⟦0a1f8d5a2⟧

TextFile

Subsystem Math_Support  
======================

Notes for release D3.0
----------------------
This bug fix release contains the following views:
          R1000_Rev1_1_1           
          R1000_Code1_1_1
those 2 views are spec compatibles with the previous release
          R1000_Rev1_0_1           
          R1000_Code1_0_1

Modifications:
 
 o package Generic_Primitive_Functions
 
    function Scale has been corrected for very large and very small scales
    
    function Remainder has been improved in the cases where arguments
    are several order of magnitudes apart

 o package Generic_Elementary_Functions
 
    The implementation has been replaced by a higher accuracy package
    developped by Peter Tang and Kenneth Dritz, from the Argonne
    National Laboratory, with minor adjustments for the R1000.
    This change fixes several errors in domain checking (Argument_Error
    was not raised, or some other exception was raised), and a big
    error in the code for tangent.
    
The code has been tested with two test suites developed by
  John Squire, from Westinghouse Electric Corporation
and
  Peter Tang, from Argonne National Laboratory.


Notes for release D2.2
----------------------

This release of the Environment features an additional subsystem called
Math_Support, located in world !Tools. This subsystem contains an
implementation of the two proposed secondary standards developed by the
ISO-IEC/JTC1/SC22/WG9 (Ada) Numerics Rapporteur Group. The two proposed
standards are:

 o Proposed Standard for a Generic Package of Primitive Functions for Ada,
    draft 1.0, 13 December 1990.

    This standard provides primitive operations to endow mathematical software,
    such as the implementation of elementary mathematical functions: 
      - functions and procedure for manipulating the fraction part and the
         exponent part of machine numbers for the generic floating-point
         type
      - functions for direct rounding to a nearby integer, for computing
         an exact remainder, and for determining the immediate neighbors of
         a floating-point number
    in a package named Generic_Primitive_Functions.

 o Proposed Standard for a Generic Package for Elementary Functions for Ada,
    draft 1.2, 12 December 1990.
    
    This standard provides certain elementary mathematical functions:
      - functions for square roots, logarithms, and exponentials
      - trigonometric functions and their inverses, for angles in radians
         or other units of measure
      - hyperbolic functions and their inverses
    in a package named Generic_Elementary_Functions.

The complete texts of both standards are provided in the subdirectory
Units.Documents, together with their "Rationale," by Kenneth W. Dritz.

This delivery of the subsystem has two parts:

 o  a nonportable implementation for the R1000 architecture, composed
     of a spec view, a load view, and a corresponding code view:
          R1000_Rev1_0_Spec      
          R1000_Rev1_0_1           
          R1000_Code1_0_1
    
    The two generic packages are provided as are instantiations for
    the standard type Float supported by the R1000.

    This R1000-specific implementation was created based upon the books:
      - "Software Manual for the Elementary Functions,"
         William J. Cody, Jr. and William Waite,
         Prentice-Hall Series in Computational Mathematics, 1980.
      - "Scientific Ada,"
         Brian Ford, Jan Kok, and Mike W. Rogers, 
         Cambridge University Press, 1986.
      - "Computer Approximations," 
         John F. Hart, et. al., 
         Krieger Publishing, 1968.
 
  o a nonportable implementation for the Motorola Mc68020 target
    with an Mc68881 floating-point coprocessor, to be used with Rational
    M68K Cross-Development Facilities.
    The directories:
           M68k_Bare_Archive      
           M68k_Hpux_Archive      
    contain archives of combined views to be used with the Mc68020_Bare
    CDF and Mc68020_Hp_Unix CDF, respectively. They can be restored with
    the command:
      Archive.Restore (Device => "!Tools.Math_Support.M68k_Bare_Archive");
      Archive.Restore (Device => "!Tools.Math_Support.M68k_Hpux_Archive");

    Note that the same source should also work with the Motorola Mc68882
    floating-point coprocessor and with the Mc68040 processor, which
    directly supports floating-point operations, but this compatibility 
    has not yet been tested.

    The two generic packages are provided as are ready-made instances for
    the two standard types Float and Long_Float supported by Rational
    Mc68020 CDFs.

For a user to modify the software, the subsystem Math_Support need to be
made Primary, with Cmvc_Maintenance.Make_Primary, then a working view must 
be created from a load view with Cmvc.Make_Path. Be sure to give another
path name to allow subsequent deliveries from Rational.
Example:
    Cmvc_Maintenance.Make_Primary (Subsystem => "!tools.Math_Support");
    Cmvc.Make_Path (From_Path => "!Tools.Math_Support.R1000_Rev1_0_1",
                    New_Path_Name => "Local_Rev1",
                    Create_Load_View => True);

Note:
Users recognize that the math-support packages provided as part of the
Rational Environment are delivered "AS IS." No warranties are given,
whether express, implied, or statutory, including implied warranties of
fitness for a particular purpose or merchantability. In no event will
Rational be liable in tort, negligence, or other liability incurred as a
result of the use of these math-support packages.

In addition, Rational does not guarantee conformance to the proposed
standards, especially in the accuracy of the results of the various math
functions.

This code can be distributed and modified freely, although Rational
would be happy to receive comments, bug reports, and suggestions for
improvements at:
                   Rational
                   3320 Scott Boulevard
                   Santa Clara, CA 95054 USA
                or by e-mail at:
                   Support@Rational.COM
DataMuseum.dk

Rational R1000/400 Tapes

⟦0a1f8d5a2⟧ TextFile

Derivation

TextFile
