Open Source Precision Timed Soft Processor for Cyber Physical System Applications
Abstract
Modern processor architectures sacrifice timing predictability to improve average performance. Branch prediction, out-of-order execution, and multi-level cache hierarchies complicate accurate execution time estimates. The timing demands of Cyber Physical Systems (CPS) have led some to propose new processor architectures, including Precision Timed (PRET) processors, which simplify analysis of execution time by removing the sources of indeterminacy. This paper presents an open source PRET processor for use in CPS applications based on the Open Fire soft processor clone of the Xilinx Micro Blaze. By interleaving instructions from multiple threads onto the pipeline all data dependencies are removed. Because of this, and with all instructions completing in identical time, accurate execution time calculations are possible. In an FPGA implementation the PRET modifications significantly improve overall throughput with only a modest increase in area.
Related Papers
- → The effect of speculatively updating branch history on branch prediction accuracy, revisited(1994)66 cited
- → Analysis of the execution time unpredictability caused by dynamic branch prediction(2004)42 cited
- → Evaluation of dynamic branch predictors for modern ILP processors(2002)4 cited
- User-guided interleaving of planning and execution(1996)
- → The Bimode++ Branch Predictor(2006)18 cited