Comparing Parallel Architectures from a Reconfigurable Computing Perspective

The sequential “stored program”model used for most computers is seriously limiting performance in several ways. This has lead to a search for more efficient ways of doing computations. One of the most important improvements is to start doing computations in parallel. This paper looks at the parallel property of reconfigurable computing and compares it with the traditional approach of extending the standard stored program model with parallel concepts. The paper concludes with the observation that reconfigurable devices naturally leads to different solutions than traditional approaches, and that the granularity concept from the field of reconfigurable computing easily extends to all kinds of parallel structures.