Software synthesis with evolutionary algorithms

This paper addresses the problem of exploiting simultaneously trade-offs between program memory, data memory and execution time requirements for algorithms specified by synchronous data flow graphs. A code generation model is used, that includes many variable parameters. The search space includes all valid schedules of a graph, different models of buffer organization and different ways of code invocation. An evolutionary algorithm (EA) is applied to this multi-objective optimization problem. This approach is evaluated and compared with existing methods