On the (Re-)Use of IP-Components in Re-configurable Platforms

When mapping high-throughput signal processing applications onto heterogeneous platforms, parts of these applications will most likely be mapped into re-configurable components. The current trend of automating such mappings assumes that one can transform application specification to implementation specification without delving deep into details. This implies that one has to rely on the (re-)use of IP components, even when it remains unknown what the final exact specification of the implementation components will be. When it comes to 1) (re-)using of IP components, 2) porting of designs from one platform (component) to another, and 3) relying on IP components in high-level (task-level) application to (re-configurable) platform (component) mapping, the design requirements and constraints are not well defined. To find out what they should be, we have conducted two case studies around the (re-)use of IP components; one focusing on the porting issue, and one dealing with the automated task-level mapping. The goal of these case studies was to identify and resolve the problems that come with the so-called completion logic or glue logic that hinders fast and accurate IP embedding in a high-level implementation specification and description.