MISAAL: Synthesis-Based Automatic Generation of Efficient and Retargetable Semantics-Driven OptimizationsRemote
Using program synthesis to select instructions for and optimize input programs is receiving increasing attention. However, existing synthesis-based compilers are faced by two major challenges that prohibit the deployment of program synthesis in production compilers: exorbitantly long synthesis times spanning several minutes and hours; and scalability issues that prevent synthesis of complex modern compute and data swizzle instructions, which have been found to maximize performance of modern tensor and stencil workloads. This paper proposes MISAAL, a synthesis-based compiler that employs a novel strategy to use formal semantics of hardware instructions to automatically prune a large search space of rewrite rules for modern complex instructions in an offline stage. MISAAL also proposes a novel methodology to make term-rewriting process in the online stage (at compile-time) extremely lightweight so as to enable programs to compile in seconds. Our results show that MISAAL reduces compilation times by up to a geomean of 16x compared to the state-of-the-art synthesis-based compiler, HYDRIDE. MISAAL also delivers competitive runtime performance against the production compiler for image processing and deep learning workloads, Halide, as well as HYDRIDE across x86, Hexagon and ARM.
Wed 18 JunDisplayed time zone: Seoul change
14:00 - 15:40 | |||
14:00 20mTalk | Functional Meaning for Parallel Streaming PLDI Research Papers DOI Pre-print | ||
14:20 20mTalk | MISAAL: Synthesis-Based Automatic Generation of Efficient and Retargetable Semantics-Driven OptimizationsRemote PLDI Research Papers Abdul Rafae Noor University of Illinois at Urbana-Champaign, Dhruv Baronia University of Illinois at Urbana-Champaign, Akash Kothari University of Illinois at Urbana-Champaign, Muchen Xu University of Illinois at Urbana-Champaign, Charith Mendis University of Illinois at Urbana-Champaign, Vikram S. Adve University of Illinois at Urbana-Champaign DOI | ||
14:40 20mTalk | First-Class Verification Dialects for MLIR PLDI Research Papers Mathieu Fehr The University of Edinburgh, Yuyou Fan University of Utah, Hugo Pompougnac Université Grenoble Alpes; Inria; CNRS; Grenoble INP, John Regehr University of Utah, Tobias Grosser University of Cambridge DOI | ||
15:00 20mTalk | [TOPLAS] A Modular Approach to Metatheoretic Reasoning for Extensible LanguagesRemote PLDI Research Papers Dawn Michaelson University of Minnesota, Gopalan Nadathur University of Minnesota, Eric Van Wyk University of Minnesota, Twin Cities DOI | ||
15:20 20mTalk | Handling the Selection Monad PLDI Research Papers DOI |