LAPACK/ScaLAPACK Development

Here is the receipt to generate a "somewhat random" symmetric positive
definite matrix:

0- fix your n
1- allocate an n-by-n array
2- set the upper part of A to random values (negative or positive)
arrange such that the elements have magnitude of 1 about
3- set the lower part of A such that A is symmetric
(Note this step is not necessary with LAPACK since LAPACK accesses
one half of the matrix and assumes symmetry.)
4- set the diagonal to positive (>0) value,
arrange such that the diagonal elements are bigger than the off-
diagonal ones. For example sqrt(n) times bigger

Julie's code will run fine with these modifications incorporated.

julien.

Thanks Julien for the explanation.

Can you just please clarify these 2 steps as I couldn't understand properly:

- arrange such that the elements have magnitude of 1 about

- arrange such that the diagonal elements are bigger than the off-
diagonal ones. For example sqrt(n) times bigger


Maho

the-shifter wrote:Thanks Julie...I passed my linking too.
One more question though...in your code you suggested an order 2 matrix, I would like to go bigger and work on variable matrices lengths.
So do you happen to have a sample algorithm that generates real symmetric positive definite matrices? or if not, can you point out from the test files source code the part which generates this matrix?

Thank you
Maho

could someone please post some help.thanks






Monitoare LCD

Thanks to Ankur. I also compiled successfully for the little code given by Julie.

If using C++ rather that C, then we should not include "CLAPACK.h" in the main program, or else it will cause some error like

"previous declaration of ‘int spotrf_(char*, integer*, real*, integer*, integer*)’ with ‘C++’ linkage
test.cpp:5: error: conflicts with new declaration with ‘C’ linkage"

thanks a lot Julie.. that code for able to compile the clapack basic program helped.