Frontier of Embodiment Informatics: ICT and Robotics, Top Global University ProjectWaseda University

Irregular applications such as big graph analysis, material simulations, molecular dynamics simulations, and finite element analysis have performance problems due to their use of sparse data　structures. Inspector-executor strategies are used to improve sparse computation performance through parallelization and data locality optimizations. An inspector re-schedules and reorders data at runtime, and an executor is a transformed version of the original computation that uses the newly reorganized data structures and schedules. Inspector-executor transformations are commonly written in a domain-specific or even application-specific fashion. Significant progress has been made in incorporating such inspector-executor transformations into existing compiler transformation frameworks, thus enabling their use with compile-time only transformations. However, composing inspector-executor transformations with each other is done on a case-by-case basis so as to reduce inspector overhead.

The Sparse Polyhedral Framework (SPF) is a compiler loop and array transformation model that can generally compose inspector-executor transformations with each other. In this talk, I review the history and current state of the art for inspector-executor strategies, review how the SPF enables the composition of inspector-executor transformations, and present current research to better balance the generality-performance tradeoff that is being done with the Sparse Polyhedral Framework.