The "Double Precision Complex" z-routines implement Complex*32 = 256bit,
when compiled with Fortran90 options /Real=64 /Double=128 /Integer=64,

This works fine with the HP.Fortran Compiler, if the source code
contains only "Real", "Double", "Complex", and "Double Complex",
rather than Real*8, Real*16, Complex*16, etc.

For some routines, the original source code was edited to replace
intrinsic function references
COMPLEX*16 by "Double Complex"
DBLE   by "Real"
DIMAG  by "aImag"
DCONJG by "Conjg"
CMPLX  by "qCmplx"

These edited versions are in [.src] and [.blas.src]

The routines required for zgBSvX have beeen tested:
  [lapack.SRC]ZgBSvX +zGBcon +zDrScl +zgbEqu +zgbRFS +zgbTRF +zgbTF2 +zgbTRS +-
              zlaCGV +zlaCON +zlaNGB +zlaNTB +zlaQGB +zlaSSQ +zlaTBS +-
              zlaDIV +dlaDIV +zlaCPY  +zlaSWP +-
  [lapack.BLAS.src]izaMax +zCopy +zaXPY +zDotC +zDotU +zdScal +zgBMV +-
                   zgEMM +zgEMV +zgERU +zsCAL +zSWAP +ztBSV +ztRSM 

CFD  17-jun-06