Now would I know non-SSE hardware if I met it in the wild?

John

2008/5/2 mabshoff <[EMAIL PROTECTED]>:
>
>  On May 2, 8:06 pm, Jason Grout <[EMAIL PROTECTED]> wrote:
>  > William Stein wrote:
>
>  Hi,
>
>
>  > Steal the CPU detection code from ATLAS and use that to test to see if
>  > the processor we're running on is the same as the processor we compiled
>  > for?  That seems like a bit much.
>
>  You really don't want to do that - believe me, I have seen and
>  improved the ATLAS cpu detection code and it requires an assembler to
>  work. Other than that it is overblown and we can cook up something
>  better and simpler with a three line bash script ;)
>
>
>  > How about compiling a generic binary (i.e., minimal optimizations)?  Is
>  > that possible with ATLAS and the other programs?
>
>  Yes, but then something else will break. Depending on the compiler you
>  use it just uses SSE2 instructions unless you specifically tell the
>  compiler not to use it. And attempting to dix that via CFLAGS and
>  CPPFLAGS is not a good idea. Somebody needs to find some pre-SSE2
>  hardware and donate it to William so we can build a "last resort"
>  binary. Anything else will likely not work.
>
>  tseug from IRC did build Sage 3.0 on some Duron laptop and it took 22
>  hours, so building from source is generally not a good idea for people
>  with low end hardware, but since we cannot and will not likely provide
>  binaries for a wide range of distributions for non-SSE2 hardware due
>  to limited and usually slow hardware it is something we will have to
>  deal with for a while. Could we use a bunch on non-SSE2 Athlons with
>  decent, i.e. 1GB RAM, this would be doable.
>
>  > Jason
>
>  Thoughts?
>
>  Cheers,
>
>  Michael
>
>
> >
>

--~--~---------~--~----~------------~-------~--~----~
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
-~----------~----~----~----~------~----~------~--~---

Reply via email to