RUSH: A Rad-Hard Unified Scalable Heterogeneous Processing Architecture

Space presents a challenging environment for computing. Extendeddevelopment times and radiation tolerance requirements leave hardwareperformance a decade or more behind the terrestrial state-of-the-artat the time of deployment. Additionally, once deployed, hardwarechanges are impractical, encouraging a trend towards increasedsoftware programmability. However, topside pressure from applicationadvancements are forcing space-based platforms to improve throughputand latency while reducing power consumption. A popular approach toaddressing the tension between these requirements is the heterogeneousprocessing architecture. By providing multiple hardware tools thatoptimally support a subset of the anticipated workload, aheterogeneous architecture can offer a diverse processing toolset tothe application developer. However, programming these systems isextremely challenging because of variations in toolsets and datasharing interfaces. As a result, data sharing and dynamic workloadscheduling across heterogeneous architectures is often suboptimal andhindered by poor scalability. Maxentric proposes to solve this problemwith RUSH, a heterogeneous processing architecture with a unifiedprogramming model for rapid development. RUSH employs a rad-hardmulticore processor as a host and an FPGA as an accelerator chip. TheRUSH software layer unifies these architectures through an innovativeprogramming model described in the proposal.