Maxima 5.16.3 release --- enourmous slowdown of gf-package

On Mon, Aug 25, 2008 at 11:09:26PM -0600, Robert Dodier wrote:
> On 8/25/08, Oliver Kullmann <O.Kullmann at swansea.ac.uk> wrote:
> 
> >  5.16.3 passed now all my tests, but an enourmous
> >  slowdown happened with package gf, which renders
> >  it unusable:
> 
> >  So for example before a single Rijndael-sbox operation
> >  took 0.2 s (which was already quite slow), while it now takes
> >  around 18 s.
> 
> If you can give an example which demonstrates the slowdown,
> that would be very helpful.
>

Basically every operation. But let's take perhaps the basic
example gf_set; first with 5.15.0:

field : [2,8,x^8+x^4+x^3+x+1];
apply(gf_set,field);
Evaluation took 0.1280 seconds (0.1272 elapsed) using 840.867 KB.
(%o3) true

and now in 5.16.3:

apply(gf_set,field);
Evaluation took 4.7883 seconds (4.9496 elapsed) using 42.714 MB.
(%o58) true

I would guess the problem is that gf_set applies now the irreducibility
check:

GF_IRREDUCIBILITY_CHECK

should regulate this, and by default is false; however it seems dysfunctional.

There is another variable, which should regulate performance, largefield,
which is true by default, but setting it to false apparently breaks
the complete package (no operation works anymore).

There are quite a few other gf-operations which do not work, and always
one gets error messages

Use `fasttimes' only on CRE polynomials with same varlists

Oliver
 

 
> There were only a few minor changes in the gf code, if I'm not
> mistaken, so I'm optimistic that we'll be able to identify the
> cause of the problem.
> 
> Robert Dodier

-- 
Dr. Oliver Kullmann
Computer Science Department
Swansea University
Faraday Building, Singleton Park
Swansea SA2 8PP, UK
http://cs.swan.ac.uk/~csoliver/