Designing embedded systems with hardware IPs is already a standard practice in the industry. These kind of IPs usually have parameters such that they can be optimized in term of power, performance or cost. We then apply this concept to the software side by proposing parametrized software IP in order to reduce embedded software design complexity. For instance, for a given system behavior, a designer just specifies requirements (energy, delay and cost), then the design optimizer generates software algorithms and assembly codes accordingly. In this paper we elaborate the proposal for parametrized embedded software IPs and show related experiments in our Design System with Soft-Core Processor.
展开▼
