Exploring multi-core solutions
Performance bottlenecks often originate from limits in computing platforms. As increasing the performance of computing platforms is no longer achievable through a rise in clock frequency, embedded systems increasingly rely on multi-core realisations to meet growing performance needs. However, the various alternatives for exploiting potential parallelism in multi-core embedded systems can yield very different performance gains.
Therefore, the alternatives must be carefully evaluated. For this purpose, design-space exploration methods are ideal for providing systematic analysis. They are an effective way to select the preferred design solution as a starting point for multi-core implementations.
Performance modelling: Developing adequate (parameterised) models for performance analysis
Calibration and validation: Calibrating the developed performance model to ensure adequacy by validating against single-core solutions
Design space exploration: Investigating alternative ways to exploit potential parallelism in embedded applications
Solution selection and realisation: Selecting a design solution with feasible performance as starting point for a multi-core implementation