Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Collaborative Research and Development Grants

McMaster University

$114,429.00

Leveraging Structure and Sparsity on SIMD and Multicolore Architectures

EF 2010-11

Résumé
Computer simulations are important for both basic science and product/process simulations. The breadth and power of these simulations have grown significantly as processor power has increased in a predictable fashion over more than a decade. But in the last several years, growth has shifted from clock-speed increases to increases in computational width as the result of design changes forced on chip designers by basic physics. Width means parallelism, either within an instruction stream (SIMD) or across multiple cores, and frequently in combination. Developing software for these new architectures is challenging because in addition to the intrinsic algorithmic difficulties, the parallelism necessary to achieve scalable performance imposes new algorithmic constraints. The Coconut Project of McMaster University is developing a new generation of user tools to make it easier for programmers to specify explicitly and implicitly parallel algorithms. The challenges are twofold: de-couple computation into pieces which can be executed in parallel, and partition the computation into pieces small enough to be addressed efficiently by future processors. In this Collaborative Research and Development project, we propose to adapt and generalize existing research into the efficient implementation of basic building blocks of scientific computation, to (1) address issues arising from the proliferation of SIMD and multi-core hardware by developing novel scheduling algorithms which incorporate resource restrictions, (2) extend partitioning algorithms designed for linear operations to the multi-linear computations which arise when simulating and identifying complex interacting systems, like networks of metabolites, or partially self-catalysing industrial or environmental reaction-diffusion systems. To validate the effectiveness of the algorithms we develop, we will generate code for the Cell BE architecture, and the latest pre-production hardware made available to us by IBM Canada. Our research results will lead to increased capabilities and increased performance for compiler products developed by IBM Canada, following on the significant 4X and 5X performance improvements represented by the MASS libraries for Cell BE generated by Coconut distributed by IBM. As important as the new ideas we will generate, we will train approximately ten students who will have an even greater impact after they graduate. They will contribute to the growing capability in compiler technology centered in southern Ontario, which in addition to IBM's major Compiler group includes compiler developers at AMD, RapidMind and Google.
Programme
Collaborative Research and Development Grants
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
385237-2009
Name nom
Anand, Christopher
Department département
Computing and Software
Organization id
27
Institution établissement
McMaster University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2010
Competition year année de concours
2009
Award amount
114429
Program id
CRDPJ
Program name en
Collaborative Research and Development Grants
Program name fr
Subventions de recherche et développement coopérative
Group en
Research partnerships programs
Group fr
Programmes de partenariat
Committee code
96
Committee name en
RPP Internal Decision Cttee
Committee name fr
Cté de décision interne - PPR
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2715
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Optimization
Research subject fr
Optimisation
Partie
2
Nb partie
3
Application title
Leveraging Structure and Sparsity on SIMD and Multicolore Architectures
Keyword
NSERC
Application summary
Computer simulations are important for both basic science and product/process simulations. The breadth and power of these simulations have grown significantly as processor power has increased in a predictable fashion over more than a decade. But in the last several years, growth has shifted from clock-speed increases to increases in computational width as the result of design changes forced on chip designers by basic physics. Width means parallelism, either within an instruction stream (SIMD) or across multiple cores, and frequently in combination. Developing software for these new architectures is challenging because in addition to the intrinsic algorithmic difficulties, the parallelism necessary to achieve scalable performance imposes new algorithmic constraints. The Coconut Project of McMaster University is developing a new generation of user tools to make it easier for programmers to specify explicitly and implicitly parallel algorithms. The challenges are twofold: de-couple computation into pieces which can be executed in parallel, and partition the computation into pieces small enough to be addressed efficiently by future processors. In this Collaborative Research and Development project, we propose to adapt and generalize existing research into the efficient implementation of basic building blocks of scientific computation, to (1) address issues arising from the proliferation of SIMD and multi-core hardware by developing novel scheduling algorithms which incorporate resource restrictions, (2) extend partitioning algorithms designed for linear operations to the multi-linear computations which arise when simulating and identifying complex interacting systems, like networks of metabolites, or partially self-catalysing industrial or environmental reaction-diffusion systems. To validate the effectiveness of the algorithms we develop, we will generate code for the Cell BE architecture, and the latest pre-production hardware made available to us by IBM Canada. Our research results will lead to increased capabilities and increased performance for compiler products developed by IBM Canada, following on the significant 4X and 5X performance improvements represented by the MASS libraries for Cell BE generated by Coconut distributed by IBM. As important as the new ideas we will generate, we will train approximately ten students who will have an even greater impact after they graduate. They will contribute to the growing capability in compiler technology centered in southern Ontario, which in addition to IBM's major Compiler group includes compiler developers at AMD, RapidMind and Google.