Re: A question about efficiency
Tim <[email protected]> Sun, 25 Jul 2004 00:44:49 +1000
| Newsgroups | gmane.comp.lib.mtl.devel |
|---|---|
| Message-ID | <[email protected]> |
That makes perfect sense. Many thanks Andrew, Tim Andrew Lumsdaine wrote: > Hi -- I think this type of operation only makes sense for triangular > matrices. I.e., if you think about matrix-vector multiplication: > > y_i = sum_j A_ij * x_j > > you need all of the values of x to compute each of the values of y. > If you change any of the values of x during the computation, you won't > get the right answer for y -- that is, if you try to update y in-place > ala an operation like y = A*y. The only exception is if A is > triangular -- then as the operation proceeds you no longer need values > x_j for j < i (or j > i depending on whether you have an upper or > lower triangular matrix). > > Best Regards, > Andrew Lumsdaine > > On Jul 23, 2004, at 2:06 AM, Tim wrote: > >> Hey guys, >> A simple question here. >> >> How do I apply a matrix transformation 'A' to a vector 'x', without >> storing the result into another vector? >> >> The only way I can think of doing it now is to do: >> "mult(A,x,y); >> copy(y,x)" >> >> I ask this because I want to be able to transform all the elements in >> a container of vectors, without getting bogged down by unnecassary >> calls to copy. >> >> Thanks in advance, >> Timothy Ling >> _______________________________________________ >> This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/ > > > _______________________________________________ > This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/ > _______________________________________________ This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/