protocol of dgeev_jlapack2_ in lapack2
'... dgeev_  n *c *c *i *d *i *d *d *d *i *d *i *d *i *i '&cd

the wiki should be updated if lapack2 is wanted


On Mon, Jun 6, 2022 at 7:17 AM Raul Miller <[email protected]> wrote:

> Today I stumbled over a regression in our lapack support.
>
> https://oldcode.jsoftware.com/wiki/Essays/Eigenvalues
>
> Changing the first lines to:
>
> require 'math/lapack2'
> ev=:[: clean_jlapack2_ 1 >@{ dgeev_jlapack2_
>
> I get a cd error from ev
>
>    ev dm >:i.20
> |domain error: cd
>
>    cder''
> 4 0
>
> (wrong number of parameters)
>
> (Oddly, also, cderx seems to produce a result for me which is unrelated
> to the result of cder. The first time I try it, I get 109 The pipe has
> been ended. but subsequent attempts get 0 The operation completed
> successfully. Meanwhile, cder will continue to give me 4 0 after one
> of these errors while cderx will at this point tell me that the
> operation completed successfully.)
>
> This behavior is consistent between J903 and J904.
>
> FYI,
>
> --
> Raul
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to