Séminaire Cambium, Inria Paris, 2 rue Simone Iff
EN LIGNE
Lundi 14 mars, 15h00
Dougal Mclaurin
Google Brain
Designing Dex: differentiation, parallelism, and index types
Dex is a new language for array-oriented numerical computing. It emphasizes
expressive types, parallelism and automatic differentiation. The core idea is
the treatment of arrays as memoized representations of functions with finite
domains. These domains can be interesting, meaningful types rather than just
subsets of the integers. Instead of bulk-array combinators, we argue for a
programming style heavy on explicit array indexing, mirroring "pointful"
function application. The language is designed to run on data-parallel hardware
like GPUs and it has automatic differentiation built in. There turns out to be
an interesting challenge in reconciling automatic differentiation with
parallelism, which we solve by adding a fine-grained effects system.
Vous pouvez vous abonner à nos annonces de séminaires:
http://cambium.inria.fr/seminar.html
Nos séminaires sont accessibles en ligne en direct:
https://webconf.math.cnrs.fr/b/fra-ryy-fjn