OK, I have now removed completely all of the Intel 11.1 compilers (C++ and Fortran) and reinstalled and I get the same error, a segmentation error, using any optimized compile (-01 to -03). The test problem runs successfully in Debug mode (-Z7). Maybe not the simplest thing to check, but it is the
PETSc solver package (ex5f.f90) from Argonne, installed with either C++ and Fortran 11.1.038 or 11.1.046. A normal configuration run followed by make all, then make test should reproduce this error. The PETSc support group could not reproduce this with Intel 10.X (they don't have 11.X yet). I didn't actually use the Visual Fortran environment for this, I just did a command line compile and run. But my own code will not converge properly using the Visual Fortran environment, in contrast to 10.X...
Windows XP Service Pack 3
M4400 Core 2 Extreme Chip Set (Q9300)
4 GB RAM
IA32
Using either 10.X MKL from Intel, or building BLAS and Lapack from source code obtained from ORNL (so the problem is not with MKL apparently)
