RFR: 8199407: Skip Rebuild Remset Phase if there are no rebuild candidates
Thomas Schatzl
tschatzl at openjdk.java.net
Mon Apr 12 13:33:41 UTC 2021
On Mon, 12 Apr 2021 11:14:20 GMT, Ivan Walulya <iwalulya at openjdk.org> wrote:
> Hi all,
>
> Please review this change that skips the rebuild remset phase and the associated updates after rebuild, If during Remark, G1 finds that there are no remembered sets to rebuild.
>
> Test: tier1-5
Changes requested by tschatzl (Reviewer).
src/hotspot/share/gc/g1/g1ConcurrentMark.cpp line 1330:
> 1328: _g1h->heap_region_iterate(&cl);
> 1329: } else {
> 1330: log_debug(gc, remset, tracking)("No Remembered Sets To Update After Rebuild");
Maybe make this `gc+phases=debug` too like the one above so that either of those shows up with the same log options. Also there does not seem to be a reason for the upper casing here. (The argument to `GCTraceTime` is a phase name, but this seems like a regular statement)
src/hotspot/share/gc/g1/g1ConcurrentMark.cpp line 1962:
> 1960: // skip the rebuild remembered set phase
> 1961: if (total_selected_for_rebuild() == 0) {
> 1962: log_debug(gc, remset, tracking)("Skipping Remembered Set Rebuild. No Regions Selected For Rebuild.");
Maybe better is `log_debug(gc, marking)` here, somewhat fitting with the other messages when running with `gc+marking=debug`. Also the upper casing the in the second sentence seems weird, `No regions selected for rebuild` seems better (`Remembered Set Rebuild` is the name of that phase, and we do the upper casing for them already. But the second statement seems to be a regular statement)
-------------
PR: https://git.openjdk.java.net/jdk/pull/3434
More information about the hotspot-gc-dev
mailing list