lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael McCandless <luc...@mikemccandless.com>
Subject Re: FST construction on jrockit, 1.6 and 1.7
Date Tue, 26 Jul 2011 12:54:03 GMT
Y axis is time to build the FST (ie, smaller is better), and X axis is
iteration?

1.6 and 1.7 are very close, but JRockit is much slower except for
iteration #2 where it's nearly the same.  Scary how hotspot can be so
wrong so "randomly".

Mike McCandless

http://blog.mikemccandless.com

On Mon, Jul 25, 2011 at 5:18 PM, Dawid Weiss <dawid.weiss@gmail.com> wrote:
> I am preparing an unrelated presentation to which I needed something
> that crunches lots of data and uses some cpu cycles. I decided to run
> FST construction on that half-gig wikipedia file. Three jits: 1.6_u26,
> 1.7 pre-release and jrockit  (newest available). The results are
> interesting (although not by any means indicative of the advantages of
> each vm, as I only did it on a single machine and with only 5 reps per
> vm): 1.6_u26 is ahead of 1.7, but watch what happens with the timing
> of jrockit (no, the bumps are not caused by background activity). I'm
> guessing the optimizer picking wrong paths and unable to revert, but
> who the hell knows these days...
>
> Dawid
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: dev-help@lucene.apache.org
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Mime
View raw message