Boost logo

Boost :

Subject: Re: [boost] [library submission] smoothsort
From: David Abrahams (dave_at_[hidden])
Date: 2008-12-14 14:24:45


on Sun Dec 14 2008, "Edouard A." <edouard-AT-fausse.info> wrote:

> Sorry for the double post, but good news everyone:
>
> I just did compilation with profiling. Smoothsort is twice faster than
> std::sort on sorted input, up to 5 times slower on random input.

That's progress.

> My benchmark output, on a vector of size 1,000,000 :
>
> std::sort running time: 0.171 s
> std::sort running time (sorted): 0.082 s
> Smoothsort running time: 0.6 s
> Smoothsort running time (sorted): 0.035 s
>
> Now I must figure out why without profiling I don’t have good performances.
> Maybe there's some built-in optimization in MS' compiler when it encounters
> its own std::sort?

Did you use -D_SECURE_SCL=0
?

Without that, anything that you do with most iterators is a dog.

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

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk