The NON-VON Supercomputer Project: Current Ideology and Three-Year Plan

Shaw, David Elliot

While we have learned a great deal during the first two years of the NON-VON Supercomputer Project, I am reluctant to commit myself at this point to anything that might be called a "position" regarding the direction and ultimate outcome of current research in the field of parallel architectures. In part, my hesitation reflects an appreciation for the difficulty of objectively assessing the state of the field as a whole while enmeshed in the "cult of personality" surrounding a particular machine. Fortunately, our local dogma has not yet become so rigid as to preclude the possibility of significant revisions of our beliefs in response to the experiences and ideas of our colleagues. At the same time, it is clear that our understanding of the essential issues of parallel machine design in general is colored by the particular challenges we have faced in the context of the NON-VON Project. In the following discussion, I will thus try to avoid any claims regarding the ideological correctness or historical inevitability of any of the architectural principles to which I now subscribe. In their place, I will attempt to list a few of our current architectural objectives, and to outline our tentative hardware implementation plans for the next three years. Software considerations will not be discussed in this document, despite the fact that they have occupied a large fraction of our time.



More About This Work

Academic Units
Computer Science
Department of Computer Science, Columbia University
Columbia University Computer Science Technical Reports, CUCS-061-83
Published Here
October 25, 2011