|
Boost : |
From: Jose (jmalv04_at_[hidden])
Date: 2005-10-17 09:06:28
Hi Arkadiy,
I've got the RML example for the new query and it is in the miliseconds
range.
There is one optimization that you would need to add so that we can compare
RML vs RTL:
before running the query you can create a 2-dimensional map with the city
locations so that when you check the distances for each city you don't use
compute the distance for all points but you limit the calculation to a range
of cities that are below a cutoff distance (where the cutoff distance is
greater than 5 km)
Regards
Jose
On 10/17/05, Arkadiy Vertleyb <vertleyb_at_[hidden]> wrote:
>
> Hi All.
>
> I updated the recently discussed RTL example to demonstrate the
> performance
> gain that can be achieved by using the RTL transaction mechanism.
>
> The original query, that involves a self-join of the ~20,000 row table on
> a
> complex condition, currently takes ~79 sec. on my setup (1.5 Ghz, MinGW,
> g++ -O2). Then a record is removed from the table, which invalidates all
> the indexes built by that point. The transaction commit repairs indexes,
> and the query is re-run.
>
> Update, commit, and re-run together take just 1 sec.
>
> Please take the example and the updated library (again, a few performance
> issues were fixed) at the usual location:
>
>
> http://www.boost-consulting.com/vault/index.php?&direction=0&order=&director
> y=RTL
>
> Please note that the directory structure of the library has been
> "boostified" so please ajust you include path accordingly (although I
> haven't yet put "boost/rtl" into #include statements, so it should go into
> the include path for now).
>
> Regards,
> Arkadiy
>
>
>
>
>
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost
>
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk