search

DEEPDIP-project / .github

0 stars 0 forks source link

Consider replacing explicit time integrators with SciML #6

Open SCiarella opened 11 months ago

SCiarella commented 11 months ago

The problem with this is how to include the pressure term: 1) solving the coupled Poisson equation (algebraic ODE) 2) replace the pressure term (it requires matrix inversion, which makes it not sparse anymore)

SCiarella commented 11 months ago

There is another problem with the spectral space in SciML: the non-linear term $\hat{u u^T}$ is usually computed as multiplication->FFT, so this requires the solver to do IFFT at each step to compute this term, but this is NOT the standard behavior of SciML.

luisaforozco commented 8 months ago

Do not know if is the right place but there is a SciMLStyle which are basically guidelines for code writing in Julia. I think is a good place to start when starting to integrate SciML in Syver's and Toby's code.

PabRod commented 8 months ago

Do not know if is the right place but there is a SciMLStyle which are basically guidelines for code writing in Julia. I think is a good place to start when starting to integrate SciML in Syver's and Toby's code.

Looks pretty relevant. Also, see this thread about parameter management in our brand new Julia Teams channel (private).