#
#This directory contains the package generic_elementary_functions.
#We have also provided a portable version of the package
#generic_primitive_functions. Because the primitives are slow
#(it does not used "represenation" facilities), the resulting
#elementary functions are not efficient. Eventually, efficient
#generic_primitive_functions should be available in each 
#system. In particular, if your system already has a tailored implementation
#of generic_primitive_functions, you should definitely use it.
#
#[Added by Dritz: That is, you should substitute your implementation of
#generic_primitive_functions for the one in GPF_b.a.
#WARNING! The body contained here is incomplete.  In particular,
#the function Remainder will not work on a hexadecimal machine; it will
#raise Program_Error.  On binary machines, it is OK.]
#
#[Another comment by Dritz: We recently discovered that the implementation of
#the elementary function EXP has a minor error in it.  The error will not not
#cause inaccurate results; rather, it will cause inefficiency in a rare case.
#Specifically, on a DEC VAX for which D-format has been enabled, and when
#GENERIC_ELEMENTARY_FUNCTIONS is instantiated with a type whose representation
#is D-format, the approximation will be carried out internally in H-format (33
#digits), using a polynomial appropriate for 15 digits, rather than in D-format
#(reported as 9 digits due to Ada's 4*B Rule, but actually 16).  This will be
corrected in future versions.]
#
#{Yet more: At one point, Dritz started to replace the raisings of
#Constraint_Error in the argument prescreening steps (used to signal inevitable
#overflow in the expansion functions, and at poles of certain other functions)
#by computations that must overflow or divide by zero.  The purpose of this was
#to make the code portable to IEEE environments where a (future) IEEE binding
#was being employed, where infinities could (would) be generated instead.
#Tang convinced Dritz not to do this until such time as an IEEE binding is
#actually available and could be used, arguing that if nonstandard modes are
#in effect, blindly overflowing or dividing by zero might not generate an
#infinity.  We cannot really know what to do until we see how the future IEEE
#binding modifies or interprets the statements about overflow and about the
#behavior of functions at poles in the GEF standard.
#
#Below is the compilation order expressed in Verdix commands.
#[See further comments by Dritz at the end.]
#
ada -v EXCP_s.a
#
ada -v GPF_s.a
ada -v GPF_b.a
#
ada -v GEF_s.a
ada -v GEF_b.a
#
# the order below is unimportant
#
#ada -v g_Aco.a
#ada -v g_Acosh.a
#ada -v g_Acoth.a
#ada -v g_Asinh.a
#ada -v g_Asn.a
#ada -v g_Atanh.a
#ada -v g_Atn.a
#ada -v g_Cos.a
#ada -v g_Cosh.a
#ada -v g_Cot.a
#ada -v g_Coth.a
#ada -v g_Exp.a
#ada -v g_Log.a
#ada -v g_Sin.a
#ada -v g_Sinh.a
#ada -v g_Sqrt.a
#ada -v g_Tan.a
#ada -v g_Tanh.a
#
ada -v g_*.a
#
#ada -v kf_AcoCy.a
#ada -v kf_AsnCy.a
#ada -v kf_Atanh.a
#ada -v kf_AtnCy.a
#ada -v kf_AtnSm.a
#ada -v kf_Cos.a
#ada -v kf_CosCy.a
#ada -v kf_CotCy.a
#ada -v kf_Em1.a
#ada -v kf_Em1Sm.a
#ada -v kf_L1p.a
#ada -v kf_L1pSm.a
#ada -v kf_LogB.a
#ada -v kf_Power.a
#ada -v kf_R1pSm.a
#ada -v kf_Sin.a
#ada -v kf_SinCy.a
#ada -v kf_Sqrt.a
#ada -v kf_TanCy.a
#
ada -v kf_*.a
#
#ada -v kp_Atn.a
#ada -v kp_Exp.a
#ada -v kp_Log.a
#ada -v kp_LogPw.a
#ada -v kp_Pi2Rd.a
#
ada -v kp_*.a
#
#
#[The library defined by this directory should be compiled first, because
# it does not depend on any other library.  The easiest way to do that on
# a Verdix-based system is:
# a.mklib
# a.make -f *.a
#]