We use cookies to make your experience of our website better. You can delete or block cookies, but some parts of this website won't work without them. By using this website you accept our use of cookies.

15/04/2019 14:04:54

Double precision, single precision and batching

- Konrad Adams, Flood Modeller and TUFLOW specialist

The discussion of single precision and double precision in Flood Modeller is frequently visited; which is most appropriate and how these are run from the user interface, particularly within the batch runner.

The first question I guess is why double precision (or what?). I'll try and not be too geeky about this, but single precision solves the numerical solution to approximately 7 significant figures. Double precision gives you 15. But surely 7 is more than enough though, I hear you say? And the answer is usually, yes, it is. However, there are certain circumstances where significant information can get lost beyond the default precision. Typical examples that immediately spring to mind are:

  • Direct rainfall modelling (particularly linked to Flood Modeller 2D and TUFLOW)
  • Reservoir modelling, especially those with large surface areas and/or high elevations above datum and/or small timesteps

When you convert rainfall to SI units of m/s (or even to m/timestep, which is typically of the same order) you can probably guess why, when compared to the other "numbers" in the numerical model.

Also, you may have even seen such reservoir model results as mentioned above - those which show a nonzero net inflow, but the water elevation doesn't change (or vice versa). This instantly suggests to me: "USE THE DOUBLE PRECISION VERSION" - see plot below for a comparison of results.

Double precision, single precision and batching - 1

Figure 1: Example of reservoir output - single vs double precision

The other question revolves around how to pick up the double precision version of Flood Modeller. There are different executable files for single precision (the default, isisf32.exe) and double precision (isisf32_DoubleP.exe). 

 

Individual Simulations:

To set the precision for an individual simulation, double-click on your simulation in the Project Panel to open the Simulation Run window. Then select the "Double precision version" radio button from the Backward compatibility tab. 

Double precision, single precision and batching - 2

NB you may need to select “View” and check the box to the left of the tab you want displayed as shown in the figure below. By default, Flood Modeller does not display all these tabs automatically.

Double precision, single precision and batching - 3

Note a checkbox on the Links tab also allows one to select to use the double precision TUFLOW version. 

Double precision, single precision and batching - 4

 

Batch Simulations:

1. If running your simulation through the Batch Runner tool, the precision will be determined by the simulations themselves. As in the above section, select the desired precision for each of your simulation (.ief) files. 

Double precision, single precision and batching - 7

2. If you wish to run batch simulations of double precision Flood Modeller, the syntax (followed by an example batch file) is as follows:

isisf32_doubleP-sd "myieffile.ief"

    NB The -sd switch forces Flood Modeller to automatically close after finishing.

Double precision, single precision and batching - 5

For individual or batched simulations, you can check whether double precision or single precision was used in the simulation in the diagnostics (*.zzd) file.

Double precision, single precision and batching - 6

 

For more support resources, including how to guides and videos, visit our Help & Support section.

Privacy & Terms

Copyright