Topics
RES = reserved by someone.
- Something relevant (to this course) that you have found
yourself; check with me first.
- How you have optimized a research code (or some other large
code).
- Intel's
hyper-threading technology.
- Tera
scale computing from Intel
- What
is new in Fortran 2003? Say something about the Co-Array
concept in Fortran 2008.
- Dual core
- More about 64-bit processing
- ATLAS (fast
dgemm).
- Another
project with the same goal.
- Recursion
in numerical linear algebra routines.
- The IBM p690
supercomputer at NIC Jülich.
- One can buy boards having 2, 4 or 8 (perhaps more) CPUs on them.
The board also has shared memory, so the board can be used as a small
shared memory computer. Find details about one such computer and give
us some details.
- It is possible to mix OpenMP- and MPI-programming. Suppose we
have a parallel computer made up of boards (as in the previous topic).
One can use OpenMP on the boards and communicate using MPI between the
boards. Here is an article
(search the web for more articles).
- If we don't have a shared memory computer but would like to use
OpenMP, shared memory can be simulated using e.g. MPI (OpenMP is
a layer on top of MPI). Find some material about such implementations
and present it. Here is a good place to start: Distributed Shared
Memory Home Pages.
- There is an article
comparing PVM and MPI. Present it.
- Fortran M
- Portable dataformats. NetCDF. Sci-data-formats
FAQ.
- SCALAPACK?
- More deatils about 3DNow and SSE etc? Look at Intel's or AMD's homepages.
- Java has builtin support for threads. Say something about it and
make some tests. How does it compare with OpenMP?
- What is specfp etc?
- How well does the Matlab the compiler,
mcc
,
work? Check first that we still have it.
- Matlab 6.5 (and later versions) has a JIT-accelerator.
Give us more details.
- Talk about the fastest
supercomputer in the world. See also this article
(2007).
- Code development, GNU's KDevelop, for example.
- Google, Altavista (and other search machines) are using powerful
computers. Indexing on the web.
- More about some unix tool such as perl, sed, awk or RCS
(a version control system). Show some nice examples.
- An alternative to RCS is CVS (Concurrent
Versions System). The system can, among other things, keep track of
different versions of a large package and such a tool is very useful
when a group of programmers are working on the code. Look also at Cervisia.
- Download PVM
and get a small program running. For someone who is used to getting
things running.
- Argonne's
parallel PDE-package, PETSc.
- The internals of Pentium 4, Xeon, Itanium 2 , Opteron or some
other modern processor. Look at Intel's- or AMD's homepage, for example.
- Some of the game-computers (Playstation,
XBOX, Game Cube) have very good floating
point performance. Present some details about the internals of these
machines. More
about the Playstation. Some people are running Linux on their XBOX.
A newer reference is BrookGPU. Here is a link
to NVIDIA's CUDA-package. Wikipedia contains an overview.
- How does one build a Beowulf
cluster (Linux-cluster) ? There are several such clusters in Sweden
(and the rest of the world).
- More about grid-computing (other link). GRID Today (news), Nordugrid.
- Something along the some lines as GRID-computing, but very
special problems. Talk about the factorization of large integers. This is a
good place to start but look at the links at the bottom of that
page as well.
- Another similar project (when it comes to the parallel
computing, but not otherwise) is SETI@home, the Search
for Extraterrestrial Intelligence.
- A few more links along the same lines as SETI.
Protein
Folding
Climate
Prediction
