On increasing the effective blocking factor of a matrix for a given cache organization
Demirhan, Atilla N.
MetadataShow full item record
Blocking (Tiling) techniques of iteration spaces to increase data reuse in the cache were reviewed. Results consistent with those previously published were experimentally obtained. The relation between the sizes of the declared matrix and the cache was studied. Based on this relations, two algorithms were presented. Both algorithms attempt to increase the critical blocking factor with no self-interference (Bc) by changing the declared matrix size. Furthermore, the execution time of the second algorithm is independent of the matrix size. Experiments based on these algorithms were performed which showed a consistent superior performance (in terms of Mflops) relative to the performance obtained using previously published algorithms for deriving Bc.
Approved for public release; distribution is unlimited
Showing items related by title, author, creator and subject.
Ozer, Erhan (Monterey, California: Naval Postgraduate School, 2013-06);Classically, the T-matrix method is a procedure to exactly compute the multiple scattering of an incident wave from a “cloud” of objects, given knowledge of the free-field scattering properties of a single object for an ...
Current, Donovan C. (Pennsylvania State University, 1974-06);Many creameries are operated by managers who, through experience, have generated rules of thumb that are very accurate for predicting product flows and input requirements to meet estimated demand. As the number of ...
Roginski, Jonathan W.; Gera, Ralucca M.; Rye, Eric C. (American Mathematical Society, 2015-10-19);The newly introduced neighborhood matrix extends the power of adjacency and distance matrices to describe the topology of graphs. The adjacency matrix enumerates which pairs of vertices share an edge and it may be summarized ...