Project 1: Xuntong Liang

[PacosLelouch]

• Repo Link
• Implemented all the required method to simulate the boids flocking and some extra credit. Completed performance analysis.
  • Implemented naive method.
  • Implemented scattered uniform grid method.
  • Implemented coherent uniform grid method.
  • Implemented grid-looping optimization to prevent a hard-coded specific number of surrounding cells.
  • Completed performance analysis and added some code to automatically output a form of data for performance analysis.
  • Run with 1 million boids at about 100 hz without visualization and 80hz with visualization on my device.
• Feedback: I have no ideas on how to add shared memory to improve the performance in a good way. And I can only analyze the function's duration in Nsight System but don't know how to analyze some other aspects such as the latency of memory accessing or launching kernels. Hope to learn more about them in the future!
• To be honest, I think the result that follows Conrad Parker's pseudocode is okay, even though when the number of boids is large, most of them will move towards the center and becomes a nebular. I guess one of the reasons for the result is not so pretty for this project is that it is so crowded and the velocity is so random, isn't it?