Other tightly coupled peripherals and accelerators require a design from scratch when switching DSP architecture, because they interface with the DSP in a very distinct manner. Such a memory controller can include sophisticated multi-channel DMAs, a cache controller, and a memory management unit (MMU).
For example, a memory controller is dependent on the specific processor, and it differs from one processor to the other. Some of these can be easily reused for a newer DSP core, by running on standard bus connectivity, while others require a redesign, sometimes from scratch. Assuming an IP DSP core is in place, it must be integrated into a System on a Chip (SoC), with a set of peripherals, memories, system interfaces, etc.
This DSP architecture framework meets the varied requirements imposed by different users and offers diverse benefits to different DSP users and applications.Ī DSP technology is not utilized in a void rather it is part of an integrated system, a total solution. As a consequence, OEMs often tend to switch DSP architectures for different applications, and migrate to a new architecture for the benefit of next generation products. These application-specific programmable DSPs, though cost effective per specific application, are not competitive platforms for use with other tasks and functions.
These different applications have created a wide range of processing demands, and have driven DSP specialists to customize their offerings and provide their customers with DSP technologies that can better perform specific tasks or applications. Eran Briman of CEVA describes reuse of a single DSP architecture for various applications and markets.Īs DSP usage becomes ubiquitous for voice compression and decompression, audio decoding and encoding applications, signal modulators and equalizers, video applications, and advanced vocoders and transcoders, the market need for DSPs is tremendous and growing.