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

CRIU BytecodeInterpreter enables DO_HOOKS #17245

Merged
merged 1 commit into from
Apr 26, 2023

Conversation

JasonFengJ9
Copy link
Member

CRIU BytecodeInterpreter enables DO_HOOKS

criuBytecodeLoopCompressed/criuBytecodeLoopFull are used when CRIU is enabled.

Signed-off-by: Jason Feng fengj@ca.ibm.com

@JasonFengJ9 JasonFengJ9 added comp:vm criu Used to track CRIU snapshot related work labels Apr 21, 2023
@JasonFengJ9 JasonFengJ9 marked this pull request as draft April 21, 2023 17:28
@JasonFengJ9 JasonFengJ9 requested a review from tajila April 21, 2023 17:29
@JasonFengJ9 JasonFengJ9 marked this pull request as ready for review April 22, 2023 00:25
@tajila
Copy link
Contributor

tajila commented Apr 22, 2023

jenkins test sanity xlinux jdk11

@tajila
Copy link
Contributor

tajila commented Apr 24, 2023

jenkins test sanity plinux,zlinux,win,alinux64 jdk11

@tajila
Copy link
Contributor

tajila commented Apr 24, 2023

jenkins test sanity amac jdk11

@tajila
Copy link
Contributor

tajila commented Apr 25, 2023

jenkins test sanity xlinux jdk17

@JasonFengJ9
Copy link
Member Author

Results for JDK=/home/dsouzai/sdks/jdk11_2023-04-12-22-00 jvmOpts=-Xint -Xshareclasses:enableBCI,name=liberty
StartupTime             avg=32399       min=32248       max=32474       stdDev=45.0     maxVar=0.7%     confInt=0.05%   samples= 25
        Outlier values:  32654.399 32778.041 33133.543 31623.374 31365.668 31638.109 31143.165
ApplicationTime         avg=21964       min=21625       max=22716       stdDev=265.1    maxVar=5.0%     confInt=0.37%   samples= 31
        Outlier values:  22945
Footprint               avg=184288      min=178728      max=190392      stdDev=3508.0   maxVar=6.5%     confInt=0.57%   samples= 32
Results for JDK=/home/dsouzai/sdks/jdk11_2023-04-13-22-00 jvmOpts=-Xint -Xshareclasses:enableBCI,name=liberty
StartupTime             avg=43427       min=43240       max=43778       stdDev=127.4    maxVar=1.2%     confInt=0.09%   samples= 32
ApplicationTime         avg=31827       min=31404       max=32914       stdDev=457.8    maxVar=4.8%     confInt=0.43%   samples= 32
Footprint               avg=180412      min=175736      max=185656      stdDev=2567.1   maxVar=5.6%     confInt=0.42%   samples= 32
Results for JDK=/home/dsouzai/sdks/jdk11_20230421-180956 jvmOpts=-Xint -Xshareclasses:enableBCI,name=liberty
StartupTime             avg=31810       min=31159       max=33434       stdDev=496.9    maxVar=7.3%     confInt=0.47%   samples= 32
ApplicationTime         avg=22132       min=21736       max=23029       stdDev=337.7    maxVar=5.9%     confInt=0.45%   samples= 32
Footprint               avg=180744      min=176536      max=191588      stdDev=3003.3   maxVar=8.5%     confInt=0.49%   samples= 32

Looks like the OpenJ9-JDK11-x86-64_linux-20230421-180956 build fixes the regression (and possibly even improves startup by ~1.5%).

@JasonFengJ9
Copy link
Member Author

Results for JDK=/home/dsouzai/sdks/jdk17_20230412-201002 jvmOpts=-Xint -Xshareclasses:enableBCI,name=liberty
StartupTime             avg=33920       min=32843       max=34935       stdDev=661.8    maxVar=6.4%     confInt=0.58%   samples= 32
ApplicationTime         avg=23517       min=22765       max=24133       stdDev=509.4    maxVar=6.0%     confInt=0.64%   samples= 32
Footprint               avg=187922      min=181396      max=196456      stdDev=5206.6   maxVar=8.3%     confInt=0.82%   samples= 32
Results for JDK=/home/dsouzai/sdks/jdk17_20230413-191340 jvmOpts=-Xint -Xshareclasses:enableBCI,name=liberty
StartupTime             avg=33476       min=32489       max=34851       stdDev=603.9    maxVar=7.3%     confInt=0.54%   samples= 32
ApplicationTime         avg=23160       min=22476       max=24125       stdDev=583.1    maxVar=7.3%     confInt=0.75%   samples= 32
Footprint               avg=187825      min=179724      max=198036      stdDev=5448.5   maxVar=10.2%    confInt=0.86%   samples= 32
Results for JDK=/home/dsouzai/sdks/jdk17_20230421-104547 jvmOpts=-Xint -Xshareclasses:enableBCI,name=liberty
StartupTime             avg=33824       min=32677       max=35211       stdDev=616.8    maxVar=7.8%     confInt=0.54%   samples= 32
ApplicationTime         avg=23479       min=22679       max=25054       stdDev=591.5    maxVar=10.5%    confInt=0.75%   samples= 32
Footprint               avg=187360      min=176404      max=197704      stdDev=4677.1   maxVar=12.1%    confInt=0.74%   samples= 32

No issues on JDK17 either.

@DanHeidinga
Copy link
Member

Should also check the footprint (RSS) when compiling an extra bytecode interpreter.

@JasonFengJ9
Copy link
Member Author

Should also check the footprint (RSS) when compiling an extra bytecode interpreter.

@dsouzai mentioned in the openj9 meeting that the Footprint perf results above are RSS.

runtime/vm/jvminit.c Outdated Show resolved Hide resolved
criuBytecodeLoopCompressed/criuBytecodeLoopFull are used when CRIU is
enabled.

Signed-off-by: Jason Feng <fengj@ca.ibm.com>
@tajila
Copy link
Contributor

tajila commented Apr 26, 2023

@keithc-ca anymore concerns

@tajila
Copy link
Contributor

tajila commented Apr 26, 2023

@pshipton we will go ahead with this change

@tajila
Copy link
Contributor

tajila commented Apr 26, 2023

jenkins compile win jdk11

@tajila tajila merged commit 7977a21 into eclipse-openj9:master Apr 26, 2023
@tajila
Copy link
Contributor

tajila commented Apr 26, 2023

@JasonFengJ9 please make a change for 0.38

@JasonFengJ9
Copy link
Member Author

@JasonFengJ9 JasonFengJ9 deleted the criubyteinterpreter branch April 26, 2023 18:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
comp:vm criu Used to track CRIU snapshot related work
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants