staikos at kde.org
Tue Nov 1 01:56:10 CET 2005
On Monday 31 October 2005 19:38, Barth Netterfield wrote:
> How much speed up do you get?
This is not easy to judge yet. It might take a profiler to get meaningful
results. Actually, I think the real gain is in lower CPU usage because we're
blocking on X again. I've removed almost all the limitations I could find on
our side now.
> > A patch to implement a short-cut interpolation in the hot spots in curve
> > plotting. It removes a function call (inlines it) and two branches, one
> > of which has a disjunction, neither of which are likely. However it's
> > slightly incomplete because it can cause invalid memory accesses if the
> > vectors have size == 1 and it tries to plot them.
> > The easy solution is to
> > plot nothing if one of the vectors has size == 1, but I suspect that's
> > undesired.
> We need to be able to plot a single x,y point.
Originally I was going to implement specifically this case and nothing else,
> But in 'line' mode, I can't see that we would ever need to plot a curve
> with only one point in one of the vectors. So we could have two
> interpolates: one for w/ lines and one for w/ points.
I was thinking of a case of X vector [0,...,n] and Y vector  which should
give a straight line at 5 from 0..n. Again, we could shortcut this case and
its complement, but maybe there are other other weird cases too. The same
applies to error bars, bar graphs, etc.
KDE Developer http://www.kde.org/
Staikos Computing Services Inc. http://www.staikos.net/
More information about the Kst