Putting multicore processing in context: Part One

Today’s embedded devices are increasingly based on multi-core designs. Systems-on-a-chip (SoCs) often contain two or more processor cores in homogeneous or heterogeneous combinations, and FPGA-based designs can include a virtually unlimited number and variety of cores. An asymmetric multiprocessing (AMP)-based RTOS is one approach to utilizing multi-core processors; symmetric multiprocessing (SMP) is another.

But from a historical perspective, multiprocessor/multicore designs have been around almost as long as the computer in general. I believe that Burroughs introduced the first commercially available one back in 1962; many more have followed. Most of them had very specific purposes in mind.

Many of the early ones were used in data centers, and for scientific applications. You can also make a case for them being in the early desktops of the late 80s with the introduction of the math co-processor and device controllers such as an Ethernet controller.

The question is: why are embedded device manufacturers becoming so enamored with multi-core devices now? One reason is that embedded devices have more and more tasks being heaped on them.

My first cell phone was a StarTek. It had a rudimentary LCD display and frankly was not very useful for anything other than making a phone call. It had a 24-hour standby time and maybe an hour talk time once the battery was broken in. That was not a real problem though since my first plan was for 30 minutes a month.

But look at today’s phones and the demands made on them: built-in cameras, video capabilities, nice graphical interfaces for browsing the Web, calendars and all sorts of games. It is no wonder that designers are looking at ways to get more performance, and better form factor while extending battery life.

Click here to read more ...