|
Geometry : |
Subject: [ggl] spacial index
From: Adam Wulkiewicz (adam.wulkiewicz)
Date: 2011-03-26 22:44:20
Bruno Lalande wrote:
> Hi Adam,
>
> Thanks for this new version of your rtree, I'm reviewing it right now.
> Not sure I can finish this week end, but so far it looks very good. The
> use of visitors for all operations separates very well structure from
> implementation, and makes the central code very concise. I also like the
> idea of tag-dispatched selectable internal implementation.
>
> Do you happen to have done any comparison with the other version in
> terms of performance and memory consumption?
For 10000 boxes with MAX=4, MIN=2 it's 2,5 times faster than version
with 2 types of nodes (on my machine 0.06s and 0.15s respectively). In
the previous rtree linear split(?) were implemented, R*tree is more
complicated and still it's faster. Nodes has size of
sizeof(std::vector)+sizeof(int).
I've commited some changes so please update.
Regards,
Adam
Geometry list run by mateusz at loskot.net