Following on from our automatic parallelisation talk. We will demonstrate our use of addressability in our system which has grown to millions of lines of code running on numerous compute cores. We will give live demonstrations of how we use addressability to provide a way to safely mutate variables that combines the rigour of functional programming with the convenience of OO. We will show how this technique also enables us to distribute computation with changes across many threads and machines without concurrency problems. Additional live demonstrations of further addressability benefits.
All Scala developers should have enough knowledge to follow this talk. We don't assume anymore knowledge than basic Scala programming.
Our objective is to share what we have learnt from the construction of this large system.
I couldn't find somewhere to write this, but there will be 2 presenters for this talk.
Morgan Stanley. I participate in a large Scala project of many millions of lines of code. We use Java to build a novel runtime which solves a number of problems we have encountered in distributed enterprise computing.
Gjeta is a project engineer working for the last year and a half in the core technology of the project. She has been profiling applications to identify reasons for cache misses and concurrency bottlenecks. She has also been tracking and analysing efficiencies across the codebase and including caching of bitemporal database queries for optimal reuse.