The computational requirements imposed by cellular standards has increased exponentially from 1G to 4G. This is due to the increased complexity of algorithms introduced to reduce the bit-error rate use of the wireless spectrum more efficiently. These algorithms have been shown to require more performance than is currently available in embedded processors. Table 1, shows the computation requirements to handle 384 Kbits/s transmission rate on a 3G WCDMA receiver[3]. The problem will persist in the future since the computation requirements are growing faster than Moore's law[3]. Power dissipation is also a major problem in battery powered mobile computing and communication devices. While the computational requirements are increasing exponentially, battery capacity is only improving at the rate of 1.03 times per year[15]. Flexibility and low time to market require the use of programmable processors for the implementation of the increasingly sophisticated digital signal processing algorithms. Power efficiency on the other hand, requires the use of a customized solution. To make 4G systems usable, the performance per unit power consumption of processors needs to improve significantly - an area in which stream processors have a significant advantage.
|