RFR: 8303244: G1: call CardTable::clear_MemRegion directly

Kim Barrett kbarrett at openjdk.org
Tue Feb 28 09:06:07 UTC 2023


On Mon, 27 Feb 2023 13:06:45 GMT, Albert Mingkun Yang <ayang at openjdk.org> wrote:

> Simple refactoring to skip a guard in card-table clearing.
> 
> Test: hotspot_gc

I think the code changes are okay.

But while looking at it I wondered what the difference between CT::clear and
CT::clear_MemRegion might be. The former has no description. The latter has a
description that seems wrong.

I also wonder why dirty_MemRegion uses an explicit loop over the cards while
clear_MemRegion uses memset. It seems like that difference has been there
since before the mercurial age.

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

PR: https://git.openjdk.org/jdk/pull/12764


More information about the hotspot-gc-dev mailing list