Chapel

Chapel is an open source programming language designed  for productive parallel computing on large scale systems. It’s development is being led by Cray Inc.

Chapel is designed to run on high-end super computers, but also, due to its portability, runs on mulitcore desktops and laptops, clusters, and the cloud. Chapel supports multithreaded execution using abstractions for data and task parallelism, concurrency and nested parallelism. Chapel also supports object-oriented programming, type inference and generic programing.

To me, coding in chapel feels like a cross between Python and C++. It does not require a main() function, like Python, allowing for rapid prototyping. However for production grade code it uses a more structured programming style by declaring modules and using a main() function as the entry point.

Chapel supports data-parallel execution using forall loops which behaves like OpenMP’s #pragma omp parallel for. Task-parallelism is achieved using coforall loops which create a new thread for every iteration. This is usefull for when each iteration does large amounts of work. It also uses begin and  cobegin to perform tasks in parallel, similar to OpenMP’s tasks and sections.

Chapel is designed to be easy to learn for users of most other languages like C, C++, Java Python, Matlab, etc. It builds on concepts and syntax from other programming languages.

Advertisements

#chapel, #parallel-computing