On Fri, Jul 16, 2010 at 3:30 AM, Dr. David Kirkby
<[email protected]> wrote:
> On 07/16/10 10:42 AM, Peter Jeremy wrote:
>>
>> On 2010-Jul-16 10:48:15 +0200, William Stein<[email protected]>  wrote:
>>>
>>> worked".  This wouldn't be likely with Microsoft's own compilers...
>>> For starters, the Singular developers don't work with those compilers,
>>> so the Singular devs will make all kinds of code changes, which would
>>> then have to be ported/fixed.
>>
>> As has been discussed previously, this effort is worthwhile in any
>> case.  Firstly, gcc does not generate especially good SPARCv9 or iA64
>> code - being able to use vendor compilers is likely to provide a boost
>> to Sage performance on those platforms.

The vast majority of people, by far, do not use SPARCv9 or iA64, and
though we shouldn't abandon support for them I'd say that getting a
Windows version up and running at all (which was the point of cephes
that started this thread) is vastly more important than optimized code
on rare platforms (though for those with different hardware and
special compilers, it's probably worth their time to try to make the
most of their resources, and we're certainly not throwing up any
roadblocks in that direction).

>> Secondly, the different checks
>> performed by different compilers are likely to turn up in Sage code and
>> correcting them will improve the reliability and robustness of Sage.
>>
>
> I agree with this very much. I think William see evidence of how the Sun
> compiler (which was being called by mistake), found two serious bugs in
> c_lib.
>
> http://trac.sagemath.org/sage_trac/ticket/6595
>
> Unfortunately, so much of the code in Sage is written in a GNU dialect of
> C/C++ and not C/C++, that building Sage with native compilers would be very
> difficult.

Yep. Apparently some coders find value in using the GNU dialects
rather than sticking to the standards. (Or it could be in part due to
ignorance, much of this code requires a deep math background rather
than a deep engineering background to write). This does not negate the
value of being able to compile on a variety of compilers, but we take
what we're given.

Currently, the (huge) problem is that there's no Windows port of Sage,
and since most upstreams components have better support for Cygwin
than the other Windows options (whether incidentally or by design).
Once we have a stable and maintainable Cygwin port those interested
can look at porting to other Windows platforms (including the
necessary upstream work).

- Robert

-- 
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to