Always click as soon as I see one of these come up!
@luiscanamarvega5 жыл бұрын
Jimmy D me too!
@StormChaser_Suedhessen5 жыл бұрын
Such beautiful Model Animations
@michaelhayward89145 жыл бұрын
I have been watching these uploads for a long time and really enjoying the journey of his to solve the problem. I have to listen it afew to understand though.
@deransadventures5 жыл бұрын
You are a legend. Really exciting stuff you're presenting. You're on a short list of people that I guess will eventually break the code. Or at the very least be directly related to it.
@marcusryczek21225 жыл бұрын
Always enjoy dr Leigh talking about his tornadoes and supercell thunderstorm model, makes me want to go out and chase
@GottaWannaDance5 жыл бұрын
I was able to make it to Dark Skies here in Madison this year. A long day, but well worth it. Leigh's talked really helped me to understand more ... but I could be wrong🤔
@mikk43095 жыл бұрын
Thankyou for your work and thankyou for sharing it with us.
@xyzct5 жыл бұрын
We'll have to pitch in and buy Leigh a new Aurora computer. I want to see 5m resolution.
@LeighOrfsThunderstormResearch5 жыл бұрын
@MarZandvliet5 жыл бұрын
Thanks so much for uploading these talks, very informative! Regarding the grid-based simulation, and the space-time resolution requirements: have you considered using sparse, possibly wavelet-compression techniques to encode the grid data on GPUs, as Space-X have been doing with their rocket propulsion simulation tools? They face similar challenges of high-resolution + lots of fields and interactions between them, and through these techniques they've been able speed up their workflow tremendously without loss of accuracy. The simulations evolve by directly manipulating compressed state. Here's a talk on the subject: kzbin.info/www/bejne/rIqkYZlsh5p-d6s (Note: I'm writing this 7 minutes in, could well be that you bring this up later, hah. I'm also a total fluid simulation newbie)
@LeighOrfsThunderstormResearch5 жыл бұрын
We are already using GPUs to do post-hoc processing from saved model data. We are chomping at the bit to use something like FastEddy (NCAR) which is a fluid code that is made specifically for GPUs. If we develop a model on GPUs it will probably use explicit finite differencing techniques similar to CM1 but because we haven't really kicked the tires with GPUs yet, we could use something like the link you shared. One of the nice advantages of using GPUs for calculation is that they can also do AI and visualization stuff with the data in memory. So GPUs are definitely on our agenda, but it is early days.
@bigthunder28605 жыл бұрын
I don't know if you read these I think you might of left out one factor electrical energy in the physics system somehow it must go together in the puzzle. I'm also from ma. Central I was DRIVING from new York when the tornado went through great Barrington thank you
@zelkuta3 жыл бұрын
Just realized the blue waters super computer runs on bulldozer chips?!??! We need to get you and your team access to fat stacks of 64 core eypc chips, like dang yo.