Hi,
I'm trying to do some modifications to a program which was written for the Sun
solaris platform. The coding also uses some calls for lapack.
The code works fine on Sun machines. I want to port it to intel/Linux. I have
installed clapack. But am not sure how the exact calls should be done.
The orginal code calls a lapack routine as:
dgels('N',data.n_ind,data.n_col,ny,data.xwork,data.n_ind,data.ywork,data.n_ind,flag);
but the clapack calls a routine in a diiferent manner and passes parameters by
reference.
So I do the clapack calls as
lapack_type='N';
dgels_(&lapack_type, &data.n_ind, &data.n_col, &ny, &data.xwork, &data.n_ind,
&data.ywork, &data.n_ind, work, work_size, flag);
The code compiles ok, but gives a segmentation error or the code just gets
stuck depending on the 'work_size'. The man pages for dgels specifies that if u
call dgels with work_size =-1; that it will return the proper work_size to be
used. Even when this is used to make the actual call the system gives a
segmentation error.
Also since fortran and C store arrays in different order, should i change the
data.xwork and data.ywork matrices now when using clapack from the earlier
sun's version of lapack?
Any help in this regard is appreciated,
Mahen/
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.cs.utk.edu/private/lapack/attachments/20061108/05b9da21/attachment.html
|