UPPSALA UNIVERSITET: Information Technology
 
 

Erik Hagersten Dean of Mathematics and Computer Science Professor in Computer Architecture

Research in computer architecture is always evolving in pace with the advancement in the state-of-the-art. New research topics are added while old are fading.  Our research focus at present are concentrated to three area we believe will be of great importance in the near future:

Intrachip multiprocessing: Recent development in chip manufacturing are making it appealing to put more than one CPU on each silicon die.  It is not well understood how to best design such multiprocessors. This will the rules for how multiprocessors are designed and open up a new research area.

Beowulf technology: It is much cheaper to buy 16 PCs or workstations than to buy a single shared-memory multiprocessor with 16 CPUs. How about if you could make the 16 PCs run a shared-memory application?

Performance tools for shared-memory programs: With the introduction of intrachip multiprocessing, multithreaded programming will become more popular and widely used in many application areas. New unexperienced programmers will be be introduced to the wonders of this paradigm. This will put new demands on the tools that will help the programmer understand why the speedup is far from ideal. Shared-memory applications will also have a higher demand for performance tuning in order to run well on our Beowolf technology.