Resources for Parallel Computing and Supercomputing
This list is maintained at
www.eecs.umich.edu/~qstout/parlinks.html
where the entries are linked to the resource.
Rather than creating a comprehensive, overwhelming, list of resources,
I have tried to be selective, pointing to the best ones that I am
aware of in each category.
You can send me an email to suggest modifications at
qstout@umich·edu
Parallel Computing
101, a tutorial for beginning and
intermediate users, managers, people contemplating purchasing
or building a parallel computer, etc.
From CI-Tutor
On-line courses about many aspects of high-performance computing
and cyberinfrastucture.
Tools:
METIS, free, a
``set of programs for partitioning graphs and for producing
fill reducing orderings for sparse matrices''.
Zoltan, free,
tools for partitioning, load balancing and data-management.
Intel's
collection of software products for tracing, analyzing and visualizing
performance of their products.
MPI,
Free, portable implementations are available at
MPICH2
and Open-MPI.
OpenMP the most
important standard for
shared-memory programming.
Contains pointers to tutorials, texts, performance analyzers, etc.
Globus, the de facto standard
for grid computing. Site contains pointers to references, tools,
free implementations, etc.
Co-Array Fortran,
a Global Address Space (GAS) language simulating shared memory on distributed
memory machines. GAS is a form of virtual shared memory.
UPC (Unified Parallel C),
a GAS language simulating shared memory on distributed
memory machines.
Available systems: Most of the large national resources available to the
general scientific community within the US are organized within the
XSEDE program.
Some additional universities also have arrangements to allow outside users on
their parallel systems.
System Rankings
Top 500
listing of most powerful computers in the world (among
those which are publically acknowledged).
Green 500, similar to Top 500 but using flops/watt as the metric, not flops.
Graph 500, ranking based on graph problems, which are more suitable benchmarks for problems such as data analytics.
