[8u] RFR: 8227006: [linux] Runtime.availableProcessors execution time increased by factor of 100

[Severin Gehwolf]

Hi Jon,

On Tue, 2020-11-10 at 13:51 +0000, Jonathan Dowland wrote:
> Original bug: https://bugs.openjdk.java.net/browse/JDK-8227006
> Original rev: https://hg.openjdk.java.net/jdk/jdk/rev/c440a6b4e096
> 
> RFR webrev: https://cr.openjdk.java.net/~shade/8227006/webrev.8u.01/

This backport looks good to me. I can sponsor the push for you and add
the label/comment.

> Please review this backport for jdk8u-dev for parity with Oracle JDK 8u281.
> 
> The original patch did not apply cleanly to jdk8u-dev. I had to untangle
> it from related changes from JDK-8232207.

Right. JDK-8232207 doesn't affect OpenJDK 8u so we shouldn't backport
it. Dropping changes to CgroupMemorySubsystem is the right call.

>  I also re-worked a call to
> "log_trace" to use "tty->print_cr" conditionally on (PrintContainerInfo)
> set, matching other uses in the file.

OK.

> One can see the cache in operation by executing the following and
> comparing the output before/after:
> 
>     java -XX:+UnlockDiagnosticVMOptions -XX:PrintContainerInfo -version

I've also run the reproducer from the bug.

Before:
-------
$ java Test
72448 calls/sec
71024 calls/sec
61751 calls/sec
59997 calls/sec
68613 calls/sec
70358 calls/sec

After:
-------
$ java Test
25218164 calls/sec
26279378 calls/sec
26165449 calls/sec
24962446 calls/sec
24157175 calls/sec
24020916 calls/sec
25469665 calls/sec

Thanks,
Severin