Concurrent Hash Tries - striving to make things simple and fast

with Phil Bagwell

Abstract:

Concurrent Hash Tries (http://lampwww.epfl.ch/papers/idealhashtrees.pdf) are a new data structure that allow non-blocking concurrent update from multiple threads. It forms the basis of a general technique to slide between the speed of mutable and the thread safety of immutable structures. Phil will cover some of the new data structure developments that enable the Scala (http://www.scala-lang.org/) parallel collections (http://infoscience.epfl.ch/record/150220/files/pc.pdf), their new concurrent hash tries, and perhaps a couple of other new structures. All this from the internal algorithm and data structure design point of view rather that the language. The main focus has been on trying to make things simpler for the developer to exploit the multi-core performance potential of modern machines, hence the title.

Bio: http://photos3.meetupstatic.com/photos/event/3/6/3/a/event_84013882.jpeg

Phil started his career as an electronics and then software engineer. For many years he was the European Services Marketing Manager for Digital Equip Co (DEC). For the last few years he has worked with Martin Odersky at EPFL (http://lamp.epfl.ch/~odersky/) where he has contributed the key data-structures used in both Scala (http://www.scala-lang.org/)'s, Clojure (http://clojure.org/)’s and Haskell (http://www.haskell.org/haskellwiki/Haskell)'s immutable collections. Recently he was a co-contributor to the new Scala parallel collections (http://infoscience.epfl.ch/record/150220/files/pc.pdf), concurrent Hash Tries (http://infoscience.epfl.ch/record/166908/files/ctries-techreport.pdf) and fast vector concatenation (http://infoscience.epfl.ch/record/169879/files/RMTrees.pdf). He is now the Marketing Manager for Typesafe (http://typesafe.com/), the Scala support company.