Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) will soon present a new chip called Swarm, which should make parallel programs more efficient and easier to write.
The chip has extra circuitry to store and manage its queue of tasks (its differentiating factor), as well as a circuit that records the memory addresses of all data its cores are currently working on. The circuitry is able to time-stamp tasks according to their priorities and begins working on the highest-priority tasks first. The scientists have proven that this time-stamping makes synchronisation between cores easier to enforce. Programmers have also confirmed that Swarm requires minimal coding; typically only one line after defining a function.
When compared to versions of six common algorithms with the best existing parallel versions, the Swarm versions were between three and 18 times as fast and typically required one-tenth as much code, or less. In one case, Swarm was faster by 75-fold on a program that computer scientists had yet to parellelise, according to MIT news.
Image source: MIT news.