PentaPico: A Pi Pico Cluster For Image Convolution
1 min read
Summary
A Raspberry Pi Pico “cluster” has been built by Willow Cunningham as a final project for Cluster Computing at the University of Maine at Orono, with a focus on imaging convolution.
The Raspberry Pi Picos are employed as a cluster of five, featuring one head node and four compute nodes, with C used for software on both node types.
The head node is connected to a workstation via USB 1.1 and controlled using a Python script, while the input image is transferred to the head node via USB and then distributed to the compute nodes via I2C.
Results are produced for compute node numbers of one, two and four, but the project finds that distributing the data requires significantly more time than the computation itself, meaning the speed of the system actually decreases as more nodes are added.
