Re: [boost] Accelerating algorithms with SIMD - Segmented iterators and alternatives

Subject: Re: [boost] Accelerating algorithms with SIMD - Segmented iterators and alternatives
From: David Abrahams (dave_at_[hidden])
Date: 2010-10-13 11:04:54

At Wed, 13 Oct 2010 03:44:37 -0700,
Jeffrey Lee Hellrung, Jr. wrote:
> 
> >> Although what you're describing is easier to make efficient for many
> >> (perhaps even the majority of cases), it doesn't sound like it
> >> actually has the same level of generality as segmented iterators do.
> >
> > Segmented iterators still allow bidirectional support, indeed (I think?).
> > But I'm note sure it can do random access as well, if you consider
> > recursively segmented local iterators.
> 
> Seems to me segmented iterators and a visitation-based/push-based
> iteration address fundamentally different problems.  At least, I don't
> see how segmented iterators help here.  The problem that
> visitation-based iteration seems to try to solve is improving
> iteration efficiency over join(t)_range-like ranges...

That's the same problem (in essence) that segmented iterators are
solving.  Did you read the Austern paper?

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com