A pipelined schedule to minimize completion time for loop tiling with computation and communication overlapping.

Nectarios Koziris Aristidis Sotiropoulos Georgios I. Goumas A pipelined schedule to minimize completion time for loop tiling with computation and communication overlapping. 1138-1151 2003 63 J. Parallel Distrib. Comput. 11 http://dx.doi.org/10.1016/S0743-7315(03)00102-3 db/journals/jpdc/jpdc63.html#KozirisSG03