PATCH: optimized libm single precision routines: erfcf, erff, expf for x86_64.

[Dmitrieva Liubov]

This assembly version is the only version we have at the moment which
is ready for release. We are ready to modify it to look better if
required, however we don’t have plans to contribute C-version because
we believe that assembly should be faster.

--
Liubov Dmitrieva

2012/2/6 Ulrich Drepper <drepper@gmail.com>:
> This code seems not to rely at all on packed floats/doubles.  Did you
> try writing the code in straight C?  With a bit of massaging you
> should be able to get code layout just like that.