Tartan Evaluating Spatial Computation for Whole Program…格子呢评估整个计划的空间计算.pdfVIP

Tartan: Evaluating Spatial Computation for Whole Program Execution Mahim Mishra,† Timothy J. Callahan,† Tiberiu Chelcea,† Girish Venkataramani,† Mihai Budiu‡ and Seth C. Goldstein† †Carnegie Mellon University ‡ Microsoft Research Pittsburgh, PA Mountain View, CA {mahim,tcal,tibi,girish,seth}@ mbudiu@ Abstract mance. In fact, demand for low power devices may mean that per- Spatial Computing (SC) has been shown to be an energy-efficient formance may no longer be the crucial metric of success. This model for implementing program kernels. In this paper we explore changing landscape has prompted many to explore alternative ar- the feasibility of using SC for more than small kernels. To this end, chitectures that harness coarse-grained parallelism; for example, we evaluate the performance and energy efficiency of entire appli- commercial multi-core chips capitalize on parallelism by replicat- cations on Tartan, a general-purpose architecture which integrates ing a number of simple cores, thereby boosting scalability. While a reconfigurable fabric (RF) with a superscalar core. Our compiler these chips are focused on server workloads which are rich in automatically partitions and compiles an application into an in- threads, it is not yet clear how single-threaded applications can struction stream for the core an