[Coin-discuss] AMPL or GAMS?

[Stefan Vigerske]

Hi,

Mario Rappi wrote:
> I am new to COIN-OR and to the list (hopefully this is the right list to use
> for my question).
> I want to start using the solvers from NEOS and am wondering whether I
> should start writing in AMPL or GAMS.
> I have had some experience with LINGO and OPL-IDE but I am not a programmer,
> just an analyst.
> If I had to start developing skills with one, AMPL or GAMS, which one would
> you recommend, based on user-friendliness, potential, power, solver
> availability, more used, documentation, etc?
> Also, will the free version for any of the two let me compile/debug a
> problem of unlimited size (I can see they will only let me solve a
> limited-size problem?

Here is my (biased) point of view:

For the NEOS solvers:
Todd Munson was so nice to have put those COIN-OR solvers with
GAMS interfaces onto NEOS. I believe he will be updating them regularly,
but if the AMPL versions have been put onto NEOS by the solver
developers, then updates of these might come faster.

>From the solver-interface side, the difference between the AMPL and the
GAMS versions is mainly that the AMPL interfaces have been written by
the solver developers, while the GAMS interfaces are written by someone
else (ok, me). Thus, the AMPL interfaces might sometimes be better
adapted to the solver needs due to the insight of the developer, while
the GAMS interfaces also attempt to take all the GAMS features,
parameters (and pitfalls ;-)) into account (output is directed into gams
channels, gams standard parameters like timelimit, gaptolerance,
iterlimit work (somehow)). The latter might make it a bit more
comfortable to use GAMS than AMPL from the users perspective.

>From the modeling language point of view, both AMPL and GAMS have their
weaknesses and strengths... ;-)

Best,
Stefan