Although it has long been considered that the price of OOP was a slower execution speed, modern scientific & computing libraries have made that wrong, and c++ is now widely used for large-scale computations. A short explanation is that c++ is both a high-level language (object-oriented), and low-level since it uses pointers and thus programming can be very close to assembly. This latter point is a boon for numerically-intensive programs.
For good examples of numerically intensive calculations using c++, you take a look at:
-> The home page for Scientific Computing in Object-Oriented Languages: http://www.oonumerics.org/
-> POOMA (http://www.acl.lanl.gov/pooma/)Parallel Object-Oriented Methods and Applications
-> Blitz++ home page (http://www.oonumerics.org/blitz/): a 'smart' array computing library which yields performance on par with that of fortran.