Improvements to powerpc32 asm code

Kevin Ryde user42@zip.com.au
Tue, 03 Jun 2003 11:50:30 +1000


Mark Rodenkirch <mrodenkirch@wi.rr.com> writes:
>
> 32             4.0290       #3.2692
> 48             4.0322       #3.2662
> 64             4.0131       #3.2816

Yep.  This is the number to work on for the main loop.

As a rule we attack the per-limb speed first, then worry about fixed
overheads, and any anomalies like for instance size==31 slower than
size==32 due to finishups or whatever.