[7u6] RFR (JAXP) : 7151118: Regressions on 7u4 b11 comp. 7u4 b06 on specjvm2008.xml.transform subbenchmark
I tested through all of the Xalan updates and found XalanJ-2271 to be the culprit. It [1] was intended to improve performance in webserver environment while resolving unescaped chars in attributes. Unfortunately, it actually resulted in a 5% drop in specjvm2008. I tried to improve the patch, by for example not making mutable copies of the CharInfo objects and etc., but none could get all of the performance loss back. Since it was a proactive update, meaning no one requested it, instead of spending too much time on this issue, I decided to roll the entire patch back. Below is the webrev [2]. [1] https://issues.apache.org/jira/browse/XALANJ-2271 [2] http://cr.openjdk.java.net/~joehw/7u6/cr7151118/webrev/ All unit/sqe/tck tests passed. Please review. Thanks, Joe
I am ok with it based on a once-over and the fact you are passing all of the tests Best Lance On Jun 5, 2012, at 11:23 PM, Joe Wang wrote:
I tested through all of the Xalan updates and found XalanJ-2271 to be the culprit. It [1] was intended to improve performance in webserver environment while resolving unescaped chars in attributes. Unfortunately, it actually resulted in a 5% drop in specjvm2008. I tried to improve the patch, by for example not making mutable copies of the CharInfo objects and etc., but none could get all of the performance loss back.
Since it was a proactive update, meaning no one requested it, instead of spending too much time on this issue, I decided to roll the entire patch back. Below is the webrev [2].
[1] https://issues.apache.org/jira/browse/XALANJ-2271 [2] http://cr.openjdk.java.net/~joehw/7u6/cr7151118/webrev/
All unit/sqe/tck tests passed.
Please review.
Thanks, Joe
Lance Andersen| Principal Member of Technical Staff | +1.781.442.2037 Oracle Java Engineering 1 Network Drive Burlington, MA 01803 Lance.Andersen@oracle.com
participants (2)
-
Joe Wang
-
Lance Andersen - Oracle