Update i386 ULPs

[Joseph S. Myers]

On Mon, 5 Mar 2012, Andreas Jaeger wrote:

> Building for i686 on my x86-64 system with gcc -m32, I noticed a few 
> differences (all ULPs are long double and have a value of 1). Also the acos 
> test missed a summary line.
> 
> Ok to commit?

I think we should consider this sort of update (where the ULPs values 
aren't too big and probably result from differences in x87 implementation 
on different processors) obvious.

-- 
Joseph S. Myers
joseph@codesourcery.com