Stepanov results with development gcc

[Joe Buck]

Brad Lucier writes:

> And on Solaris 2.8, I get:
> 
> banach-7% /pkgs/gcc-2.95.2/bin/g++ -O2 -o stepanov stepanov_v1p2.C
> banach-8% ./stepanov 
> 
> test      absolute   additions      ratio with
> number    time       per second     test0
> 
>  0        0.63sec    79.37M         1.00
>  1        0.32sec    156.25M         0.51
...

I filed a GNATS report about this one a while back, see problem
report c/859 in GNATS.  The compiler does a horrible job on test 0;
this is a C as well as a C++ problem.