(eng) Minimizing communications when mapping affine loop nests onto distributed memory parallel computers has already drawn a lot of attention. This paper focuses on the next step : as it is generally impossible to obtain ...