Reduction-Level Parallel Computations of Gr.obner and Janet Bases

Abstract

In previous papers we presented algorithm for parallel calculation of Gr.obner and Janet bases that works in terms of parallel normal forms computations. The realization was quite promising but faced problem of "starvation" (e.g. in some moments of time only few processors was fully loaded). In this talk one approach to raise scalability and avoid "starvation" will be presented. Experimental results of parallel computations on eight core SMP machine will be shown. Work was partially supported by the RFBR grant 07-01-00660 and by the grant 1027.2008.2 of the Ministry of Education and Science of the Russian Federation.