About running time

[Lin Yan]

Hello,

I calculated simplified contour trees with different simplification levels (based on persistence, 0.12 means 12% of maximum persistence) using several datasets. The running time is as the following (1st column: dataset, 2nd column: simplification level, 3rd: running time (s)):

I think the running time should decrease as we increase the simplification threshold, which is true for most of the datasets. But it is not true for the Viscous Fingering dataset (IEEE Scientific Visualization Contest dataset). Any ideas on why it happens?

Any help is appreciated!

Thanks,

Lin

[Julien Tierny]

Dear Lin,

thanks for your email and your interest in TTK.

> I calculated simplified contour trees with different simplification levels
> (based on persistence, 0.12 means 12% of maximum persistence) using several
> datasets. The running time is as the following (1st column: dataset, 2nd
> column: simplification level, 3rd: running time (s)):

Can you tell us what version of TTK you are using?
Also, can you tell us what do these timings correspond to?
- the simplification procedure?
- the contour tree computation?
- the sum of the two?
- do you include any other parts? (IO, preconditioning, etc.)

Also, these numbers seem pretty high to me.
If you are using the viscous fingering dataset from the ttk-data repository, you should have the simplification computed under 1 second and the contour tree in around 2 seconds. This is what I observe on my laptop.

> Any ideas on why it happens?

There could be many reasons, depending on your answers to the above questions.
For instance, it may seem counter-intuitive, but the computation time of the simplification procedure actually increases for increasing simplification thresholds (see Figure 7 of the following paper: https://arxiv.org/pdf/2009.00083)

Thanks for your feedback.

Best,
--
Dr Julien Tierny
CNRS Researcher
Sorbonne Universite
http://lip6.fr/Julien.Tierny

> *[image: Screen Shot 2022-05-13 at 9.49.13 AM.png]*
>
>