JC wrote: > How large is "large scale matrices"
Depends on the size of your memory, but a vector with a dimension of several million and SPARSE matrices with millions of rows AND columns (but few non- zero entries) should be possible. Of course, dense matrices with one million rows and columns need too much memory. > and do you have any performance > figures that you can share? Not yet, but if you use dense matrices and vectors only, many operations can be done by cblas routines and should be as fast. Frank
