[Kst] Interesting performance results

[Barth Netterfield]

All as expected.

On December 15, 2004 05:54 pm, George Staikos wrote:
> On Wednesday 15 December 2004 17:47, Barth Netterfield wrote:
> > How many points in the vectors?  I expect to be >100,000 before curves
> > start to dominate (and its processing the curves that matters.) 
> > Otherwise, these numbers are consistent with mine.
>
>   I have 25000 points plotted (visible), with points, and points are taking
> 11-12 times as long as the curve to draw.  

Curves-lines compress points to max and min per pixel.  Curves-Points can't do 
that, so they are much slower.

> In the PSD, with points, there 
> is no significant change.  I think points are too expensive for dense
> curves, and maybe they should be disabled after a certain point.

Hard to say.  Maybe someone wants to see the 4 points in a spike, in 100,000 
samples....  We could warn (?)

> Plotting curve 0: 92ms
>     Without locks: 91ms
>             Lines: 7ms
>            Points: 84ms
>
> Removing points but leaving lines at thickness 2 drops the overall time to
> ~10-15ms again.
>
> PSD is still impressively fast, even with points enabled.

It is just a curve with a small number of lines.  Should be really fast.

I don't think there are any huge problems in curves.  I will play with 
optimizing it over Break.

We should be looking for unneccisary updates (both plots and data), locks, 
etc....