Changes from NumLib 2.0.2 to 2.1.0
----------------------------------
Only minor changes have been done on the STABLE_2_0 tree as only minor
issues have been found.
Main development is going on in the head branch, but the GNU Compiler yields
crazy compile resource requirements: Some rather easy vector code using the
new TBCI framework made the compiler fail after compiling for more than an
hour (CPU time on Athlon-800) and allocating more than 600MB of memory, if
-O2 was used. With -g, it compiles just fine within 10 seconds ...
The author is now busy investigating this, so the development is stalled a
bit. It may have to do with the flow analysis being obscured by mutables or
the bad string compile performance of gcc.
(DEC CXX BTW does not show these sort of problems.)

Back to the 2.1.0 changes:
* Some more files have been documented with doxygen
* The stopwatch file has better support for Win32 timing functions
* Lots of std:: prefixes have been added; no using std::cout,cerr,...
  should be needed to compile the TBCI stuff any longer
* The automated generation of the guiding decls and inst lists via
  auto_decl have been changed in the Makefile in order to allow parallel
  makes (make -j2); however, this does only work, if the autogenerated files
  are not generated via Makefiles from different directories.
  Otherwise, you may want to call make -jX in one of the subdirs first ...
* BVector <-> Matrix conversions.
* Some minor cleanups for gcc-2.97/3.0-pre snapshots.
* (B)Vectors have now a size of type long, allowing larger objects on 64bit
  archs. Something similar holds for Matrix, F?<atrix and BdMatrix: the no
  of columns and rows still needs to fit into the range of an int; the total
  no. of elements does only need to fit into a long. This change may have
  backward compatibility issues; therefore the name 2.1.0 was chosen in
  favour of 2.0.3.

			Kurt Garloff <K.Garloff@Phys.TUE.NL>, 2001-03-15
			
PS: It may very well be that this is the last 2.x release.