A method for capturing state data on dynamically reconfigurable processors

In a preemptive multitasking environment, when a task is preempted, necessary state information must be correctly preserved in order for the task to be resumed later. For hardware tasks executing on a coarse-grained dynamically reconfigurable processing array (DRPA), a greate amount of state data are usually distributed on many different storage elements. Besides, DRPAs have different architectures using a variety of development tools. This paper addresses such problems and propose a method for capturing the state data of hardware tasks. Based on resource usage analysis, algorithms for identifying preemption points and inserting preemption states subject to user-specified preemption latency are proposed. Also, the integaration of the proposed steps into the system design flow is discussed. The performance degradation caused by preemption is minimized by allowing preemption only at predefined points where demanded resources are small. The evaluation result using a model based on NEC Electronics DRP-1 shows that the proposed method could allow preemption for a certain task satisfying a given preemption latency with reasonable hardware overhead (from 6% to 15%).