Explore the innovative approach to scalable parallel programming in this ACCU 2017 conference talk. Delve into the motivations behind Chapel, a general-purpose language designed to address the shortcomings of current parallel programming paradigms. Learn about Chapel's key features, including its focus on productivity, portability, and open-source nature. Discover how Chapel aims to make parallel programming more accessible and modern while supporting distributed-memory computing. Examine code examples demonstrating Chapel's base language features, task parallelism, locality, and distributed data parallelism. Gain insights into the challenges faced by the Chapel development team and their strategies for overcoming them. Compare Chapel's performance against other languages in benchmarks and real-world applications. Whether you're an HPC enthusiast or simply curious about the future of parallel programming, this talk offers a thought-provoking look at making scalable parallelism a first-class language feature.
The Audacity of Chapel - Scalable Parallel Programming Done Right
Overview
Syllabus
Intro
Fair Warning Statement
Cray Market Segments
I don't really care about HPC programming... CRAY
What does Productivity mean to you?
"The Audacity of Chapel"?
Scalable Parallel Programming Concerns
STREAM Triad: Chapel
HPC's Status Quo: SPMD Programming
SPMD by Example (in code)
Chapel's Goal
The Challenge
Chapel is Portable
Chapel is Open-Source
Chapel Community R&D Efforts
Chapel language feature areas
Base Language Features, by example
Task Parallelism and Locality, by example
Parallelism and Locality: Distinct in Chapel
Higher-Level Features
Distributed Data Parallelism, by example
Domain Maps
Chapel's Domain Map Philosophy
Chapel's Multiresolution Philosophy
Two Other Multiresolution Features
Computer Language Benchmarks Game CG CRAY
CLBG: Chapel vs. 9 key languages
RA Performance: Chapel vs. MPI
Suggested Reading (short attention spans)
Taught by
ACCU Conference
