Chapel (programming language)
| Company type | Open source project |
|---|---|
| Industry | Software, Programming languages |
| Predecessor | Cray Cascade initiative |
| Founded | 2009 |
| Founder | Cray |
| Products | Parallel programming language |
| Owner | Hewlett Packard Enterprise |
| Website | chapel-lang |
Chapel, known as the Cascade High Productivity Language, is a parallel programming language initially created by Cray and subsequently maintained by Hewlett Packard Enterprise following its acquisition of Cray. It originated from the Cray Cascade initiative under the DARPA High Productivity Computing Systems (HPCS) program, which aimed to improve supercomputer productivity by the year 2010. Chapel is distributed as an open source project under the terms of the Apache License version 2.
The Chapel compiler is developed using C and C++ (specifically C++14), with its optimization backend built upon LLVM technology. Certain optional utilities, such as the testing framework and c2chapel—a utility for creating C interface bindings—depend on Python version 3.7 or later. While Chapel primarily produces binary executables, it also has the capability to generate C source code directly, circumventing LLVM. The language supports building libraries that can be invoked from C, Fortran, and Python environments. Furthermore, Chapel facilitates GPU programming through code generation compatible with NVIDIA and AMD GPU architectures.
Goals
The primary objective of Chapel is to improve the ease of programming on parallel computers, with a focus on the Cascade architecture, by providing a higher-level language that distinctly separates algorithmic logic from the underlying data structure implementations.
The creators designed Chapel to serve as a bridge between conventional high-performance computing (HPC) developers—who mostly use Fortran, C, or C++ combined with OpenMP and MPI—and newer programmers accustomed to languages such as Java, Python, or Matlab. The language strives to merge the productivity advantages of modern programming languages with the performance demands inherent in HPC applications.
Features
Chapel offers a sophisticated multithreaded parallel programming framework that includes abstractions for data parallelism, task parallelism, and nested parallelism. It enhances data locality and computational efficiency by providing abstractions for data distribution alongside data-driven placement of subcomputations. The language encourages code reuse and versatility through its support for object-oriented and generic programming paradigms, including constructs for defining locales.
The parallel programming constructs in Chapel are heavily influenced by concepts from High Performance Fortran (HPF), ZPL, and extensions to Fortran and C found in the Cray MTA system.
See also
References
Further reading
- Panagiotopoulou, K.; Loidl, H. W. (2016). "Transparently Resilient Task Parallelism for Chapel" Parallel and Distributed Processing Symposium Workshops (IPDPSW), 2016 IEEE International Symposium, Chicago, IL.