Core Library Version 1.3 released.

[Chee Yap]

===================== SOFTWARE RELEASE ANNOUNCEMENT ==========================

Core Library Version 1.3 is now available for free download at
	
		http://cs.nyu.edu/exact/core/

The Core Library (CORE) is a collection of C/C++ classes to support 
exact computation with constructible real numbers (+,-,x,/,sqrt).

Numerical nonrobustness is a widely acknowledged problem.  It has proven
particularly intractible in the context of geometric algorithms where
numerical data and combinatorial data are intermixed in a strongly constrained
manner.  Recent research in the computational geometry community has
demonstrated a variety of techniques that can address such problems.
A basic goal of our library is to make such techniques easily accessible to 
the wider community of programmers and researchers.

Basic CORE Features:

   -- ease of use:
	Any C/C++ programmer can now write numerical or geometric code
	that are fully robust.

   -- ease of migration:
	Many existing C/C++ programs can be converted into robust programs
	with minimal effort.

   -- natural and novel numerical accuracy model:
	Users can choose and get the numerical accuracy that best fit
	their applications.  

   -- state of art technology:
	Precision-driven approach to exact geometric computation,
	best known root bounds, etc, will be incorporated into the library
	as the field progresses.  In this way, the user's application program
	will automatically be upgraded (at the cost of re-compilation).

   -- small system:
	Just over 2MB (including source and documentation, compressed).
	It can serve as the "core" for your own applications.

   -- tested on Sun Sparc and Linux platforms:
	Earlier CORE 1.2 was also tested on SGI and Windows.

What is new with CORE 1.3 ?

   -- improved speed:
	The runtime for all the samples programs in
	this distribution takes 55.0 seconds under CORE 1.3.
	Under CORE 1.2.2, the time is 1040.38 seconds.  The test
	was performed on a Sun UltraSPARC 10, 440 MHz machine,

   -- improved root bounds:
	This yield great speedups for the important class of expressions
	expressions with division (e.g., expressions with rational number
	or floating point number inputs).

   -- adoption of LiDIA's number kernel:
	This allows the easy plug-in of alternative big number packages
	such as Gnu's gmp.  CORE Distribution comes with CLN, one of the
	fastest big number packages (it implements the Schonhage-Strassen
	multiplication).  Because of the move to LiDIA, we now support only
	the g++ compiler. 

   -- improved numerical output:
	Only significant digits are printed, and a choice of 
	scientific or positional formats.

   -- improved numerical input:
	With input from strings, the user can control precision of
	conversion, including forcing exact input. Again, choice of
	scientific or positional formats.

   -- more sample programs:
	Examples show how one can easily use CORE to compute transcendental
	constants (Pi) or functions (e.g., sin, log) to any an desired
	precision.  An improved randomized prover for ruler-and-compass 
	theorems, etc.

For further information
	http://cs.nyu.edu/exact/
	mailto://exact@cs.nyu.edu.
We welcome your comments and input.

-- Chee Yap (yap at cs.nyu.edu)
   Chen Li (chenli at cs.nyu.edu)
   September 22, 2000

                                | Department of Computer Science
                                | Courant Institute of Mathematical Sciences
                                | New York University
                                | 251 Mercer Street
                                | New York, NY 10012, USA

===================== SOFTWARE RELEASE ANNOUNCEMENT ==========================


-------------
The compgeom mailing lists: see
http://netlib.bell-labs.com/netlib/compgeom/readme.html
or send mail to compgeom-request at research.bell-labs.com with the line:
send readme
Now archived at http://uiuc.edu/~sariel/CG/compgeom/maillist.html.