CoCoA » CoCoALib

Short answer: Yes and Yes!

ex-Normaliz2-overflow.in is really large, the purpose was to give an example where overflow occurs. The smart CoCoALib cone detects the overflow and starts it with gmp again.
It is a single simplex with 10^9 points that have to be computed. And at the moment normaliz will first create all these vectors, which is too much for the memory. This is a weak point that we will attack in the future.

One solution could be to use a other computation type, SupportHyperplane instead of HilbertBasis should work. (Didn't tested it.)

I have attached a smaller example that also causes an overflow but should run without problems with gmp. But here the overflow could be avoided, so it might happen that future versions of normaliz do not cause an overflow.

The CoCoA example reads only the matrix from the file, everything else is ignored. I added the last line to run it with "native" normaliz.
The decision what to compute is also in the ex-Normaliz2.C, there you have to replace HilbertSeries by SupportHyperplanes in line 62.

Should I make a new example (ex-Normaliz3) which computes the support hyperplanes; then the "big" input can be renamed as ex-Normaliz3.in.

Or is there really no point in doing so?

Remember these are supposed to be clear examples of how to use features in CoCoALib; they are not intended to be thorough tests.

Since it is just to demonstrate the behaviour when overflow occurs, I think it is okay to simply change it to SupportHyperplanes and maybe keep both example inputs.
In ex-Normaliz1.C all the possible computations are shown.

With Normaliz 2.11 this will no longer be an issue of memory. But the computation time could still be too big.