Boost logo

Ublas :

From: choon (teochoonhui_at_[hidden])
Date: 2007-07-04 03:06:12


I tried some examples and the conclusion I got was that M*B is much slower
than M*b, where M is a (r-by-c) matrix, B a (r-by-1) matrix and b a vector
of length r.

Since ublas is not runtime polymorphism, and in certain applications, I
would like to have a runtime polymorphic matrix/vector which is determined
only when we "look" at the problem/data.

Is there a way to close the performance gap between 1D matrix and vector? Is
possible, then every object will be of type matrix and this make things much


View this message in context:
Sent from the Boost - uBLAS mailing list archive at