Budapest Hackathon 2016/Talks
The following are the planned talks for this years event. All the slides will be uploaded to the events github page.
András Kovács - A tour of GHC 8 features
Abstract: GHC 8 brought us a number of new features. This talk focuses on the changes to the type system (new kind system, explicit type applications, injective families) and presents some use cases and idioms that were impossible or unwieldy with previous GHC releases. We also touch on Strict Haskell, stack tracing, custom type errors, record field overloading, and their current usability.
Dániel Berényi - Selected use cases of structured recursion schemes
Abstract: We show how the simplest structured recursion schemes can drive generic tree manipulations. Catamorphisms perform bottom-up, while Anamorphisms do the opposite (top-down) recursive traversals. The transformations applied to the trees are completely separated from the traversal in the form of algebras and coalgebras. We show how these methods can be used to do simple manipulations to a small Embedded Domain Specific Language. We briefly review their potential applications in High-Performance Computing applications.
Dániel is an associate of the GPU Lab at Wigner Research Centre for Physics.
Dániel Vigovszky - Haskell at Prezi
Abstract: Prezi uses Haskell in its document model related projects. In this talk first we will explain the domain and show the problems we wanted to solve. Following that we will first show the unique build environment we created to be able to work efficiently on complex Haskell codebases. Then we will look into our extensive use of QuickCheck, how it's used and what we learned.
Máté Karácsony - Zeldspar: The Road to Epiphany
Abstract: Zeldspar is a domain-specific language to implement digital signal processing pipelines. It's an implementation of the Ziria language (by Microsoft Research) on the top of Feldspar (an EDSL written in Haskell for signal processing, developed at Chalmers in Göteborg, Sweden). In this talk I will present how Zeldspar is built on the top of a deep language stack, and I also show how it can be compiled to run efficiently on many-core devices such as Adapteva's Epiphany chips and Parallella boards.