From: Roy S. <roy...@ic...> - 2008-12-17 15:53:49
|
On Wed, 17 Dec 2008, Tim Kroeger wrote: > Why does NumericVector::localize_to_one() write its results into an > std::vector<T> rather than into another NumericVector<T>, like most of > the localize() methods do? We only really use that method for data output, for which it's pointless to create a PETSc or some such linear algebra vector. > Is there some easy workaround for that? Don't we have NumericVector constructors that take a std::vector? Not that that's efficient, but it's easy. Otherwise you may have to write an overloaded localize_to_one(). What do you need a NumericVector on one processor for? --- Roy |