RFR: JDK-8317683: Add JIT memory statistics [v5]

[Thomas Stuefe]

On Fri, 13 Oct 2023 15:30:27 GMT, Brice Dutheil <bdutheil at openjdk.org> wrote:

> > anon mappings were unnamed
> 
> > this type of investigation workflow:
> > 
> > * there is a report of increased RSS
> > * we check first the `pmap -X` output to find if C1/C2 compiler threads anon mappings and diff them from the good version
> > * if we have an increase there we go down into the rabbit hole with the JIT statistics you provided
> 
> I can relate to that in more ways than only compiler threads.
> 
> Using `prctl` / `PR_SET_VMA_ANON_NAME` approach looks interesting though albeit unavailable on other than Linux. I never used this mechanism before but using this may imply possible issues with glibc's ability to merge adjacent VMAs ; could it be avoided if the name is the "same", i.e. making `malloc` use the the same mmaped area under the hood and reported as such in `pamp -X`.
> 
> _Note: This is just exposing unproven ideas at this point, maybe opening ticket makes more sense to discuss this if the idea is indeed interesting / doable._

Yes, issues with VMA folding, and possibly interfere with THPs as well (should different names preclude folding).

But yeah, lets keep these discussions out of this PR, this is better discussed with the glibc people anyway.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/16076#issuecomment-1761722186