Technology Opportunities

The interactions between multiple parallel processors incur performance and power overheads. These overheads are partly a result of data synchronization, entailing exchange between a sequential core and the other cores at the commencement and conclusion of a concurrent section of the workload. While cycles lost on data exchange may be hidden, the power consumption spent on such exchange remains and may significantly reduce the power efficiency of the overall system. GP-SIMD ‎ is 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.