[Coin-ipopt] looking to beta testers of new IPOPT release

[Andreas Waechter]

Hello, and Happy New Year! :)

We had difficulties with our Watson mail servers, and I had trouble
posting message to the mailing list.  Tonight let's try it again (knock on
wood...)

---------- Forwarded message ----------
Date: Fri, 19 Dec 2003 17:43:22 -0500 (EST)
From: Andreas Waechter <andreasw at bigopt.watson.ibm.com>
To: coin-ipopt mailing list <coin-ipopt at www-124.ibm.com>
Subject: looking to beta testers of new IPOPT release

Hello,

I have been working on a new release of IPOPT for a while now, and it is
finally taking shape.  Most of the improvements are related to the full
space version of the optimization method (which is the default option,
i.e. if you use IPOPT within AMPL, this is what you would use).  The new
implementation is more efficient and in particular more robust than the
current release, at least in the tests that I have performed.

It would be very helpful to me if a few people would be willing to try the
beta version of the new release, before I officially release it.

Here a few main features of the new version:

- I tried to simplify the installation procedure significantly.  In
  principle, all that one has to do now is to download necessary third
  party source code (as before), and then type 'configure' and 'make
  install'.  It is no longer necessary to use GNU make or to try to
  understand the makefiles - the configuration script should find out
  everything for you.

- It is no longer necessary to obtain the third party component TRON - a
  new restoration phase which is more efficient and robust has been
  implemented.  (You would still need TRON if for some reason you need to
  use the reduced-space version - if you don't know what that is it means
  that you don't need TRON :)

- IPOPT uses dynamic memory allocation now, i.e. it is no longer necessary
  to guess how much memory is need before IPOPT is called.

- The call for IPOPT has changed a little bit (the IPOPT function call has
  a few more parameters now, and it is no longer necessary to call
  INIT_PARAMS before calling IPOPT).  At this point I have not implemented
  an optional backwards-compatible interface - if someone has strong
  feelings about this, I will be happy to add this feature.

- Thanks for Kirk Abbott, also a C interface is included, which should
  simplify hooking up IPOPT do a program that has been written in C or
  C++.

I will wait with the `official' release of the new version until a few
people have tried it.  If you are interested in trying the beta version
and/or giving feedback, please send an email to `andreasw at watson.ibm.com'.

Thanks, and Happy Holidays,

Andreas