Concurnas is a JVM programming language designed for concurrent and parallel computing. It provides an alternative concurrency model to that of threads and locks found in Java. Much of the work which inspired this concurrency model was conducted in the 1970s but largely forgotten due to chip architectures being single cored in nature.
But things changed recently, now multi-core CPUs are commonplace and developers are obliged to implement systems utilizing concurrency in order to obtain optimal performance. Furthermore, we have also recently seen GPUs take center stage for powering general-purpose parallel computing, developers have yet another tool at their disposal for solving computational problems.
In this session, a high-level overview of Concurnas, focusing on concurrency and GPU programming will first be provided. Then we shall deep dive into the inspiration for and implementation of the Concurnas concurrency model within the language, compiler, and runtime. Finally, we'll explore how GPU computing is implemented within Concurnas and how this integrates with the core concurrency model.
Почему этот доклад в программе
Чем хорош спикер: Jason Tatton worked many years on high-frequency trading systems for different well-known banks before he decided to spend several years on the design and implementation of Concurnas.
Чем хороша тема: Concurrency is hard and taming concurrency in heterogeneous CPU/GPU environments is even harder! This is also true for Java with its built-in support for concurrent programming. Concurnas is a new and fresh approach to tackle this problem on the JVM.
Кому будет полезно: This talk is for everybody interested in new language trends or seeking new ideas for building reliable high performance concurrent, distributed and parallel systems.
Почему здесь и сейчас: Come and hear directly from the Concurnas author himself about Concurnas' basic features, design principles and technical implementation details.