[Patch] Nathan's improvements to basic_string

[Martin Sebor]

Nathan Myers wrote:
...
> Numerics people I talk to, by the way, say that each C compiler
> release since 2.7 has generated slower code than the last, and 
> they're starting to get very testy about it.

That's been our general experience as well (although it's not
necessarily specific to floating point computations). See my
post comparing the performance of 2.95.2 vs 3.2 on SPARC here
http://gcc.gnu.org/ml/gcc-bugs/2002-12/msg01432.html. IIRC,
we saw a similar regression on i86.

Regards
Martin