Re: [boost] [SORT] Parallel Algorithms

Subject: Re: [boost] [SORT] Parallel Algorithms
From: Paul A. Bristow (pbristow_at_[hidden])
Date: 2015-01-14 06:59:17

> -----Original Message-----
> From: Boost [mailto:boost-bounces_at_[hidden]] On Behalf Of Steven Ross
> Sent: 14 January 2015 02:53
> To: boost_at_[hidden]
> Subject: Re: [boost] [SORT] Parallel Algorithms
> 
> Francisco,
> 
> On Tue Jan 13 2015 at 12:43:00 PM Francisco José Tapia <fjtapia_at_[hidden]>
> wrote:
> 
> > Hi Steven,
> >
> >  This Christmas I had been working hard. I developed a new version of
> > the merge sort about 10-15% faster than the previous, using an
> > additional memory only ½ of the memory used by the data. The GCC and
> > TBB algorithms use an additional memory as the size of the data.

<snip>

> > Yes, indirect sorting is clearly superior for large data types, and it
> would be good to make it as easy as possible to use.  It'd be nice to also have a
> simple wrapper that can be use the spreadsort algorithm with indirect sorting too.

Would Boost.Multiprecision (50 decimal digits floating-point say) provide a simple demo type?

> I am in the process of integrating the first version of the boost sort library in
> https://github.com/boostorg/sort.  It would be good to integrate your testing with
> the tests in that library.

Perhaps a git branch off the new /sort/develop branch would be useful?

I note that Steven Ross is making a separate namespace for spreadsort

boost::sort::spreadsort

Although this seems long (slow typists can provide an alias or a using ... ;-) 

I strongly suggest that you should follow this example, perhaps

boost::sort::tbb

though I don't like tbb at all - far too acronymic?  Think of a better name, even if longer?

boost::sort::parallel?

Paul

PS The new Sort/Spreadsort docs are almost ready for release.

I will add Quickbook documentation (from plain text notes) for your parallel sort version if you write to me off-list.

---
Paul A. Bristow
Prizet Farmhouse
Kendal UK LA8 8AB
[phone]