WOLFHPC: Fourth International Workshop on
Domain-Specific Languages and High-Level Frameworks for High Performance Computing

November 17, 2014

Full-day workshop in conjunction with

SC14: The International Conference for High Performance Computing, Networking, Storage, and Analysis

New Orleans, Louisiana, USA

Held in cooperation with ACM SIGHPC


| Home | Advance Program |

Advance Program

Session 1:

9:00 Opening remarks
9:00-10:00 Compiler technology for solving PDEs with performance portability [Talk]
Keynote Speaker: Paul Kelly
Imperial College London
Abstract: We have become used to the idea that higher-level languages supporting a higher level of abstraction come with a price in performance. In contrast, we should expect that the more information the compiler has about the structure and properties of our code, the more scope it should have for optimisation. This talk is about our experience in the Firedrake project of trying to make good on this idea. We are building software tools for a solving PDEs on unstructured meshes, mainly using the finite element method. We support a concise high-level programming model, based on the FEniCS Project's Unified Form Language, while mapping onto a high-performance implementation entirely automatically via an intermediate representation for loops over the mesh, called PyOP2. Our compiler is based around three layers of domain-specific program representation, each supporting different optimisations. The resulting software tools, implemented in Python, achieve higher performance than established C++ and Fortran codes.

Break: 10:00-10:30

Session 2: Building Domain-Specific Languages

10:30-11:00 Patrick McCormick, Christine Sweeney, Nick Moss, Dean Prichard, Samuel Gutierrez, Kei Davis and Jamaludin Mohd-Yusof
"Exploring the Construction of a Domain-Aware Toolchain for High-Performance Computing" [Talk]
11:00-11:30 Alejandro Fernández, Vicenç Beltran, Sergi Mateo, Tomasz Patejko and Eduard Ayguadé
"A Data Flow Language to Develop High Performance Computing DSLs" [Talk]
11:30-12:00 Anthony Danalis, George Bosilca, Aurelien Bouteiller, Thomas Herault and Jack Dongarra
"PTG: an abstraction for unhindered parallelism" [Talk]

Lunch Break: 12:00-1:30

Session 3: Framework for Loop Transformations

1:30-2:30pm The Exascale Materials Co-design Center and Lessons Learned from the 2014 LLNL Co-design Summer School [Talk]
Keynote Speaker: James Belak
Lawrence Livermore National Laboratory
Abstract: The DOE Office of Science, Advanced Scientific Computing Research program formed Application Co-design Centers to provide a workload for the developers of future exascale computer systems to design against. At the same time, the lessons learned from engagements with hardware and software developers are used to re-design the application codes, that is, Co-design. In this talk, we review the co-design process as executed in the Exascale Co-design Center for Materials in Extreme Environments (ExMatEx) and, in particular, review the lessons learned from the 2014 LLNL Co-design Summer School. To date, the application center has focused on exposing the work as we currently express it. Recently however, there has emerged an increased awareness of the need for application codes to express the application work in a manner that is less dependent on specific machine (worker) configurations. This change in focus from worker to work is fundamental to task-based programming. Core to these models is a reliance on the runtime system to schedule this "task decomposition" and achieve good data locality and avoid resource conflicts (race conditions). The purpose of the summer school was manifold: the creation of materials science application codes in these programming models to guide model developers, and to assess the models for the expression of materials science applications.
2:30-3:00 Christos Kartsaklis, Eunjung Park and John Cavazos
"HSLOT: The HERCULES Scriptable Loop Transformations Engine" [Talk]

Break: 3:00-3:30

Session 4: Grid Abstractions

3:30-4:00 Christian Schmitt, Sebastian Kuckuk, Frank Hannig, Harald Köstler and Jürgen Teich
"ExaSlang: A Domain-Specific Language for Highly Scalable Multigrid Solvers" [Talk]
4:00-4:30 Richard Membarth, Philipp Slusallek, Marcel Köster, Roland Leißa and Sebastian Hack
"Target-Specific Refinement of Multigrid Codes" [Talk]
4:30-5:00 Ist&vacute;an Reguly, Gihan Mudalige, Mike Giles, Dan Curran and Simon McIntosh-Smith
"The OPS Domain Specific Abstraction for Multi-Block Structured Grid Computations" [Talk]
5:00-5:30 Bruce Palmer, William Perkins, Yousu Chen, Shuangshuang Jin, David Callahan, Kevin Glass, Ruisheng Diao, Mark Rice, Stephen Elbert, Mallikarjuna Vallem, Zhenyu Huang
"GridPack: A Framework for Developing Power Grid Simulations on High Performance Computing Platforms" [Talk]
5:30 Closing Remarks