Linus Torvalds (torvalds.delete@this.linux-foundation.org) on April 6, 2014 9:55 am wrote:
> Michael S (already5chosen.delete@this.yahoo.com) on April 5, 2014 10:45 am wrote:
> >
> > That's incorrect.
> > Recent "big" Intel cores handle some cases of denormals with
> > no penalty, but not all cases and even not all common cases.
>
> Ok. I went by Agner Fog who tends to be really good, and reported that Sandy Bridge had no penalty
> for "many of these cases". He may have happened to check mostly the ones that work.
>
> Linus

Here is the post I was looking for:
http://www.realworldtech.com/forum/?threadid=137844&curpostid=138710

Finding it took nontrivial effort (nontrivial for me, first, because in general I'm not in that sort of scripting and second, because I jumped on opportunity to learn a little about Windows powershell, for somebody else it would be trivial).