CoCoA » CoCoALib

This student translated the code in CoCoA prototype-GBZZ.cpkg5 into cocoalib.
And also added some changed found in a paper by Christian Eder (arxiv:1811.05736 warning contains several typos)

Elisa and Michele: want to run some tests and contribute to make it usable?

John Abbott wrote:

A student (Khalil Loukhnati) worked on this last semester. He has produced working code.
It still needs considerable polishing.

Performance is disappointing (but the profiler reported that 50% CPU time was spent doing downcasting!!)

It seems that we have several implementations: some in C++, and some in CoCoA-5.
The C++ versions ought to be quicker (but perhaps not by much if arithmetic costs dominate).
I would like to compare Loukhnati's impl with that by the student Anna mentioned in comment 5.

One important point is that if a constant element of the ideal is found, it can be used to reduce all coeffs of all other polys (might need to be careful if the LC reduces to zero?). Loukhnati's code produced over 700 constants in one ideal: the smallest was only 5 digits, while other had many thousands of digits...

The CoCoA-5 version is in *prototype-GBZZ.cpkg5*.
I tried just one test, and the CoCoA-5 code was noticeably faster (maybe 100x). Must investigate!

John Abbott wrote:

The CoCoA-5 version is in *prototype-GBZZ.cpkg5*.
I tried just one test, and the CoCoA-5 code was noticeably faster (maybe 100x). Must investigate!

Elisa, Michele: are you going to write a paper for JSAG?

After that, or with that, we could write a proper translation in CoCoALib.

That is the idea. We just have to understand who will be the authors and what should be done before submitting. I believe the interested people are me, Michele, John and Florian (who is in contact with John).

Florian will finish his PhD this summer, and then start a real job. He says he does not need to be named as a co-author.
He is also willing to answer questions.

John Abbott wrote:

Florian will finish his PhD this summer, and then start a real job. He says he does not need to be named as a co-author.
He is also willing to answer questions.

Thanks, Florian!

I am not sure exatly how the algorithm works, but do believe that it is important to handle monomial reducers specially. In particular, if we have an element which is an integer then it can be used to reduce all coefficients. We may get a succession of ever smaller integers, so there may be several such reductions.