John Turner, particles with Coulomb and and gravity, Internet better practices for simulations and global collaborations
John Turner: N-Body Simulation with 70,000+ Particles at https://www.youtube.com/watch?v=fSeEg5XwGGo
John Turner, I can think of many ways to make this easier for humans to see, to monitor for useful event and situations, to visual new things. First, you are showing position. You can also show velocity and acceleration, you can summarize statistic on changes, you can give data on the interaction, you can give data on acceleration.
You did not give a link to any paper, nor a link to your code, so I cannot see if your algorithms is efficient. If a real universe, you have to include magnetic effects, they are often much stronger and more permanent than Coulomb effects. Because gravity has hysteresis effects (matter accumulates and sorts it self, stars ignite in certain places and follow specific ignition then spreading processes (photon, electron, protons, ions, mass ejections, photo-ionization, lightning, magnetic re-connection. There are many things that can happen and do happen. But humans have limited ability to follow many small motions. Grass growing is fine, so we use time lapse. All the frames look very similar so we calculate rates of change or difference between frames. We use false colors liberally. And we use statistics massively. This would be a fun statistical mechanics problem. It would be a great machine learning problem. But you did not give enough information to quickly and precisely find and use your algorithms (exactly what equipment, software). You did not share your raw simulation results (record the data at each step in an open compact format). you did not give summaries of the timings, and efficiencies of the calculations.
Is it beautiful? Yes nice eye candy. At these low resolutions, and lossy video format, NOT a true representation of the results from your simulation. Raw data, the algorithm itself, or statistics are better. A simple filter for event that you record at high resolution, would be much better. Run it and look for close triple and quadruple interaction.
Richard Collins, The Internet Foundation
I did find the entry for “Chapter 31. Fast N-Body Simulation with CUDA from GPU Gems 3” by Lars Nyland, Mark Harris, and Jan Prins at the link you posted. Since John Turner is not mentioned, I am not sure of your connection. I am looking at this video because it came up as a random recommendation from YouTube. The section “31.2 All-Pairs N-Body Simulation”.and “31.3 A CUDA Implementation of the All-Pairs N-Body Algorithm” are useful. But they are “text” not tools. not simulations, no results or data. So not very satisfying or useful. I have worked on orbital mechanics, gravitational potentials, and n-body simulations for just over 50 years now. I have seen a lot of attempts. Most of them are like yours – they run the thing and do not share the results in ways that are useful for analyzing the quality. If you think people like running things just to be running them, yes there are a few people, But most of the 5 billion people using the Internet are trying to compare with real things, real galaxies, real clusters. Helping them solve their problems by better understanding their goals, probably is more important than handing them brute force tools. Richard Collins, The Internet Foundation
