High Performance Computing (Master Course)

Contents: 

We will start with an introduction into basic principles and algorithms of hardware aware and parallel computing followed by transferring selected algorithms onto many-core architectures. 

 

Lecturer:

Prof. Gundolf Haase, Heinrichstr. 36, Zi 506, Tel. 5178,

Appointments:

Monday 8:15 - 9:45 in Heinrichstr. 36, SR 11.33

Exercises (Folllow the links in the pdf-document):  

• Exercise I (sequential programming, BLAS).
  Benchmark results: Haase (smaller parameters), Smertnig, Kraft, Grossar, Aichmayr/Dauschan, Pichler, Neic, Boiger, Innerwinkler, Anyyeva, Herbst
  
• Exercise II  (shared memory programming, OpenMP). Template.
  Benchmark results: Pichler, Anyyeva, Haase, Kraft, Smertnig, Grossar (max, poly), Neic, Aichmayr/Dauschan, Innerwinkler, Boiger, Herbst
  
• Exercise III  [included MPI course] (distributed memory programming, MPI).
  Results: Kraft
  

Hardware (login from outside KFU only via  VPN):

• corei7 (4 cores+HT: Core i7-920, 2.67 GHz; 6 GB memory):        20
  
• quadquad (16 cores: 4 Opteron 8347, 1.9GHz; 32 GB memory): 201
  
• dualhex (12 cores: 2 Opteron 2427; 2.2 GHz; 28 GB memory):     16

Material:

• Using BLAS.
  
• OpenMP, Summary, Tutorial, nice paper, guide into OpenMP
  
• OpenMPI
• Intel optimized BLAS/Lapack: MKL(local) with PBLAS(local)
  
• AMD optimizied BLAS: ACML (good docu)
  
• valgrind on MPI, see §4.8 of manual.
  

Further Links

• MultiCoreInfo, GPGPU
  
• LAPACK on GPU (Info, Tools )
• Comparison GPU/CPU
• Wiener Supercomputer [Nov. 27, 2009]
• Chinease GPU-supercomputers [May 31, 2010]
• Aubrey Isle / Knights Ferry co-processor card by Intel [June 1, 2010; c't 13/2010, p.20],
  
• Core i7-3960X (40 GB/s bandwidth, 141 GFLOPS in Linpack [89%])
  
• Plan: Supercomputers in Europe [Feb 2012]}
  

  11.05.2012