2028: Cores' law, swarm computing, and people power
|November 2028 is the 40th anniversary of ESD. Click here read other 2028 lookbacks.|
Looking back over the past year, it's hard to ignore the significant advances that embedded processing has brought to the world. In particular, four prominent accomplishments come to mind:
For as long as anyone can remember, chip developers have longed for simple design tools. The "holy grail" has been to enable a bright 12-year-old to design a deep sub-nanometer processor using a hodge-podge of open-source drawing programs and spoken commands. Unfortunately, this vision has yet to pan out, but Microsoft subsidiary Google has surely delivered the next best thing: PowerPoint 2027 includes a PPT-to-Netlist converter, which allows marketing and sales representatives to create a graphical block-based ASIC floorplan in four easy steps (and millions of brilliant colors). Each block is automatically linked by label to a dizzying array of lower-level synthesized CPU elements, memories, and peripherals. The whole process takes about four hours, not including any animations or complex shading schemes.
While there has been some resistance to this approach from the silicon design community, the general impression is that the improvement in tape-out dates is more than enough to offset the effects of disgruntled designers, most of whom are returning to grad school to procure MBA degrees.
As we all know, 20 years ago the pundits were eulogizing Moore's Law, lamenting that transistor performance had finally peaked and that the only way to achieve faster speeds was to adopt multicore approaches. Well, that worked well for a while, until it sparked "Cores' Law"--that is, the number of cores on a processor platform doubles every 24 months. Ultimately, it was the development-tools vendors who revolted, after 32-core embedded processors taxed the limits of how many windows could simultaneously be useful and viewable on the IDDE (integrated development and debugging environment) screen.
Luckily, optical transistors hit the mainstream just in time. Once maligned for its relatively high power requirements and dependence on physical fibers, optical computing received a facelift with MEMS-based light coupling. This led to extremely dense optical transistor arrangements, which, in turn, injected new life into an embedded marketplace struggling for the next big performance leap.