5th ACM SIGPLAN International Workshop on Libraries, Languages and Compilers for Array Programming

Philadelphia, USA - June 19, 2018

DEADLINE: April 15, 2018

ARRAY 2018 is part of PLDI 2018

39th Annual ACM SIGPLAN Conference on Programming Language Design and Implementation

Keynote: Albert Cohen

Tensor Comprehensions: deep learning as a polyhedral compiler's killer app

Deep learning models with convolutional and recurrent networks analyze massive amounts of audio, image, video, text and graph data, with applications to automatic translation, speech-to-text, scene understanding, ranking user preferences, ad placement, etc. Competing frameworks for building these networks such as TensorFlow, Chainer, CNTK, Torch/PyTorch, Caffe1/2, MXNet and Theano, explore different tradeoffs between usability and expressiveness, research or production orientation and supported hardware. They operate on a DAG of computational operators, wrapping high-performance libraries such as CUDNN (for NVIDIA GPUs) or NNPACK (for various CPUs), and automate memory allocation, synchronization, distribution. Custom operators are needed where the computation does not fit existing high-performance library calls, usually at a high engineering cost. Such operators suffer a severe performance penalty, which limits the pace of innovation. Furthermore, existing library primitives often do not offer optimal performance in a particular network architecture, missing optimizations between operators as well as specialization to the size and shape of data.

Accepted Papers

Call for Papers

About

Array-oriented programming offers a unique blend of programmer productivity and high-performance parallel executions. As an abstraction, array programming directly mirrors high-level mathematical abstractions common in the sciences as well as in many compute-intensive applied areas. The data-parallel nature of array programming facilitates advanced analyses and with it, compiler-driven code generation for modern massively parallel hardware platforms.

This workshop is intended to bring together researchers from many different communities, including language designers, library developers, compiler researchers and practitioners who are using or working on numeric, array-centric aspects of programming languages, libraries and methodologies from all domains: imperative or declarative; object-oriented or functional; interpreted or compiled; strongly typed, weakly typed or untyped.

Topics

Application experiences of array programming, from productivity to parallel performance

Array, graph, and tensor abstractions

Compilers and libraries for array and graph programs on potentially massively parallel computers

Building-blocks for (dense and sparse) matrix/tensor algorithms

Compiler transformations and intermediate languages for array computations

Systematic array notation, including axis- and index-based approaches,

Representation of mathematical structure, including sparsity, rank, and hierarchy

Array programming is at home in many communities, reaching from domain experts that primarily use array languages, to computer scientists that research the various aspects of array-language design, analysis, implementation and infrastructure. ARRAY is intended as a forum where these communities can exchange ideas on the construction of computational tools for arrays.

Important Dates

Submissions

Manuscripts may fall into one of the following categories:

• research papers on any topic related to the focus of the workshop;
• tool descriptions reporting on a tool relevant to the workshop area.

Submissions should be 4-8 pages for research papers and 4-6 pages for tool descriptions. All papers should be formatted in conformance with the ACM SIGPLAN proceedings style.

In the case of a tool description, the workshop presentation should include a demo of the tool, and the submission should include a short appendix summarizing the demo. This appendix is for the information of the PC only and will not be part of the published paper, nor will it be counted as part of the six-page limit.

Clearly mark your submission as either a "research paper" or a "tool description" in the paper’s subtitle.

Submissions must be in PDF format, printable in black and white on US Letter sized paper, and interpretable by Ghostscript. Papers must adhere to the standard SIGPLAN conference format: two columns, nine-point font on a ten-point baseline, with columns 20pc (3.33in) wide and 54pc (9in) tall, with a column gutter of 2pc (0.33in). A suitable document template for LaTeX is available at http://www.sigplan.org/Resources/Author/.

Papers must be submitted using EasyChair: https://easychair.org/conferences/?conf=array2018.

As in previous years, accepted papers will be published in the ACM Digital Library.

Organizing Committee

Program Committee

ARRAY 2018 is sponsored by ACM SIGPLAN. Presenters and authors of papers are eligible to apply for SIGPLAN PAC funding.