Re: [Help-gsl] Re: General CBLAS question

[Hongzheng Wang]

On Tue, Sep 14, 2010 at 5:36 PM, Rodney Sparapani <address@hidden> wrote:
> Oh, and getting back to the original question.  Am I reading the
> BLAST Forum standard correctly in stating that the correctness of
> y = 1 A y + 0 y is not guaranteed?  On page 26, in subsection 2.3.2,
> it states:  Correctness is only guaranteed if output arguments are not
> aliased with any other arguments.  The fact that y is both the
> first and second vector passed to dgemv would imply an alias
> between the arguments, correct?
Yes, using y in such a way introduces an alias and may results in
undefined error.
>
> By the way, there must be an Errata somewhere since on that page in
> the above subsection 2.2.6, there are some obvious typos such as
> x(i)).
Agree.