Skip to content

Commit d2dc055

Browse files
dongjoon-hyundongjoon-hyun
committed
[SPARK-53238][CORE][TESTS] Improve SorterBenchmark to include Arrays.parallelSort
### What changes were proposed in this pull request? This PR aims to improve `SorterBenchmark` to include `Arrays.parallelSort` additionally. ### Why are the changes needed? To compare multi-core CPU performance together. ### Does this PR introduce _any_ user-facing change? No, this is a benchmark improvement. ### How was this patch tested? Manual review with the generated benchmark result. ### Was this patch authored or co-authored using generative AI tooling? No. Closes #51964 from dongjoon-hyun/SPARK-53238. Authored-by: Dongjoon Hyun <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]>
1 parent ba14965 commit d2dc055

File tree

3 files changed

+51
-20
lines changed

3 files changed

+51
-20
lines changed

‎core/benchmarks/SorterBenchmark-jdk21-results.txt

Lines changed: 13 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -2,25 +2,28 @@
22
key-value pairs sort
33
================================================================================================
44

5-
OpenJDK 64-Bit Server VM 21.0.7+6-LTS on Linux 6.11.0-1014-azure
5+
OpenJDK 64-Bit Server VM 21.0.8+9-LTS on Linux 6.11.0-1018-azure
66
AMD EPYC 7763 64-Core Processor
77
key-value pairs sort 25000000: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
88
------------------------------------------------------------------------------------------------------------------------
9-
Tuple-sort using Arrays.sort() 17292 17584 413 1.4 691.7 1.0X
10-
KV-sort using Sorter 21245 21340 134 1.2 849.8 0.8X
9+
Tuple-sort using Arrays.sort() 16360 16496 192 1.5 654.4 1.0X
10+
Tuple-sort using Arrays.parallelSort() 7796 8631 1180 3.2 311.9 2.1X
11+
KV-sort using Sorter 19344 19436 130 1.3 773.8 0.8X
1112

1213

1314
================================================================================================
1415
primitive int array sort
1516
================================================================================================
1617

17-
OpenJDK 64-Bit Server VM 21.0.7+6-LTS on Linux 6.11.0-1014-azure
18+
OpenJDK 64-Bit Server VM 21.0.8+9-LTS on Linux 6.11.0-1018-azure
1819
AMD EPYC 7763 64-Core Processor
19-
primitive int array sort 25000000: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
20-
-------------------------------------------------------------------------------------------------------------------------------
21-
Java Arrays.sort() on non-primitive int array 14231 14561 468 1.8 569.2 1.0X
22-
Java Arrays.sort() on primitive int array 2141 2145 6 11.7 85.6 6.6X
23-
Sorter without key reuse on primitive int array 8607 8612 7 2.9 344.3 1.7X
24-
Sorter with key reuse on primitive int array 10602 10621 27 2.4 424.1 1.3X
20+
primitive int array sort 25000000: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
21+
-------------------------------------------------------------------------------------------------------------------------------------
22+
Java Arrays.sort() on non-primitive int array 14783 14862 111 1.7 591.3 1.0X
23+
Java Arrays.parallelSort() on non-primitive int array 6336 6423 124 3.9 253.4 2.3X
24+
Java Arrays.sort() on primitive int array 2122 2122 0 11.8 84.9 7.0X
25+
Java Arrays.parallelSort() on primitive int array 833 846 21 30.0 33.3 17.7X
26+
Sorter without key reuse on primitive int array 8637 8651 21 2.9 345.5 1.7X
27+
Sorter with key reuse on primitive int array 10726 10733 10 2.3 429.0 1.4X
2528

2629

‎core/benchmarks/SorterBenchmark-results.txt

Lines changed: 13 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -2,25 +2,28 @@
22
key-value pairs sort
33
================================================================================================
44

5-
OpenJDK 64-Bit Server VM 17.0.15+6-LTS on Linux 6.11.0-1014-azure
5+
OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Linux 6.11.0-1018-azure
66
AMD EPYC 7763 64-Core Processor
77
key-value pairs sort 25000000: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
88
------------------------------------------------------------------------------------------------------------------------
9-
Tuple-sort using Arrays.sort() 20545 20560 21 1.2 821.8 1.0X
10-
KV-sort using Sorter 26696 26797 143 0.9 1067.8 0.8X
9+
Tuple-sort using Arrays.sort() 16765 16984 308 1.5 670.6 1.0X
10+
Tuple-sort using Arrays.parallelSort() 10521 10746 318 2.4 420.8 1.6X
11+
KV-sort using Sorter 22863 23017 219 1.1 914.5 0.7X
1112

1213

1314
================================================================================================
1415
primitive int array sort
1516
================================================================================================
1617

17-
OpenJDK 64-Bit Server VM 17.0.15+6-LTS on Linux 6.11.0-1014-azure
18+
OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Linux 6.11.0-1018-azure
1819
AMD EPYC 7763 64-Core Processor
19-
primitive int array sort 25000000: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
20-
-------------------------------------------------------------------------------------------------------------------------------
21-
Java Arrays.sort() on non-primitive int array 15325 15510 263 1.6 613.0 1.0X
22-
Java Arrays.sort() on primitive int array 2088 2141 76 12.0 83.5 7.3X
23-
Sorter without key reuse on primitive int array 8254 8262 10 3.0 330.2 1.9X
24-
Sorter with key reuse on primitive int array 10184 10186 3 2.5 407.3 1.5X
20+
primitive int array sort 25000000: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
21+
-------------------------------------------------------------------------------------------------------------------------------------
22+
Java Arrays.sort() on non-primitive int array 14334 14344 14 1.7 573.3 1.0X
23+
Java Arrays.parallelSort() on non-primitive int array 6734 7146 582 3.7 269.4 2.1X
24+
Java Arrays.sort() on primitive int array 2177 2178 2 11.5 87.1 6.6X
25+
Java Arrays.parallelSort() on primitive int array 1001 1001 0 25.0 40.0 14.3X
26+
Sorter without key reuse on primitive int array 8363 8365 2 3.0 334.5 1.7X
27+
Sorter with key reuse on primitive int array 10189 10191 2 2.5 407.6 1.4X
2528

2629

‎core/src/test/scala/org/apache/spark/util/collection/SorterBenchmark.scala

Lines changed: 25 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -57,6 +57,15 @@ object SorterBenchmark extends BenchmarkBase {
5757
timer.stopTiming()
5858
}
5959

60+
benchmark.addTimerCase("Tuple-sort using Arrays.parallelSort()") { timer =>
61+
val kvTupleArray = new Array[AnyRef](numElements)
62+
System.arraycopy(kvTuples, 0, kvTupleArray, 0, numElements)
63+
timer.startTiming()
64+
util.Arrays.parallelSort(kvTupleArray, (x: AnyRef, y: AnyRef) =>
65+
x.asInstanceOf[(JFloat, _)]._1.compareTo(y.asInstanceOf[(JFloat, _)]._1))
66+
timer.stopTiming()
67+
}
68+
6069
// Test Sorter where each element alternates between Float and Integer, non-primitive
6170
val keyValues = {
6271
val data = new Array[AnyRef](numElements * 2)
@@ -106,6 +115,14 @@ object SorterBenchmark extends BenchmarkBase {
106115
timer.stopTiming()
107116
}
108117

118+
benchmark.addTimerCase("Java Arrays.parallelSort() on non-primitive int array") { timer =>
119+
val intObjectArray = new Array[Integer](numElements)
120+
System.arraycopy(intObjects, 0, intObjectArray, 0, numElements)
121+
timer.startTiming()
122+
util.Arrays.parallelSort(intObjectArray, (x: Integer, y: Integer) => x.compareTo(y))
123+
timer.stopTiming()
124+
}
125+
109126
benchmark.addTimerCase("Java Arrays.sort() on primitive int array") { timer =>
110127
val intPrimitiveArray = new Array[Int](numElements)
111128
System.arraycopy(ints, 0, intPrimitiveArray, 0, numElements)
@@ -114,6 +131,14 @@ object SorterBenchmark extends BenchmarkBase {
114131
timer.stopTiming()
115132
}
116133

134+
benchmark.addTimerCase("Java Arrays.parallelSort() on primitive int array") { timer =>
135+
val intPrimitiveArray = new Array[Int](numElements)
136+
System.arraycopy(ints, 0, intPrimitiveArray, 0, numElements)
137+
timer.startTiming()
138+
util.Arrays.parallelSort(intPrimitiveArray)
139+
timer.stopTiming()
140+
}
141+
117142
benchmark.addTimerCase("Sorter without key reuse on primitive int array") { timer =>
118143
val intPrimitiveArray = new Array[Int](numElements)
119144
System.arraycopy(ints, 0, intPrimitiveArray, 0, numElements)

0 commit comments

Comments
 (0)