Job Opening - Senior Physics Engineer 🎉
Are you passionate about rigid/soft body physics, fluids, and other kinds of complex sims that can be achieved in real-time for highly dynamic and malleable game worlds for future games at @EA?
https://t.co/CxI0Kr4h66
#gamedev#jobs
So yeah, thanks to #Putin I can't really work on the liquid sim right now. And I was planning to work on some performance improvements, oh well.
For anyone worried, I'm fine, and currently in western Ukraine.
So for now I urge everyone to #StandWithUkriane to stop this war!
@nice_byte Going beyond rigid body physics has many difficult implications for how the rest of the game works & is rendered so is hard to have as a general engine feature. Overall I think we are in a deep local minimum and will be for at least a few more years.
@nice_byte Algorithms for real-time physics have barely changed in two decades, and the middleware everyone uses has had a lot of time to optimize all the many code paths, particularly in collision detection. Makes it very hard for a new competitor to arise.
SEED is well-represented at SIGGRAPH 2021. This week, we have three presentations showcasing some of our most exciting recent work: Cloth simulation, a direct delta mush skinning model, and global illumination based on surfels.
Check them out: https://t.co/HkvogHCCka
@tuxedolabs I don't have numbers for this particular scenario but it generally runs in a ms or so total in my test bed at 128x128x256. Most of that time is spent expanding the distance field outwards which could be made faster. It's pretty unoptimized over all (unreleased prototype)
@tuxedolabs@ProgrammerLin Going from density to distance field does smooth it too much in my experience, what I found I had to do was compute an accurate short range distance field directly from the particles. That has enough detail to render, especially if you use good interpolation.
Presentation for SIGGRAPH 21 by SEED’s Chris Lewin. Swish is a real-time machine-learning-based cloth simulation technique for games. It’s used to generate realistic cloth deformation and wrinkles for NFL player jerseys in Madden NFL 21. https://t.co/PlxXOHyYIn
Two weeks to go before Women in Machine Learning (#WiML) Workshop and #NeurIPS19 in #Vancouver! Want to know how SEED uses ML to drive innovation & creativity for future games at @EA? From SEED Monica and Yui will be there to tell you more :)
#WeAreSEED#WeAreEA#MachineLearning
@milesmacklin Great paper, shame I can't be there for the presentation! Do you think Newton-style methods hold promise for the real-time regime? (where you may only have the budget for a few iterations)
@davidwuzhi Are you using the constraint derivatives written in the white paper? Unfortunately they are wrong, you may have to derive them yourself. But it could also be another problem. If you use point-point collision does it work?
@MaxCWhitehead I think I was trying to 'simplify' the algebra for the technical report and made a bunch of mistakes. I would strongly suggest you calculate these derivatives yourself though, because I don't have a high confidence that they are completely correct.