[Cuis-dev] Ideal gas and real gas simulation

[Luciano Notarfrancesco]

Very interesting!
However, how big has a simulation have to be in order to be useful for
them? I had the impression that these things are typically huge (millions
or billions of particles) and run in parallel on supercomputers with GPUs.
I’d like to know if small simulations running on a CPU have realistic
applications. Or perhaps it could be more of an educational tool? For some
computations you could take advantage of the FloatArray primitives to
compute in bulk (I’m using this for audio DSP and they are really helpful).
There’s also the possibility to use BLAS libraries via FFI. And in Cuis you
could also use OpenCL. Keep us posted, this is really interesting.


On Sun, Dec 7, 2025 at 17:01 Hilaire Fernandes via Cuis-dev <
cuis-dev at lists.cuis.st> wrote:

> Hi,
>
> When discussing with Physics teacher at my school, its appears gas
> simulation is something desired. There are some simulation out there, but
> they keep mentioning me it is a nice tool to have. So I gave a shoot and
> start the implementation of a DKM[1] for the Dybo project.
>
> A first iteration was a very naive, kind of billiard:
> https://mamot.fr/@drgeo/115577138308536197
>
> Then hack some energy transfer from the wall:
>
> https://mamot.fr/@drgeo/115598616477391933
> https://mamot.fr/@drgeo/115600731271105530
>
> However this implementation was not correct from the thermodynamic
> perspective, so I implement a real diffuse energy transfer model:
> https://mamot.fr/@drgeo/115623761342500850
>
> This gave a more or less realistic simulation of ideal gas.
>
> The next challenge was to have inter-atomic potential to be considered:
> atom can be attracted or repulsed from each other. The Lennard-Jones law
> was used. The tricky part was really the scale and simulation step time:
>
> Atom can get crazy unrealistic velocity
> https://mamot.fr/@drgeo/115669231450451191 because of not small enough
> simulation time.
>
> With a fined tuned simulation time and Morph step time of 100Hz, the
> result are quite interesting: https://mamot.fr/@drgeo/115669511921472246
>
> The final model has both diffuse energy transfer from the wall and
> inter-atomic potential calculus.
>
> Again, the Cuis environment proved to be extremely helpful to fine tune
> the algorithm.
>
> Of course this DKM can be used from a Dybo document, as seen there with
> another DKM: https://mamot.fr/@drgeo/115400315017925601
>
> The aims is develop hundreds of such modules.
>
> Hilaire
>
> [1]
> https://github.com/Dynamic-Book/doc/tree/main/4-Explanations/200-Dynamic-Knowledge-Models
>
> -- http://mamot.fr/@drgeo
>
> --
> Cuis-dev mailing list
> Cuis-dev at lists.cuis.st
> https://lists.cuis.st/mailman/listinfo/cuis-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cuis.st/mailman/archives/cuis-dev/attachments/20251207/d7744be9/attachment.htm>