In order to clarify the interface between the Lattice C Compiler and
its associated libraries, I have put together a list of all "CX" routines
used by the compiler.  These routines are functions that are contained in
the Lattice libraries, but are not callable by the C programmer.  Calls to
these routines are automatically generated by the compiler when their
functionality is needed.

*** Do not attempt to call any of these functions in a C program. ***
These routines assume arguments in specific registers where the compiler
knows to place them, and return results in specific registers where the
compiler expects them.



The following internal CX routines are contained in the Lattice
standard C library in all forms.

        CXBRK    - Default Control-C break handler
        CXD22    - Divide an unsigned long by another
        CXD33    - Divide a signed long by another
        CXFERR   - Floating point error handler (historical)
        CXM22    - Multiply two unsigned longs
        CXM33    - Multiply two signed longs

The following internal CX routines are contained in the Lattice
math library in all forms.

        CXA55    - Add two doubles
        CXADJ4   - Adjust float after multiply/divide
        CXADJ5   - Adjust double after multiply/divide
        CXC55    - Compare two doubles
        CXD55    - Divide one double by another
        CXFERR   - Floating point error handler
        CXFNM5   - Normalize double in significand form
        CXFXT5   - Extra exponent/significand for double
        CXIND4   - Float Indefinite handler
        CXIND5   - Double Indefinite handler
        CXINF4   - Float Infinity handler
        CXINF5   - Double Infinity handler
        CXM55    - Multiply two doubles
        CXN5     - Negate a double
        CXNAN4   - Not-A-Number float handler
        CXNAN5   - Not-A-Number double handler
        CXNRM4   - Normalize a float
        CXNRM5   - Normalize a double
        CXOVF4   - Float overflow handler
        CXOVF5   - Double overflow handler
        CXREN4   - Renormalize denormalized float
        CXREN5   - Renormalize denormalized double
        CXRTN4   - Repack and return float
        CXS55    - Subtract one double from another
        CXT5     - Test double against zero
        CXTAB5   - Double normalization table
        CXUNF4   - Float underflow handler
        CXUNF5   - Double underflow handler
        CXV24    - Convert unsigned long to float
        CXV25    - Convert unsigned long to double
        CXV34    - Convert signed long to float
        CXV35    - Convert signed long to double
        CXV42    - Convert float to unsigned long
        CXV43    - Convert float to signed long
        CXV45    - Convert float to double
        CXV52    - Convert double to unsigned long
        CXV53    - Convet double to signed long
        CXV54    - Convert double to float
        CXVDF    - Convert decimal string to double
        CXVFD    - Convert double to decimal string
        CXVSD    - Convert decimal string to double
        CXZER4   - Float handler for zero
        CXZER5   - Double handler for zero


In all current Lattice compiler releases, these functions are in the
ANSI user name space.  The next Lattice release will have an underscore
prepended to each name to remove these functions from the user name space.