Re: [Boost-users] boost accumulators - kth central moments? [and slow compiles]

Subject: Re: [Boost-users] boost accumulators - kth central moments? [and slow compiles]
From: Eric Niebler (eric_at_[hidden])
Date: 2009-02-09 12:35:40

Gordon Anderson wrote:
> 
> I have a couple Qns about boost accumulators in 1_37_0 -
> 
> 1) Central moments ? -
> 
> There are accumulators for simple kth moments - 
> boost/accumulators/statistics/moment.hpp
> 
> Are central kth moments [relative to the mean] there somewhere?
> 
> Obviously you can construct one from the other with a weighted sum of 
> the k moments on the same data.

I don't think it's there yet. I'm cc'ing Matthias in case he has some 
kth moment accumulators lying around somewhere. If not, I'd be happy to 
accept patches.

> 2) Slow build ! -
> 
> It is a bit slow to compile code using eg. moment<6> accumulators..
> It takes around 12 seconds on my machine to build sample code with 
> 'moment<12>'
> 
> exponential recursive template evaluation??
> 
> I'm using gcc 4.2.4

There shouldn't be any dependence of compile time on the "N" in 
moment<N>. Are you seeing that?

I've experienced long compile times with accumulators, too. It certainly 
has never been optimized for compile times, and there's probably a lot 
of room for improvement here. I unfortunately don't have time to look into.

The emphasis with accumulators has always been to make runtime as 
efficient as possible, FWIW.

-- 
Eric Niebler
BoostPro Computing
http://www.boostpro.com