Scientific Thingamabobs

Papers And Presentations

Diploma thesis "Domain Specific Language for Specifying Access Controls" (April 2007)
Thesis (pdf, english)
FMEA (Analysis) my group did for Total Quality Management/ISO9001 Certification (January 2006)
Presentation (pdf, german)
Software Deployment incl. a DarwinPorts Overview (December 2005)
Presentation (pdf, english)
Paper (pdf, english)
Lightning talk about open source software licensing (October 2005)
Presentation (pdf, english)
Paper (pdf, english)
Presentation about the security and system architecture of our project server (June 2005)
Presentation (pdf, german)
XML-Signatures, from the course Internet Security (April 2005)
Presentation (pdf, english)
Paper (pdf, english)
Parallelizing compilers, from the course Parallel Computing (December 2004)
Presentation (pdf, german)
Paper (pdf, german)
Short Introduction to Java Generics (May 2003)
Presentation (quicktime, german)
Final status presentation of the EPTA project my team did in February 2003
Presentation (pdf, german)
An introduction to Microsoft .NET, presentation from November 2002
Presentation (quicktime, german),
Handout (pdf, german)
A Paper about Wrapper Building Frameworks; presentation was in Winter 2001 (you may also find this Paper in the Arbeitsbericht Heterogene Informationssysteme from September 2002)
Paper (pdf, german)
Presentation about 3D-API Comparison, Direct3D vs. OpenGL from Winter 2000
Presentation (quicktime, german),
Handout (pdf, german)"

Parallel Computing (January 2005)
Distributed PI Calculator, using Monte Carlo method - uses PVM and GMP (C source, tar.bz2)
libpmsort (libpmsort-0.3.tar.bz2) is a C implementation of the parallel mergesort algorithm. It is interface compatible with qsort(3) from the standard C-library, so you may easily swap pmsort(3) for it, if needed. The function uses posix threads.
Though mergesort is at average slower than quicksort by a constant factor and requires twice the memory of heapsort, you may well gain an tremendous advantage in speed if you have multiple processors or if your compare-function is I/O-bound. As a plus, mergesort is a stable sorting algorithm, so you can sort for multiple attributes consecutively.
On a 4x PPC970 system, pmsort(3) yields an advantage of round about 100% over the systems qsort(3) when sorting pure integers. In this case pmsort consumes about twice the CPU time but scales very well with multiple processors. Remind however, that sorting pure integers is the worst case for pmsort. When sorting larger objects the memory bandwidth wasted by pmsort decreases as pmsort only requires twice the space for the pointers to the objects. You also can expect pmsort going further in the lead when sorting with I/O bound compare functions, e. g. ones that require RPC calls - in this ocasion even on single processor systems.