Boost logo

Boost Users :

Subject: Re: [Boost-users] [accumulators] Kahan algorithmic variants of sum_impl, mean_impl?
From: Eric Niebler (eric_at_[hidden])
Date: 2011-09-30 13:49:12

On 9/30/2011 7:28 AM, Rhys Ulerich wrote:
>>From a look at boost::accumulators::impl::sum_impl [1] it appears a
> straightforward, fast 'sum += value' is used to accumulate the running
> sum. That can fall prey to numerical errors if the sum is large and
> the value is small. A slower but more precise summation is possible
> using the Kahan summation algorithm [2].
> Has anyone implemented a Kahan-based variant for either sum_impl or
> mean_impl [3]?


You can get a kahan mean by putting both sum_kahan and mean in the
accumulator set, since sum_kahan satisfies the sum dependency.

sum_kahan is new in 1.47, thanks for Gaetano Mendola and Simon West.

Eric Niebler
BoostPro Computing

Boost-users list run by williamkempf at, kalb at, bjorn.karlsson at, gregod at, wekempf at