Re: Strange Behaviour (with Test Case)

From: Stefan Koehler <contact_at_soocs.de>
Date: Thu, 6 Aug 2015 01:04:04 +0200 (CEST)
Message-ID: <1278008383.2463.1438815844454.JavaMail.open-xchange_at_app04.ox.hosteurope.de>

Hi Tanel,
maybe a little OT, but i don't want to miss the chance as you already have given a basic description about rowsets.

Martin Bach asked me about the rowset spec in the column projection information some time ago and i could not figure out (or find) any details about it. The question was in a different context (FTS in his case), but the only stuff i could figure out was event 10055 and some hidden parameters to control it. You can find my test case here: http://www.soocs.de/public/research/150120_dbms_xplan_count_rowset.txt

Do you have any more details about the implementation of this pipeline optimization (rowsets)?  

Thank you very much.

Best Regards
Stefan Koehler

Freelance Oracle performance consultant and researcher Homepage: http://www.soocs.de
Twitter: _at_OracleSK  

> Tanel Poder <tanel_at_tanelpoder.com> hat am 5. August 2015 um 23:37 geschrieben:
>
> Just in case you're wondering what the heck this is - rowsets are a yet another execution plan pipeline optimization to save a few CPU cycles,
> which is apparently buggy. The CASE statements are evaluated in EVA functions, thus 0x200 should help (helped on my exadata test).

--
http://www.freelists.org/webpage/oracle-l

Received on Thu Aug 06 2015 - 01:04:04 CEST