RFR: 8302215: G1: Last-ditch Full GC should do serial compaction for tail regions in per thread compaction points.

[Albert Mingkun Yang]

On Mon, 13 Feb 2023 10:00:20 GMT, Ivan Walulya <iwalulya at openjdk.org> wrote:

> Hi all, 
> 
> Please review this change to call `G1FullCollector::phase2c_prepare_serial_compaction` when `do_maximal_compaction`. For maximum compaction, we re-prepare all objects in the "tail" regions of the per thread compaction points.
> 
> Testing: Tier 1-3.
> 
> Thanks

Changes requested by ayang (Reviewer).

src/hotspot/share/gc/g1/g1FullCollector.cpp line 382:

> 380:   // point. To get maximum compaction and reduce fragmentation, we sort the regions
> 381:   // by hrm_index so that we compact objects to one end of the heap.
> 382:   serial_cp->sort_regions();

I think the intention can be made more explicit by adding higher-index regions to `serial_cp` in order, sth like:


for (w : workers) {
  w->cp->filter(< lowest_current_hr);
}

for i in [lowest_current_hr..max) {
  r = region_at(i);
  if r->is_pinned
    continue;
  serial_cp->add(i);
}

src/hotspot/share/gc/g1/g1FullGCPrepareTask.inline.hpp line 118:

> 116:     // We skip objects compiled into the first region or
> 117:     // into regions not part of the serial compaction point.
> 118:     if (target_hr->hrm_index() <= _bottom_index) {

Instead of comparing region-index, I wonder if one can use addresses directly, sth like `if (forwardee < serial_cp_limit)`.

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

PR: https://git.openjdk.org/jdk/pull/12529