A Reconfigurable SIMD-MIMD Processor Architecture for Embedded Vision Processing Applications

In designing vision processors for advanced automotive safety applications, the developer must address various issues, including true real-time performance and power efficiency as well as software flexibility to account for various applications, situations and environments.

While parallelization through SIMD (single instruction/ multiple data) architectures has been a proven solution to speed up the initial processing of an image, SIMD processors have limited efficiency for the final image-processing steps that are mainly serialized or require floating-point arithmetic, where a MIMD (multiple instructions/multiple data) architecture would be beneficial.

The paper introduces the architecture of the IMAPCAR-XC® core, a highly parallel processor that incorporates in its design the flexibility to handle all of these requirements efficiently.