Hi, I have a couple of suggestions for the benchmarks.
> I would suggest something along the following lines for the MVPoly > benchmark: > > Have a matrix of test cases: > > *number of indeterminates: > - small (3 indeterminates) > - medium (10 indeterminates) > - large (25 indeterminates) Are you talking about the number of variables in the ring or in a single polynomial here? In either case I think 25 is pretty small. At least for the applications I have in mind, a factor of 10 - 100 is considered reasonably large. > *length of polynomials > - small (3 monoids) > - medium (10 monoids) > - large (25 monoids) Again, I think 25 is pretty smart. Consider e.g. polynomials during a GB calculations, these can blow up significantly. However, you don't e.g. multiply those polynomials in a GB calculation anyway. > * Ring/Algebra > - Z (no very interesting - at least to me :)) > - Z2 (special case for certain fields like crypto research - I do > care about that) + the QuotientRing modulo the "field ideal". > - Zp small, i.e. p=32003 > - Zp huge i.e. p=something prime beyond 2^32 (not everybody has that > implemented, at least not efficiently) > - Q small > - Q large > - Weyl Algebras, non-commutative Algebras in general - rather exotic, > not present everywhere > * Term Ordering - not sure about the impact of that - this might > disappear: > - DegRevLex > - DegLex > - Lex > - Ordering Matrix > > Depending on how many options you select and after computing the > cartesian product of all selected options you end up with lots of > tests. To graph the result I really like what was done for FLINT by > plotting the cases in a plane with colored circles of different size > signifying who was faster by what factor. > > Operations: > - addition > - mutiply > - power small > - power huge > - GCD > > adding should be pretty boring, small powers are probably, too. Can > anybody else suggest more operations? How about operations on monomials like LCM? Also, S-polynomials (even though probably fairly boring) are relevant. If I had to choose a benchmarking suite, I would step through lets say F4 and take every monomial or polynomial operation and benchmark that ... well, we should probably step through some classical Buchberger to include reductions. Martin PS: Dear list, please let us know if you are not interested in this discussion anymore and we take it off list. -- name: Martin Albrecht _pgp: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x8EF0DC99 _www: http://www.informatik.uni-bremen.de/~malb _jab: [EMAIL PROTECTED] --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/ -~----------~----~----~----~------~----~------~--~---
