README for SCC

This is extracted from the beginning of the scc.texinfo file
which is part of the scc.tar.gz distribution.
It is available on

ftp.ma.utexas.edu:pub/maxima/scc.tar.gz


Description
===========

This Manual documents how to use and install SCC which is a precompiler
offering extensions for Algebraic Programming in C.  SCC is intended to
be used in conjunction with a standard C compiler (either ANSI C or old
style), making possible efficient convenient computations with algebraic
objects other than just floating point and small integers.  It also
supports input and output for such objects in a natural way using printf and
scanf.

Because it is difficult for humans to learn new languages, and because
we feel C is an excellent programming language, we have tried to
minimize the differences with standard C.  The only data types built
into the compiler are those of standard C.  What SCC is capable of doing
however, is being told how to apply operators such as cast or @code{+}
to user defined types.  SCC also allows overloading function names,
making it possible for different functions to be invoked depending on
argument types.

One test of this was done with Rafael de la Llave.  In about 100 lines of
user SC code, we were able to convey sufficient information to SCC about
the Brent multiprecision floating point library (in FORTRAN), so as
compile 380 files from the numerical recipes library distribution (C
version).  The resulting code ran correctly, with the extra precision,
as far as we were able to determine.  We made no changes to the original
sources, except for correcting one error in the source and automatically
adding an include file to each numerical recipes file, before its
compilation by SCC.  The numerical recipes code exercises many features
of C.  It certainly requires printf and scanf to work correctly.  SCC
includes an extended stdio library, to allow users to easily define the
stdio of C to work correctly with user defined types.  This includes
handling of constants in source code, so that if we had constants in the
source which contained 50 significant digits they would be read
correctly.  The user defined scan function may be used to read such
constants.

Another application was to link with the PARI library code, which contains
many mathematical objects, including extended precision integers as well
as floats.   

Other applications involve adding complex numbers and their operations,
adding interval arithmetic, and vector operations.    There is no limit
on the number of different types handled.   All knowledge about them is
conveyed by include files, with none built into the compiler itself.

Eventually it is hoped that a large number of library routines will be
available.  We feel the portability, ease of linkage with existing code,
speed, and the fact that it is available with sources and without
charge, will make it attractive to a number of users.