Skip to content

Use XxHash3 for calculating page checksums #28116

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wendigo wants to merge 6 commits into
base: master
Choose a base branch
from
Open

Use XxHash3 for calculating page checksums #28116

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
9b299bb
Update aircompressor v3 to 3.5
wendigo Feb 4, 2026
f7b6004
Use XxHash3 for calculating page checksums
wendigo Feb 4, 2026
d328c05
Calculate array position links using xxhash3
wendigo Feb 4, 2026
78cd801
Pool hashers
wendigo Feb 5, 2026
aaf3ed9
Use xxhash3 only for sizes > 16K
wendigo Feb 5, 2026
0cfeffd
Rename static variable
wendigo Feb 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 21 additions & 4 deletions core/trino-main/src/main/java/io/trino/execution/buffer/PagesSerdeUtil.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,8 @@
package io.trino.execution.buffer;

import com.google.common.collect.AbstractIterator;
import io.airlift.compress.v3.xxhash.XxHash3Hasher;
import io.airlift.compress.v3.xxhash.XxHash3Native;
import io.airlift.slice.Slice;
import io.airlift.slice.SliceInput;
import io.airlift.slice.SliceOutput;
Expand All @@ -39,6 +41,8 @@

public final class PagesSerdeUtil
{
private static ThreadLocal<XxHash3Hasher> HASHERS = ThreadLocal.withInitial(XxHash3Native::newHasher);

private PagesSerdeUtil() {}

static final int SERIALIZED_PAGE_POSITION_COUNT_OFFSET = 0;
Expand Down Expand Up @@ -77,11 +81,24 @@ static Page readRawPage(int positionCount, SliceInput input, BlockEncodingSerde

public static long calculateChecksum(List<Slice> pages)
{
XxHash64 hash = new XxHash64();
for (Slice page : pages) {
hash.update(page);
int size = pages.stream().mapToInt(Slice::length).sum();
long checksum;
if (size > 16384) {
Copy link
Member

@martint martint Feb 5, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why?

Copy link
Contributor Author

@wendigo wendigo Feb 5, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It seems that for small slices the cost of calling native outweights the benefits of a faster hashing function

Copy link
Member

@findepi findepi Feb 6, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

could the magic live in aircompressor ?

Copy link
Member

@findepi findepi Feb 6, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

also, if we cannot move this switch there, the code as it stands now absolutely requires code comment

XxHash3Hasher hasher = HASHERS.get();
hasher.reset();
for (Slice page : pages) {
hasher.update(page.byteArray(), page.byteArrayOffset(), page.length());
}
checksum = hasher.digest();
}
long checksum = hash.hash();
else {
XxHash64 hasher = new XxHash64();
for (Slice page : pages) {
hasher.update(page);
}
checksum = hasher.hash();
}

// Since NO_CHECKSUM is assigned a special meaning, it is not a valid checksum.
if (checksum == NO_CHECKSUM) {
return checksum + 1;
Expand Down
11 changes: 6 additions & 5 deletions core/trino-main/src/main/java/io/trino/operator/join/ArrayPositionLinks.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,9 +13,11 @@
*/
package io.trino.operator.join;

import io.airlift.slice.XxHash64;
import io.airlift.compress.v3.xxhash.XxHash3Hasher;
import io.airlift.compress.v3.xxhash.XxHash3Native;
import io.trino.spi.Page;

import java.lang.foreign.MemorySegment;
import java.util.Arrays;
import java.util.List;

Expand Down Expand Up @@ -63,11 +65,10 @@ public PositionLinks create(List<JoinFilterFunction> searchFunctions)
@Override
public long checksum()
{
long hash = 0;
for (int positionLink : positionLinks) {
hash = XxHash64.hash(hash, positionLink);
try (XxHash3Hasher hasher = XxHash3Native.newHasher()) {
hasher.update(MemorySegment.ofArray(positionLinks));
return hasher.digest();
}
return hash;
}
};
}
Expand Down
2 changes: 1 addition & 1 deletion pom.xml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -605,7 +605,7 @@
<dependency>
<groupId>io.airlift</groupId>
<artifactId>aircompressor-v3</artifactId>
<version>3.4</version>
<version>3.5</version>
</dependency>

<dependency>
Expand Down