The Research On Platform-Based SoC System Synthesis

In order to cope with the increasing complexity of integrate circuits and the pressure of Time-To-Market, the platform-based design methodology which emphasizes platform level re-use becomes a major focus. But the traditional platform-based SoC design flow is too experimental to adopt enough automatic techniques. In traditional design flow, designer chooses platform by experience firstly, then adds or removes IPs to get new SoC implementation finally, the two stages can not be fulfilled automatically. The research in this dissertation focuses on the automatic PSSS (Platform-based SoC System Synthesis) techniques including specification and optimization algorithm.The main original works in this thesis are following.1. The proposed PSSS techniques in this thesis can be implemented automatically.2. The two-stage mapping model can avoid a lot of invalid codes during the system synthesis.3. To speed up the convergence of the GA, it is presented using two-stage memory model to modify the mutation operator, and a two-stage memory model based hill climbing genetic algorithm is proposed. The two methods are proved valid during the experience.4. The plaguing premature convergence problem has always been the hindrance for GAs to be applied in wide optimization problems. The past research works focus on the crossover operator and the mutation operator of the GAs and ignore the influence of the selection operator. This thesis proposes a novel selection operator, modified ranking selection operator (MRS). MRS includes the Static Linear Ranking Selection operator (S-MRS (L)), the Dynamic Linear Ranking Selection operator (D-MRS(L)), the Static Exponential Ranking Selection operator(S-MRS(E)) and the Dynamic Exponential Ranking Selection operator(D-MRS(E)).At last, the proposed PSSS are proved to be valid through an example.The proposed PSSS technique is the first automatic technique supporting the platform-based SoC design.