Changes

[http://www.umiacs.umd.edu/~vishkin/ Uzi Vishkin]

[http://www.umiacs.umd.edu/~vishkin/ Uzi Vishkin]

<br>

<br>

Thinking algorithmically in parallel provides an interesting ``alien intellectual culture'', relative to previous algorithmic theories. Current UMD researchers have had a strong presence in this field for over two decades, as evidenced by the recent recognition of 2 parallel algorithms researchers: Clyde Kruskal and Uzi Vishkin, as [http://www.isihighlycited.com Highly Cited Researchers] -- Kruskal and Vishkin are 2 of a total of 15 active UMD faculty members to be recognized by ISI Thompson, publishers of the science citation index. The PRAM-On-Chip project at UMD is a direct outgrowth of our theoretical work. The project offers a concrete agenda for challenging the 1946 von-Neumann architecture, through streamlining the massive knowledge base developed by the parallel algorithms community with the roadmap for CMOS VLSI.

Thinking algorithmically in parallel provides an interesting ``alien intellectual culture", relative to previous algorithmic theories. Current UMD researchers have had a strong presence in this field for over three decades, as evidenced by the recognition of 2 parallel algorithms researchers: Clyde Kruskal and Uzi Vishkin, as [http://www.isihighlycited.com Highly Cited Researchers]. The PRAM-On-Chip project at UMD is a direct outgrowth of our theoretical work. The project offers a concrete agenda for challenging the 1946 von-Neumann architecture, through streamlining the massive knowledge base developed by the parallel algorithms community with the roadmap for CMOS VLSI. Using our hardware and software prototypes, we recently demonstrated parallel speedups that beat all current platforms by orders of magnitude on the most advanced parallel algorithms in the literature, including for max-flow, graph biconnectivity and graph triconnectivity. This stress test validates our 30-year old hypothesis that the PRAM algorithmic theory can save parallel computing from its programming woes, while traversing an architecture narrative that has belittled for decades the PRAM theory, stipulating that it is too simplistic for any existing or future parallel machine. The demonstrated advantage is especially important for the so-called, irregular, fine-grained programs that all commercial platforms to date fail to support effectively.  

<br><br>

<br><br>