Project 1: Aaron Tian

[AaronTian-stack]

Repo Link

• Implemented boid rules for naive, uniform grid, and coherent grid implementations.
• Extra Credit: Grid Looping Optimization
  • Cells to iterate over are calculated by calculating the min and max overlapping cells from the neighborhood radius, then iterating over that range. This way only the cells the neighborhood overlaps with are iterated over. I'm not sure why this is extra credit since I don't see how else you would determine which <=8 or <=27 cells to check, but I'll take it!
• Extra Performance Analysis in GPUView

Feedback

It might be useful to remind students that it is possible to mistakenly do invalid out of range array read/writes in the uniform/coherent grid implementations depending on how they implement it. You could point them to the CUDA Compute Sanitizer that would be able to quickly triage these issues.

[shehzan10]

It might be useful to remind students that it is possible to mistakenly do invalid out of range array read/writes in the uniform/coherent grid implementations depending on how they implement it. You could point them to the CUDA Compute Sanitizer that would be able to quickly triage these issues.

@AaronTian-stack Thats fantastic feedback. Can I ask you to open a new pull request into Project 1 that adds the consice instructions / tips for using CUDA Compute Sanitizer? I'll then copy the same into Project 0 as well, which can help students learn before Project 1.

[AaronTian-stack]

@shehzan10 Sure, I'll write a guide/explanation in the README for Project 1 and open a PR.

[AaronTian-stack]

@shehzan10 I have opened the PR here: #55

There are instructions for basic usage of memcheck. I have linked official docs for further info of using the tools in the sanitizer.