[issue2819] Full precision summation

[Jean Brouwers]

Jean Brouwers <MrJean1 at Gmail.com> added the comment:

Attached is revision 11 of the mathmodule.c patch for Python 2.6a3.  

This one includes Raymond's full precision summation, Mark's rounding 
partials addition and correct non-finites and error handling.

However, intermediate overflow will raise an OverflowError and a 
FLT_RADIX not equal 2 cause a NotImplementedError.

An updated test_math_sum11.py script is also attached.  All test cases 
pass on 5 different builds of Python 2.6a3:

- 32-bit MacOS X 10.4.11 (Intel) with gcc 4.0.1
- 32-bit MacOS X 10.3.9 (PPC) with gcc 3.3
- 64-bit RHEL 3 update 7 (Opteron) using gcc 4.1.2
- 32- and 64-bit Solaris 10 (Opteron) with Sun C 5.8.

/Jean Brouwers

Added file: http://bugs.python.org/file10405/mathmodule11.c.2.6a3.diff

__________________________________
Tracker <report at bugs.python.org>
<http://bugs.python.org/issue2819>
__________________________________