LaMeres' Research Overview
Exploiting Reconfigurable Fabrics to Optimize Computing
Reconfigurable digital circuitry has always provided inherent design flexibility. Historically, this flexibility has come at the expense of performance. However, with advances in IC fabrication technology, programmable logic performance has advanced to the point where it is meeting the computation needs of modern applications. This has created a paradigm shift in the way digital circuitry can be implemented. If programmable logic cells can perform as well as dedicated IC blocks, then the standard VLSI design flow will need to be reinvented. Reprogrammable fabrics now have the capacity to contain not only custom hardware but multiple soft processor cores. This capability enables computing approaches such as single-chip hardware accelerated processors, dynamically scalable parallel processing, and reconfigurable computing. The capability that now exists in computing hardware makes the effective partitioning between hardware and software a difficult challenge due to the endless implementation possibilities.
Reconfigurable fabrics also have enabled novel architectures to address the robustness of a computer system. Redundancy (both static and dynamic) can be used to detect and recover from faults and spatial avoidance of faults can be used to extend lifetime of a part. These opportunities for fault tolerance are of great interest to the military and aerospace industry due to their unique need for robust computing platforms.
Currently, research is being conducted in the ECE department at MSU in the area of effective hardware/software partitioning using soft processors on FPGAs. Research is also being conducted on the design of a radiation tolerant computing system for the aerospace industry which exploits partial reconfiguration of an FPGA to spatially move soft processors to different locations on the FPGA in order to avoid radiation strikes. Reconfiguration is also used to dynamically recover from a non-damaging radiation strike.
The picture shown here is a Radiation Tolerant Many Core Computing System implemented on a Xilinx Artix-7 FPGA. This system was developed for NASA to help increase reliability in interplanetary flight systems. This system contains 16 soft processors. At any given time, 3 of the processors are used in Triple Modular Redundancy (TMR) to check for faults due to radiation. Upon a soft radiation strike, the TMR system reboots and resynchronizes the faulted processors. If the fault occurs in the reconfiguration RAM of the FPGA, the system performs partial reconfiguration on the effected processor in order to recover. If the fault is unrecoverable, the system brings on a new spare processor to form the TMR configuration and marks the damaged area as unusable.
Shown here is an implementation of the Radiation Tolerant Computer that is currently on the International Space Station.