A short YouTube video (Two Minute Papers-style) describing recent improvements that allow running the Material Point Method (MPM, called NPM in the transcript) efficiently on graphics cards. The presenters claim up to a 10× speedup over previous work, demonstrate a granular scene with >6.5 million particles running at ~40 seconds per frame, and a gelatin scene running in <7 seconds per frame. The technique also claims improved numerical stability for simulating grains, fluids, snow, cloth and other materials. The video ends with a Patreon call-to-action.
Questions this video answers
Up to 10 times improvement.
More than 6.5 million particles.
Around 40 seconds per frame.
Less than 7 seconds per frame.
April 2, 2019.
Chapters
Dear fellow scholars, this is two minute.
Papers with know that I see a piece of fluid and I can't resist making videos about it.
I just can't.
Oh my goodness, look at that. These animations were created using the material Point method, or NPM in short, which is a hybrid simulation method which is able to simulate not only substances like.
Water and honey, but it can also.
Simulate snow, granular solids, cloth, and many, many other amazing things that you see here. Before you ask, the hybrid part means.
That it both uses particles and grids during the computations.
Unfortunately, it is very computationally demanding, so it takes forever to get these simulations ready. And typically in my simulations after this.
Step is done, I almost always find that the objects did not line up.
Perfectly, so so I can start the whole process again.
Ah, well, this technique has multiple stages, uses multiple data structures in many of them, and often we have to wait for the results of one stage to.
Be able to proceed to the next.
This is not that much of a.
Problem if we seek to implement this on our processor, but it would be.
Way, way faster if we could run.
It on the graphics card as long as we implement these problems on them properly.
However, due to these stages waiting for.
Each other, it is immensely difficult to.
Use the heavily parallel computing capabilities of the graphics card.
So, here you go.
This technique enables running NPM on your.
Graphics card efficiently, resulting in an up.
To 10 times improvement over previous works.
As a result, this granulation scene has.
More than 6.5 million particles on a.
Very fine grid and can be simulated.
In only around 40 seconds per frame.
And not only that, but the numerical stability of this technique is also superior to previous works, and it is thereby able to correctly simulate how the individual.
Grains interact in this block of sand.
Here is a more detailed breakdown of.
The number of particles, grid resolutions, and.
The amount of computation time needed to simulate each step.
I am currently in the middle of.
A monstrous fluid simulation project and oh man, I wish I had these numbers.
For the computation time.
This gelatin scene takes less than seven seconds per frame to simulate with a.
Similar number of particles.
Look at that heavenly gooey thing.
It probably tastes like strawberries.
And if you enjoyed this video and.
You wish to help us teach more.
People about these amazing papers, please consider.
Supporting us on Patreon.
In return, we can offer you early.
Access to these episodes, or you can also get your name in the video.
Description of every episode as a key supporter.
You can find us at patreon.com twominutepapers thanks for watching and for your generous support and I'll see you next time.
