Seminar: Reflections on 25 Years of Parallel Computing: Hardware transformations, programming incrementalism, and the tenacity of Chapel

_Abstract

Over the past 25 years, the capabilities of parallel computers have grown immensely due to technology trends such as multicore processors, high-radix/low-bandwith networks, and GPU computing. Such trends have also made parallel computing abundant after decades of being fairly niche: today, parallelism is readily available via our laptops, smartphones, single-board computers, the cloud, and AI datacenters. However, despite these massive improvements in capability and ubiquity, the programming notations used for scalable parallel programming are much the same as they were 25 years ago. Meanwhile, mainstream programming has seen the rise of new languages that provide productivity and/or safety benefits, such as Python or Rust.

In this talk, I'll motivate and introduce Chapel, an open-source programming language developed by my team at Cray Inc. and HPE over the past 25 years. Chapel supports general-purpose parallel programming via code that's similar to Python in clarity, yet with performance and scalability that can rival or beat conventional HPC notations. I'll describe how Chapel's approach of expressing parallelism and locality independently of specific hardware features or mechanisms has made it resilient to the changes in parallel system architecture that have taken place during its lifetime. Along the way, I'll also characterize the diversity of applications that users have written in Chapel and touch on some of the social and practical challenges involved in getting a new language like Chapel adopted.

For those who'd like a more in-depth or hands-on introduction to the language, shortly after this seminar's conclusion I'll be giving an interactive Chapel tutorial.

_Biography

Brad Chamberlain is a Distinguished Technologist at Hewlett Packard Enterprise (formerly Cray Inc.) who has spent his career focused on user productivity for HPC systems, particularly through the design and development of the Chapel parallel programming language
(https://chapel-lang.org).  He received his Ph.D in Computer Science & Engineering from the University of Washington in 2001 where he focused on the ZPL parallel array language, and he remains associated with the department as an affiliate professor of the Paul G. Allen School.