42
edits
Changes
→Assignment 2
=== Assignment 2 ===
During our initial attempt at parallelizing a Rubiks Cube solver, we came across issues with compiling with the CUDA compiler. The issue turned out to be compiling CUDA code with C. Our first attempt at resolving this issue was by adding extern "C" to all affected files. While that resolved some linking errors, we were still unable to properly link it. We decided to instead turn our attention to Theo's colour sorter that he posted for Assignment 1. We were able to successfully parallelize it with a single thread. The performance increase from serial to CUDA was extremely significant to the point to where we had to increase n to 10^10 to actually measure the difference. As we progress to part 3 of these assignments, we see a constraint of the sorter being fixed to digits from 0 to 2. We hope to remove this restriction in conjunction with optimizing it.
=== Assignment 3 ===
