Re: de-tuned i386 bn code

> A while ago, Bill Simpson mentioned that there had been a factor of
> three slowdown in the bignum code in libcrypto between 1.6 and now.
> 
> I've traced this down to the fact that it appears that the hand tuned
> machine language bignum implementation for the i386 port was removed
> during the 0.9.7b import.
> 
> A factor of three is a pretty significant issue for many people. Are
> there any ideas on how we could go about bringing this code back?

	done recovering bits from attic.  pls try.

itojun