Re: [Help-glpk] different results when complied by different version of gcc (fwd)

[Andrew Makhorin]

> Can anyone tell me what the problem is ? Look upto the end
> of the mail. Please..... say what could be the problem with gcc version 
> 3.0.4.

It is difficult to say what exactly happens. The simplex falls into
infinite loop, because a termination criterion is not satisfied
(this, unfortunately, may sometimes happen, because the implementation
is still not robust). In the latter case it looks like some differences
in the object code and inexact arithmetic caused the simplex to use
other path to the optimum, or basic solution components were computed
more accurately, or something else happened, due to which the criterion
was satisfied. In any case it was just a lucky chance that the search
was finished. Try not to use scaling (--noscale); this sometimes helps
to prevent cycling once the optimum has been reached.

Andrew Makhorin