On Oct 13, 9:28 am, Christian Hilberg <hilb...@unix-ag.org> wrote:
> Hi,
>
> I'm experiencing instruction set issues with the SAGe Linux i686
> binary tarball from
>
> http://mirror.switch.ch/mirror/sagemath/linux/32bit/\
> sage-4.1.1-linux-Debian_GNU_Linux_5.0_lenny-i686-Linux.tar.gz
>
> I've checked on two different platforms which have an almost
> identical Debian/Lenny installation: an Athlon-XP <athlon> and a
> Pentium-M <pentium>. (Platform details: see below, if more
> information is needed I'll be happy to provide.)
>
> Compiling entirely from source worked for <athlon>, but I didn't try
> full source compilation on <pentium>.
>
> After extracting the binary tarball, startup of SAGe yields:
>
> <athlon>
> --------------------------------------------------------------------
> | Sage Version 4.1.1, Release Date: 2009-08-14 |
> | Type notebook() for the GUI, and license() for information. |
> --------------------------------------------------------------------
>
> **********************************************************************
> WARNING! This Sage install was built on a machine that supports
> instructions that are not available on this computer. Sage will
> likely fail with ILLEGAL INSTRUCTION errors! The following processor
> flags were on the build machine but are not on this computer:
>
> sse2
This issue comes up many times. I wonder if it would be preferable to
not include the sse2 instruction set. I don't know how much of an
impact on performance that might have. Wolfram Research had a similar
problem with the Solaris x86 version of Mathematica 6. It was only
supported on AMD, not Intel CPUs. I thought at first that was a bit of
a stupid decision to make, but I now believe they had not tested it on
Intel CPUs and only noticed the problem after the release.
Another option for Sage might be to compile the two bits of code twice
- one with SSE and once without. Then when Sage is run for the first
time, it automatically copies over the right binaries. It would
obviously make the binary distibution larger, but that might be a
small price to pay for not having the issue of producign a 'fat' and a
'think' build, and getting numerous reports of this error.
Dave
--~--~---------~--~----~------------~-------~--~----~
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
-~----------~----~----~----~------~----~------~--~---
