On Mon, Apr 4, 2011 at 2:15 AM, Aaron S. Meurer <asmeu...@gmail.com> wrote: > Well, how sparse does a matrix have to be for a sparse implementation to be > faster than the dense representation, in general?
I haven't really tested this but most of the matrices I use in my day job only have a few percentage of non-zeros (for example on the order of 10 - 20 entries for rows at 100K). > > There's also the line somewhere in between numerical matrices and symbolic > matrices, which is matrices with exact numeric values (instances of > Rational), to consider. Not sure what you are getting at here. > > I remember people posting to this list with symbolic matrices in the past. > Maybe a search will show how people are using them. The only things I could find were with small matrices. -- Andy > > Aaron Meurer > > On Apr 3, 2011, at 9:10 AM, Andy Ray Terrel wrote: > >> Hello, >> >> We currently have several students interested in symbolic sparse >> matrix calculations for GSOC. IMHO, we could use a linear algebra >> interface more compatible with scipy.linalg but I don't have a use for >> symbolic sparse matrices. >> >> Since I believe in not adding unnecessary features to a library, I >> would like to know who the customers of a symbolic sparse matrix >> format would be and what applications, point to code please, we need >> to speed up. >> >> Also how big are the matrices you are working with. You have to >> remember that sparse matrices usually don't make sense on typical >> numerical algorithms until you hit about 1000x1000, because optimized >> BLAS can stream so much better. I would guess the number is much >> lower with symbolics, but the storage will not be an issue until that >> size since we only have a pointer to S(0). >> >> Having this information would be quite beneficial in organizing what >> the student projects should implement. >> >> -- Andy >> >> -- >> You received this message because you are subscribed to the Google Groups >> "sympy" group. >> To post to this group, send email to sympy@googlegroups.com. >> To unsubscribe from this group, send email to >> sympy+unsubscr...@googlegroups.com. >> For more options, visit this group at >> http://groups.google.com/group/sympy?hl=en. >> > > -- > You received this message because you are subscribed to the Google Groups > "sympy" group. > To post to this group, send email to sympy@googlegroups.com. > To unsubscribe from this group, send email to > sympy+unsubscr...@googlegroups.com. > For more options, visit this group at > http://groups.google.com/group/sympy?hl=en. > > -- You received this message because you are subscribed to the Google Groups "sympy" group. To post to this group, send email to sympy@googlegroups.com. To unsubscribe from this group, send email to sympy+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sympy?hl=en.
