The custom reduction in the TSQR used by the power basis is currently segfaulting in parallel with the Intel compilers/MPI.