Minutes of the meeting of the Parallelisation subgroup held on 6th April 2006

Present Ian Watson, Ian Rogers, Andrew Dinn, Jisheng Zhao, Mohammad Ansari, Chris Kirkham

JZ presented some more results comparing "adaptive searching" and "recursive distribution". These were discussed. The conclusion seemed to be that for Linpack recursive distribution was better when the numbers of processors was a power of 2, but that adaptive searching was other better. AD suggested doing an experiment with several competing threads - this might better illustrate the advantage of adaptive search. Perhaps it would be useful to mix up sequential and parallel phases of code in this, and the jBYTEmark benchmark might provide some relevant code.

CCK wrote these minutes.

Apr0606

Minutes of the meeting of the Parallelisation subgroup held on 6th April 2006