Subject: Re: [ublas] numeric bindings
From: Rutger ter Borg (rutger_at_[hidden])
Date: 2010-11-01 03:47:51
On 2010-10-29 12:07, Andreas Hehn wrote:
> the parameters n,m,k reported to the actual blas function call (e.g.
> DGEMM) are all 0. If however I change the typedef to
> typedef std::ptrdiff_t size_type;
> everything works just fine. But this doesn't really represent the size
> type of my matrix type.
> The same also happens for the stride() functions.
> What's the right way to go?
IIRC, BLAS/LAPACK expect signed integers, so your approach with
ptrdiff_t is correct. These size/stride functions are returning by
value, so the conversion happens right there, without any computational