/sys$common/syshlp/helplib.hlb FORTRAN, /ASSUME, ACCURACY_SENSITIVE  *Conan The Librarian (sorry for the slow response - running on an old VAX)

  Controls whether code transformations that affect the results of
  floating-point operations are allowed.  These changes can affect
  the correctness of the program's results.

  If you specify NOACCURACY_SENSITIVE, the compiler is free to
  reorder floating-point operations, based on algebraic identities
  (inverses, associativity, and distribution).  This allows the
  compiler to generate optimized forms of sum reductions (such as dot
  products) and to move divide operations outside loops, improving
  performance.

  The default, ACCURACY_SENSITIVE, directs the compiler to use only
  naive scalar rules for calculations.  This setting can prevent some
  optimizations.