Techniques and Technologies
By Henry Neeman
The OU Supercomputing Center for Education and Research (
OSCER)
This is the final module in a trilogy. It is preceeded by
"HPC on a Single Thread", and concludes with
"Multithreading and Multiprocessing". Together these three modules comprise much of the core material from the
2010 Blue Waters Undergraduate Petascale Institute. It is intended that these materials may be readily adapted and adopted by undergraduate
faculty to serve as the core content for an undergraduate course on scientific parallel
computing. These materials were, in turn, adapted from the
"Supercomputing in Plain English" materials originally developed at OSCER for OU campus and regional education, outreach and
training.
Links to the module resources follow the content description below.
* High Throughput Computing
This submodule focuses on Condor and similar technologies. Topics include: what is HTC?
(definition, throughput vs. performance, throughput on a cluster); tightly vs. loosely
coupled; Monte Carlo methods; opportunistic computing (supercomputing at night, BOINC); Condor
(Condor vs. BOINC, features, limitations, running a job, batch script, Linux Condor on Windows
via coLinux); grid computing.
* GPGPU: Number Crunching in Your Graphics Card
This submodule focuses on General Purpose Graphics Programming Units (GPGPU), emphasizing the
practical aspects of GPGPU programming and its implications for scientific computing. Topics
include: accelerators; GPUs; GPGPU programming (CUDA, StreamSDK/Brook+, OpenCL, Portland
Group's accelerator compiler directives); NVIDIA Tesla in depth (current Tesla card and server
specifications, x86 vs Tesla, disadvantages of GPU computing, PCI-e slot performance
bottleneck, community code speedup examples); CUDA programming (thread hierarchy: kernels,
threads, warps, blocks and grids, intrinsic variables and keywords, memory hierarchy: global,
shared, local, constant, texture); matrix-matrix multiply.
* Grab Bag: Scientific Libraries, I/O, Visualization
This submodule concludes the series with some smaller issues of interest. Topics include:
scientific computing pipeline; scientific libraries; I/O libraries (native binary vs. text,
portable binary); visualization (contour lines, slice planes, isosurfaces, streamlines, volume
rendering).
Resources:
Presentation: High-Throughput Computing : High-Throughput Computing presentation in PPT format.
Presentation: GPGPU: Number Crunching in Your Graphics Card : GPGPU presentation in PPT format.
Presentation: Grab Bag: Scientific Libraries, I/O, Visualization : Grab bag presentation in PPT format.
Exercise: Building GotoBLAS : Building GotoBLAS exercise in DOC format.
Exercise: Building High Performance Linpack : Building High Performance Linpack (HPL) with GotoBLAS exercise in DOC format.
Exercise: Running HPL : Running HPL exercise in DOC format.
Exercise: Building NetCDF : Building NetCDF exercise in DOC format.
Exercise: Using NetCDF : Using NetCDF exercise in DOC format.