http://www.nothings.org/computer/judy/
Sean Barrett made some good points regarding the lack of benchmark tests vs
other algorithms, as well as the impact of Judy's use of double the Pentium
cache-line size.
However, the comparison is between a traditional Hash table and JudyL (an
optimised form of tree if I understood the documentation correctly).
I was wondering if anyone has done similar tests (speed and space) with
JudyHS vs a mainstream STL hash table implementation on a decent optimising
compiler - at the algorithm level this would effectively be comparing the
hash bucket implementations.
In addition, a big research area of the past few years has been in
concurrent algorithms. Has anyone tried to apply such wait-free/lock-free
techniques or even tried to make an offset-based structure that can be
shared (shared "sections" a.k.a "memory mapped files" under Windows) between
processes (or persisted to disk as-is)?
The main advantage of hash tables over tree-based structures is their
transparency of implementation: this leads to the ability to very quickly
work out what the lowest granularity of mutual exclusion is, where to apply
it, and how to replace pointers with offsets.
However, this does not mean that hash tables are better in terms of
speed/space, just that they are easier to apply such transformations to
unless you are the original author of the source code.
Regards,
Toni.
----
Toni Cassisi
Tovica Ltd
http://www.tovica.com <http://www.tovica.com/>
Tel: +44 (0) 7971 874 054
IM: AOL/Yahoo/MSN: tcassisi
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Judy-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/judy-devel
