Litcius/Paper detail

Graphene: An IR for Optimized Tensor Computations on GPUs

Bastian Hagedorn, Bin Fan, Hanfeng Chen, Cris Cecka, Michael Garland, Vinod Grover

202328 citationsDOIOpen Access PDF

Abstract

Modern GPUs accelerate computations and data movements of multi-dimensional tensors in hardware. However, expressing optimized tensor computations in software is extremely challenging even for experts. Languages like CUDA C++ are centered around flat buffers in one-dimensional memory and lack reasonable abstractions for multi-dimensional data and threads. Existing tensor IRs are not expressive enough to represent the complex data-to-thread mappings required by the GPU tensor instructions.

Topics & Concepts

Computer scienceCUDAComputationParallel computingThread (computing)Tensor (intrinsic definition)Computational scienceSoftwareTheoretical computer scienceProgramming languageMathematicsGeometryParallel Computing and Optimization TechniquesTensor decomposition and applicationsAdvanced Data Storage Technologies