Title: Domain Specific Languages for Programming Many-Core Processors (GPUs)
Abstract: As part of the Stanford Pervasive Parallelism Laboratory,
we are developing three domain-specific languages for parallel computing
which should be of interest to researchers in biocomputing.
The first is Liszt, which is targeted at solving partial differential
equations on meshes. The second is Kore, which is a data parallel
language inspired by k, a language for massive column-store databases.
The third is a language to support probabilistic computation.
In this talk, I will outline the approach we are taking. I will
concentrate on Liszt, but briefly described the other languages.
As an introduction, I will also present an overview of the techniques
used by GPUs to achieve high-performance, since many-core architectures are
one of the targets for these languages.
