On 08/ 6/10 04:06 AM, Jason B Hill wrote:
Has sympow been tested on Itanium Linux much?
Itanium versions exist in the repositories for Debian (Lenny/Sid), Ubuntu
and Fedora. So, the short answer is "yes."
Thank you.
If I'm not mistaken, that script will include x86 based code for the
floating point processor (in the file fpu.c) if the system is linux running
Itanium. Would you agree? (I think ia64 = Itanium).
I agree, for the most part. (See below)
But Itanium CPUs are not x86 compatible.
It depends on which ia64. As I understand it, earlier Itaniums had a x86
mode that would theoretically support any x86 (meaning 32-bit at the time)
binary. They could even support a full x86-based OS, and interrupts, etc.,
at the time. That was done simply to make legacy support available, but was
later dropped because it was incredibly inefficient (it was basically a
register translator at the hardware level). Newer Itaniums use a software
emulation layer for the same sort of support, but I have absolutely no clue
as to how well this is known or used. As well, parts of the emulation layer
carry a proprietary Intel license... which may cast some serious doubts on
integration into various OSs.
In your opinion, is it better to leave the current behavior of including fpu.c
on Itanium Linux systems, or just remove that since it will be safer?
Bearing in ming the author's desire not to use the standard maths library for
copyright reasons (see below), I doubt he would personally have wanted to make
use of propriety Intel licenses!
II'll change the notice to indicate one does not have an x86 Linux system.
I think this is a good idea.
Yes, that's the easy bit. Actually getting this to run properly on Solaris seems
difficult. It's a very strange bit of code, on several counts. For example,
README says:
"SYMPOW also does not use the standard math libraries, thus eliminating possible
license conflicts".
The program behaves totally different if I change compiler optimisations.
SYMPOW's Configure script checks for the existance of some programs like rm,
grep, uname and other standard commands defined by POSIX. It does that at one
point using 'which' - a command not defined by POSIX. IMHO, It would have been
simpler to just assume 'rm', 'grep', uname and other similar programs just exist.
The C code is not the easiest to follow either.
Anyway, I assume SYMPOW plays a useful role in Sage, so I'll have to work out
why it will not play ball on Solaris x86.
Jason
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
