On Sun, Apr 18, 2010 at 10:51 PM, Thierry Dumont <tdum...@math.univ-lyon1.fr> wrote: > It is certainly possible to copy atlas libs from one machine to another > (and it is what you do when you install sage binaries, or when you install > atlas in a linux distribution (debian, ubuntu with binary packages), and > this will change nothing if both machines have the same processor. But it is > quite sure that the machines will have different processors and different > cache size: thus performances will not be optimal. The installation of atlas > consist in compilation plus determination of optimal block sizes for > matrix.matrix and matrix.vector products, by a set of tests. This is very > efficient (on my 3.1 GHZ machine, the product of 2 matrix(RDF,1000) takes > 0.2 second: this is 5 Gigaflops!). > yours > t.
John was specifically asking about building ATLAS once as part of Sage on a specific machine (t2.math.washington.edu), then making sure he never has to build ATLAS from source again on *that* same machine, when he builds new versions of Sage. I think ATLAS takes > 12 hours to build from source on t2.math.washington.edu. -- William > > Le 19/04/2010 07:24, William Stein a écrit : >> >> On Sun, Apr 18, 2010 at 10:08 PM, John H Palmieri >> <jhpalmier...@gmail.com> wrote: >>> >>> >>> On Apr 18, 2:55 pm, William Stein<wst...@gmail.com> wrote: >>>> >>>> On Sun, Apr 18, 2010 at 2:50 PM, John H Palmieri<jhpalmier...@gmail.com> >>>> wrote: >>>> >>>>> If I've built atlas once on a particular machine and if it took a long >>>>> time (e.g. t2.math, but also on various linux boxes), if I want to >>>>> build Sage again from scratch, are there just some files I can copy so >>>>> I can just touch spkg/installed/atlas... to skip it the next time? >>>> >>>> Yes, just do: >>>> >>>> 1. touch spkg/installed/atlas?? >>>> 2. cp the following files to local/lib: libatlas.* >>>> libcblas.* libf77blas* liblapack* >>> >>> I'm guessing I also need local/include/atlas/* and the files cblas.h >>> and clapack.h from local/include. Is that right? >> >> I didn't in my testing, but it certainly couldn't hurt. >> >>> >>>> It would be great if you could add this to the README.txt for Sage... >>> >>> If I have the time and if I can figure out a good place to put it, >>> I'll do it. >>> >>> -- >>> John >>> >>> >>> -- >>> 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 >>> >> >> >> > > -- > 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 > -- William Stein Professor of Mathematics University of Washington http://wstein.org -- 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
