Releases: Niceno/FortranOpenACC
Matrix of Cubes as Obstacles
Solves Navier-Stokes Equations Without Obstacles
This is the last version before the implementation of obstacles. Obstacles are ready, I just wanted to mark this version before I push the changes up.
Simulates Stokes Flow
This is the first time it simulates Stokes flow, the lid driven cavity, to be more specific.
However, it only works on the host, no attempt have been made yet to run the Stokes flow on GPUs. I need to clean the sources a little bit before porting them to GPUs.
Gradient Calculation Works
I got the gradient calculation running properly, after third or fourth attempt. Nothing magic about it, I think I simply wasn't too focused in previous attempts but, nonetheless, it is probably a good idea to mark a version in which gradient calculation is working.
Although the gradient calculation as implemented here works in the same way as in T-Flows, it does not seem to be a version which is suitable for GPU parallelization with OpenACC, since the gradient calculation on GPUs create race conditions. This will have to be addressed in the future.
Conjugate Gradient Method Works!
This is the first version in which CG method works. It is diagonally preconditioned variant, it is still too early for more sophisticated preconditioners.
After this release I will probably also ditch the dense matrix examples, after which the Sparse_Mod will become just the Matrix_Mod, and I will probably get rid of the Vector_Mod since we do not have it in T-Flows.
Clearly, after this I move in a more streamlined way towards the numerical models in T-Flows.