|
|
Ublas : |
From: Michael Stevens (mail_at_[hidden])
Date: 2006-04-21 11:50:12
On Thursday, 20. April 2006 17:20, Andreas Dolfen wrote:
> > A couple of things you could look at. uBLAS has bench1-4. If you are only
> > using Dense matrices then it would be sufficient to just compile up
> > bench1 for both versions and compare results. This may give us a better
> > idea of which operations are suffering.
>
> I just did that. The results are attached. It seems that the bench1
> program shows the same performance degradation. I used:
> export TOOLS=vacpp
> bjam "-sBUILD=release"
> for both boost libraries only changing the include-dirs.
The timing resolution is very low with most times around 0.01s. You can run
bench1 with an argument to increase the number of times everything is
executed.
Although it does look like the 1_33_1 are slower the results are a bit odd.
The c_array results are around 0s, but maybe this is just a resolution
problem.
Michael
--
___________________________________
Michael Stevens Systems Engineering
34128 Kassel, Germany
Phone/Fax: +49 561 5218038
Navigation Systems, Estimation and
Bayesian Filtering
http://bayesclasses.sf.net
___________________________________