c/performance_comparison_[3_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/64f5032e5881f0001b4b68e58057db538199b76c/performance_comparison_[4_4]/report.html
AARCH64
https://s3.amazonaws.com/clickhouse-test-reports/35961/64f5032e5881f0001b4b68e58057db538199b76c/performance_comparison_aarch64_[1_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/64f5032e5881f0001b4b68e58057db538199b76c/performance_comparison_aarch64_[2_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/64f5032e5881f0001b4b68e58057db538199b76c/performance_comparison_aarch64_[3_4]/report.html
https://s3.amazonaws.com/clickhouse-test-reports/35961/64f5032e5881f0001b4b68e58057db538199b76c/performance_comparison_aarch64_[4_4]/report.html

[kitaisreal]

Insert into MergeTree improvements.

Setup table:

CREATE TABLE test_table_insert (id Decimal64(8), value1 UInt64, value2 UInt64, value3 UInt64) ENGINE=MergeTree ORDER BY id;

Before:

INSERT INTO test_table_insert SELECT (rand64() % 50000) as a, a, a, a
FROM system.numbers LIMIT 50000000;
0 rows in set. Elapsed: 6.032 sec. Processed 50.33 million rows, 402.60 MB (8.34 million rows/s., 66.74 MB/s.)

After:

INSERT INTO test_table_insert SELECT (rand64() % 50000) as a, a, a, a FROM system.numbers LIMIT 50000000;
0 rows in set. Elapsed: 3.040 sec. Processed 50.33 million rows, 402.60 MB (16.56 million rows/s., 132.45 MB/s.)

Before:

INSERT INTO test_table_insert SELECT (rand64() % 500000) as a, a, a, a
FROM system.numbers LIMIT 50000000;
0 rows in set. Elapsed: 7.593 sec. Processed 50.33 million rows, 402.60 MB (6.63 million rows/s., 53.02 MB/s.)

After:

INSERT INTO test_table_insert SELECT (rand64() % 500000) as a, a, a, a
FROM system.numbers LIMIT 50000000;
0 rows in set. Elapsed: 4.644 sec. Processed 50.33 million rows, 402.60 MB (10.84 million rows/s., 86.69 MB/s.)

[kitaisreal]

Sorting already sorted data improvements:

Setup table:

CREATE TABLE sequential_UInt64(key UInt64, value UInt64) Engine = Memory;
INSERT INTO sequential_UInt64 SELECT number, number FROM numbers(500000000);

Before:

SELECT key, value FROM sequential_UInt64 ORDER BY key FORMAT Null
0 rows in set. Elapsed: 1.269 sec. Processed 500.00 million rows, 8.00 GB (394.04 million rows/s., 6.30 GB/s.)

SELECT key, value FROM sequential_UInt64 ORDER BY key, value FORMAT Null
0 rows in set. Elapsed: 1.332 sec. Processed 500.00 million rows, 8.00 GB (375.26 million rows/s., 6.00 GB/s.)

SELECT key, value FROM sequential_UInt64 ORDER BY key DESC FORMAT Null
0 rows in set. Elapsed: 2.143 sec. Processed 500.00 million rows, 8.00 GB (233.33 million rows/s., 3.73 GB/s.)

SELECT key, value FROM sequential_UInt64 ORDER BY key DESC, value DESC FORMAT Null
0 rows in set. Elapsed: 2.438 sec. Processed 500.00 million rows, 8.00 GB (205.05 million rows/s., 3.28 GB/s.)

After:

SELECT key, value FROM sequential_UInt64 ORDER BY key FORMAT Null
0 rows in set. Elapsed: 0.647 sec. Processed 500.00 million rows, 8.00 GB (772.51 million rows/s., 12.36 GB/s.)

SELECT key, value FROM sequential_UInt64 ORDER BY key, value FORMAT Null
0 rows in set. Elapsed: 0.657 sec. Processed 500.00 million rows, 8.00 GB (761.02 million rows/s., 12.18 GB/s.)

SELECT key, value FROM sequential_UInt64 ORDER BY key DESC FORMAT Null
0 rows in set. Elapsed: 1.480 sec. Processed 500.00 million rows, 8.00 GB (337.84 million rows/s., 5.41 GB/s.)

SELECT key, value FROM sequential_UInt64 ORDER BY key DESC, value DESC FORMAT Null
0 rows in set. Elapsed: 1.631 sec. Processed 500.00 million rows, 8.00 GB (306.63 million rows/s., 4.91 GB/s.)