2010 Poster Sessions : Liszt a DSL for Mesh-Based PDEs

Student Name : Maria Montserrat Medina Martinez
Advisor : Patrick Hanrahan
Research Areas: Graphics/HCI
Abstract:
To achieve high performance, general-purpose programming languages require that solvers be written at low level for a specific architecture. Thus, the state-of-the-art solvers produced by PSAAP are limited to one platform. New platforms such as GPGPUs promise large performance gains but require a different programming model.
Rather than commit to re-architecting the code for every new programming model, we propose writing the code at a higher level. Liszt is a domain specific language that exposes a high-level interface for building mesh-based solvers of PDEs. Domain specific optimizations allow Liszt to provide similar performance to low-level code without committing to a particular architecture.
Liszt code runs either on a cluster or an SMP. A GPU runtime is in progress. Liszt frees scientist to write simulation code without worry about architecture-specific tuning.

Bio:
Montse Medina is a 2nd year Ph.D. student at the Institute of Computational and Mathematical Engineering of Stanford University. She is advised by Pat Hanrahan. She is working on Liszt, domain specific language for scientific computing. Montse's broad interests include scientific and parallel computing and data mining. Montse holds an MS in Aeronautics and Astronautics from Stanford University.