caa a22 4500 46320302X CHVBK 20180405153117.0 cr unu---uuuuu 170326e20070501xx s 000 0 eng 10.1007/s00200-007-0038-9 doi (NATIONALLICENCE)springer-10.1007/s00200-007-0038-9 When cache blocking of sparse matrix vector multiply works and why [Elektronische Daten] [Rajesh Nishtala, Richard Vuduc, James Demmel, Katherine Yelick] We present new performance models and more compact data structures for cache blocking when applied to sparse matrix-vector multiply (SpM × V). We extend our prior models by relaxing the assumption that the vectors fit in cache and find that the new models are accurate enough to predict optimum block sizes. In addition, we determine criteria that predict when cache blocking improves performance. We conclude with architectural suggestions that would make memory systems execute SpM × V faster. Springer-Verlag, 2007 Performance optimization nationallicence Sparse matrix multiplication nationallicence Memory hierarchies nationallicence Performance modeling nationallicence Nishtala Rajesh Computer Science Division, University of California at Berkeley, 575 Soda Hall, 94720, Berkeley, CA, USA aut Vuduc Richard Computer Science Division, University of California at Berkeley, 575 Soda Hall, 94720, Berkeley, CA, USA aut Demmel James Computer Science Division, University of California at Berkeley, 575 Soda Hall, 94720, Berkeley, CA, USA aut Yelick Katherine Computer Science Division, University of California at Berkeley, 575 Soda Hall, 94720, Berkeley, CA, USA aut Applicable Algebra in Engineering, Communication and Computing Springer-Verlag 18/3(2007-05-01), 297-311 0938-1279 18:3<297 2007 18 200 https://doi.org/10.1007/s00200-007-0038-9 text/html Onlinezugriff via DOI 1 research-article jats NATIONALLICENCE 856 40 https://doi.org/10.1007/s00200-007-0038-9 text/html Onlinezugriff via DOI NATIONALLICENCE 700 1- Nishtala Rajesh Computer Science Division, University of California at Berkeley, 575 Soda Hall, 94720, Berkeley, CA, USA aut NATIONALLICENCE 700 1- Vuduc Richard Computer Science Division, University of California at Berkeley, 575 Soda Hall, 94720, Berkeley, CA, USA aut NATIONALLICENCE 700 1- Demmel James Computer Science Division, University of California at Berkeley, 575 Soda Hall, 94720, Berkeley, CA, USA aut NATIONALLICENCE 700 1- Yelick Katherine Computer Science Division, University of California at Berkeley, 575 Soda Hall, 94720, Berkeley, CA, USA aut NATIONALLICENCE 773 0- Applicable Algebra in Engineering, Communication and Computing Springer-Verlag 18/3(2007-05-01), 297-311 0938-1279 18:3<297 2007 18 200 Metadata rights reserved Springer special CC-BY-NC licence nationallicence BK010053 XK010053 XK010000 NATIONALLICENCE NATIONALLICENCE NL-springer