Re: [ublas] Possible defect in mapped_vector

Subject: Re: [ublas] Possible defect in mapped_vector
From: Benson Margulies (bimargulies_at_[hidden])
Date: 2008-10-20 20:17:41

I did later open a trac with a test case that fails, I believe.

Advice-wise, here's my issue: I don't need linear algebra at all. I
need a big, sparse, 'map', and std::map has awful performance.

It turns out that a co-worker implemented something that looks a lot
like mapped_array, only without the issue I reported in trac.

On Mon, Oct 20, 2008 at 5:48 PM, Gunter Winkler <guwi17_at_[hidden]> wrote:
> Am Samstag, 18. Oktober 2008 04:49 schrieb Benson Margulies:
>> I am trying to use mapped_vector with map_array, as per doc.
>>
>> The first thing I do is resize, and it immediately sets of a
>> consistency check in the map_array which throws.
>
> Could you please attach a small failing example to the trac task?
>
>>
>> Both of the iterators are null, as is 'data' in the mapped_vector.
>>
>
> Please consider using compressed_vector or coordinate_vector. (At least
> one of them should perform better than mapped_vector.)
>
> @others: Is anyone using mapped_vector for large applications?
>
> mfg
> Gunter
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
>
>