Single/double precision in 5.0
Dear developers,
Hacking around with the code of version 5.0 beta and having version 4.4 as baseline, I noticed that some functions previously dedicated to single precision have been removed.
In line with this, the help file of DualSPHysics shows that the option posdouble (0: neither use nor store positions in double; 1: use and do not store; 2: use and store) has been substituted by saveposdouble just for saving.
So it looks like in 5.0 the calculations of positions are always done in double precision. As of today, this change is not mentioned in the list of new features at https://github.com/DualSPHysics/DualSPHysics/wiki/9.-New-in-DualSPHysics#new-in-dualsphysics-v50
Would you please confirm that this is a new stable feature of DualSPHysics? And share your considerations behind this choice?
Thanks in advance for addressing these questions.
Comments
Yes, you are right, thanks for noting this.
We will include the information in the documentation.
Really appreciated.
We will also reply soon about your other issues since we are right now working on them as well.
Regards
@Alex Thanks for considering this/these in turn. (Perhaps redundantly said, please keep the replies separate, so future readers can stay up-to-date with the evolution of each thread.)
@any-developer-who-knows Just getting back to question above: are there specific reasons or a vision why single precision for position has been phased out? This would be interesting to know in order to put this change in the broader context and line of development of DualSPHysics. Thanks for considering this.