What is the difference between DSP digital signal processing and general-purpose CPU?
[Copy link]
From the perspective of microprocessor resolution, there seems to be no difference between the two. But the areas they focus on are completely different. General-purpose CPUs focus on transaction processing, and their support for real-time performance is much worse than that of DSPs. Although some general-purpose CPUs are also adding some high-performance computing support, they are still far behind DSPs. For example, current general-purpose DSPs can complete a 1024-point complex FFT within a few us, and general-purpose processors can never reach this level (as it is now), although their system clock may be an order of magnitude higher than that of DSPs. But the difference between completing multiple complex operations in one cycle and completing one operation in multiple cycles is very large. DSPs complete algorithm-intensive transactions, and the work may be relatively simple, but the real-time requirements are very high. All required operations must be completed within a certain period of time (possibly just a few us). General-purpose CPUs are very good at transaction management, The ability in this aspect is much better than that of DSP. Therefore, most high-performance systems are a combination of DSP and general-purpose CPU. Make full use of their respective advantages. Because DSP pays special attention to high performance, its structure is much more complicated than general-purpose CPU under the same conditions, and it is more difficult to design. Therefore, its price is a little more expensive than general-purpose CPU (only compared with general-purpose DSP). Some DSPs with very single functions are very cheap, even cheaper than general-purpose CPU. Because general-purpose DSP has to consider generality, its structure is also different from that of special-purpose DSP.
|