Structural Aware Quantitative Interprocedural Dataflow Analysis

A key challenge in system design whether for high performance computing or in embedded systems is to automatically partition software on the thread level for target architectures like multi-core, heterogeneous, or even hardware/software co-design systems. Similar to the techniques employed on the instruction level, the inter-procedural data flow in a software system is an essential property for identifying global data dependencies and hence extracting tasks or exploiting coarse grained parallelism. Also, detailed insight into the global data flow in quantity and its dynamics is vital to precisely estimate interconnect resources demands (e.g. bandwidth, latency) in distributed or embedded, heterogeneous target environments. This paper presents a novel approach to quantitative, global dataflow analysis based on a virtual instruction set architecture, which allows precise investigation of interprocedural dataflow over time of complex software system from the atomic to the transactional level. The introduced analysis system LLILA is part of the Synphony HW/SW co-design framework and can be used to evaluate the quality of system partitioning.