Used to be that there were processors (of the “regular” kind) and there were DSPs. It’s no longer enough to be a DSP: you have to be the right kind. At least that’s how CEVA has rolled out their offering, with one family for communications, one for video and imaging, and one for audio and voice.
They recently announced the latest version of the latter, their TeakLite family. As in other areas that used to seem so simple and innocent, voice and audio processing have become increasingly sophisticated, with multiple microphones helping quash ambient noise (which we’ll talk more about in an upcoming feature), and even “beam forming”: an array of mikes that can zero in on an individual in a crowd – without moving the array, and with no noise. Creepy much? (It’s typically used for sports, but we all know that’s just a gateway stalk…)
Anyway, CEVA has announced four new cores. Two are for stand-alone DSP chip use, with one optimized for small area (single 32×32-bit MAC or dual 16×16-bit MAC), the other for performance (double the MACs plus optional audio instructions). The other two are for integration with a CPU on an SoC; they add cache controllers and an AXI interface to the first two.
Feeds and speeds can be found in their release…