Hi Clement, On 13 Jul., 15:25, Clement Pernet <clement.per...@gmail.com> wrote: > ... > but the wrapping in sage is missing. So I think it might probably > be simpler to wrapp it rather > than creating a new spkg with MeatAxe.
I am not so sure if that would be easier FOR ME. If I understand correctly, finishing the LinBox wrapper requires C++ knowledge, which I don't really have. But I do have a wrapper for an old (but tuned) version of MeatAxe, and I do have a wrapper for a more recent (but untuned) version of MeatAxe. The former is part of my optional cohomology-spkg. Currently, it seems that in that way I would get most easily a good performance for the functions that I need: 1. Matrix times Matrix: MeatAxe is only slightly slower over prime fields than LinBox. Over non-prime fields, it is faster than a naive implementation of the Boothby-Bradshaw trick. 2. Matrix times column: Over prime fields, the native Sage implementation is clearly fastest, MeatAxe is slower, LinBox is VERY slow. Over non-prime fields, MeatAxe is fastest, the naive implementation of the "trick" is 5 times slower. 3. Row times matrix: MeatAxe is clearly fastest both over prime and over non-prime fields -- unless someone has already wrapped a specialised multiplication command of LinBox?? What I mainly need is 1. and 3., both over prime and over non-prime fields. It seems to me that extending my MeatAxe wrapper is (for me) easier than extending the LinBox wrapper, and gives reasonably good performance. I'll go for it. QUESTION: According to their web site, "The MeatAxe should run on most UNIX platforms including Linux, NetBSD, Ultrix, Solaris, HP-UX, and on Windows NT." So, a MeatAxe wrapper might actually be (eventually) a candidate for a standard package, isn't it? OTOH, the current MeatAxe licence is GPL 2. Does that disqualify MeatAxe? Would GPL 2+ be required? > PS: weird, my reply by email has been rejected and I'm now posting via > the web interface. I do that all the time... Cheers, Simon -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
