On Tue, 27 Jul 1999, Frank E Harrell Jr wrote:
> There are two areas for which I would appreciate hearing about the current
> status.
>
> 1. I have been seeing a fair amount of negative comments about the
> speed of 5.x. 5.x was hailed as a major improvement in speed when
> it was launched. Part of the reason for my asking is that I am
> getting some requests to modify my Hmisc and Design libraries to
> run in 5.x. I am not inclined to do so until the speed issue has
> been settled or until 5.x is available on Windows/NT. I have seen
> posts where 5.x is very slow compared with 3.4, and others where
> there is a slight improvement. What I would like to know is an
> approximate "average" comparison for routine analyses and
> bootstrapping, for example.
Almost all those comments were about obselete versions, that is prior to
5.1. I would not say 5.1 is faster (the claims for very large problems,
perhaps unrealistically large ones) but it is not much slower (e.g. for
bootstrapping). I can't give exact comparisons as some things work in
different versions, but for the V&R3 scripts the ratio of speed 3.4:5.1 is
from about 1:0.25 to 1:1.2. As a ballpark figure, I'd say 5.2 is 2.5x
slower, using twice as much memory.
Remember that some users (e.g. Linux) only have access to 5.x.
Brian
--
Brian D. Ripley, ripley@stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272860 (secr)
Oxford OX1 3TG, UK Fax: +44 1865 272595
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
|