Program killed as soon as it starts to run

[Paul Smith]

On Thu, May 8, 2008 at 2:35 PM, egbirgin <no-reply-gw at fcp.surfsite.org> wrote:
> Dear Paul,
>
> Well, ok, if your problem is very large then algencan will use a lot of memory. However, if your problem is small, you can very easily solve it using algencan.
>
> Just go to the file sources/algencan/dim.par and, with care, remove a couple of zeros from mmax, nmax, jcnnzmax, hnnzmax and nwma27.
>
> The current version of dim.par looks like
>
> C     PARAMETERS
>
>      integer mmax,nmax,nsmax,nwma27,hnnzmax,jcnnzmax
>
>      parameter ( mmax      =   500000 )
>      parameter ( nmax      =   500000 )
>      parameter ( jcnnzmax  = 10000000 )
>      parameter ( hnnzmax   = 10000000 )
>      parameter ( nwma27    = 10000000 )
>      parameter ( nsmax     =     1000 )
>
> Your modified version should looks like
>
> C     PARAMETERS
>
>      integer mmax,nmax,nsmax,nwma27,hnnzmax,jcnnzmax
>
>      parameter ( mmax      =   5000 )
>      parameter ( nmax      =   5000 )
>      parameter ( jcnnzmax  = 100000 )
>      parameter ( hnnzmax   = 100000 )
>      parameter ( nwma27    = 100000 )
>      parameter ( nsmax     =   1000 )
>
> After doing this modification, compile and run again. Everything will work fine. (Fortran 77 has no dynamic memory allocation.)

Thanks, Ernesto, for your e-mail. As you saw, one of the discussants
did check the amount of memory allocated by Algencan, and the result
was: a large amount of memory.

At home, Algencan runs well, because here I have a newer and more
powerful (and with much more RAM) computer than at work.

At work, Algencan (compiled with the simple toyprob.f) does not work
at all, because it is killed while loading.

I will not have access to my computer at work before next week. So, I
cannot test your suggestion now. I will let you know about the results once
I have tested it.

I think Algencan would become more popular if there was a mailing list
where users could discuss all issues regarding Algencan. Many people
(like me) are always very reluctant to write directly to the authors
of the packages, as they always think they may be bothering the
authors. A mailing list would overcome this problem. Other solvers do
have a mailing list (e.g., Ipopt).

Moreover, I think the number of the current available version of
Algencan should be made explicit at the web-site. This would reduce
the number of download trials in vain. (Unfortunately, one only sees
the number of the current version when one is about to download it,
after having filled the form.)

Best regards, Paul