RFR: Out-of-cycle Degenerated GC should process references and unload classes

[Roman Kennke]

Good.

Am 29. August 2018 10:49:55 MESZ schrieb Aleksey Shipilev <shade at redhat.com>:
>Hopefully the comment describes the rationale.
>
>diff -r 90a268e66f81 src/hotspot/share/gc/shenandoah/shenandoahHeap.cpp
>--- a/src/hotspot/share/gc/shenandoah/shenandoahHeap.cpp        Tue Aug
>28 07:58:08 2018 -0400
>+++ b/src/hotspot/share/gc/shenandoah/shenandoahHeap.cpp        Wed Aug
>29 10:22:59 2018 +0200
>@@ -1800,10 +1800,21 @@
>       op_final_traversal();
>       op_cleanup_traversal();
>       return;
>
>     case _degenerated_outside_cycle:
>+      // We have degenerated from outside the cycle, which means
>something is bad with
>+      // the heap, most probably heavy humongous fragmentation, or we
>are very low on free
>+      // space. It makes little sense to wait for Full GC to reclaim
>as much as it can, when
>+      // we can do the most aggressive degen cycle, which includes
>processing references and
>+      // class unloading, unless those features are explicitly
>disabled.
>+      //
>+      // Note that we can only do this for "outside-cycle" degens,
>otherwise we would risk
>+      // changing the cycle parameters mid-cycle during concurrent ->
>degenerated handover.
>+      set_process_references(ShenandoahRefProcFrequency != 0);
>+      set_unload_classes(ClassUnloading);
>+
>       if (heuristics()->can_do_traversal_gc()) {
>// Not possible to degenerate from here, upgrade to Full GC right away.
>         cancel_gc(GCCause::_shenandoah_upgrade_to_full_gc);
>         op_degenerated_fail();
>         return;
>
>Testing: tier3_gc_shenandoah (running)
>
>Thanks,
>-Aleksey

-- 
Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.