On Thu, Jul 16, 2009 at 6:20 AM, Bjarke Hammersholt
Roune<bjarke.ro...@gmail.com> wrote:
>
> Frobby is currently an optional component of Sage, which performs
> computations related to monomial ideals. In particular, it can compute
>
> * Multigraded Hilbert series
> * Alexander dual of monomial ideals
> * Maximal standard monomials of monomial ideals
> * Irreducible decomposition of monomial ideals
> * Optimization of any linear function over the maximal standard
> monomials of a monomial ideal using branch-and-bound.
>
> Sage currently is not able to do any of this. Sage does have the less
> general (1,..., 1)-graded Hilbert series using Singular, but it
> doesn't support arbitrarily large exponents as Frobby does.
>
> Applications of the last item above include Frobenius numbers for
> instances with very large numbers (with 4ti2), as demonstrated at Sage
> Days 16, and described in (1) below. Another application is the
> integer programming gap (also with 4ti2) of a matrix where the right-
> hand-side is allowed to vary as described in (2).
>
> This is put up for a vote now since I wrote a cython interface to
> Frobby at Sage Day 16, and I'm told this requires Frobby to be a
> standard component of Sage.
Who told you that? I don't agree with that at all. Just because
something has a cython interface doesn't mean it has be a standard
component of Sage. That would be a pretty sad limitation for Sage.
That said, I *do* think it is a good idea to considering getting
Frobby into standard Sage, simply because it provides much new
optimized functionality. That said -- I want to ask a question of
people who are voting +1 to this proposal: have you ever used Frobby's
capabilities? Do you expect to ever use them? Do you know people who
will?
> Frobby is fastest at items 2-4 listed above as documented in (3)
> below, by factors of up to 1000x, with the exception of specially-
> constructed inputs (in particular taking the dual of a dual to recover
> the original ideal). Item 1 is Hilbert series, where CoCoALib might be
> faster right now, since the algorithm I use is for now unpublished,
> and I haven't compared it to CoCoALib yet. In any case I will also
> implement the Bigatt et.al. algorithm that CoCoALib uses, though this
> is not done yet.
>
> Frobby has an extensive test-suite, which includes running Frobby
> under valgrind to detect memory leaks, and is supported for Mac OS
> 10.5, Linux and Cygwin. It compiles using MS Visual Studio Express,
> though I haven't tested it on that platform since I couldn't get GMP
> to build on Windows. GMP is the only dependency Frobby has other than
> a C++ compiler. The build system is make-based. I am the upstream
> contact, and Frobby is licensed as GPL version 2.0 or later.
It is very easy to build MPIR on Windows (especially if you do a
C-only build, i.e., no assembler).
Can you test building Frobby on Windows using MPIR instead of GMP?
http://mpir.org/
By the way, frobby takes about 30 seconds to build from source, so
doesn't add too much to the Sage build time.
-- William
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
