Litcius/Paper detail

Sparseloop: An Analytical, Energy-Focused Design Space Exploration Methodology for Sparse Tensor Accelerators

Yannan Nellie Wu, Po-An Tsai, Angshuman Parashar, Vivienne Sze, Joel Emer

202120 citationsDOI

Abstract

This paper presents Sparseloop, the first infrastructure that implements an analytical design space exploration methodology for sparse tensor accelerators. Sparseloop comprehends a wide set of architecture specifications including various sparse optimization features such as compressed tensor storage. Using these specifications, Sparseloop can calculate a design's energy efficiency while accounting for both optimization savings and metadata overhead at each storage and compute level of the architecture using stochastic tensor density models. We validate Sparseloop on a well-known accelerator design and achieve ~99% accuracy in terms of runtime activities (e.g., compressed memory accesses). We also present a case study that highlights the key factors (e.g., uncompressed traffic, data density) that affect sparse optimization features' impact on energy efficiency. Tool available at: https://github.com/NVlabs/timeloop.

Topics & Concepts

Computer scienceTensor (intrinsic definition)MetadataUncompressed videoDesign space explorationSet (abstract data type)Overhead (engineering)Computational scienceComputer engineeringKey (lock)Sparse matrixArchitectureTheoretical computer scienceParallel computingEmbedded systemComputer hardwareProgramming languageMathematicsOperating systemArtPure mathematicsVisual artsQuantum mechanicsVideo processingGaussianVideo trackingPhysicsParallel Computing and Optimization TechniquesLow-power high-performance VLSI designTensor decomposition and applications
Sparseloop: An Analytical, Energy-Focused Design Space Exploration Methodology for Sparse Tensor Accelerators | Litcius