MULTIPLE-OBJECTIVE OPTIMIZATION IN A BEHAVIORAL SYNTHESIS SYSTEM

The paper describes the implementation of an 'intelligent' silicon com piler that provides the ability to optimise a design, given as a behav ioural description, with respect to multiple objectives. The user subm its goals or objectives to the system, named MOODS, which automaticall y finds an optimal solution in the context of the user-specified const raints. The MOODS system provides a sound basis for multiple objective optimisation and allows the user to characterise and explore the desi gn space. An accurate representation of the implementations in the des ign space is ensured by feeding up technology-dependent information fr om a cell library. This data is used in a global cost function which g uides the application of transformations on a multilevel representatio n of the design. The results obtained show that designs can be produce d which meet users' objectives and that a varied set of implementation s can be synthesised from a single behavioural specification. The resu lting characteristics of the set of implementations enable the design space to be characterised.