Pushing parallel processing

The adoption of multicore processing by Intel and AMD has spurred indirectly the digital signal processor (dsp) start up market. So argues Jeff Bier, president of dsp analysis and development firm BDT. “There are now 20 start ups with massively parallel processing architectures and many of these are multicore,” he said. Mainstream processing’s embrace of multiple core has lent legitimacy to the approach, resulting in an influx of venture capital to dsp start ups after several years of neglect. One dsp start up benefiting from the VCs’ attention is Silicon Valley based Stream Processors. As implied by the company’s name, its architecture uses a computing approach known as stream processing. Operating on a stream of data, the approach applies intensive parallel processing to each of the stream’s segments or data blocks. An example stream could be video or broadband wireless data. A data block is processed – for example using a transform – and sent on. This contrasts with traditional processing, where data is cached, with the expectation that it will be needed again.