Matrix multiplication (MM) is a computationally-intensive operation in many algorithms used in scientific computations. Not only one of the kernels in numerical linear algebra, the problem of matrix multiplication is also fundamental for almost all matrix problems such as least square and eigenvalues problem. The performance analysis of the MM needs to be re-evaluated to find out the best-practice algorithm on novel architectures. This motivated the analysis which is presented in this article and which is carried out by means of the new modelling framework that the authors have already introduced (L. D’Amore et al. On a Mathematical Approach for Analyzing Parallel Algorithms, 2015). The model exploits the knowledge of the algorithm and the multilevel parallelism of the target architecture and it could help the researchers for designing optimized MM implementations.
Mathematical approach to the performance evaluation of matrix multiply algorithm / D'Amore, Luisa; Mele, Valeria; Laccetti, Giuliano; Murli, Almerico. - 9574:(2016), pp. 25-34. (Intervento presentato al convegno 11th Conference on Parallel Processing and Applied Mathematics, PPAM 2015 tenutosi a Krakow, Poland nel September 6-9, 2015) [10.1007/978-3-319-32152-3_3].
Mathematical approach to the performance evaluation of matrix multiply algorithm
D'AMORE, LUISA;MELE, VALERIA;LACCETTI, GIULIANO;MURLI, ALMERICO
2016
Abstract
Matrix multiplication (MM) is a computationally-intensive operation in many algorithms used in scientific computations. Not only one of the kernels in numerical linear algebra, the problem of matrix multiplication is also fundamental for almost all matrix problems such as least square and eigenvalues problem. The performance analysis of the MM needs to be re-evaluated to find out the best-practice algorithm on novel architectures. This motivated the analysis which is presented in this article and which is carried out by means of the new modelling framework that the authors have already introduced (L. D’Amore et al. On a Mathematical Approach for Analyzing Parallel Algorithms, 2015). The model exploits the knowledge of the algorithm and the multilevel parallelism of the target architecture and it could help the researchers for designing optimized MM implementations.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
