Paul Barrett

Paul Barrett is an Associate Research Professor in the Department of Physics at The George Washington University. His studies focus on multi-wavelength observations of Cataclysmic Variable stars, primarily at radio frequencies. He also has over forty years of scientific software experience in astronomy including the development of Scientific Python beginning thirty years ago. He is currently a lead developer for several Julia programming language projects and a member of the JuliaAstro steering council.


Sessions

11-01
13:00
120min
An Introduction to the Julia Programming Language
Paul Barrett

The tutorial will give a brief overview of the high-performance scientific programming language Julia. It will show how Julia solves the two language problem by compiling command-line expressions directly to machine code. It will touch on performing calculations using measurements (i.e., value-error pairs), units, symbolic manipulation, and GPU programming. It will conclude by showing how the combination of multiple dispatch and abstract types result in concise, generic, and high-performance software. The tutorial assumes a basic knowledge of scientific programming.

Topical Computing, Software and Algorithms
Seminar Room 1
11-02
15:48
12min
AURIS.jl: A High Performance Radio Interferometry Software Suite
Paul Barrett

Most radio astronomy software was designed and developed over 30 years ago before parallel processing was the norm. Therefore, it does not take advantage of the high-performance computing resources that are available today such as new software technology in the form of on-the-fly compilation and advanced numerical algorithms like auto-differentiation. AURIS (A*stronomical Utilities for Radio Interferometry and Simulation) is a Julia software suite designed to use the latest technologies whether that be now or in the future. This enables AURIS to process large volumes of data efficiently and accurately. All algorithms are designed to take advantage of the inherently parallel nature of radio interferometry data by distributing computations across multicore CPUs and GPUs, when available. In addition, AURIS by default uses half-precision (16-bit) floats for most calculations, which reduces the memory footprint and processing time by half without a decrease in accuracy. As a result, the AURIS calibration pipeline is over fifty times faster than the NRAO CASA calibration pipeline for VLA data, reducing calibration time from several hours to several minutes. This talk will discuss how the various features of the Julia programming language and its associated packages are used to implement a modular, high-performance software suite that can quickly and easily adapt to the increasing volume and wide variety of radio data, such as the next-generation VLA (ngVLA) and the Event Horizon Telescope* (EHT).

Building and operating science platforms and workflows in the Petabyte Era
Conference Foyer 1