HBASE-29995 Improve existing hash implementations by reading 4 bytes at once

[jinhyukify]

Jira https://issues.apache.org/jira/browse/HBASE-29995

[Apache9]

Please add a dummy change in pom.xml so we can run all the tests in HBase to see if there are any problems with the changes? once we confirm there is no problem, you can do a force push to remove the dummy commit and then merge the PR.

And why not read 8 bytes at once?

[jinhyukify]

@Apache9 Thank you for checking!

And why not read 8 bytes at once?

The existing hashes (Jenkins, Murmur, Murmur3) are all 4-byte based, so they operate on 4-byte chunks. There is currently no place in the algorithms where 8-byte reads are used.

In theory we could read 8 bytes and split them into two ints, but given the current structure it would not provide much benefit. If we add hash functions that operate on 8-byte words in the future, we could revisit this.

[junegunn]

So we're adding a non-trivial amount of code here, and it seems to overlap with the existing get(int offset) implementation to some degree. Would it make sense to refactor get to call this new method instead, so we can reduce code duplication? We could consider that approach as long as the performance overhead is negligible.

[jinhyukify]

Thank you for checking.
I changed to use assembleCrossingLE method in get method.

[Copilot]

Pull request overview

This PR targets HBASE-29995 by improving hashing throughput (Bloom filter-related hash keys and hash functions) via little-endian 4-byte loads, reducing per-byte get() calls in the hot loops.

Changes:

• Add HashKey#getIntLE and switch Murmur/Murmur3/Jenkins inner loops to load 4 bytes at a time.
• Introduce LittleEndianBytes (with optional Unsafe acceleration) and extend UnsafeAccess with LE int helpers.
• Update RowBloomHashKey/RowColBloomHashKey implementations and add new unit tests for LE conversions and ROWCOL bloom key layout.

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
hbase-common/src/main/java/org/apache/hadoop/hbase/util/HashKey.java	Adds getIntLE contract for faster 4-byte little-endian reads.
hbase-common/src/main/java/org/apache/hadoop/hbase/util/ByteArrayHashKey.java	Implements getIntLE for byte[]-backed keys.
hbase-common/src/main/java/org/apache/hadoop/hbase/util/RowBloomHashKey.java	Implements getIntLE using row fast-path reads.
hbase-common/src/main/java/org/apache/hadoop/hbase/util/RowColBloomHashKey.java	Adds getIntLE fast paths + crossing-boundary assembly; caches total length.
hbase-common/src/main/java/org/apache/hadoop/hbase/util/LittleEndianBytes.java	New LE int read/write utility (Unsafe-accelerated when available).
hbase-common/src/main/java/org/apache/hadoop/hbase/util/UnsafeAccess.java	Adds toIntLE/putIntLE helpers for byte[] and ByteBuffer.
hbase-common/src/main/java/org/apache/hadoop/hbase/util/MurmurHash.java	Uses hashKey.getIntLE for 4-byte block loads.
hbase-common/src/main/java/org/apache/hadoop/hbase/util/MurmurHash3.java	Uses hashKey.getIntLE for 4-byte block loads.
hbase-common/src/main/java/org/apache/hadoop/hbase/util/JenkinsHash.java	Uses hashKey.getIntLE for 4-byte block loads.
hbase-common/src/test/java/org/apache/hadoop/hbase/util/TestLittleEndianBytesBase.java	New base tests covering LE int conversions and Cell row/qualifier reads.
hbase-common/src/test/java/org/apache/hadoop/hbase/util/TestLittleEndianBytes.java	Tagged JUnit5 runner for the base tests (unsafe-enabled path).
hbase-common/src/test/java/org/apache/hadoop/hbase/util/TestLittleEndianBytesWoUnsafe.java	Tagged JUnit5 runner that mocks unaligned=false (pure-Java path).
hbase-common/src/test/java/org/apache/hadoop/hbase/util/TestRowColBloomHashKey.java	New tests validating ROWCOL bloom virtual layout and getIntLE consistency.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.