> Excellent. I'm all for more cooperation between CoCoaLib and Linbox,
> since both will be standard components in SAGE.
>
Ok, when Martin and I initially compared the mvpolynomial performance
of singular vs. CoCoALib singular was 3 times faster for something
like xyz*xyz, when squaring a polynomial with two monomials that
dropped down to a factor of 1.5. Now I have figured out how to avoid a
copy constructor and a couple other optimizations so that we can give
singular a run for its money ;). I will get back to the code after
2.5. The more interesting things in CoCoALib are the approximate
Buchberger Möller and Module/Ideal operations.
> > Regarding Linbox: The svn snapshot in 2.5.0alpha2 is about 6 weeks old
> > and juding from the svn log it seems to be a good idea to update.
>
> The last two times I tried the svn version was broken. The version
> included in SAGE was the result of Clement Pernet being at SAGE Days 3 and
> spending several days fixing bugs and build issues. As soon as linbox
> development
> moved forward past then, it became broken in various way fairly
> quickly. I reported
> all the issues I had, so maybe they are fixed now.
Ok, I tried 1.1.3r2701 on three different platforms and all tests
passed, Clement fixed a bug in 1.1.3r2699 that caused problems in test-
rank. Do you use your own testcases or how do you determine that there
were issues? If you have your own code you can send them to me and I
can take a look. I have been getting more and more familiar with the
code and I am currently testing 1.1.3r2701 with valgrind. My goal is
to fix all know issues that valgrind points out over the next months
or so that I can do the port of Linbox to MSVC 2005. I am reluctant to
port code to Windows when there are known issues because I prefer
debugging on Linux :)
>
> > Another taks I would suggest is also adding "make check" to the build
> > script because that would obviously greatly expand the testing base
> > for the Linbox code. This would increase the build time of that spkg
> > very much, but at least we can catch build issues and/or bugs much
>
> ^^^^^^^^^^^^^^
>
> Running make check will not be the *default* for builds, but I have been
> planning to introduce some form of this for a while. However, I intend
> to have optional scripts spkg-check in each sage package, that would
> be run after spkg-install succeeds and before the temp build directory
> is deleted. Then if one builds using
> "make buildcheck"
> SAGE would build each package and run spkg-check. If one builds
> with just plan "make" then none of the spkg-check's get run.
>
Great. Can we also add something that builds the packages without
optimizations and with debug info so that debugging would get easier?
Is there also a way to finish the compilation and installation of a
spkg that had a build failure? I am currently stuck on cygwin with
libsingular and I can build it manually and run spkg-install, but when
I restart the build process with "make" it starts rebuild singular
from scratch :(. If I get it installed manually I can move forward
while waiting on somebody else to fix the build.
> > earlier. There is also now two google groups for Linbox: linbox-use
> > and linbox-devel, so if you are interested please check them out.
>
> Excellent!
>
> > Another package which should also do a "make check" is gmp because it
> > is famous for being miscompiled, especially on MacOSX.
>
> > I can package those two spkgs with the suggested changes if William
> > would like me to. I am not sure if there are dedicated maintainers to
> > individual spkgs in SAGE at the moment.
>
> Best would be for you to send me an spkg-check script for each.
>
Will do. I am still working on an updated gdmodule that no longer uses
different Setup.py[.cygwin]
> William
Cheers,
Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@xxxxxxxxxxxxxxxx
To unsubscribe from this group, send email to
sage-devel-unsubscribe@xxxxxxxxxxxxxxxx
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/
-~----------~----~----~----~------~----~------~--~---
|
