SWIRL: High-performance many-core CPU code generation for deep neural networks. (November 2019)

• Record Type: Journal Article
• Authors: Venkat, Anand; Rusira, Tharindu; Barik, Raj; Hall, Mary; Truong, Leonard
• Other Names: Dongarra Jack guest-editor.; Tourancheau Bernard guest-editor.
• Is Part Of: International journal of high performance computing applications. Volume 33:Number 6(2019)
• Page Start: 1275
• Page End: 1289
• Publication Date: 2019-11
• Subjects: Compilers -- code generation -- optimization -- deep neural networks -- code transformations; High performance computing -- Periodicals; Supercomputers -- Periodicals; 004.1105
• DOI: 10.1177/1094342019866247
• Languages: English
• ISSNs: 1094-3420

Abstract

Deep neural networks (DNNs) have demonstrated effectiveness in many domains including object recognition, speech recognition, natural language processing, and health care. Typically, the computations involved in DNN training and inferencing are time consuming and require efficient implementations. Existing frameworks such as TensorFlow, Theano, Torch, Cognitive Tool Kit (CNTK), and Caffe enable Graphics Processing Unit (GPUs) as the status quo devices for DNN execution, leaving Central Processing Unit (CPUs) behind. Moreover, existing frameworks forgo or limit cross layer optimization opportunities that have the potential to improve performance by significantly reducing data movement through the memory hierarchy. In this article, we describe an alternative approach called SWIRL, a compiler that provides high-performance CPU implementations for DNNs. SWIRL is built on top of the existing domain-specific language (DSL) for DNNs called LATTE . SWIRL separates DNN specification and its schedule using predefined transformation recipes for tensors and layers commonly found in DNN layers. These recipes synergize with DSL constructs to generate high-quality fused, vectorized, and parallelized code for CPUs. On an Intel Xeon Platinum 8180M CPU, SWIRL achieves performance comparable with Tensorflow integrated with MKL-DNN; on average 1.00× of Tensorflow inference and 0.99× of Tensorflow training. It also outperforms the original LATTE compiler on average by 1.22× and 1.30× on inference and training, respectively.