Optimal allocation of I/O device parameters in hardware and software codesign methodology

For a programmable I/O device controller, the allocation of device parameters on I/O registers affects the code size and execution time of its associated I/O device driver. In traditional design flow, the development of device drivers can not begin until the allocation is fixed. This paper presents a new design methodology that allows a designer to seek an allocation that reduces the software or hardware cost concurrently with developing device drivers. The software cost means the code size or execution time and the hardware cost the number of I/O registers. The exact allocation with the minimum cost under constraints is formulated as zero-one integer linear programming problem. Heuristic algorithms based on iterative refinement are also proposed. The proposed design methodology was implemented in C language. Compared with current industrial designs, the approach can obtain design alternatives that reduce both software and hardware costs. Furthermore, the experimental results also investigate design spaces for various application features. It turns out that the HW/SW codesign approach is favorable in development of embedded systems.