Final project created for CMU's spring 15-418, Parallel Computer Architecture & Programming course.
Primary goal of the project was to investigate the usage of MPI, specifically OpenMPI, in the context of a real world setting. Our real world setting is composed by trying to identify potholes given the depth image from a Kinect V1 (Xbox 360), and then clustering points corresponding to particular potholes. Originally, the goal of the project was to utilize Raspberry Pi's in a scatter-gather pattern to speed up our implementation of k-means using Lloyd's Method. Logistically, we found this difficult to implement since we had to face issues outside of the project scope (power, wire connections, ethernet controllers). Instead, we focused on the analysis of an MPI implementation on one of our laptops utilizing an Intel Core i5-4200M, with 2 physical cores, each hyperthreaded. We ran tests using our identical implementation against other k-means implementations, with results included in our report.
Our implementation of k-means is written in C, with a wrapper in Python for us to communicate with the workers.
Our findings and a more in-depth explanation are found in our poster file:
Some detections:
In addition, a video highlighting some of our real life evaluations is provided:
Built off work from an MPI implementation of kmeans.c designed for another purpose. We added the capability to handle work not divisible by processors, communication with a master Python process, improved scatter and gather uses, and some OpenMP pragmas such as explicit vectorization (still done regardless with -O3) link