Hi Florian,
This is very strange. Just curious if Olaf or Drosos from the PARDISO project
are able to reproduce the issue or give you any insight into what is causing
this?
Ray
> On Oct 25, 2015, at 9:40 AM, Florian Schaefer
> <[email protected]> wrote:
>
> Dear Mr. Zimmerman,
>
> I tried to install the PARDISO Solver. Even though I thought it would be
> correctly installed, matpower refuses to use Pardiso. I figured out that
> this part of have_fcn() is the reason:
>
> A = sparse([1 2; 3 4]);
> b = [1;1];
> % Summary PARDISO 5.1.0: ( reorder to reorder )
> info = pardisoinit(11, 0);
> info = pardisoreorder(A, info, false);
> info = pardisofactor(A, info, false);
> [x, info] = pardisosolve(A, b, info, false);
> pardisofree(info);
> if any(x ~= [-1; 1])
> %here is the error -> x is [-0.1111 ; 0.3333] instead of -1 and 1.
>
> The terminal from which matlab is called displays "Intel MKL ERROR:
> Parameter 2 was incorrect on entry to DGER"
>
> I'm using gcc/gfortran 4.7.4 on an Ubuntu 15.04 machine and these
> commands to compile the PARDISO library:
>
> mex -cxx CXX=g++ CC=g++ LD=g++ -L/usr/lib/pardiso -largeArrayDims
> -lpardiso -lmwlapack -lmwblas -lgfortran -lm -output pardisofactor
> common.cpp matlabmatrix.cpp sparsematrix.cpp pardisoinfo.cpp
> pardisofactor.cpp
> mex -cxx CXX=g++ CC=g++ LD=g++ -L/usr/lib/pardiso -largeArrayDims
> -lpardiso -lmwlapack -lmwblas -lgfortran -lm -output pardisofree
> common.cpp matlabmatrix.cpp sparsematrix.cpp pardisoinfo.cpp pardisofree.cpp
> mex -cxx CXX=g++ CC=g++ LD=g++ -L/usr/lib/pardiso -largeArrayDims
> -lpardiso -lmwlapack -lmwblas -lgfortran -lm -output pardisoinit
> common.cpp matlabmatrix.cpp sparsematrix.cpp pardisoinfo.cpp pardisoinit.cpp
> mex -cxx CXX=g++ CC=g++ LD=g++ -L/usr/lib/pardiso -largeArrayDims
> -lpardiso -lmwlapack -lmwblas -lgfortran -lm -output pardisoreorder
> common.cpp matlabmatrix.cpp sparsematrix.cpp pardisoinfo.cpp
> pardisoreorder.cpp
> mex -cxx CXX=g++ CC=g++ LD=g++ -L/usr/lib/pardiso -largeArrayDims
> -lpardiso -lmwlapack -lmwblas -lgfortran -lm -output pardisosolve
> common.cpp matlabmatrix.cpp sparsematrix.cpp pardisoinfo.cpp
> pardisosolve.cpp
>
> There are no compilation errors. Unfortunately Pardiso seems to deliver
> wrong solutions.
> Since the PARDISO Integration in matlab is fairly new, I couldn't find
> any solutions in the mailing list. Has anybody faced similar problems
> with this? Is there any solution?
>
> Best regards
> Florian Schaefer
>
>
>
