NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
Hi Maohai, > unfortunately point size is not the only issue. Making the > pointSize larger as you suggested was the first thing I > tried. It worked (reducing the speed by a factor of > 20 though), but there is only one shape of point in this > mode. I am asked to be able to plot several data set on one > plot with different shape of points. So I need 1) to have > multiple shapes, 2) to have the shapes at very small > cost in memory and speed for data sets of million > points, 3) the shapes need not change often. There must be some way around this, because the user cannot come close to distinguishing 1 million shapes in a single screen. One thought is to plot the million points as points, and to create another data object selecting only points not near to other points and draw them as shapes. > By the way I have written a plotting application from > scratch using Graphics.drawPolyline(...). It beats the > fastest speed of all existing packages by a > factor of ~10. It doesn't have multiple shapes, but > the connecting lines show out-lying points nicely. > It also shows that there is space for speed improvement > in existing packages. Absolutely. The default DataRenderers are very general (more than any other visualization system), but its hard for them to optimize every case. So we also support custom DataRenderers. A good example is visad/bom/ImageRendererJ3D.java, which renders images and image sequences much faster than the default DataRenderers. In order to make it easier for folks to write their own DataRenderers, we have the DataRenderer tutorial, and the new visad-renderer@xxxxxxxxxxxxx mailing list. Because of the system's generality, there are lots of other tricks that applications can play, short of writing custom DataRenderers. Cheers, Bill
visad
archives: