Chapter

Euro-Par 2009 Parallel Processing

Volume 5704 of the series Lecture Notes in Computer Science pp 809-820

Out-of-Core Computation of the QR Factorization on Multi-core Processors

  • Mercedes MarquésAffiliated withDepto. de Ingeniería y Ciencia de Computadores, Universidad Jaume I (UJI)
  • , Gregorio Quintana-OrtíAffiliated withDepto. de Ingeniería y Ciencia de Computadores, Universidad Jaume I (UJI)
  • , Enrique S. Quintana-OrtíAffiliated withDepto. de Ingeniería y Ciencia de Computadores, Universidad Jaume I (UJI)
  • , Robert van de GeijnAffiliated withDepartment of Computer Sciences, The University of Texas at Austin

* Final gross prices may vary according to local VAT.

Get Access

Abstract

We target the development of high-performance algorithms for dense matrix operations where data resides on disk and has to be explicitly moved in and out of the main memory. We provide strong evidence that, even for a complex operation like the QR factorization, the use of a run-time system creates a separation of concerns between the matrix computations and I/O operations with the result that no significant changes need to be introduced to existing in-core algorithms. The library developer can thus focus on the design of algorithms-by-blocks, addressing disk memory as just another level of the memory hierarchy. Experimental results for the out-of-core computation of the QR factorization on a multi-core processor reveal the potential of this approach.

Keywords

Dense linear algebra out-of-core computation QR factorization multi-core processors high performance