Dear all, i made a new version of my wrapper that is based on the most recent MeatAxe.
You can find it at http://sage.math.washington.edu/home/SimonKing/MeatAxe.tar.gz The tar'd folder contains: - a part of the unaltered source files of MeatAxe 2.4, - additional (very simplistic) C-code in MoreFeatures.c and .h, - my wrapper comprising mtx.pxd and mtx.pyx, - a Makefile, - and some files such as p005.zzz that you can ignore. On Sage.Math, you simply say "make" to build mtx.so, which can be loaded into Sage. On a different machine, you probably have to alter the hard-coded path to sage-python in the Makefile. Concerning p005.zzz and p007.zzz: Files like this will be written to your current directory when you handle MTX over Z/5 or Z/7. I don't like it, but that is how MeatAxe works. I made some timings on Sage.Math, that you find in http://sage.math.washington.edu/home/SimonKing/MTXtiming.txt This also provides some examples how MTX matrices work. On Mar 25, 5:02 pm, Simon King <[EMAIL PROTECTED]> wrote: > On Mar 25, 6:31 am, Clement Pernet <[EMAIL PROTECTED]> wrote: > > I still did not look at the code of Meat axe, but I remember having been > > really impressed a presentation at MSRI last year about MeatAxe. > > The timings were really impressive especially the matmul ones. > > So I am really surprised by your experience with slow matmul. Were these timings about matmul? Then it is really surprising. According to my timings, in fact matrix multiplication is slow: It was 453 ms with Sage matrices but 937 ms with MTX. Since MeatAxe's MatMul alters the first input matrix, i need one additional copy operation. But the copy works very fast, less than 1 ms. So, that can't be the reason. Surprisingly, MTX.nullspace(), which essentially yields what .kernel().basis() does for Sage matrices, is fast: 46.57 s with Sage matrices, but 4.74 s for the corresponding MTX matrix. I understood that Sage caches part of the result - MTX does not, yet. > But i don't need these executables, my wrapper does all these things > in memory. Only exception: I need the executable "maketab"... No, i don't. mtx.so suffices, no additional executable besides sage is needed. My opinion is: Personally, I have reasons to use MTX in my project. But the little progress (if there is any) would probably not justify to include yet another matrix type into Sage. Yours Simon --~--~---------~--~----~------------~-------~--~----~ 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://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
