Marius Pirvu

Limit the amount of memory used by JITServer AOT cache

1

Currently there is no upper bound on the amount of memory used by the JITServer AOT cache. Add a configurable limit and stop adding new entries once the limit has...

`-XX:+JITServerLocalSyncCompiles` is a JITServer specific option that allows the client JVM to perform synchronous compilations locally, at cold opt level, and follow up with remote recompilations at higher opt levels....

User provided OPENJ9_JAVA_OPTIONS env var overrides OpenJ9 SCC options

8

https://github.com/AdoptOpenJDK/openjdk-docker/pull/338 added the ability to embed a shared-class-cache (SCC) in OpenJ9 containers. This OpenJ9 JVM feature is enabled by setting the following env var: ``` ENV OPENJ9_JAVA_OPTIONS="-Xshareclasses:name=openj9_system_scc,cacheDir=/opt/java/.scc,readonly,nonFatal" ``` However, if...

Allow JITServer AOT cache to work without a shared class cache at the client

30

Currently, the JITServer AOT cache feature requires the client JVM to use a shared class cache (SCC) with some empty space and write permissions. This is because the AOT code...

When deploying JITServer with the helm chart I see the following warning ``` W0513 13:23:01.453840 84779 warnings.go:70] spec.template.spec.affinity.nodeAffinity.requiredDuringSchedulingIgnoredDuringExecution.nodeSelectorTerms[0].matchExpressions[0].key: beta.kubernetes.io/arch is deprecated since v1.14; use "kubernetes.io/arch" instead ``` We should make...

JLM (java lock monitor) gives us a summary of activity on monitors, whereas MonitorContended events are triggered for every monitor individually and they also can be used to compute the...

Changes to the code cache repository allocation code

2

In order to eliminate helper trampolines, OpenJ9 tries to allocate the code cache in the vicinity of the JIT dll. This intention is signalled by using a `preferredStartAddress` when calling...

Existing code disclaims the cold code between `coldCodeAlloc` and `coldCodeAllocBase`. However, if the amount of cold code is larger than the amount of warm code, we will ask to disclaim...

Eliminate redundant vlog lock

2

Changed TR_VerboseLog::writeLineLocked() into TR_VerboseLog::writeLine() because we are already in a TR_VerboseLog::CriticalSection

7% AcmeAirEE8 throughput regression in 1P mode

19

I spotted a 7% throughput regression when running AcmeAirEE8 in containers limited to 1 CPU. I limited the regression to the following two nightly builds: OpenJ9-JDK17-x86-64_linux-20240725-231127.tar.gz ``` openjdk version "17.0.12-internal"...