Re: JNI 3d

Posted by Thomas Boudier on Feb 16, 2008; 10:11am
URL: http://imagej.273.s1.nabble.com/JNI-3d-tp3697181p3697183.html

Hi,

I agree with Wilhem and dscho, java can be quite fast also, my jni
plugin has two purposes : first to serve as a sample for people
interested in developing such programs, and second I mainly use it for
fast 3D morphological filtering such as tophat that need large
neighborhood (radius=7).  I do not have accurate benchmarks but my first
tests show that :
it is not suitable for 2D filtering, IJ is faster.
it may be suitable for 3D filtering since it is faster than my java 3D
filtering
it is highly suitable for 3D morphological filtering such as tophat (5 s
radius=7 on T1-head, on my  AMD64 3200)

Thomas

> Thomas,
>
> it would be interesting to know what "quite fast" means. Do you have any benchmark figures that compare the performance of your library against the pure Java implementation? I did some tests myself a few months ago to find out that the speed advantage of my JNI implementation was only around 20%.
>
> Wilhelm
>
>
> ________________________________
>
> Von: ImageJ Interest Group im Auftrag von Thomas Boudier
> Gesendet: Fr 15.02.2008 19:44
> An: [hidden email]
> Betreff: JNI 3d
>
>
>
> Hi,
>
> I would like to announce a plugin for 3D filtering done with JNI so it
> is quite fast.
> It can be downloaded at :
> http://imagejdocu.tudor.lu/imagej-documentation-wiki/Members/tboudier/plonearticle.2008-02-15.5454604181
>
> Thomas
>
> --
> /*****************************************************/
>  Thomas Boudier, MCU Université Pierre et Marie Curie
>  UMR 7101 / IFR 83. Bat A 328, Campus Jussieu
>  Tél : [phone]  Fax : [phone]
> /*****************************************************/
>
>
>  

--
/*****************************************************/
 Thomas Boudier, MCU Université Pierre et Marie Curie
 UMR 7101 / IFR 83. Bat A 328, Campus Jussieu
 Tél : [phone]  Fax : [phone]
/*****************************************************/