compiler comparison report available for preview

[David G. Hough at validgh]

I've decided to redo that report before publishing it further, for these
reasons:

1) There are newer releases of all the compilers studied.

2) By compiling with the "-Xsave" option, the Apogee compilers can handle
most of the programs that were troublesome unoptimized without it; the
difference is due to Apogee's default assumption that undeclared Fortran
variables are AUTOMATIC; this permits slightly more efficient code,
but it doesn't work on a lot of old Fortran programs that rely on undeclared
SAVE behavior.    Debugging those old programs is no fun, although I've done
it for perfect.track, perfect.spec77, herwig57, and jetset74, and I now
understand why most customers won't do it; they would just compile
with "-Xsave" and take the performance hit.  So I'll do the same.

3) Due to a sh script bug and my lack of awareness that -V -O means something
rather different on GCC than on SPARCcompilers or Apogee, my SPARC results with
GCC+F2C on Fortran programs were probably all compiled unoptimized and so
are optimistic with respect to compile
time and optimizer correctness, and pessimistic with respect to
run time.   Results for C programs were not affected,
nor were x86 results.

When I have the revised results I'll notify the persons who requested the
first version so they can take another look if they are still interested.