RFR: Remove racy assert in ShenandoahResetNextMarkBitmapTraversalTask

Aleksey Shipilev shade at redhat.com
Tue Sep 11 21:13:33 UTC 2018


On 09/11/2018 06:04 PM, Roman Kennke wrote:
> Right. The following makes both the order in cleanup and the
> reset-next-markbitmap loop consistent with adaptive:
> 
> Incremental:
> http://cr.openjdk.java.net/~rkennke/fixtamsassert/webrev.01.diff/
> Full:
> http://cr.openjdk.java.net/~rkennke/fixtamsassert/webrev.01/

Looks good.

Seems to me ShenandoahResetNextBitmapTraversalTask should never experience trashed regions now.
Which means we can simplify it like below, and this will also solidify the expectation that trashed
regions recycling should happen before resetting the bitmaps:
  http://cr.openjdk.java.net/~shade/shenandoah/traversal-resetbitmaps-1.patch

This passes tier3_gc_shenandoah on x86_64.

-Aleksey




More information about the shenandoah-dev mailing list