Many of the linear algebra decompositions need to operate on submatrices of the original matrix so it wouldn't make sense to allocate a new matrix.
To do what you want, you can allocate a new matrix and call gsl_matrix_memcpy On Dec 5, 2014 11:26 PM, Matthias Sitte <[email protected]> wrote: > > Forgot to CC the list... > > An implementation-related questions: Why does gsl_matrix_submatrix() > return a view of a matrix on the stack (which just wraps the underlying > gsl_matrix struct) and not a pointer to a newly created gsl_matrix > struct (which I would have to free manually, but I have to free most of > GSL's structs by hand anyway)??? > > If there is no quick answer, that's fine, I was just wondering... > > Matthias > > > -------- Forwarded Message -------- > Subject: Re: [Help-gsl] strange 'data' pointers using more than one > submatrix simultaneously > Date: Sat, 6 Dec 2014 00:28:28 +0100 > From: Frank Reininghaus <[email protected]> > To: Matthias Sitte <[email protected]> > > Hi, > > 2014-12-05 17:38 GMT+01:00 Matthias Sitte <[email protected]>: > > Oh my ... I totally missed that. Corrected that and it works! :D > > > > An implementation-related questions: Why does gsl_matrix_submatrix() return > > a view of a matrix on the stack (which just wraps the underlying gsl_matrix > > struct) and not a pointer to a newly created gsl_matrix struct (which I > > would have to free manually, but I have to free most of GSL's structs by > > hand anyway)??? Just wondering ... > > I have no time to look into that right now. Please always include the > mailing list in your replies - that makes it much more likely that > anyone will answer your question soon. > > Regards, > Frank > >
