search

CliMA / Oceananigans.jl

🌊 Julia software for fast, friendly, flexible, ocean-flavored fluid dynamics on CPUs and GPUs
https://clima.github.io/OceananigansDocumentation/stable
MIT License
989 stars 194 forks source link

Development milestones and order? #207

Closed ali-ramadhan closed 5 years ago

ali-ramadhan commented 5 years ago

Not supposed to be concrete but just to lay out some goals for big new features and prioritize them based on our needs. Also in case I forgot about them.

Most of these can be done independently so the version ordering is mostly for priority.

vchuravy commented 5 years ago

I would recommend using Milestones instead of a big meta issue.

ali-ramadhan commented 5 years ago

Ah yes good point. We're kind of already using milestones (only up to v0.7 right now) but this was kind of to discuss which milestones to tackle and in which order. I made the title a bit clearer.

johncmarshall54 commented 5 years ago

LES closure is a very high priority.

On Wed, Apr 24, 2019, 5:39 PM Ali Ramadhan notifications@github.com wrote:

Not supposed to be concrete but just to lay out some goals for big new features and prioritize them based on our needs. Also in case I forgot about them.

Most of these can be done independently so the version ordering is mostly for priority.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/climate-machine/Oceananigans.jl/issues/207, or mute the thread https://github.com/notifications/unsubscribe-auth/AKXUEQUFOFFJEHO2C7FNEB3PSDHR3ANCNFSM4HIIDSNA .

ali-ramadhan commented 5 years ago

I've bumped LES closures to v0.6 then so it'll be the next feature we implement.

glwagner commented 5 years ago

+1 for LES!

We may want to look into WENO finite difference stencils. This can be thought of as another LES scheme, and may also be useful for larger-scale simulations in which ordinary LES closures are suspect.

Getting Float32 to work could perhaps be a lower priority.

I'm not sure what the status of the code comparisons are, but we may want to figure out how to do a c comparison with other LES codes when an LES model is implemented.

ali-ramadhan commented 5 years ago

Ah interesting, I thought WENO was a family of higher-order advection schemes.

I renamed "More conservative time-stepping for Float32" to "Machine precision mass conservation" as it's an issue with Float64 as well but probably does not matter for simulations with fewer than 1~10 million time steps. Might matter for the mixed layer seasonal cycle simulations which go for more than 10 million time steps.

glwagner commented 5 years ago

I believe you are correct that WENO is a family of higher order advection schemes — that permit accurate, relatively non-dissipation simulations without explicitly modeled dissipation terms. Maybe the milestone could be broadened to “high-order advection schemes”.

Mass conservation is definitely important.

ali-ramadhan commented 5 years ago

Ah I see, would be interesting to see if it does better than an LES closure seeing as AMD can get computationally expensive (but maybe not too expensive if efficiently implemented on a GPU).

I added a "Higher-order advection schemes" milestone which we can prioritize as necessary. I'm sure which advection schemes to use/implement will be a hot topic but for now I'm thinking these might be interesting:

cc @sandreza this issue may of interest to you.

ali-ramadhan commented 5 years ago

Closing as this issue has served its purpose (for me at least). Development milestones are clearer now. Further discussion should occur in more focused issues.