RFR(xs): 8244733: Add ResourceHashtable::compute_if_absent

[Thomas Stüfe]

Hi Ioi,

Thank you. I planned adapt CDS next :) but if possible I'd like to get this
patch in since it already ran through our CI in this form.

The first caller changed will be
https://bugs.openjdk.java.net/browse/JDK-8244777, see:
http://cr.openjdk.java.net/~stuefe/webrevs/8244777--reduce-classloaderstats-vm-operation-runtime/webrev.00/webrev
<http://cr.openjdk.java.net/~stuefe/webrevs/8244777--reduce-classloaderstats-vm-operation-runtime/webrev.00/webrev/src/hotspot/share/classfile/classLoaderStats.cpp.udiff.html>
.
, but since that may require more JFR-specific changes which have nothing
to do with ResourceHashTable I do it in a separate change.

Cheers, Thomas

On Tue, May 12, 2020 at 6:35 AM Ioi Lam <ioi.lam at oracle.com> wrote:

> Hi Thomas,
>
> This is something that I've been wanting.
>
> Would it be possible for you to change the VM code so we will have at
> least one example of using it for real? How about here :-)
>
>    DumpTimeSharedClassInfo* find_or_allocate_info_for(InstanceKlass* k) {
>      DumpTimeSharedClassInfo* p = get(k);
>      if (p == NULL) {
> assert(!SystemDictionaryShared::no_class_loading_should_happen(),
>               "no new classes can be loaded while dumping archive");
>        put(k, DumpTimeSharedClassInfo());
>        p = get(k);
>        assert(p != NULL, "sanity");
>        p->_klass = k;
>      }
>      return p;
>    }
>
>
> Thanks
> - Ioi
>
> On 5/11/20 9:25 PM, Thomas Stüfe wrote:
> > Hi all,
> >
> > May I please have reviews for this enhancement:
> >
> > JBS: https://bugs.openjdk.java.net/browse/JDK-8244733
> > Webrev:
> >
> http://cr.openjdk.java.net/~stuefe/webrevs/8244733--add-resourcehashtable--compute_if_absent/webrev.00/webrev/
> >
> > A common pattern when using ResourceHashTable is:
> >
> > if (table.get(k) == NULL) table.add(k,v);
> >
> > This runs the lookup code twice. By providing a new method which combines
> > these two operations we would save one lookup.
> >
> > The new method is named ResourceHashTable::compute_if_absent() and
> behaves
> > very similar to the similarly named variant in j.u.Map.
> >
> > Note: Actually replacing caller code will happen in a separate RFE. The
> > first caller to use this will be the ClassLoaderStats VM operation - I
> was
> > able to significantly reduce runtime for that vm op. But since that is
> JFR
> > territory and this is more of a runtime topic I wanted to keep the review
> > separate.
> >
> > Thanks, Thomas
>
>