Linus Torvalds (torvalds.delete@this.linux-foundation.org) on August 22, 2012 10:16 am wrote:
> This is
> one of the reasons why I was arguing so strongly at one point (both here and to
> some Intel engineers) for the transaction model to use "success prediction"
> (possibly actually re-using the branch prediction hardware, but possibly with a
> separate predictor). Because part of "really cheap failure modes" is to not even
> bother trying if it's unlikely to work.


Wouldn't you want a combination of both hardware and software as the ideal? Same thing as the hints to branch prediction used in Linux - "if (unlikely(condition)) ..." There will be some transactions with little contention you feel would almost never fail, and vice versa.