Add Deprecated warning for using Xlp, and Xlp<size> using Java 11+

[AlenBadel]

Adding warning for any users using -Xlp, and -Xlp<size> using Java 11+.
Per Docs, these options are slated to be deprecated in all versions above Java 8. Since the functionality needs to be supported for Java 8, this should be a warning until support for Java 8 officially ends.

See: https://www.eclipse.org/openj9/docs/xlp/

Signed-off-by: AlenBadel [email protected]

[AlenBadel]

FYI @pshipton

[pshipton]

What is the equivalent option to -Xlp?

Not sure if there are still tests which use these options. If there are, the tests may start failing due to the warning.

[AlenBadel]

What is the equivalent option to -Xlp?

Well the doc mentions to use -Xlp:codecache:pagesize=<size>, and -Xlp:objectheap:pagesize=<size>. Which is equivalent to -Xlp<size>.
However, -Xlp has no equivalent.
The proposed -XX:+UseLargePages is the closest impact. See: #4930.

Not sure if there are still tests which use these options. If there are, the tests may start failing due to the warning.

I'll double check sanity on my side to verify this.

[pshipton]

Seems we should get -XX:+UseLargePages merged, and update any tests using -Xlp to use it before starting to print a warning.

[pshipton]

I'll double check sanity on my side to verify this.

We also have extended.functional testing that may use it.

[andrewcraik]

-Xlp:codecache:pagesize=, and -Xlp:objectheap:pagesize= are the same? They are talking about the JIT codecache and the heap - I didn't think they were the same and could be adjusted independently... FYI @vijaysun-omr

[AlenBadel]

-Xlp:codecache:pagesize=, and -Xlp:objectheap:pagesize= are the same? They are talking about the JIT codecache and the heap - I didn't think they were the same and could be adjusted independently... FYI @vijaysun-omr

The parsing code translates -Xlp<size> to exactly -Xlp:codecache:pagesize=<size> and -Xlp:objectheap:pagesize=<size>. So they should be exactly equivalent.

NOTE: On Z/OS the equivalent assortment would be -Xlp:codecache:pagesize=<size>,pageable and -Xlp:objectheap:pagesize=<size>,nonpageable.

-Xlp itself sets the page size to a series of pre-configured sizes if that page is provisioned on the system. This option only applies to the objectheap, and has no impact on the codecache. While the proposed -XX:+UseLargePages would equivalently change the objectheap size, additionally changing the codecache page size in the same fashion.

The only other difference is that on Z/OS. Where -Xlp1M utilized fixed 1M pages over pageable. That behaviour would cease with the deprecation of this option. Since the proposed -XX:+UseLargePages would favor pageable over fixed.

[dmitripivkine]

The parsing code translates -Xlp<size> to exactly -Xlp:codecache:pagesize=<size> and -Xlp:objectheap:pagesize=<size>. So they should be exactly equivalent.

This is not accurate statement. For some platforms (ZOS for instance) -Xlp<size> and '-Xlp:objectheap:pagesize=' behaves differently.

[pshipton]

@joransiu fyi

[AlenBadel]

This is not accurate statement. For some platforms (ZOS for instance) -Xlp<size> and '-Xlp:objectheap:pagesize=' behaves differently.

Could you elaborate?

[dmitripivkine]

This is not accurate statement. For some platforms (ZOS for instance) -Xlp<size> and '-Xlp:objectheap:pagesize=' behaves differently.

Could you elaborate?

Sorry, don't mind please. I thought about omitted parts of -Xlp like <size> or pageable. The only formal difference on ZOS is -Xlp:objectheap:pagesize= must have [non]pageable modifier which old format -Xlp does not have

[AlenBadel]

Not sure if there are still tests which use these options. If there are, the tests may start failing due to the warning.

functional sanity/extended system sanity/extended seems to pass on my end with the warning.

This is simply a warning that echos what has already been published in our docs. I don't think we have any strong justification to wait on merging this given our tests haven't broken.

[pshipton]

It doesn't seem reasonable to print a warning for a deprecated option when the replacement option -XX:+UseLargePages is not yet available. What action to you expect someone to take when they see the warning?

[genie-openj9]

Can one of the admins verify this patch?