RFR (M): (7u60): Backport of 8027476: Improve performance of Stringtable unlink, 8027455: Improve symbol table scan times during gc pauses (original) (raw)

Thomas Schatzl thomas.schatzl at oracle.com
Thu Jan 30 12:05:25 UTC 2014


Hi,

On Wed, 2014-01-29 at 11:55 -0800, Jon Masamitsu wrote:

Thomas,

Changes look good.

thanks for the review :)

Please run some simple test (does enough work to exercise your changes) with -XX:+UseDynamicNumberOfGCThreads .

G1 seems to crash in all recent builds for 7u60, 8 and 9 when enabling UseDynamicNumberOfGCThreads at the first (young) GC. I filed a bug for that (https://bugs.openjdk.java.net/browse/JDK-8033228).

I did not assign it to you, because this issue seems somewhat unrelated though. E.g. 7u60 and 8 do not have these recent changes at all. I will look into it today, and if I find anything interesting I will ping you.

The crash in debug mode seems to be caused by a general problem with the logging (g1gcphasetimes) we have: it does not support the case that when measuring the separate per-thread timings one thread does not report any time. The other issue that shows up in product builds I have no idea about right now.

Thomas



More information about the hotspot-gc-dev mailing list