Don't store or find heap size hints when heap fully expanded - #19305
Conversation
This removes the GC stress on shared cache, that may otherwise exist if there are lots of JVMs, while the hints don't do anything useful. Signed-off-by: Aleksandar Micic <Aleksandar_Micic@ca.ibm.com>
|
@hangshao0 fyi |
|
Not sure if you want to mention this in the documentation: https://eclipse.dev/openj9/docs/xxusegcstartuphints/ |
dmitripivkine
left a comment
dmitripivkine
left a comment
There was a problem hiding this comment.
If we decide to adjust something else except initial heap size this decision can be revisited.
|
Jenkins test sanity xLinux jdk21 |
This is a performance optimization, and not changing behaviour/functionality. I don't feel a need to document that. Especially, because it will apply only after a specific release, and I don't want to go into those details. That said, this sentence is not completely clear: Perhaps, we want to add that even if -Xms is set, the overridden value can only be larger than -Xms (what also implies that if heap is fully expanded, it will not be overridden, but that part might be relatively obvious and does not have be mentioned). Does anyone (including @pshipton) think that extra explanation is useful/needed? |
|
Jenkins test sanity pLinux jdk21 |
|
Jenkins test sanity xLinux jdk21 |
|
Clearer documentation is better, I don't see why we wouldn't want to add that "the overridden value can only be larger". |
|
We do record performance highlights for each release on the web site, we may want to add a sentence to explain the improvement. |
|
eclipse-openj9/openj9-docs#1319 Yes, I'm ok with mentioning the perf improvement in release notes. |
|
When you are ready, pls write a sentence for the release notes, or open an issue at https://github.com/eclipse-openj9/openj9-website/issues/ with the detail. |
4 participants
This removes the GC stress on shared cache, that may otherwise exist if there are lots of JVMs, while the hints don't do anything useful.