I've been staring at it for the last few minutes trying to work out why the
larger set of figures isn't exactly 5 times the smaller set of figures.
One is too large, the other too small
fact_0: 3 -> 13 (short by 2)
fact_dim_1: 4 -> 28 (long by 8)
Some effects of delayed block cleanout / consistent read possibly - how many
times did you run the test, does the order of the tests affect the result.
Some effects of buffer pinning for the synchronised walks of three indexes
maybe - but that would surely be highly coincidental in a small data set.
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
@jloracle
________________________________
From: oracle-l-bounce@xxxxxxxxxxxxx [oracle-l-bounce@xxxxxxxxxxxxx] on behalf
of Charles Schultz [sacrophyte@xxxxxxxxx]
Sent: 06 January 2016 21:14
To: Sayan Malakshinov
Cc: ORACLE-L
Subject: Re: Trying to understand the BITMAP AND operation a bit better
Sayan,
That is an excellent observation - I knew I had been staring at this for way
too long! :) And now I feel kinda stupid.... oh well. :)
Thanks much.
On Wed, Jan 6, 2016 at 3:08 PM, Sayan Malakshinov
<xt.and.r@xxxxxxxxx<mailto:xt.and.r@xxxxxxxxx>> wrote:
Hi Charles,
Note that 7-11 lines have 5 "Starts" in the first plan, because previous
steps(NL to FTS of DIMENSION2) returned 5 rows, but there is only 1 "Start" in
second plan for these operations.
So each of BITMAP INDEX SINGLE VALUE in first plan executed 5 times instead of
1 in the second plan.
--
Best regards,
Sayan Malakshinov
http://orasql.org
--
Charles Schultz