The largest modern supercomputers are increasingly built using high-diameter networks with scalable topologies such as meshes and tori. These networks are cost effective in that they reduce the cost of switching hardware and are easily expandable. However, these topologies also make application performance sensitive to the placement of application processes on the physical network. Existing techniques to optimize applications for networks such as these do not take complex internal communication structure or application phases into account. However, these features could significantly reduce the complexity of the search for an optimal mapping. Moreover, they allow for more intuitive attribution of performance data to application constructs. This paper describes a communication measurement framework to record performance data pertaining to application phases and application communication structure using instrumentation provided by application developers. We show that visualizing data collected using our framework can provide valuable insights into the performance of large-scale application codes, and that such data may eventually be used to guide automatic node-mapping for future petascale and exascale applications.