Skip to content
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.

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

Performance degradation for the same code with IBM Semeru Runtime Open Edition 11.0.13.0 #44

Open
VincentSalewyn opened this issue Nov 30, 2022 · 4 comments

Comments

@VincentSalewyn
Copy link

VincentSalewyn commented Nov 30, 2022

We have 2 different releases for SPSS Modeler Premium.

18.3:
Windows: 8u265-b01 openj9
Mac: 8u265-b01 Hotspot
Linux: IBM JRE 8.0.6.25

18.4:
IBM Semeru Runtime Open Edition 11.0.13.0
Download link: https://github.com/ibmruntimes/semeru11-binaries/releases/tag/jdk-11.0.13%2B8_openj9-0.29.0

We did the following test on Mac machine: Text Analytics Scoring for a big corpora and huge set of categories.

First test in 18.4 (so using JRE 11):
Stream execution complete, Elapsed=567.0 sec, CPU=2017.1 sec

Second test on 18.4 replacing shipped JRE (JRE 11) by the one from 18.3 (JRE 8):
18.4 JRE 8: Stream execution complete, Elapsed=437.0 sec, CPU=1560.87 sec

What can explain such a difference on CPU time ?

@pshipton
Copy link
Member

What processor is in the Mac machine running the test? Is it x86 or Aarch64/M1?

@hanzhhan
Copy link

hanzhhan commented Dec 1, 2022

x86

@pshipton
Copy link
Member

pshipton commented Dec 1, 2022

The concern is that on Mac x86, Hotspot jdk8 (July 2020 - 20_03) is faster than OpenJ9 jdk11 (0.29 / 11.0.13 - 21_04). There is a newer OpenJ9 version (0.35 / 11.0.17 - 22_04) but I don't know if it would improve the performance.
@0xdaryl @knn-k

@0xdaryl
Copy link

0xdaryl commented Dec 2, 2022

Just to be clear, you are comparing an x64 Semeru JDK 11 on macOS vs an x64 HotSpot JDK 8 ?

It is difficult to explain the results generally without looking at what the application or workload is doing. There have been some performance degradations observed between JDK 8 and 11 for a variety of reasons (maybe on the order of 10%), but that doesn't account for your full difference.

Is this application and workload something I can run to analyze and provide you with a more meaningful answer? I think you're an IBMer so you can reach out to me (Daryl Maier) directly internally and we can discuss.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants