Thomas:

Thx Hong. I tried the ex148, but I doesn't work as I have complex petsc scalars. Instead I tried ex154, where I did the following modification to be able to compile it

ex148.c only works with real precision: #if defined(PETSC_USE_COMPLEX) SETERRQ(PETSC_COMM_WORLD,PETSC_ERR_SUP, "This example requires real numbers"); #endif As you see, these examples were written last summer, and have not been checked/tested with FFTW since then, therefore messy and buggy. I'll check what you reported below and get back to you soon. Thanks for reporting the problems, Hong

hg diff diff -r b800705fa966 src/mat/examples/tests/ex154.c --- a/src/mat/examples/tests/ex154.c Tue Jul 10 21:23:47 2012 -0500 +++ b/src/mat/examples/tests/ex154.c Wed Jul 11 11:06:03 2012 +0200 @@ -46,7 +46,7 @@ ierr = VecGetSize(input,&vsize);CHKERRQ(ierr); printf("The vector size of output from the main routine is %d\n",vsize);

- ierr = InputTransformFFT(A,input,x);CHKERRQ(ierr); + ierr = VecScatterPetscToFFTW(A,input,x);CHKERRQ(ierr); // ierr = VecAssemblyBegin(x);CHKERRQ(ierr); // ierr = VecAssemblyEnd(x);CHKERRQ(ierr); // ierr = VecView(x,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); @@ -54,7 +54,9 @@ ierr = MatMult(A,x,y);CHKERRQ(ierr); ierr = MatMultTranspose(A,y,z);CHKERRQ(ierr);

- ierr = OutputTransformFFT(A,z,output);CHKERRQ(ierr); + ierr = VecScatterFFTWToPetsc(A,z,output);CHKERRQ(ierr); + + /* Check accuracy */ fac = 1.0/(PetscReal)N; ierr = VecScale(output,fac);CHKERRQ(ierr);

Running this program with a single node (mpirun -n 1 ex154) works but with multiple (mpirun -n 2 ex154) I get a segfault http://pastie.tk/view.php?id=29C95711

ex143 with -use_FFTW_interface YES and-use_FFTW_interface NO works for both single and multiple nodes. ex145 only for more than one node (as expected).

Regards Thomas