PDA

View Full Version : Material Point Method

kotsoft
04-27-2009, 06:39 PM
The material point method is another way of simulating fluids that I recently implemented. It's one of those hybrid grid and particle based methods.

Videos:
http://www.youtube.com/watch?v=UcJtn0j08sw
http://www.youtube.com/watch?v=6Q5t6RJhQF0
lafe
04-27-2009, 06:59 PM
The material point method is another way of simulating fluids that I recently implemented. It's one of those hybrid grid and particle based methods.

Videos:
http://www.youtube.com/watch?v=UcJtn0j08sw
http://www.youtube.com/watch?v=6Q5t6RJhQF0

thats pretty sweet! :)
i was checking out those vids during study-hall :D (after i did my hw ofcourse)

are you going too release this?
kotsoft
04-27-2009, 11:13 PM
yeah, i'll probably release it after i clear up some of the bugs and extend everything a little.
meow-muffin
04-27-2009, 11:58 PM
Cool, nice to see you developing more stuff. :)
kotsoft
04-28-2009, 09:18 PM
http://www.youtube.com/watch?v=glCv2q7R0RU
this is a test with gooey material. this one ran at 30fps on my dual core, which means 60 fps for the quad core people at 26664 particles.
lafe
04-28-2009, 09:41 PM
http://www.youtube.com/watch?v=glCv2q7R0RU
this is a test with gooey material. this one ran at 30fps on my dual core, which means 60 fps for the quad core people at 26664 particles.

OMG, thats insane!
its THAT optimized!
LemonScented
04-29-2009, 02:04 AM
Cool stuff :)

How have you found this comparing to some of the previous methods (SPH, or Viscoelastic) in terms of speed, stability, flexibility etc? I'm still looking into other approaches since my viscoelastic stuff, although it's improved a lot, still doesn't get much beyond 20 FPS with 5000 particles on a single core...
kotsoft
04-29-2009, 02:25 AM
well, speed-wise, this is pretty fast, and easily parallelizable. and for the most part pretty easy to implement (when i finally got myself to do it, i did it in an hour). stability is mostly pretty good, there are a few problems in my current implementation. i'm implementing basically from the original mpm stuff, but there are definitely papers available which show changes to the original formulation to improve the stability and accuracy. anyways, this simulator is still young so i can't really discuss it to much depth yet.
zalo
04-29-2009, 04:10 AM
I Have a quad core (Vista) so if you ever want
to see if Quad cores really do quadruple performance
over single core, I'm here.

It's 64 Bit, 2.33 Ghz (each core), with 6 Gigabytes of
ram incase you wanted to know.

But it really does look amazing, speed wise.
antotabo
05-01-2009, 01:15 AM
Are you going to remake or complete your poluted planet game?? It was cool! It seems to me that you made progress in your simulation so do you plan putting your work on a game soon??
Nate The Great
05-10-2009, 05:53 PM
cool stuff

so since you have experience in this field what method for real time fluids would you say is the most effective and easiest alongside rigid body verlet physics?
lafe
05-10-2009, 05:54 PM
cool stuff

so since you have experience in this field what method for real time fluids would you say is the most effective and easiest alongside rigid body verlet physics?

Judigng by what kotsoft uses most often, probably SPH
LemonScented
05-12-2009, 12:55 AM
Far be it from me to second guess kotsoft (for all I know, the Material Point Method is the ultimate solution to all things fluid), but I'd say some form of SPH as well. There was an article in one of the Game Programming Gems books (I forget which one - 3, I think?), about exactly that: rigid bodies represented with Verlet integration being modelled as springs and particles within a similarly Verlet-based SPH system.