Date Added: Jun 2014
The authors present GP-SIMD, a novel hybrid general purpose SIMD computer architecture that resolves the issue of synchronization by in-memory computing, through combining data storage and massively parallel processing. The GP-SIMD's arithmetic, logic and associative characteristics and its intercommunication network are discussed. GP-SIMD employs modified SRAM storage cells. An analytic performance model of the GP-SIMD architecture is presented, comparing it to Associative Processor and to conventional SIMD architectures. Cycle-accurate simulation supports the analytical comparison. Assuming a moderate die area, GP-SIMD architecture outperforms both the Associative Processor and the SIMD co-processor architectures by almost an order of magnitude while consuming less power.