From: Michael Stevens (mail_at_[hidden])
Date: 2005-08-12 12:05:18
On Tuesday 02 August 2005 02:45, Paul C. Leopardi wrote:
> Hi all,
> I have been testing uBLAS CVS using GluCat and g++ 4.0.1. If I compile
> without inlining (eg. -O1), I see the following warning.
> Line 42 of operation_sparse.hpp has:
> vector<value_type> temporary (e2 ().size2 ());
> Using -O3 the warning disappears, leading me to guess that with -O3 the
> optimizer is getting more information from the inlined call to
> temporary.clear() on line 43 of operation_sparse.hpp.
> Can and should the uBLAS code in operation_sparse.hpp be changed to prevent
> this warning?
Any ideas how such a change could be done. I can't see a way to convince the
compiler that the temporary is initialised.
Since you are using sparse_prod I have a question. Any idea what it is
supposed to do. The code is littered with issues and possible bug. For
example the dense 'temporary' seems very odd for a "sparse" function.