Tiramisu: A Code Optimization Framework for High Performance Systems
Authors:
Preprint
Abstract
This paper introduces Tiramisu, a polyhedral framework designed to generate high performance code for multiple platforms including multicores, GPUs, and distributed machines. Tiramisu introduces a scheduling language with novel extensions to explicitly manage the complexities that arise when targeting these systems. The extensions include explicit communication, synchronization, and mapping buffers to different memory hierarchies. Tiramisu relies on a flexible representation based on the polyhedral model and explicitly uses a well-defined four-level IR that allows full separation between the algorithms, loop transformations, data-layouts, and communication. This separation simplifies targeting multiple hardware architectures with the same algorithm. We evaluate Tiramisu by writing a set of image processing and stencil benchmarks and compare it with state-of-the-art compilers. We show that Tiramisu matches or outperforms existing compilers on different hardware architectures, including multicore CPUs, GPUs, and distributed machines.Keywords
All Related Versions
This publication has 21 references indexed in Scilit:
- A Unified Backend for Targeting FPGAs from DSLs2018
- PENCIL: A Platform-Neutral Compute Intermediate Language for Accelerator Programming2015
- POLLY — PERFORMING POLYHEDRAL OPTIMIZATIONS ON A LOW-LEVEL INTERMEDIATE REPRESENTATIONParallel Processing Letters, 2012
- Loop Transformation Recipes for Code Generation and Auto-Tuning2010
- isl: An Integer Set Library for the Polyhedral Model2010
- POET: Parameterized Optimizations for Empirical Tuning2007
- Semi-Automatic Composition of Loop Transformations for Deep Parallelism and Memory HierarchiesInternational Journal of Parallel Programming, 2006
- New Complexity Results on Array Contraction and Related ProblemsJournal of Signal Processing Systems, 2005
- Automatic storage management for parallel programsParallel Computing, 1998
- Dataflow analysis of array and scalar referencesInternational Journal of Parallel Programming, 1991