[PATCH 3/5] math: New generic fma implementation

[Wilco Dijkstra]

Hi Adhemerval,

This looks like a major improvement. There is likely further scope for 
performance (the call to scalbn takes ~30% of the time on AArch64,
and the 64-bit int to double is a library call on most 32-bit targets),
but the gains so far are pretty good. In the future we could streamline
it further.

Reviewed-by: Wilco Dijkstra  <Wilco.Dijkstra@arm.com>

Cheers,
Wilco