SPECjvm
This is SPECjvm2008 modified to run with JDK24. Scores are ops/min, more is better. All benchmarks have been run on a AWS c5.9xlarge instance.
JVM flags: -Xms8g -Xmx8g -XX:+AlwaysPreTouch
Notes: The CryptoRsa benchmark shows a legit regression of ~2%. I confirmed this by running 50 times (most benchmarks are run 3 times). The regression disappears when improving oopDesc::klass().
Benchmark | Baseline | -UCOH | % | +UCOH | % |
|---|---|---|---|---|---|
Compiler.compiler | 1127.077 | 1127.628 | 0.04% | 1150.953 | 2.11% |
Compiler.sunflow | 3121.071 | 3190.964 | 2.23% | 3202.471 | 2.60% |
Compress | 3100.924 | 3114.255 | 0.42% | 3037.949 | -2.04% |
CryptoAes | 214.615 | 215.491 | 0.40% | 212.788 | -0.86% |
CryptoRsa | 10284.167 | 10094.019 | -1.85% | 10110.913 | -1.69% |
CryptoSignVerify | 48591.761 | 48541.003 | -0.11% | 48107.972 | -1.00% |
Derby | 3646.723 | 3620.906 | -0.71% | 3598.845 | -1.32% |
MpegAudio | 1012.491 | 1011.096 | -0.14% | 1005.387 | -0.71% |
ScimarkFFT.large | 278.913 | 278.937 | 0.00% | 278.613 | -0.11% |
ScimarkFFT.small | 3348.048 | 3374.245 | 0.78% | 3276.573 | -2.14% |
ScimarkLU.large | 21.737 | 21.771 | 0.15% | 21.640 | -0.45% |
ScimarkLU.small | 5876.355 | 5870.420 | -0.11% | 5867.935 | -0.15% |
ScimarkMonteCarlo | 17635.373 | 17767.207 | 0.74% | 17845.029 | 1.18% |
ScimarkSOR.large | 324.000 | 322.590 | -0.44% | 322.137 | -0.58% |
ScimarkSOR.small | 1647.129 | 1647.330 | 0.01% | 1648.043 | 0.05% |
ScimarkSparse.large | 196.737 | 196.639 | -0.05% | 196.934 | 0.10% |
ScimarkSparse.small | 1305.586 | 1294.917 | -0.82% | 1289.868 | -1.21% |
Serial | 48075.242 | 48503.947 | 0.89% | 48583.323 | 1.05% |
Sunflow | 667.590 | 673.791 | 0.92% | 701.846 | 5.13% |
XmlTransform | 2005.798 | 2016.622 | 0.53% | 1965.465 | -2.02% |
XmlValidation | 4503.275 | 4486.646 | -0.37% | 4485.866 | -0.39% |
Renaissance
This is renaissance-jmh-0.15.0, excluding Als, ChiSquare, DecTree, GaussMix, LogRegression, MovieLens, NaiveBayes, PageRank and DbShootout, which are not compatible with JDK24 due to SecurityManager restriction. Scores are ms/op, less is better. All benchmarks have been run on a AWS c5.9xlarge instance.
JVM flags: -Xms8g -Xmx8g -XX:+AlwaysPreTouch
Benchmark | Baseline | -UCOH | % | +UCOH | % |
|---|---|---|---|---|---|
AkkaUct | 799.526 | 796.987 | -0.32% | 792.829 | -0.84% |
Reactors | 15192.454 | 15070.100 | -0.81% | 14533.405 | -4.34% |
FjKmeans | 1048.697 | 1049.511 | 0.07% | 953.387 | -9.09% |
FutureGenetic | 2011.925 | 1991.429 | -1.02% | 2072.172 | 2.99% |
Mnemonics | 2760.609 | 2739.520 | -0.77% | 2824.556 | 2.31% |
ParMnemonics | 2270.166 | 2286.626 | 0.72% | 2304.885 | 1.52% |
Scrabble | 53.150 | 53.525 | 0.70% | 55.052 | 3.57% |
Neo4jAnalytics | 1482.061 | 1479.183 | -0.20% | 1511.736 | 2.00% |
RxScrabble | 104.344 | 103.626 | -0.69% | 105.156 | 0.77% |
Dotty | 697.111 | 700.798 | 0.52% | 700.191 | 0.44% |
ScalaKmeans | 175.545 | 177.320 | 1.01% | 188.345 | 7.29% |
Philosophers | 5598.596 | 5645.268 | 0.83% | 6379.457 | 13.94% |
ScalaStmBench7 | 1010.489 | 1018.077 | 0.75% | 999.638 | -1.08% |
FinagleChirper | 3643.971 | 3629.694 | -0.40% | 3681.006 | 1.01% |
FinagleHttp | 3392.307 | 3338.891 | -1.58% | 3269.203 | -3.63% |
SPECjbb2015
SPECjbb2015 in composite mode, run on a AWS c5.9xlarge instance, 10 times in a row, scores averaged over all runs. More is better.
JVM flags: -Xms20g -Xmx20g -XX:+AlwaysPreTouch
Score | Baseline | -UCOH | % | +UCOH | % |
|---|---|---|---|---|---|
max-jops | 49818 | 49700 | -0.24% | 51652 | 3.68% |
critial-jops | 39139 | 38726 | -1.06% | 41219 | 5.31% |
GC Pause times
Young GC time, measured over 10 runs of SPECjbb2015 each, the total number of cycles is given, too.
Young GC time, baseline: avg: 71.4059 ms, #cycles: 22909Young GC time, -UCOH: avg: 73.2457 ms, #cycles: 23032Young GC time, +UCOH: avg: 69.8878 ms, #cycles: 20619
Concurrent marking time, measured over a run of SPECjvm compiler.compiler benchmark.
Concurrent marking time, baseline: avg: 1757.43 ms, #cycles: 69Concurrent marking time, -UCOH: avg: 1724.96 ms, #cycles: 78Concurrent marking time, +UCOH: avg: 1664.67 ms, #cycles: 62
SPECjbb2015 (run at Red Hat)
The following tests were run at Red Hat on a bare metal RHEL9 x64 i7-4770 8core machine, with the benchmark process isolated on 6 cores, 10 consecutive runs, all scores averaged. Perf statistics were done for the full benchmark process run, therefore include warmup and report generation.
JVM flags: -Xshare:off -Xlog:gc* -XX:+UseG1GC -Xms4g -Xmx4g
| -COH | +COH | % | |
| maxjops | 10136.6 | 10823.6 | +6.78% |
| critjops | 3742 | 3995.6 | +6.78% |
| Number of GCs | 4971.8 | 3469.8 | -30.21% |
| GC Real Times, Sum, seconds | 350.34 | 273 | -22.08% |
| GC User Times, Sum, seconds | 2031.83 | 1588.63 | -21.81% |
| GC Sys Times, Sum, seconds | 3.07 | 2.2 | -28.34% |
| L1 Misses | 1294363538563.4 | 1109038645193.3 | -14.32% |
| L1 Loads | 25511385965443.6 | 21610503370719.1 | -15.29% |
| LLC Misses | 273833668865.4 | 212658994859.6 | -22.34% |
| LLC Loads | 655211047461.7 | 555606059480 | -15.20% |
| TLB Misses | 37718612782.9 | 32378833894.3 | -14.16% |
| TLB Loads | 25308734883537.9 | 21437105901958.6 | -15.30% |
| Instructions | 83966573365463.8 | 72477839817765.8 | -13.68% |
| Branches | 15749934748078.2 | 13181107354563.9 | -16.31% |
To measure the reduction in liveset size, the same benchmark was repeated with a periodic Full GC being scheduled every 5 minutes, and the liveset size was measured after every full GC:
| -COH | +COH | Percent | Gains | |
| Usage pre GC, MB | 3894.9 | 3216.6 | 82.58 | - 17.42 |
| Usage post GC, MB | 1071.7 | 885.9 | 82.66 | - 17.34 |
Overview
Content Tools
ThemeBuilder