Episode 171: Scala Update with Martin Odersky

Parallel Challenges

Parallel programming in Scala faces significant challenges, especially when scaling to handle many cores effectively. Martin Odersky highlights the increasing demand for compute power beyond graphics, such as big data and precise models, which necessitates more parallelization 1. He argues that while tools like actors and software transactional memory (STM) are useful, they don't fully address the complexities of parallelism, which involves using many cores for sequential tasks 1.

You will have to paralyze more and more things and nobody knows how to do that.

--- Martin Odersky

Odersky emphasizes the need for running sequential applications in parallel without burdening programmers with concurrency issues, distinguishing parallelism from concurrency 2.

   

Concurrency

Concurrency in Scala is evolving, with a focus on embedding domain-specific languages rather than extending the language with concurrent constructs. Martin Odersky explains that the goal is to handle the increasing number of processor elements efficiently, as hardware developments continue to advance 3. He notes that the introduction of parallel collections in Scala 2.9 allows operations to be interpreted in parallel, enhancing performance without requiring programmers to manage concurrency directly 4.

We take the collections framework and we make to give you optionally the possibility to interpret all operations in parallel.

--- Martin Odersky

This approach aims to balance the need for parallelism with the simplicity of sequential programming, leveraging tools like fork join thread pools to optimize execution 4.