[KinoSearch] KinoSearch 0.30_01

[Marvin Humphrey]

On Thu, Jun 25, 2009 at 09:32:41PM -0500, Peter Karman wrote:

> w00t!!

:)

> any chance to update the benchmarks to quantify the improvements?

The core has gotten faster -- here are some results from the indexing
benchmark test, which uses a whitespace tokenizer:

    KinoSearch 0.165       31.60 seconds
    KinoSearch 0.20_051    14.76 seconds
    KinoSearch 0.30_01     10.71 seconds

However, in practice most people will see a larger speedup than that because
most apps use a PolyAnalyzer which includes a Stemmer, and Stemmer is much
faster in 0.30_01.  

As far as updating the comparative benchmarking page on rectangular.com... I'm
not planning to invest the time in that until we get to release candidates for
0.30_xx replacing the stable branch.  I'll have to update the Lucene app, and
doing scientifically responsible benchmarking -- as opposed to
"benchmarketing" -- is time-consuming hard work.

Marvin Humphrey

------------------------------------------------------------
1    Secs: 10.546  Docs: 19043
2    Secs: 10.834  Docs: 19043
3    Secs: 10.711  Docs: 19043
4    Secs: 10.545  Docs: 19043
5    Secs: 11.620  Docs: 19043
6    Secs: 10.753  Docs: 19043
------------------------------------------------------------
KinoSearch 0.30_01 
Perl 5.10.0
Thread support: no
Linux 2.6.18-92.1.6.el5PAE i686
Mean: 10.835 secs 
Truncated mean (4 kept, 2 discarded): 10.711 secs
------------------------------------------------------------


------------------------------------------------------------
1    Secs: 15.66  Docs: 19043
2    Secs: 14.87  Docs: 19043
3    Secs: 14.97  Docs: 19043
4    Secs: 14.59  Docs: 19043
5    Secs: 14.63  Docs: 19043
6    Secs: 14.52  Docs: 19043
------------------------------------------------------------
KinoSearch 0.20_051 
Perl 5.10.0
Thread support: no
Linux 2.6.18-92.1.6.el5PAE i686
Mean: 14.87 secs 
Truncated mean (4 kept, 2 discarded): 14.76 secs
------------------------------------------------------------


------------------------------------------------------------
1    Secs: 32.76  Docs: 19043
2    Secs: 31.61  Docs: 19043
3    Secs: 31.24  Docs: 19043
4    Secs: 31.75  Docs: 19043
5    Secs: 30.92  Docs: 19043
6    Secs: 31.81  Docs: 19043
------------------------------------------------------------
KinoSearch 0.165 
Perl 5.10.0
Thread support: no
Linux 2.6.18-92.1.6.el5PAE i686
Mean: 31.68 secs 
Truncated mean (4 kept, 2 discarded): 31.60 secs