New development branch: ia64-fp-model-branch

[Paolo Bonzini]

 > Implementation of #pragma _USE_SF will definitely require front-end
 > changes, as GCC currently has no support for lexically-scoped #pragma.
 > (This feature is desirable for C99 and OpenMP support as well.)
 > Assuming some mechanism for propagating the information all the way
 > through the tree optimizers, the backend needs only check the state of
 > the #pragma in expanders and choose the appropriate pattern.

This is slightly different from OpenMP, where all the lowering is done 
way before the backend are reached.

But anyway I'd like it if Joseph's parser were placed on a branch, and 
gomp-branch (maybe ia64-fp-model-branch as well) rebranched off there. 
I don't like the current situation where we get a 150k patch each week.

WRT this, Joseph wrote:

 > As regards the suggestion of working on gomp-branch, I don't want to
 > mix these changes with any not really required by them, as the aim is
 > for a minimal patch to apply to mainline; and before there is full
 > testsuite coverage with no regressions, it would destabilise
 > gomp-branch, but once there is full coverage with no regressions, it
 > is ready for mainline.

Subbranching would fix the first problem, and as to destabilising 
gomp-branch, I think subbranching makes it easier too (because you only 
merge when you have a small number of regressions or zero).

Paolo