-
-
Notifications
You must be signed in to change notification settings - Fork 1.2k
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
add java-17-openjdk #5610
add java-17-openjdk #5610
Conversation
1e090bc
to
94b5654
Compare
I think you need
Since ZGC is not supported on ARM other than AARCH64. And for some reason Fedora only build zero on PPC platforms. |
Thanks @arnebjarne, I have already fixed the build for 32-bit ARM (but not yet pushed, since I try to remove obsolete x-libraries first) |
@arnebjarne in java-11-openjdk package we disabled ZGC for all packages and built the zero variant. Is there any reason to build another jvm variant than zero? IFAICR we choose the zero variant to minimize the package size. EDIT:
|
--with-jvm-features=shenandoahgc,-zgc Is a leftover from the Java 11 package. I cannot remember why i used the shenandoahgc GC - maybe its faster on SoC chips like ARM etc. I need to research that... |
https://wiki.openjdk.org/display/shenandoah/Main#Main-JDKSupport Ah, it was optional in Java 11. In Java 17 its standard so
should be omitted. |
@hgy59, I've installed the
I also tested some of the simple Java programs from https://www.cs.uic.edu/~sloan/CLASSES/java/ and they worked file:
All in all it seems good to go. |
It would be interresting to see some execution times. I remember when i made the Java 11 version that speed was very different depending if I build server,hotspot or zero variant. |
Do you have a specific test methodology in mind? |
Here are the results of a benchmark test of the latest build on my DS218+ with DSM 7.1.1-42962 Update 3 and 8GB RAM
But I Have no reference to compare the results... |
Adding to the results from my VirtualDSM:
|
The results with the current Java 11 OpenJDK (on the same NAS as above):
|
@arnebjarne you are right, the server variant has much more performance than the zero variant: still on the same NAS with the Zero VM:
|
Maybe snatch some building tricks from https://github.com/adoptium/temurin-build ;) |
Its worth to notice that Adoptium use the correct supported GCC 10.3 to compile Java and their default is to build the hotspot variant. |
Updating feedback for current install... (log -- https://pastebin.com/uWN3szYq)
Performance is fairly similar...
|
Speed comparison DS418 Compiled with target hotspot and feature=server java --version openjdk 17.0.6 2023-01-17 LTS java -classpath scimark2lib.jar jnt.scimark2.commandline SciMark 2.0a Composite Score: 205.78632413749855 java.vendor: N/A compiled with target hotspot, feature zero: java --version Composite Score: 8.368952648493377 java.vendor: N/A So staying away from zero is the way to go BTW: @hgy59 --with-vendor-name=SynoCommunity (hint sh ./work-*/jdk17u-jdk-17.0.6-10/build/.configure-support/generated-configure.sh --help for compile time arguments) |
I have added all the Debian patches and it compiles Check |
- remove obsolete x-libraries - fix build for 32-bit arm archs (shenandoahgc not supported) - remove configure warnings: move arguments from environment to command line
- fix build of ARMv7 - remove build of jvm variant zero
- this gives a little bit more performance (2% with SciMark 2.0a)
(cherry picked from commit 79c239d)
I finally updated my ds115j to DSM 7 - and I can confirm that #4912 is fixed with this OpenJDK 17 package.
|
Description
create a package for
Java 17 OpenJDK
the current LTS version.Closes #5605, Closes #4912
Checklist
all-supported
completed successfullyType of change