[vtkusers] speed question

[Kevin Teich]

I've written a quick and dirty viewer/editor for volumetric data. It uses
a custom subclass of StructuredPointsSource with a function for setting a
value directly (uses SetComponent() on the scalars data). The app has a
minimum of two volumes on screen, both using lookup tables to map values
to colors. This is written in Tcl/Tk and running on Linux boxes and nVidia
graphics cards, geForce3s and 4 2 Gos with lots of RAM. The volumes are
256^3 unsigned chars. It's much slower than I expected.  When the user
clicks in the view to edit a value, the update takes 1+ second to show.

Should this be something that should run quickly? What should be the first
thing I try to optimize? I'm guessing the update extent that the SPS
subclass passes upwards through the pipeline on an edit. Is this
considered a 'large data set'?

Thanks.

-- 
Kevin Teich