SSE fix 12 - argument alignment PR pending/8212

[Jan Hubicka]

> On Fri, Oct 18, 2002 at 11:40:30AM +0200, Jan Hubicka wrote:
> > I've attempted to play safe game and also added abort
> > when something gets different than 4 byte alignment and is not SSE.
> 
> Will fail with user-specified alignments.

I was thinking about that already.  How do we want to handle user
alignment?  Previously we simply ignored it.  Perhaps I can keep the
current situation just with removing the non-SSE check?
Then we will mandate it for structures and ignore for scalars that is
not very sane too...
I will try to think about this a bit.

Honza