10th Seminar on High Performance Numerical Computing
La Maison de la simulation organise le 28 juin 2013 un séminaire ouvert à tous sur le thème "High performance numerical computing" qui sera animé par les professeurs Kunle Olukotun (Stanford university) et Nahid Emad (Université de Versailles).
- Date : 28/06/2013
- Lieu : Maison de la simulation (bâtiment Digiteo - 565, CEA Saclay)
- Intervenant(s) : Kunle Olukotun et Nahid Emad
- Organisateur(s) : Maison de la Simulation
“High Performance Embedded Domain Specific Languages”
Prof. Kunle OLUKOTUN (Pervasive Parallelism Lab., Stanford University).
Today, all high-performance computer architectures are parallel and heterogeneous; a combination of multiple CPUs, GPUs and specialized processors. This creates a complex programming problem for application developers. Domain-specific languages (DSLs) are a promising solution to this problem because they provide an avenue for high-level application-specific abstractions to be mapped directly to low level architecture-specific programming models providing both high programmer productivity and high execution performance. In this talk I will describe our approach to building high performance DSLs, which is based on embedding in Scala, light-weight modular staging and a DSL infrastructure called Delite. I will describe how we transform DSL programs into efficient first-order low-level code using domain specific optimization, parallelism optimization, locality optimization, scalar optimization, and architecture-specific code generation. I will also briefly explain how these techniques can be used to generate specialized hardware. All optimizations and transformations are implemented in an extensible DSL compiler architecture that minimizes the programmer effort required to develop new DSLs.
Kunle Olukotun is a Professor of Electrical Engineering and Computer Science at Stanford University. Olukotun is a pioneer in chip multiprocessor (CMP) design and lead the Stanford Hydra CMP research project. Olukotun founded Afara Websystems to develop high-throughput, low-power server systems with CMP technology. The Afara microprocessor, called Niagara, was acquired by Sun Microsystems. Niagara derived processors now power all Oracle SPARC-based servers. Olukotun currently directs the Stanford Pervasive Parallelism Lab (PPL) which seeks to proliferate the use of parallelism in all application areas using Domain Specific Languages (DSLs). Olukotun is an ACM Fellow and IEEE fellow.
“Programming Model for Sustainable Numerical Libraries for Extreme Scale Computing”
Prof. Nahid EMAD (PRiSM and “Maison de la Simulation” laboratories, University of Versailles).
The promise of computer systems with very large orders of processing elements cannot be realized without an effective solution that targets the programming model with a suitable programming environment. Nowadays, it is necessary to identify and rapidly make available robust software technologies to enable high-end computer applications to run efficiently on these emerging systems, and to enable the development of more complex and capable simulation codes for scientific and engineering applications. We present a modular and multi-level parallelism approach that has a strict separation between computational operations, data management and communication. We illustrate how under this model we are able to design more scalable algorithmic implementations that better exploit their functionalities, scalability and code reusability. We also show how this model even enables the formulation of hybrid numerical scheme to be run efficiently on multi-level parallel systems with a large number of heterogeneous processing units without confining the parallelism to the programming model of the communication library. We use the multiple restarted Arnoldi methods as our test cases and our experiments include comparisons with state-of-the-art of numerical libraries on high-end computing systems.
Nahid Emad is Professor of Computer Science at University of Versailles. She has Ph.D. and MS in Applied Mathematics from Pierre et Marie Currie University (Paris VI). She leads the Intensive Numerical Computation group of Computer Science department at the University of Versailles. She is the advisor of more than 10 Ph.D. and authored more than 70 papers in international journals and conferences. She is specialized in numerical algorithms in linear algebra, parallel and distributed computing and software engineering for parallel and distributed numerical computing.