The ADD_ATOMIC function adds the value of an expression  to  the
  value  of  a  variable,  stores  the  newly  computed value, and
  returns the previous value.

  Syntax:

    ADD_ATOMIC(e,v)

  The type of the expression 'e'  must  be  assignment  compatbile
  with  that  of  the  variable  'v'.  The variable 'v' must be an
  INTEGER, UNSIGNED, INTEGER64, or UNSIGNED64 variable and must be
  allocated  on  a  natural boundary (ie, longword for INTEGER and
  UNSIGNED and quadword for INTEGER64 and UNSIGNED64).  The result
  of ADD_ATOMIC is the same type as the variable 'v'.

  Overflow and  subrange  checking  are  never  performed  on  the
  ADD_ATOMIC  operation,  even  if these options are in effect for
  the rest of the function or compilation unit.