Model Based FPGA Design Space Exploration and Optimization

Object-oriented hardware design techniques are gaining popularity as a means generate robust FPGA implementations. They potentially allow subject matter experts to convert high-level algorithms models directly into robust FPGA programs without the assistance of a hardware design specialist. Other potential benefits include increased productivity, reduced development costs, and fewer design errors. These model-based design techniques typically offer less efficient implementations and currently offer no means for the subject matter expert to tailor a design to meet specific size, weight, or power (SWAP) constraints. We propose a methodology that will provide the subject matter expert with FPGA Design Space Exploration and Optimization (DeSEO) capability in the domains of performance, power, and size. A meta-model, or object-oriented representation, of the algorithm will be used to provide a link between the high-level model and low-level FPGA implementation tools. Tobject-oriented hardware design techniques are gaining popularity as a means generate robust FPGA implementations. They potentially allow subject matter experts to convert high-level algorithms models directly into robust FPGA programs without the assistance of a hardware design specialist. Other potential benefits include increased productivity, reduced development costs, and fewer design errors. These model-based design techniques typically offer less efficient implementations and currently offer no means for the subject matter expert to tailor a design to meet specific size, weight, or power (SWAP) constraints. We propose a methodology that will provide the subject matter expert with FPGA Design Space Exploration and Optimization (DeSEO) capability in the domains of performance, power, and size. A meta-model, or object-oriented representation, of the algorithm will be used to provide a link between the high-level model and low-level FPGA implementation tools. This intermediate representation offers greater code portability while providing greater flexibility for design optimization and code generation.