Actor Oriented Programming in Chapel

Programming for scalable multi-cores with more focus on locality has increased the interest in message passing models of programming. One of the well known theoretical and practical models of message passing is the actor model. There are a multitude of actor oriented libraries and languages, each of which implement some variant of actor semantics. Among the languages and the libraries based on the actor semantics we can name Erlang, SALSA, E language and Axum. Chapel is a global view language and have no notion of actor and message passing. However, it tends to be a general language for parallel and concurrent programming and wants to provide simple means for the programmer to implement a parallel program. In this work we want to study and analyze some questions regarding actors in the Chapel language. In particular we want to answer the following questions: • Why do we need actor semantics in Chapel? In particular we want to see if we really need something like actors in Chapel. • What are the advantages and disadvantages of having actor semantics implemented in Chapel? By answering this question we will see what is the profit of having actors for programmer. Also we try to see what is the cost of having the actor semantics implemented in chapel. • Can we do actor oriented programming with current Chapel language constructs? This question tries to reveal how well or poorly the current constructs in Chapel support actor style programming. • What changes are needed to support true actor semantics in Chapel? The answer to this question helps us to realize the changes required to the compiler and run-time in order to have a true actor semantics in Chapel. This study would be valuable to the Chapel community because it provides a fair amount of knowledge on the cost and benefits of having actors in Chapel and can help them with starting points for implementing actor semantics in the language.