Research Interests - Peter Strazdins

• Computer Architecture for High Performance Computing:
  • multicore computing design and organization, including energy considerations
  • heterogeneous multicore design for efficient OS and virtualization support
  • network design for parallel (cluster) computers; SMP and cc-NUMA memory system design
  • performance analysis, performance evaluation and simulation of shared and distributed memory parallel processors
• Systems Development for High Performance Computers
  • global address space paradigms for clusters (e.g. OpenMP, global arrays, service-oriented approaches)
  • virtualization support for high performance and cloud computing
  • Operating Systems design and implementation, for SMP and cluster parallel computers; scheduling
  • Distributed Java Virtual Machine implementation
• Computational Science Applications: Weather and climate science, electromagnetic compatibility analysis and data mining
• Numerical algorithms on parallel and high-performance computers
  • Parallel BLAS and LAPACK
    • BLAS for GPUs and multicore processors
    • parallel algorithm design, parallel numerical libraries: design and implementation, interface and portability issues.
  • Symmetric indefinite solvers and data mining
  • Sparse direct solvers
  • Matrix factorization algorithms
    
    • techniques including storage and algorithmic blocking, lookahead and `panel scattering'
    • LU, LLT, QR and LDLT factorizations
  • Singular Value Decomposition Algorithms
    • floating point operations balance and reduction methods, optimizations on register, cache and external processor levels of the memory hierarchy.
• Programmable Systolic Arrays :
  • design and semantics, microprogramming, efficient control structures for the Instruction Systolic Array (ISA).

last Modified: August 2010