Re: [ublas] Sorry, lost track. Is uBLAS CVS the same as Boost CVS?

From: Paul C. Leopardi (leopardi_at_[hidden])
Date: 2005-07-17 08:23:15

Hi Michael, Thanks for you prompt response. Reply below.
On Sun, 17 Jul 2005 08:19 pm, Michael Stevens wrote:
> On Sunday 17 July 2005 10:56, Paul C. Leopardi wrote:
> > Hi all,
> > I feel it is time for me to test GluCat with uBLAS CVS.
>
> Excellent idea!! I suspect there are a few source incompatibilities since
> the merge of the uBLAS_pure branch. I just did a new release of my Bayes++
> project in anticipation of Boost 1.33.0
[...]
> Hopefully the lock will go away soon. It is worth noting that if you did a
> check of Boost CVS once before then you don't need to checkout again. A cvs
> update will do the job.

I have done the CVS update and have re-compiled GluCat 0.1.9 with uBLAS CVS.
In my other email message entitled "Changes in uBLAS between Boost 1.31 and 
Boost 1.32?" I noted that the Boost 1.32.0 version of uBLAS caused errors in 
the GluCat tests. I have re-run the GluCat tests with the CVS version of 
uBLAS and the test output is correct.

In summary, the GluCat tests give:
1.31 OK
1.32 errors
CVS  OK

I did have to make one change to matrix_multi_imp.h, where operator==() now 
has
      return ublas::norm_inf(m_matrix - rhs.m_matrix) == 0;
instead of
      return ublas::equals(m_matrix, rhs.m_matrix);

I don't yet know how this affects performance.
Best regards, Paul Leopardi