This week we would like to highlight the following article about a pretty nifty new device.
Whilst completing his PhD at the University of California, Berkeley, Martin Maas and colleagues designed a compact accelerator unit requiring only a small amount of chip area and power. The purpose of this unit is to take over and more efficiently execute the task of garbage collection from the CPU. This means that while software is running, it no longer needs to use any CPU performance for this task, but can instead keep using the available memory. Effectively this should increase the general productivity and efficiency of computers!
This way of liberating the CPU from tasks is something reflected in several of current technology developments, for instance with Apple’s iPhone chips also becoming clusters of many units each with distinct specialties (CPU, GPU, neural net processor for FaceID etc).
As one professional jokingly pointed out to us: “Now maybe Python will finally get to be as fast as C++!”
