(PLDI 2020) Predictable Accelerator Design with Time-Sensitive Affine Types
While field-programmable gate arrays (FPGAs) provide an opportunity to co-design applications with hardware accelerators, they remain difficult to program. High-level synthesis (HLS) tools promise to raise the level of abstraction by compiling C or C ++ to accelerator designs. Repurposing legacy software languages, however, requires complex heuristics to automatically map unrestricted imperative code onto hardware structures. We find that the black-box heuristics in HLS tools can be unpredictable: changing parameters in the program that should improve performance can counterintuitively yield slower and larger FPGA implementations. This paper proposes a type system that restricts HLS to programs that can predictably compile to hardware accelerators. The key idea is to model consumable hardware resources with a time-sensitive affine type system that prevents conflicting simultaneous uses of the same hardware structure. We implement the type system in Dahlia, a programming language that compiles to HLS C ++ , and evaluate how its type system can reduce the size of HLS parameter spaces while accepting Pareto-optimal designs.
Wed 15 JunDisplayed time zone: Pacific Time (US & Canada) change
13:30 - 14:50 | Domain Specific LanguagesSIGPLAN Track at Cockatoo Chair(s): Zachary Tatlock University of Washington | ||
13:30 20mTalk | (POPL 2021) Petr4: Formal Foundations for P4 Data Planes SIGPLAN Track Ryan Doenges Cornell University, Mina Tahmasbi Arashloo Cornell University, Santiago Bautista Univ Rennes, ENS Rennes, Inria, IRISA, Alexander Chang Cornell University, Newton Ni Cornell University, Samwise Parkinson Cornell University, Rudy Peterson Cornell University, Alaia Solko-Breslin Cornell University, Amanda Xu Cornell University, Nate Foster Cornell University | ||
13:50 20mTalk | (PLDI 2020) Predictable Accelerator Design with Time-Sensitive Affine Types SIGPLAN Track Rachit Nigam Cornell University, Sachille Atapattu Cornell University, USA, Samuel Thomas Cornell University, USA, Theodore Bauer AWS Inc, Apurva Koti Cornell University, USA, Zhijing Li Cornell University, USA, Yuwei Ye Cornell University, USA, Adrian Sampson Cornell University, Zhiru Zhang Cornell University, USA | ||
14:10 20mTalk | (POPL 2022) Dependently-Typed Data Plane Programming SIGPLAN Track Matthias Eichholz Technical University of Darmstadt, Eric Campbell Cornell University, Matthias Krebs TU Darmstadt, Nate Foster Cornell University, Mira Mezini TU Darmstadt | ||
14:30 20mTalk | (POPL 2022) Safe, Modular Packet Pipeline Programming SIGPLAN Track |