I have long had problems with zgelsd. Quite often it fails with the message: "forrtl: error(73): floating divide by zero."
I have now captured a matrix, 1024 x 1024, that causes this behaviour consistently.
I have written a test routine that uses, in turn,
1. zgetrf and zgetrs
2. zgetrf and zgetri plus some extra code
4. zgesvd plus some extra code
5. zgelss plus some extra code
All of these produce the same answer set, except 3. which gives the above error.
Does anyone have any knowledge or suggestions of what might be wrong?
I am running on an Alpha machine, with Redhat Linux 7.2 and the last Compaq Fortran/C suite. I also use the Compaq linker.
Malcolm M Bibby