Fast Eulerian Fluid Simulation In Games Using Poisson Filters (SCA 2020 Showcase)

[@PeterSikachev]

This is one of the best research I've ever tackled. Clear, concise, no academic obfuscation, and yet dealing with one of the mathematically hardest problems in 3D graphics.

[@SazLowify]

Even the artifacts are interesting!! Great work. Can't wait to see a player character hold this torch.

[@SirKillo0o]

Great explanation on the math, the video basically solves 3-4 months of hard work!

[@時計草-g9z]

CompactPoissonFiltersのgithubレポジトリを復活してもらえませんでしょうか？

[@UncoveredTruths]

i just started learning CFD, i thought the separability trick was very neat! will have to try it with my GPU implementation

[@gamurarandrei2657]

amazing

[@xayah3375]

Brilliant work

[@gregoryjohnson460]

Excellent work

[@krinodagamer6313]

thats wild!!!!!!!!

[@Thomason1005]

very interesting solution! though i could not replicate it, it looks really good and solid!

[@ARVash]

CC called it "Possum filters" which killed me.

[@EekSaidTheCreeper]

Any chance the github repository could be put back up? Great stuff!

[@shahinrab]

We are releasing the code for the follow up work "Compact Poisson Filters ​for Fast Fluid Simulation" SIGGRAPH 2022 paper in a few days.

[@xortdsc]

Very nice ! Though, if I understood it correctly, the trick can not be applied to solve systems with variable coefficients like they occur in fluid simulations with varying density. Is that correct ? Also I don't understand how this method works together with obstacles. Wouldn't that require dynamic values of alpha/beta which makes precomputing the kernels impossible ? Also the paper does not really mention this issue. I'd really appreciate if you could shed some light on this. :)

[@shahinrab]

We have recently submitted a paper that explains all your questions in detail, including the 3D sim and dealing with obstacles. In a nutshell, this is a drop-in replacement for the linear solver used in the projection step and diffusion (not addressed in the above work), so it does not really care about what is solving for (density/temperature diffusion or mass preservation). But if your simulation parameters need to be variable in real-time (diffusivity or cell size), you would need to pre-generate a bunch of filters beforehand. I will put a link to the new paper in the description whenever it becomes available.

[@xortdsc]

@@shahinrab thank you ! can't wait :)

[@Science4Ever]

Does anyone know what software was used to create those slides with all the equations and figures in this presentation?

[@shahinrab]

good old PowerPoint for slides, LatexIt for equations, and our own prototyping engine for rendering.

[@allstar4065]

This is why computer science degrees require higher level math.

[@jhanolaer8286]

much better if you teach to simulate fluid

[@krinodagamer6313]

this is used in GTA6