Vyasa: A High-Performance Vectorizing Compiler for Tensor Convolutions on the Xilinx AI Engine
Preprint
Abstract
Xilinx's AI Engine is a recent industry example of energy-efficient vector processing that includes novel support for 2D SIMD datapaths and shuffle interconnection network. The current approach to programming the AI Engine relies on a C/C++ API for vector intrinsics. While an advance over assembly-level programming, it requires the programmer to specify a number of low-level operations based on detailed knowledge of the hardware. To address these challenges, we introduce Vyasa, a new programming system that extends the Halide DSL compiler to automatically generate code for the AI Engine. We evaluated Vyasa on 36 CONV2D and 6 CONV3D workloads, and achieved geometric means of 7.6 and 23.3 MACs/cycle for 32-bit and 16-bit operands (which represent 95.9% and 72.8% of the peak performance respectively). For 4 of these workloads for which expert-written codes were available to us, Vyasa demonstrated a geometric mean performance improvement of 1.10x with 50x smaller code relative to the expert-written codes.Keywords
All Related Versions
This publication has 18 references indexed in Scilit:
- T2S-Tensor: Productively Generating High-Performance Spatial Hardware for Dense Tensor Computations2019
- Caffeine: Toward Uniformed Representation and Acceleration for Deep Convolutional Neural NetworksIEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2018
- The ARM Scalable Vector ExtensionIEEE Micro, 2017
- Automatically scheduling halide image processing pipelinesACM Transactions on Graphics, 2016
- PolyMage2015
- Learning Hierarchical Features for Scene LabelingIEEE Transactions on Pattern Analysis and Machine Intelligence, 2012
- Automatic SIMD vectorization of fast fourier transforms for the larrabee and AVX instruction sets2011
- AnySPACM SIGARCH Computer Architecture News, 2009
- SODA: A High-Performance DSP Architecture for Software-Defined RadioIEEE Micro, 2007
- Automated empirical optimizations of software and the ATLAS projectParallel Computing, 2001