A general constraint-centric scheduling framework for spatial architectures

Abstract

Specialized execution using spatial architectures provides energy efficient computation, but requires effective algorithms for spatially scheduling the computation. Generally, this has been solved with architecture-specific heuristics, an approach which suffers from poor compiler/architect productivity, lack of insight on optimality, and inhibits migration of techniques between architectures. Our goal is to develop a scheduling framework usable for all spatial architectures. To this end, we expresses spatial scheduling as a constraint satisfaction problem using Integer Linear Programming (ILP). We observe that architecture primitives and scheduler responsibilities can be related through five abstractions: placement of computation, routing of data, managing event timing, managing resource utilization, and forming the optimization objectives. We encode these responsibilities as 20 general ILP constraints, which are used to create schedulers for the disparate TRIPS, DySER, and PLUG architectures. Our results show that a general declarative approach using ILP is implementable, practical, and typically matches or outperforms specialized schedulers.

On the Scalability of Constraint Solving for Static/Off-Line Real-Time Scheduling

“…Our paper also develops over previous work on applying exact solving techniques (ILP/SMT/CP, more classical branch-and-cut, model checking) to off-line real-time scheduling problems [12,6,8]. These previous results all focus on either the (improved) modeling of a given scheduling problem or on the improvement of the solver algorithm.…”

“…We do not consider in this category the many other papers which discard exact solving techniques based on theoretical considerations, rather than experimental ones, because the issue is indeed to consider the potential gap between the two. -A few recent papers reporting the successful application of modern constraint solving tools to realistic real-time scheduling and compilation problems [12,6], due to increase in solver efficiency.…”

Exploring the potential of heterogeneous von neumann/dataflow execution models

“…We specifically extend a general scheduling framework for spatial architectures [27] with the ability to model instruction bundling. We formally describe this next and present brief results.…”

Constraint centric scheduling guide

“…To address these deficiencies, Nowatzki et al have designed a spatial scheduling framework, formulating the problem using integer linear programming (ILP) [8]. Their solution is advantageous in that the formulation is general enough to be applied to a variety of architectures, and their constraint-based approach finds solutions with guaranteed optimality bounds.…”

“…Their solution is advantageous in that the formulation is general enough to be applied to a variety of architectures, and their constraint-based approach finds solutions with guaranteed optimality bounds. This paper describes, from a developer's perspective, the spatial scheduling tool we have created, which leverages the framework of [8]. Written in the GAMS language [1], this tool's source is at http://www.cs.wisc.edu/vertical/ilp-scheduler, and this paper serves as its usability guide.…”