RFR: JDK-8222333: fastdebug build broken after JDK-8221393
Ao Qi
aoqi at loongson.cn
Thu Apr 11 12:33:09 UTC 2019
On Thu, Apr 11, 2019 at 8:09 PM Zhengyu Gu <zgu at redhat.com> wrote:
>
>
>
> On 4/11/19 7:51 AM, Ao Qi wrote:
> > Thanks Zhengyu. Updated:
> > http://cr.openjdk.java.net/~aoqi/8222333/webrev.01/ .
> > linux-x86_64-server-release hotspot:tier1_gc tested.
>
> Sorry if I did not explain clearly.
>
> (2) just an example. You would have to do the same for phases that start
> with following comments:
>
> /* Per-thread timer block, should have "roots" counters in consistent
> order */
>
> such as, scan_roots, update_roots, init_evac and etc.
Sorry, my mistake. Updated: http://cr.openjdk.java.net/~aoqi/8222333/webrev.02/
Is other modifications needed?
$ grep "Per-thread timer block"
src/hotspot/share/gc/shenandoah/shenandoahPhaseTimings.hpp
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
/* Per-thread timer block, should have "roots" counters in
consistent order */ \
$ grep "stringtable_roots"
src/hotspot/share/gc/shenandoah/shenandoahPhaseTimings.hpp
f(scan_stringtable_roots, " S: String
Table Roots") \
f(update_stringtable_roots, " U: String
Table Roots") \
f(evac_stringtable_roots, " E: String
Table Roots") \
f(final_update_stringtable_roots, " UR: String
Table Roots") \
f(degen_gc_update_stringtable_roots, " DU: String
Table Roots") \
f(init_traversal_gc_stringtable_roots, " TI: String
Table Roots") \
f(final_traversal_gc_stringtable_roots, " TF: String
Table Roots") \
f(final_traversal_update_stringtable_roots, " TU:
String Table Roots") \
f(full_gc_stringtable_roots, " F: String
Table Roots") \
$ grep "resolved_method_table_roots"
src/hotspot/share/gc/shenandoah/shenandoahPhaseTimings.hpp
f(scan_resolved_method_table_roots, " S: Resolved
Table Roots") \
f(update_resolved_method_table_roots, " U: Resolved
Table Roots") \
f(evac_resolved_method_table_roots, " E: Resolved
Table Roots") \
f(final_update_resolved_method_table_roots, " UR: Resolved
Table Roots") \
f(degen_gc_update_resolved_method_table_roots, " DU: Resolved
Table Roots") \
f(init_traversal_gc_resolved_method_table_roots, " TI: Resolved
Table Roots") \
f(final_traversal_gc_resolved_method_table_roots, " TF: Resolved
Table Roots") \
f(final_traversal_update_resolved_method_table_roots, " TU:
Resolved Table Roots") \
f(full_gc_resolved_method_table_roots, " F: Resolved
Table Roots") \
>
> Thanks,
>
> -Zhengyu
>
>
> >
> > On Thu, Apr 11, 2019 at 7:25 PM Zhengyu Gu <zgu at redhat.com> wrote:
> >>
> >> Hi Ao Qi,
> >>
> >> 1) Style does not match
> >> f(StringTableRoots, "StringTable Roots(ms):") \
> >> + f(Resolved_Method_TableRoots, "Resolved_Method_Table Roots(ms):")
> >>
> >> 'Resolved_Method_TableRoots' => ResolvedMethodTableRoots
> >>
> >>
> >> 2) This is a sub-phase, you also need corresponding phases, For example:
> >>
> >> f(scan_jfr_weak_roots, " S: JFR Weak Roots") \
> >> f(scan_jni_weak_roots, " S: JNI Weak Roots") \
> >> f(scan_stringtable_roots, " S: String Table Roots") \
> >> f(scan_resolved_method_table_roots, " S: Resolved Table Roots") \
> >> f(scan_vm_weak_roots, " S: VM Weak Roots") \
> >>
> >> Thanks,
> >>
> >> -Zhengyu
> >>
> >> On 4/11/19 7:07 AM, Ao Qi wrote:
> >>
> >> Hi,
> >>
> >> Bugs:
> >> https://bugs.openjdk.java.net/browse/JDK-8222333
> >>
> >> Webrev:
> >> http://cr.openjdk.java.net/~aoqi/8222333/webrev.00/
> >>
> >> fastdebug build is broken after JDK-8221393. It seems that
> >> phase_mapping[] doesn't match enum Phase in WeakProcessorPhases
> >> anymore. Is this fix ok? I am not familiar with shenandoah.
> >> linux-x86_64-server-release hotspot:tier1 tested.
> >>
> >> Thanks,
> >> Ao Qi
More information about the hotspot-gc-dev
mailing list