ODIn/spark-instrumented-optimizer
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

[SPARK-12856] [SQL] speed up hashCode of unsafe array

Browse source 
We iterate the bytes to calculate hashCode before, but now we have `Murmur3_x86_32.hashUnsafeBytes` that don't require the bytes to be word algned, we should use that instead.

A simple benchmark shows it's about 3 X faster, benchmark code: https://gist.github.com/cloud-fan/fa77713ccebf0823b2ab#file-arrayhashbenchmark-scala

Author: Wenchen Fan <wenchen@databricks.com>

Closes #10784 from cloud-fan/array-hashcode.
This commit is contained in:
Wenchen Fan 2016-01-16 00:38:17 -08:00 committed by Davies Liu
parent 242efb7546
commit 2f7d0b68a2
1 changed files with 2 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

7
sql/catalyst/src/main/java/org/apache/spark/sql/catalyst/expressions/UnsafeArrayData.java
View file

@ -25,6 +25,7 @@ import org.apache.spark.sql.catalyst.util.ArrayData;
import org.apache.spark.sql.types.*;
import org.apache.spark.unsafe.Platform;
import org.apache.spark.unsafe.array.ByteArrayMethods;
import org.apache.spark.unsafe.hash.Murmur3_x86_32;
import org.apache.spark.unsafe.types.CalendarInterval;
import org.apache.spark.unsafe.types.UTF8String;
@ -299,11 +300,7 @@ public class UnsafeArrayData extends ArrayData {
@Override
public int hashCode() {
int result = 37;
for (int i = 0; i < sizeInBytes; i++) {
result = 37 * result + Platform.getByte(baseObject, baseOffset + i);
}
return result;
return Murmur3_x86_32.hashUnsafeBytes(baseObject, baseOffset, sizeInBytes, 42);
}
@Override