RFR: JDK-8213746: GC/C2 abstraction for C2 matcher
Roman Kennke
rkennke at redhat.com
Mon Nov 12 20:43:47 UTC 2018
Ha! Good point!
Updated webrevs accordingly:
Incremental:
http://cr.openjdk.java.net/~rkennke/JDK-8213746/webrev.01.diff/
Full:
http://cr.openjdk.java.net/~rkennke/JDK-8213746/webrev.01/
Thanks for pointing this out!
Roman
Am 12.11.18 um 20:39 schrieb Milan Mimica:
> Hi
>
> matcher.cpp
> + bool cont = false;
> + if (!gc_handled) {
> + cont = find_shared_visit(mstack, n, nop, mem_op, mem_addr_idx);
> + }
> + if (cont) {
> + continue;
> + }
>
> It could just continue immediately, without the local variable.
>
>
>
> pon, 12. stu 2018. u 20:30 Roman Kennke <rkennke at redhat.com
> <mailto:rkennke at redhat.com>> napisao je:
>
> There's a bunch of GC specific code in matcher.cpp that should better
> be abstracted into BarrierSetC2 and GC specific subclasses.
>
> Some notes about the proposed change:
> - It hooks GC up to both the 'visit' and 'post-visit' phase of the node
> iteration loops. ZGC currently inserts into visit, and Shenandoah
> requires both vist and post-visit. The latter is needed because
> Shenandoah provides its own set of CAS nodes which need to be
> binary-treed.
> - There's also a hook in post_store_load_barrier() which Shenandoah
> needs, again for its own set of CAS nodes
> - The switches are factored out into their own methods to make the GC
> hookup clearer and easier to manage.
>
> Bug:
> https://bugs.openjdk.java.net/browse/JDK-8213746
> Webrev:
> http://cr.openjdk.java.net/~rkennke/JDK-8213746/webrev.00/
>
> Testing: passes tier
>
> Thoughts? Reviews?
>
> Thanks,
> Roman
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20181112/163a122e/signature-0001.asc>
More information about the hotspot-compiler-dev
mailing list